| Home |
| Search |
| Today's Posts |
|
#1
|
|||
|
|||
Improved AM Detector
This one is in RDH4. I'm a bit surprised no one has referred to it,
although I've not watched the posts too carefully recently. I know several out there have a copy of RDH4. Please excuse if someone has already referenced this circuit. See it at ABSE & ABPR Cheers, John Stewart |
|
#2
|
|||
|
|||
|
John Stewart wrote: This one is in RDH4. I'm a bit surprised no one has referred to it, although I've not watched the posts too carefully recently. I know several out there have a copy of RDH4. Please excuse if someone has already referenced this circuit. See it at ABSE & ABPR Cheers, John Stewart What page in RDH4 has the improved detector? Many of us *do* have RDH4. I can't see anything about it at ABSE or ABPR. Patrick Turner. |
|
#3
|
|||
|
|||
|
Patrick Turner wrote: John Stewart wrote: This one is in RDH4. I'm a bit surprised no one has referred to it, although I've not watched the posts too carefully recently. I know several out there have a copy of RDH4. Please excuse if someone has already referenced this circuit. See it at ABSE & ABPR Cheers, John Stewart What page in RDH4 has the improved detector? Many of us *do* have RDH4. I can't see anything about it at ABSE or ABPR. Patrick Turner. It shows up on page 1074 of my copy of RDH4. This is an original I bought around 1956 while I worked at the research division of Ferranti Electric. The publish date is February 1954. Cheers, John Stewart |
|
#4
|
|||
|
|||
|
John Stewart wrote: Patrick Turner wrote: John Stewart wrote: This one is in RDH4. I'm a bit surprised no one has referred to it, although I've not watched the posts too carefully recently. I know several out there have a copy of RDH4. Please excuse if someone has already referenced this circuit. See it at ABSE & ABPR Cheers, John Stewart What page in RDH4 has the improved detector? Many of us *do* have RDH4. I can't see anything about it at ABSE or ABPR. Patrick Turner. It shows up on page 1074 of my copy of RDH4. This is an original I bought around 1956 while I worked at the research division of Ferranti Electric. The publish date is February 1954. Cheers, John Stewart Yes, I see that schematic OK in the Book. The 6SJ7 driver tube ahead of a 6V6 output has NFB applied from 6V6 anode to 6SJ7 cathode via a 150k and 1 k divider, so the SJ7 cathode signal is nearly the same as the SJ7 grid signal, which comes from the wiper on a volume control pot which is the current sinking R from the detector caps. The grid of the 6SJ7 is biased from the 1k cathode R, so the input resistance into the 6SJ7 is perhaps at least several times 1M, which thus presents a high AC coupled load. I might add that in real circuits, the value of R2 is quite critical for lowest thd, and it should be estabished experimentally for lowest thd; too high a value will have terrible cut off distortion on the positive peaks of the audio, and too low a value will dissallow AVC voltage to be made, allowing too much IF amp current, and there will be terrible distortions. But this circuit still has the diode of the detector powered via a vigh impedance circuit of the secondary of the IFT, and its still not a best possible outcome. Better of course would be to have a CF tube to accept the IF envelope, and the low impedance output from the CF can then power a crystal diode, or a tube diode in a variety of ways I have previously explained in post on the matter. Patrick Turner. |
|
#5
|
|||
|
|||
|
In article , Patrick Turner
wrote: I might add that in real circuits, the value of R2 is quite critical for lowest thd, and it should be estabished experimentally for lowest thd; too high a value will have terrible cut off distortion on the positive peaks of the audio, "Positive peaks of the audio"? Don't you mean negative modulation peaks of the audio? But this circuit still has the diode of the detector powered via a vigh impedance circuit of the secondary of the IFT, and its still not a best possible outcome. Better of course would be to have a CF tube to accept the IF envelope, and the low impedance output from the CF can then power a crystal diode, or a tube diode in a variety of ways I have previously explained in post on the matter. You still haven't explained how this added cathode follower, to drive the detector, helps matters? Many experts even make the claim that a finite source resistance can be beneficial in reducing distortion, especially high frequency distortion. I can see where a cathode follower could be beneficial if we were trying to build a radio with an IF as narrow as possible, in which case it would help keep the Q of the transformer secondary as high as possible, but we are talking about a radio with wide band audio, and are probably talking about adding loading resistors across the transformers anyway, so why the cathode follower, why not just let the load of the detector diode do the job? A cathode follower after the IFT seems like a waste to me, better to use it after the detector, with a negative cathode supply voltage, to buffer the detector from the AGC and audio lines. While many AM receivers have been designed in a cost conscious way, there have also been a few where no expense was spared, and parts were freely used, and yet I have never seen a cathode follower used as you propose in a commercial design, I would think if it were beneficial someone would have used it commercially, anyone know of any examples? Some designs add other relatively expensive parts to the detector circuit, one trick I have seen whose effects might be worth looking into is replacing the second capacitor in the peak detector & RF filter network with a series LC network tuned to 455 kHz. I don't know how this circuit actually works, but I assume that the idea is to improve the tradeoff of the total peak detector capacitance vs. tangential clipping at high frequencies. This is something I will have to look into further. There are other detector circuit subtleties like this that may, or may not, be worth while. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#6
|
|||
|
|||
|
John Byrns wrote: In article , Patrick Turner wrote: I might add that in real circuits, the value of R2 is quite critical for lowest thd, and it should be estabished experimentally for lowest thd; too high a value will have terrible cut off distortion on the positive peaks of the audio, "Positive peaks of the audio"? Don't you mean negative modulation peaks of the audio? But this circuit still has the diode of the detector powered via a vigh impedance circuit of the secondary of the IFT, and its still not a best possible outcome. Better of course would be to have a CF tube to accept the IF envelope, and the low impedance output from the CF can then power a crystal diode, or a tube diode in a variety of ways I have previously explained in post on the matter. You still haven't explained how this added cathode follower, to drive the detector, helps matters? Many experts even make the claim that a finite source resistance can be beneficial in reducing distortion, especially high frequency distortion. IN my case there *is* a finite source resistance which is the 100k R across each IFT winding. I can see where a cathode follower could be beneficial if we were trying to build a radio with an IF as narrow as possible, in which case it would help keep the Q of the transformer secondary as high as possible, but we are talking about a radio with wide band audio, and are probably talking about adding loading resistors across the transformers anyway, so why the cathode follower, why not just let the load of the detector diode do the job? A cathode follower after the IFT seems like a waste to me, better to use it after the detector, with a negative cathode supply voltage, to buffer the detector from the AGC and audio lines. I do things to suit the desire for wide as possible AF bw, and the R loading of the IFTs helps achieve that end. I don't want severe selectivity and IFT gain; that only belongs in Z grade AM radios and communications sets. Try using a CF buffer to power a detector with a germanium diode, you'll hear the difference! Measurements will confirm the improvement. While many AM receivers have been designed in a cost conscious way, there have also been a few where no expense was spared, and parts were freely used, and yet I have never seen a cathode follower used as you propose in a commercial design, I would think if it were beneficial someone would have used it commercially, anyone know of any examples? I have NEVER seen any ancient commercially produced radio or audio product where the sound quality was not compromised, often severely, with many lies told by the market cowboys, after the maker had reduced the parts count to reduce costs, to be able to compete. Some designs add other relatively expensive parts to the detector circuit, one trick I have seen whose effects might be worth looking into is replacing the second capacitor in the peak detector & RF filter network with a series LC network tuned to 455 kHz. I don't know how this circuit actually works, but I assume that the idea is to improve the tradeoff of the total peak detector capacitance vs. tangential clipping at high frequencies. A series LC tuned to 455 kHz needs to be driven by a low impedance to get a decent Q to reject the 455 kHz ripple, but it simply is far easier to achive in well known ways with R&C. Usually, the CRC arrangement of 100pF, 47k, and 100pF is entirely adequate for removing RF detector ripple voltage. This is something I will have to look into further. There are other detector circuit subtleties like this that may, or may not, be worth while. You need to use a soldering iron to find out about what I am promoting about AM detection. There is no other way. Patrick Turner. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#7
|
|||
|
|||
|
John Stewart wrote:
This one is in RDH4. Saw it. In a radio where I already split off the AVC function off the audio detector, I'm trying a fixed positive bias of 300mV on the "bottom" of the volume control. Simulations indicate lower distortion on many various levels of carrier, and it seems borne out in an actual set. But the only cap I'm working against is the coupling cap between the volume control and the 12AV6 grid. Which is essentially a fixed value. I'll give it some time and thought to be sure that this is in fact a good thing or not.... |
|
#8
|
|||
|
|||
|
In article , Patrick Turner
wrote: John Byrns wrote: In article , Patrick Turner wrote: Better of course would be to have a CF tube to accept the IF envelope, and the low impedance output from the CF can then power a crystal diode, or a tube diode in a variety of ways I have previously explained in post on the matter. You still haven't explained how this added cathode follower, to drive the detector, helps matters? Many experts even make the claim that a finite source resistance can be beneficial in reducing distortion, especially high frequency distortion. IN my case there *is* a finite source resistance which is the 100k R across each IFT winding. Yes, but the diode is driven from the low source impedance of the cathode follower, not something on the order of 100k. Why not eliminate the cathode follower and choose the diode detector load so that it looks like 100k to the IFT at 455 kHz? I can see where a cathode follower could be beneficial if we were trying to build a radio with an IF as narrow as possible, in which case it would help keep the Q of the transformer secondary as high as possible, but we are talking about a radio with wide band audio, and are probably talking about adding loading resistors across the transformers anyway, so why the cathode follower, why not just let the load of the detector diode do the job? A cathode follower after the IFT seems like a waste to me, better to use it after the detector, with a negative cathode supply voltage, to buffer the detector from the AGC and audio lines. I do things to suit the desire for wide as possible AF bw, and the R loading of the IFTs helps achieve that end. I don't want severe selectivity and IFT gain; that only belongs in Z grade AM radios and communications sets. Yes of course, I was simply trying to point out one situation where a cathode follower driving the diode might be useful. Try using a CF buffer to power a detector with a germanium diode, you'll hear the difference! Like others, yourself included, I am just prejudiced against some ideas, and a cathode follower between the IFT and detector is just something that I have little intention of trying. Measurements will confirm the improvement. Or they may only confirm that the cathode follower helps with your detector design for some as yet unexplained reason, but doesn't help in the general case, see my comments further along. Have you measured identical detectors with and without the cathode follower? The same diode driven at the same level, with the same DC bias applied, and the same total load reflected to the IFT secondary at 455 kHz? While many AM receivers have been designed in a cost conscious way, there have also been a few where no expense was spared, and parts were freely used, and yet I have never seen a cathode follower used as you propose in a commercial design, I would think if it were beneficial someone would have used it commercially, anyone know of any examples? I have NEVER seen any ancient commercially produced radio or audio product where the sound quality was not compromised, often severely, with many lies told by the market cowboys, after the maker had reduced the parts count to reduce costs, to be able to compete. There were certainly commercially produced AM tuners where the maker didn't reduce the parts count at all in order to reduce the costs, the sound quality may or may not have been compromised by your standards, but if it was, it was due to a poor use of the parts, rather than to a lowered parts count. There are certainly commercial AM tuner designs that use significantly more parts than your tuner uses. Some designs add other relatively expensive parts to the detector circuit, one trick I have seen whose effects might be worth looking into is replacing the second capacitor in the peak detector & RF filter network with a series LC network tuned to 455 kHz. I don't know how this circuit actually works, but I assume that the idea is to improve the tradeoff of the total peak detector capacitance vs. tangential clipping at high frequencies. A series LC tuned to 455 kHz needs to be driven by a low impedance to get a decent Q to reject the 455 kHz ripple, This is patent nonsense, if anything just the opposite is true, if a series LC to ground were driven by a very low impedance it would have virtually no effect. but it simply is far easier to achive in well known ways with R&C. Usually, the CRC arrangement of 100pF, 47k, and 100pF is entirely adequate for removing RF detector ripple voltage. Just goes to show that those manufacturers that included the series LC weren't among the ones you are speaking of that reduced the parts count wherever they could. This is something I will have to look into further. There are other detector circuit subtleties like this that may, or may not, be worth while. You need to use a soldering iron to find out about what I am promoting about AM detection. There is no other way. Actually I am fairly certain that there are several other ways, like me, you are simply prejudiced towards your own ways. I think I finally understand what you are doing with the bias on your detector, and how it works. I believe I have misunderstood what your biased detector was all about and you haven't explained it. You have made statements like "This method means that detection of weak signal lower than the forward voltage of the Ge diode of 0.27v peak approx are not subject to the non linear turn on of the diode, ie, there is no clipping by the diode." This lead me to believe that you were using the bias to somehow "linearize" the diode, which I didn't understand. The figure from the RDH4 which John Stewart posted finally made me realize that you were doing exactly the same thing with your bias as the RDH4 figure, except that you used a fixed bias and left out the tracking feature. The distortion reduction you claim makes sense in that context, because your receiver as described by the schematic you posted has an extremely poor AC/DC load ratio and I am sure the distortion is extreme without the bias. Your bias scheme presumably partially compensates for the poor AC/DC load ratio, rather than somehow improving the "non linear turn on of the diode" as I had erroneously assumed from what you have said. Now the only remaining question is, does the apparently pointless cathode follower driving the diode also compensate in some way for another unnoticed design flaw? Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#9
|
|||
|
|||
|
John Byrns wrote: In article , Patrick Turner wrote: John Byrns wrote: In article , Patrick Turner wrote: Better of course would be to have a CF tube to accept the IF envelope, and the low impedance output from the CF can then power a crystal diode, or a tube diode in a variety of ways I have previously explained in post on the matter. You still haven't explained how this added cathode follower, to drive the detector, helps matters? Many experts even make the claim that a finite source resistance can be beneficial in reducing distortion, especially high frequency distortion. IN my case there *is* a finite source resistance which is the 100k R across each IFT winding. Yes, but the diode is driven from the low source impedance of the cathode follower, not something on the order of 100k. Why not eliminate the cathode follower and choose the diode detector load so that it looks like 100k to the IFT at 455 kHz? Because a CF does it better. With 100k source impedance, one still gets distortion to the audio signal where the diodes conduct to charge the 100 pF C1 of thre CRC detector filter. I prefer to go the extra country mile at every part of the audio chain, and it all adds up to a good total outcome, rather than one which is medicocre along the way, which then adds up to a poor total outcome. I will NEVER build anything because " we only ever did it that way before ..." or any other equally gutless and stupid reason. I can see where a cathode follower could be beneficial if we were trying to build a radio with an IF as narrow as possible, in which case it would help keep the Q of the transformer secondary as high as possible, but we are talking about a radio with wide band audio, and are probably talking about adding loading resistors across the transformers anyway, so why the cathode follower, why not just let the load of the detector diode do the job? A cathode follower after the IFT seems like a waste to me, better to use it after the detector, with a negative cathode supply voltage, to buffer the detector from the AGC and audio lines. I do things to suit the desire for wide as possible AF bw, and the R loading of the IFTs helps achieve that end. I don't want severe selectivity and IFT gain; that only belongs in Z grade AM radios and communications sets. Yes of course, I was simply trying to point out one situation where a cathode follower driving the diode might be useful. Try using a CF buffer to power a detector with a germanium diode, you'll hear the difference! Like others, yourself included, I am just prejudiced against some ideas, and a cathode follower between the IFT and detector is just something that I have little intention of trying. Try something different, and try abandoning your prejudice, just for an hour it takes to make something; maybe you hear something good. Measurements will confirm the improvement. Or they may only confirm that the cathode follower helps with your detector design for some as yet unexplained reason, but doesn't help in the general case, see my comments further along. Have you measured identical detectors with and without the cathode follower? I don't need to make the comparisons with measurements. Its so plain obvious with CRO experiments and dual trace tracing of the envelope shape against the detected signal. The same diode driven at the same level, with the same DC bias applied, and the same total load reflected to the IFT secondary at 455 kHz? While many AM receivers have been designed in a cost conscious way, there have also been a few where no expense was spared, and parts were freely used, and yet I have never seen a cathode follower used as you propose in a commercial design, I would think if it were beneficial someone would have used it commercially, anyone know of any examples? I have NEVER seen any ancient commercially produced radio or audio product where the sound quality was not compromised, often severely, with many lies told by the market cowboys, after the maker had reduced the parts count to reduce costs, to be able to compete. There were certainly commercially produced AM tuners where the maker didn't reduce the parts count at all in order to reduce the costs, the sound quality may or may not have been compromised by your standards, but if it was, it was due to a poor use of the parts, rather than to a lowered parts count. There are certainly commercial AM tuner designs that use significantly more parts than your tuner uses. 99% of surviving old radios are crap, and were crap when they were foisted onto a gullible public. I have never seen anything here which remotely was above the lowest common denominator, except perhaps the Quad AM tuner I aquired. But the one I built measures and sounds better on the BCB. Sure some AM tuners use more parts than I have, and I recently posted a few SS design schematics at ABSE and ABPR to let folks know there is more to AM reception than RDH4 ideas and a few tubes. Some designs add other relatively expensive parts to the detector circuit, one trick I have seen whose effects might be worth looking into is replacing the second capacitor in the peak detector & RF filter network with a series LC network tuned to 455 kHz. I don't know how this circuit actually works, but I assume that the idea is to improve the tradeoff of the total peak detector capacitance vs. tangential clipping at high frequencies. A series LC tuned to 455 kHz needs to be driven by a low impedance to get a decent Q to reject the 455 kHz ripple, This is patent nonsense, if anything just the opposite is true, if a series LC to ground were driven by a very low impedance it would have virtually no effect. I suggested low impedance, not zero ohms impedance. You would find that with a high source impedance that the attenuation curve is too broad, and as you reduce Rg, the attenuation of the series LC becomes sharper and deeper, but below a certain Rg the attenuation is restricted. I have tried this sort of idea for a 9 kHz whistle filter for better DX listening. I found the shape of the null, or the Q of the null was very variable with Rg, and a CF with a series R was best to find the right sort of null. R can't be too small, lest the filter overload the tube. But a broad null was needed to not only reduce the 9 kHz whistle but also reduce the monkey chatter a bit. Nothing was really effective 100%. Another type of much better null filter is a bridged LC type, with two caps and an L, and an R to 0V from the CT between the two caps, and this is less dependant on any critical value of Rg, although for a deep null, the two C have to be matched, and the R value to 0V is very critical. But in the case of detectors, there in no need for null filters or traps, and in any case, if you wanted good rejection of 455 kHz, with LC, you'd use a critically damped low pass filter using LC which would not only reject 455 kHz very adequately, but any other RF noise. but it simply is far easier to achive in well known ways with R&C. Usually, the CRC arrangement of 100pF, 47k, and 100pF is entirely adequate for removing RF detector ripple voltage. Just goes to show that those manufacturers that included the series LC weren't among the ones you are speaking of that reduced the parts count wherever they could. Depends how they used the parts which were not "rationalised out of the package to save $$" The 47k could be replaced with a choke in the CRC filter, but the CRC does enough to reduce the 455 kHz ripple to such low levels that any remaining ripple at say -40 dB will not cause a bother in the audio amp. And the audio amp will have almost no gain at 455 kHz. This is something I will have to look into further. There are other detector circuit subtleties like this that may, or may not, be worth while. You need to use a soldering iron to find out about what I am promoting about AM detection. There is no other way. Actually I am fairly certain that there are several other ways, like me, you are simply prejudiced towards your own ways. I prefer my own methods, and I welcome anyone to post a schematic which works better. I think I finally understand what you are doing with the bias on your detector, and how it works. I believe I have misunderstood what your biased detector was all about and you haven't explained it. You have made statements like "This method means that detection of weak signal lower than the forward voltage of the Ge diode of 0.27v peak approx are not subject to the non linear turn on of the diode, ie, there is no clipping by the diode." This lead me to believe that you were using the bias to somehow "linearize" the diode, which I didn't understand. In a tubed diode detector used so often, the value of ripple voltage varies a lot between the bottom of a detected sine wave to the top of the crests of the wave, and its a distortion mechanism, worst when modulation % is high, which it is, on most transmitted AM signals these days; since efficient use of a carrier is wanted. In my biased detector, the variation in ripple voltage along all parts of the audio wave form is very nearly the same amplitude. The figure from the RDH4 which John Stewart posted finally made me realize that you were doing exactly the same thing with your bias as the RDH4 figure, except that you used a fixed bias and left out the tracking feature. My CF plus following Ge diode is totally different to anything in RDH4. The distortion reduction you claim makes sense in that context, because your receiver as described by the schematic you posted has an extremely poor AC/DC load ratio and I am sure the distortion is extreme without the bias. Your bias scheme presumably partially compensates for the poor AC/DC load ratio, rather than somehow improving the "non linear turn on of the diode" as I had erroneously assumed from what you have said. The diode detector schematic I did post does have a poor AC/DC load ratio, but still works fine to make a few volts without any wave clipping or added distortion because of the ratio. Even better results with capacity for a much higher undistorted output voltage is possible with the same first CF and Ge diode and CRC filter, but then directly coupled to a second CF, which is the other half of a twin triode, and behold, there is zero AC loading on the detector. If the load resistances of the CF are exchanged for transistor CCS the bulk of the small thd caused by the CF transfer character is also reduced to perhap 0.02% at 10v of audio output, depending on the triode used. 6DJ8 would be ideal... 12AU7 is OK, 6AQ8, ECC85, 6201, 12AT7 are probably better, since the gain reduction with 12AT7 is around say 35 times for a given load R, so if the open loop thd was 1%, its reduced to 1/35 % by the NFB effect of the follower. A CCS reduces the thd at least several times more. Now the only remaining question is, does the apparently pointless cathode follower driving the diode also compensate in some way for another unnoticed design flaw? I give up. If you cannot see the benefits of intelligent buffering, and you won't try an idea before roundly condemning it, then it becomes pointless for me to provide any more justification than I already have. Patrick Turner. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#10
|
|||
|
|||
|
Just curious here. I've been following the debate. I have a radio I built
that uses a separate AVC amplifier to separate the AVC function from the detector, and give me some control over AVC action. I used an infinite impedance detector so as to not load the last IF secondary. So I pretty much have the cathode follower and would only need to add a diode to convert to the discussed detector. would there be any noticeable improvement over the infinite impedance detector? thanks ------ Bob La Rocca Lindenhurst, NY "Patrick Turner" wrote in message ... John Byrns wrote: In article , Patrick Turner wrote: John Byrns wrote: In article , Patrick Turner wrote: Better of course would be to have a CF tube to accept the IF envelope, and the low impedance output from the CF can then power a crystal diode, or a tube diode in a variety of ways I have previously explained in post on the matter. You still haven't explained how this added cathode follower, to drive the detector, helps matters? Many experts even make the claim that a finite source resistance can be beneficial in reducing distortion, especially high frequency distortion. IN my case there *is* a finite source resistance which is the 100k R across each IFT winding. Yes, but the diode is driven from the low source impedance of the cathode follower, not something on the order of 100k. Why not eliminate the cathode follower and choose the diode detector load so that it looks like 100k to the IFT at 455 kHz? Because a CF does it better. With 100k source impedance, one still gets distortion to the audio signal where the diodes conduct to charge the 100 pF C1 of thre CRC detector filter. I prefer to go the extra country mile at every part of the audio chain, and it all adds up to a good total outcome, rather than one which is medicocre along the way, which then adds up to a poor total outcome. I will NEVER build anything because " we only ever did it that way before ...." or any other equally gutless and stupid reason. I can see where a cathode follower could be beneficial if we were trying to build a radio with an IF as narrow as possible, in which case it would help keep the Q of the transformer secondary as high as possible, but we are talking about a radio with wide band audio, and are probably talking about adding loading resistors across the transformers anyway, so why the cathode follower, why not just let the load of the detector diode do the job? A cathode follower after the IFT seems like a waste to me, better to use it after the detector, with a negative cathode supply voltage, to buffer the detector from the AGC and audio lines. I do things to suit the desire for wide as possible AF bw, and the R loading of the IFTs helps achieve that end. I don't want severe selectivity and IFT gain; that only belongs in Z grade AM radios and communications sets. Yes of course, I was simply trying to point out one situation where a cathode follower driving the diode might be useful. Try using a CF buffer to power a detector with a germanium diode, you'll hear the difference! Like others, yourself included, I am just prejudiced against some ideas, and a cathode follower between the IFT and detector is just something that I have little intention of trying. Try something different, and try abandoning your prejudice, just for an hour it takes to make something; maybe you hear something good. Measurements will confirm the improvement. Or they may only confirm that the cathode follower helps with your detector design for some as yet unexplained reason, but doesn't help in the general case, see my comments further along. Have you measured identical detectors with and without the cathode follower? I don't need to make the comparisons with measurements. Its so plain obvious with CRO experiments and dual trace tracing of the envelope shape against the detected signal. The same diode driven at the same level, with the same DC bias applied, and the same total load reflected to the IFT secondary at 455 kHz? While many AM receivers have been designed in a cost conscious way, there have also been a few where no expense was spared, and parts were freely used, and yet I have never seen a cathode follower used as you propose in a commercial design, I would think if it were beneficial someone would have used it commercially, anyone know of any examples? I have NEVER seen any ancient commercially produced radio or audio product where the sound quality was not compromised, often severely, with many lies told by the market cowboys, after the maker had reduced the parts count to reduce costs, to be able to compete. There were certainly commercially produced AM tuners where the maker didn't reduce the parts count at all in order to reduce the costs, the sound quality may or may not have been compromised by your standards, but if it was, it was due to a poor use of the parts, rather than to a lowered parts count. There are certainly commercial AM tuner designs that use significantly more parts than your tuner uses. 99% of surviving old radios are crap, and were crap when they were foisted onto a gullible public. I have never seen anything here which remotely was above the lowest common denominator, except perhaps the Quad AM tuner I aquired. But the one I built measures and sounds better on the BCB. Sure some AM tuners use more parts than I have, and I recently posted a few SS design schematics at ABSE and ABPR to let folks know there is more to AM reception than RDH4 ideas and a few tubes. Some designs add other relatively expensive parts to the detector circuit, one trick I have seen whose effects might be worth looking into is replacing the second capacitor in the peak detector & RF filter network with a series LC network tuned to 455 kHz. I don't know how this circuit actually works, but I assume that the idea is to improve the tradeoff of the total peak detector capacitance vs. tangential clipping at high frequencies. A series LC tuned to 455 kHz needs to be driven by a low impedance to get a decent Q to reject the 455 kHz ripple, This is patent nonsense, if anything just the opposite is true, if a series LC to ground were driven by a very low impedance it would have virtually no effect. I suggested low impedance, not zero ohms impedance. You would find that with a high source impedance that the attenuation curve is too broad, and as you reduce Rg, the attenuation of the series LC becomes sharper and deeper, but below a certain Rg the attenuation is restricted. I have tried this sort of idea for a 9 kHz whistle filter for better DX listening. I found the shape of the null, or the Q of the null was very variable with Rg, and a CF with a series R was best to find the right sort of null. R can't be too small, lest the filter overload the tube. But a broad null was needed to not only reduce the 9 kHz whistle but also reduce the monkey chatter a bit. Nothing was really effective 100%. Another type of much better null filter is a bridged LC type, with two caps and an L, and an R to 0V from the CT between the two caps, and this is less dependant on any critical value of Rg, although for a deep null, the two C have to be matched, and the R value to 0V is very critical. But in the case of detectors, there in no need for null filters or traps, and in any case, if you wanted good rejection of 455 kHz, with LC, you'd use a critically damped low pass filter using LC which would not only reject 455 kHz very adequately, but any other RF noise. but it simply is far easier to achive in well known ways with R&C. Usually, the CRC arrangement of 100pF, 47k, and 100pF is entirely adequate for removing RF detector ripple voltage. Just goes to show that those manufacturers that included the series LC weren't among the ones you are speaking of that reduced the parts count wherever they could. Depends how they used the parts which were not "rationalised out of the package to save $$" The 47k could be replaced with a choke in the CRC filter, but the CRC does enough to reduce the 455 kHz ripple to such low levels that any remaining ripple at say -40 dB will not cause a bother in the audio amp. And the audio amp will have almost no gain at 455 kHz. This is something I will have to look into further. There are other detector circuit subtleties like this that may, or may not, be worth while. You need to use a soldering iron to find out about what I am promoting about AM detection. There is no other way. Actually I am fairly certain that there are several other ways, like me, you are simply prejudiced towards your own ways. I prefer my own methods, and I welcome anyone to post a schematic which works better. I think I finally understand what you are doing with the bias on your detector, and how it works. I believe I have misunderstood what your biased detector was all about and you haven't explained it. You have made statements like "This method means that detection of weak signal lower than the forward voltage of the Ge diode of 0.27v peak approx are not subject to the non linear turn on of the diode, ie, there is no clipping by the diode." This lead me to believe that you were using the bias to somehow "linearize" the diode, which I didn't understand. In a tubed diode detector used so often, the value of ripple voltage varies a lot between the bottom of a detected sine wave to the top of the crests of the wave, and its a distortion mechanism, worst when modulation % is high, which it is, on most transmitted AM signals these days; since efficient use of a carrier is wanted. In my biased detector, the variation in ripple voltage along all parts of the audio wave form is very nearly the same amplitude. The figure from the RDH4 which John Stewart posted finally made me realize that you were doing exactly the same thing with your bias as the RDH4 figure, except that you used a fixed bias and left out the tracking feature. My CF plus following Ge diode is totally different to anything in RDH4. The distortion reduction you claim makes sense in that context, because your receiver as described by the schematic you posted has an extremely poor AC/DC load ratio and I am sure the distortion is extreme without the bias. Your bias scheme presumably partially compensates for the poor AC/DC load ratio, rather than somehow improving the "non linear turn on of the diode" as I had erroneously assumed from what you have said. The diode detector schematic I did post does have a poor AC/DC load ratio, but still works fine to make a few volts without any wave clipping or added distortion because of the ratio. Even better results with capacity for a much higher undistorted output voltage is possible with the same first CF and Ge diode and CRC filter, but then directly coupled to a second CF, which is the other half of a twin triode, and behold, there is zero AC loading on the detector. If the load resistances of the CF are exchanged for transistor CCS the bulk of the small thd caused by the CF transfer character is also reduced to perhap 0.02% at 10v of audio output, depending on the triode used. 6DJ8 would be ideal... 12AU7 is OK, 6AQ8, ECC85, 6201, 12AT7 are probably better, since the gain reduction with 12AT7 is around say 35 times for a given load R, so if the open loop thd was 1%, its reduced to 1/35 % by the NFB effect of the follower. A CCS reduces the thd at least several times more. Now the only remaining question is, does the apparently pointless cathode follower driving the diode also compensate in some way for another unnoticed design flaw? I give up. If you cannot see the benefits of intelligent buffering, and you won't try an idea before roundly condemning it, then it becomes pointless for me to provide any more justification than I already have. Patrick Turner. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#11
|
|||
|
|||
|
BTW,
I think one of the original posts had a schematic. Can you repost it? Thanks, ------ Bob La Rocca Lindenhurst, NY "Bob" wrote in message t... Just curious here. I've been following the debate. I have a radio I built that uses a separate AVC amplifier to separate the AVC function from the detector, and give me some control over AVC action. I used an infinite impedance detector so as to not load the last IF secondary. So I pretty much have the cathode follower and would only need to add a diode to convert to the discussed detector. would there be any noticeable improvement over the infinite impedance detector? thanks ------ Bob La Rocca Lindenhurst, NY "Patrick Turner" wrote in message ... John Byrns wrote: In article , Patrick Turner wrote: John Byrns wrote: In article , Patrick Turner wrote: Better of course would be to have a CF tube to accept the IF envelope, and the low impedance output from the CF can then power a crystal diode, or a tube diode in a variety of ways I have previously explained in post on the matter. You still haven't explained how this added cathode follower, to drive the detector, helps matters? Many experts even make the claim that a finite source resistance can be beneficial in reducing distortion, especially high frequency distortion. IN my case there *is* a finite source resistance which is the 100k R across each IFT winding. Yes, but the diode is driven from the low source impedance of the cathode follower, not something on the order of 100k. Why not eliminate the cathode follower and choose the diode detector load so that it looks like 100k to the IFT at 455 kHz? Because a CF does it better. With 100k source impedance, one still gets distortion to the audio signal where the diodes conduct to charge the 100 pF C1 of thre CRC detector filter. I prefer to go the extra country mile at every part of the audio chain, and it all adds up to a good total outcome, rather than one which is medicocre along the way, which then adds up to a poor total outcome. I will NEVER build anything because " we only ever did it that way before ..." or any other equally gutless and stupid reason. I can see where a cathode follower could be beneficial if we were trying to build a radio with an IF as narrow as possible, in which case it would help keep the Q of the transformer secondary as high as possible, but we are talking about a radio with wide band audio, and are probably talking about adding loading resistors across the transformers anyway, so why the cathode follower, why not just let the load of the detector diode do the job? A cathode follower after the IFT seems like a waste to me, better to use it after the detector, with a negative cathode supply voltage, to buffer the detector from the AGC and audio lines. I do things to suit the desire for wide as possible AF bw, and the R loading of the IFTs helps achieve that end. I don't want severe selectivity and IFT gain; that only belongs in Z grade AM radios and communications sets. Yes of course, I was simply trying to point out one situation where a cathode follower driving the diode might be useful. Try using a CF buffer to power a detector with a germanium diode, you'll hear the difference! Like others, yourself included, I am just prejudiced against some ideas, and a cathode follower between the IFT and detector is just something that I have little intention of trying. Try something different, and try abandoning your prejudice, just for an hour it takes to make something; maybe you hear something good. Measurements will confirm the improvement. Or they may only confirm that the cathode follower helps with your detector design for some as yet unexplained reason, but doesn't help in the general case, see my comments further along. Have you measured identical detectors with and without the cathode follower? I don't need to make the comparisons with measurements. Its so plain obvious with CRO experiments and dual trace tracing of the envelope shape against the detected signal. The same diode driven at the same level, with the same DC bias applied, and the same total load reflected to the IFT secondary at 455 kHz? While many AM receivers have been designed in a cost conscious way, there have also been a few where no expense was spared, and parts were freely used, and yet I have never seen a cathode follower used as you propose in a commercial design, I would think if it were beneficial someone would have used it commercially, anyone know of any examples? I have NEVER seen any ancient commercially produced radio or audio product where the sound quality was not compromised, often severely, with many lies told by the market cowboys, after the maker had reduced the parts count to reduce costs, to be able to compete. There were certainly commercially produced AM tuners where the maker didn't reduce the parts count at all in order to reduce the costs, the sound quality may or may not have been compromised by your standards, but if it was, it was due to a poor use of the parts, rather than to a lowered parts count. There are certainly commercial AM tuner designs that use significantly more parts than your tuner uses. 99% of surviving old radios are crap, and were crap when they were foisted onto a gullible public. I have never seen anything here which remotely was above the lowest common denominator, except perhaps the Quad AM tuner I aquired. But the one I built measures and sounds better on the BCB. Sure some AM tuners use more parts than I have, and I recently posted a few SS design schematics at ABSE and ABPR to let folks know there is more to AM reception than RDH4 ideas and a few tubes. Some designs add other relatively expensive parts to the detector circuit, one trick I have seen whose effects might be worth looking into is replacing the second capacitor in the peak detector & RF filter network with a series LC network tuned to 455 kHz. I don't know how this circuit actually works, but I assume that the idea is to improve the tradeoff of the total peak detector capacitance vs. tangential clipping at high frequencies. A series LC tuned to 455 kHz needs to be driven by a low impedance to get a decent Q to reject the 455 kHz ripple, This is patent nonsense, if anything just the opposite is true, if a series LC to ground were driven by a very low impedance it would have virtually no effect. I suggested low impedance, not zero ohms impedance. You would find that with a high source impedance that the attenuation curve is too broad, and as you reduce Rg, the attenuation of the series LC becomes sharper and deeper, but below a certain Rg the attenuation is restricted. I have tried this sort of idea for a 9 kHz whistle filter for better DX listening. I found the shape of the null, or the Q of the null was very variable with Rg, and a CF with a series R was best to find the right sort of null. R can't be too small, lest the filter overload the tube. But a broad null was needed to not only reduce the 9 kHz whistle but also reduce the monkey chatter a bit. Nothing was really effective 100%. Another type of much better null filter is a bridged LC type, with two caps and an L, and an R to 0V from the CT between the two caps, and this is less dependant on any critical value of Rg, although for a deep null, the two C have to be matched, and the R value to 0V is very critical. But in the case of detectors, there in no need for null filters or traps, and in any case, if you wanted good rejection of 455 kHz, with LC, you'd use a critically damped low pass filter using LC which would not only reject 455 kHz very adequately, but any other RF noise. but it simply is far easier to achive in well known ways with R&C. Usually, the CRC arrangement of 100pF, 47k, and 100pF is entirely adequate for removing RF detector ripple voltage. Just goes to show that those manufacturers that included the series LC weren't among the ones you are speaking of that reduced the parts count wherever they could. Depends how they used the parts which were not "rationalised out of the package to save $$" The 47k could be replaced with a choke in the CRC filter, but the CRC does enough to reduce the 455 kHz ripple to such low levels that any remaining ripple at say -40 dB will not cause a bother in the audio amp. And the audio amp will have almost no gain at 455 kHz. This is something I will have to look into further. There are other detector circuit subtleties like this that may, or may not, be worth while. You need to use a soldering iron to find out about what I am promoting about AM detection. There is no other way. Actually I am fairly certain that there are several other ways, like me, you are simply prejudiced towards your own ways. I prefer my own methods, and I welcome anyone to post a schematic which works better. I think I finally understand what you are doing with the bias on your detector, and how it works. I believe I have misunderstood what your biased detector was all about and you haven't explained it. You have made statements like "This method means that detection of weak signal lower than the forward voltage of the Ge diode of 0.27v peak approx are not subject to the non linear turn on of the diode, ie, there is no clipping by the diode." This lead me to believe that you were using the bias to somehow "linearize" the diode, which I didn't understand. In a tubed diode detector used so often, the value of ripple voltage varies a lot between the bottom of a detected sine wave to the top of the crests of the wave, and its a distortion mechanism, worst when modulation % is high, which it is, on most transmitted AM signals these days; since efficient use of a carrier is wanted. In my biased detector, the variation in ripple voltage along all parts of the audio wave form is very nearly the same amplitude. The figure from the RDH4 which John Stewart posted finally made me realize that you were doing exactly the same thing with your bias as the RDH4 figure, except that you used a fixed bias and left out the tracking feature. My CF plus following Ge diode is totally different to anything in RDH4. The distortion reduction you claim makes sense in that context, because your receiver as described by the schematic you posted has an extremely poor AC/DC load ratio and I am sure the distortion is extreme without the bias. Your bias scheme presumably partially compensates for the poor AC/DC load ratio, rather than somehow improving the "non linear turn on of the diode" as I had erroneously assumed from what you have said. The diode detector schematic I did post does have a poor AC/DC load ratio, but still works fine to make a few volts without any wave clipping or added distortion because of the ratio. Even better results with capacity for a much higher undistorted output voltage is possible with the same first CF and Ge diode and CRC filter, but then directly coupled to a second CF, which is the other half of a twin triode, and behold, there is zero AC loading on the detector. If the load resistances of the CF are exchanged for transistor CCS the bulk of the small thd caused by the CF transfer character is also reduced to perhap 0.02% at 10v of audio output, depending on the triode used. 6DJ8 would be ideal... 12AU7 is OK, 6AQ8, ECC85, 6201, 12AT7 are probably better, since the gain reduction with 12AT7 is around say 35 times for a given load R, so if the open loop thd was 1%, its reduced to 1/35 % by the NFB effect of the follower. A CCS reduces the thd at least several times more. Now the only remaining question is, does the apparently pointless cathode follower driving the diode also compensate in some way for another unnoticed design flaw? I give up. If you cannot see the benefits of intelligent buffering, and you won't try an idea before roundly condemning it, then it becomes pointless for me to provide any more justification than I already have. Patrick Turner. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#12
|
|||
|
|||
|
Bob wrote: Just curious here. I've been following the debate. I have a radio I built that uses a separate AVC amplifier to separate the AVC function from the detector, and give me some control over AVC action. I used an infinite impedance detector so as to not load the last IF secondary. So I pretty much have the cathode follower and would only need to add a diode to convert to the discussed detector. would there be any noticeable improvement over the infinite impedance detector? Try building, measuring, and listening, and then you can decide which is best. Patrick Turner. thanks ------ Bob La Rocca Lindenhurst, NY "Patrick Turner" wrote in message ... John Byrns wrote: In article , Patrick Turner wrote: John Byrns wrote: In article , Patrick Turner wrote: Better of course would be to have a CF tube to accept the IF envelope, and the low impedance output from the CF can then power a crystal diode, or a tube diode in a variety of ways I have previously explained in post on the matter. You still haven't explained how this added cathode follower, to drive the detector, helps matters? Many experts even make the claim that a finite source resistance can be beneficial in reducing distortion, especially high frequency distortion. IN my case there *is* a finite source resistance which is the 100k R across each IFT winding. Yes, but the diode is driven from the low source impedance of the cathode follower, not something on the order of 100k. Why not eliminate the cathode follower and choose the diode detector load so that it looks like 100k to the IFT at 455 kHz? Because a CF does it better. With 100k source impedance, one still gets distortion to the audio signal where the diodes conduct to charge the 100 pF C1 of thre CRC detector filter. I prefer to go the extra country mile at every part of the audio chain, and it all adds up to a good total outcome, rather than one which is medicocre along the way, which then adds up to a poor total outcome. I will NEVER build anything because " we only ever did it that way before ..." or any other equally gutless and stupid reason. I can see where a cathode follower could be beneficial if we were trying to build a radio with an IF as narrow as possible, in which case it would help keep the Q of the transformer secondary as high as possible, but we are talking about a radio with wide band audio, and are probably talking about adding loading resistors across the transformers anyway, so why the cathode follower, why not just let the load of the detector diode do the job? A cathode follower after the IFT seems like a waste to me, better to use it after the detector, with a negative cathode supply voltage, to buffer the detector from the AGC and audio lines. I do things to suit the desire for wide as possible AF bw, and the R loading of the IFTs helps achieve that end. I don't want severe selectivity and IFT gain; that only belongs in Z grade AM radios and communications sets. Yes of course, I was simply trying to point out one situation where a cathode follower driving the diode might be useful. Try using a CF buffer to power a detector with a germanium diode, you'll hear the difference! Like others, yourself included, I am just prejudiced against some ideas, and a cathode follower between the IFT and detector is just something that I have little intention of trying. Try something different, and try abandoning your prejudice, just for an hour it takes to make something; maybe you hear something good. Measurements will confirm the improvement. Or they may only confirm that the cathode follower helps with your detector design for some as yet unexplained reason, but doesn't help in the general case, see my comments further along. Have you measured identical detectors with and without the cathode follower? I don't need to make the comparisons with measurements. Its so plain obvious with CRO experiments and dual trace tracing of the envelope shape against the detected signal. The same diode driven at the same level, with the same DC bias applied, and the same total load reflected to the IFT secondary at 455 kHz? While many AM receivers have been designed in a cost conscious way, there have also been a few where no expense was spared, and parts were freely used, and yet I have never seen a cathode follower used as you propose in a commercial design, I would think if it were beneficial someone would have used it commercially, anyone know of any examples? I have NEVER seen any ancient commercially produced radio or audio product where the sound quality was not compromised, often severely, with many lies told by the market cowboys, after the maker had reduced the parts count to reduce costs, to be able to compete. There were certainly commercially produced AM tuners where the maker didn't reduce the parts count at all in order to reduce the costs, the sound quality may or may not have been compromised by your standards, but if it was, it was due to a poor use of the parts, rather than to a lowered parts count. There are certainly commercial AM tuner designs that use significantly more parts than your tuner uses. 99% of surviving old radios are crap, and were crap when they were foisted onto a gullible public. I have never seen anything here which remotely was above the lowest common denominator, except perhaps the Quad AM tuner I aquired. But the one I built measures and sounds better on the BCB. Sure some AM tuners use more parts than I have, and I recently posted a few SS design schematics at ABSE and ABPR to let folks know there is more to AM reception than RDH4 ideas and a few tubes. Some designs add other relatively expensive parts to the detector circuit, one trick I have seen whose effects might be worth looking into is replacing the second capacitor in the peak detector & RF filter network with a series LC network tuned to 455 kHz. I don't know how this circuit actually works, but I assume that the idea is to improve the tradeoff of the total peak detector capacitance vs. tangential clipping at high frequencies. A series LC tuned to 455 kHz needs to be driven by a low impedance to get a decent Q to reject the 455 kHz ripple, This is patent nonsense, if anything just the opposite is true, if a series LC to ground were driven by a very low impedance it would have virtually no effect. I suggested low impedance, not zero ohms impedance. You would find that with a high source impedance that the attenuation curve is too broad, and as you reduce Rg, the attenuation of the series LC becomes sharper and deeper, but below a certain Rg the attenuation is restricted. I have tried this sort of idea for a 9 kHz whistle filter for better DX listening. I found the shape of the null, or the Q of the null was very variable with Rg, and a CF with a series R was best to find the right sort of null. R can't be too small, lest the filter overload the tube. But a broad null was needed to not only reduce the 9 kHz whistle but also reduce the monkey chatter a bit. Nothing was really effective 100%. Another type of much better null filter is a bridged LC type, with two caps and an L, and an R to 0V from the CT between the two caps, and this is less dependant on any critical value of Rg, although for a deep null, the two C have to be matched, and the R value to 0V is very critical. But in the case of detectors, there in no need for null filters or traps, and in any case, if you wanted good rejection of 455 kHz, with LC, you'd use a critically damped low pass filter using LC which would not only reject 455 kHz very adequately, but any other RF noise. but it simply is far easier to achive in well known ways with R&C. Usually, the CRC arrangement of 100pF, 47k, and 100pF is entirely adequate for removing RF detector ripple voltage. Just goes to show that those manufacturers that included the series LC weren't among the ones you are speaking of that reduced the parts count wherever they could. Depends how they used the parts which were not "rationalised out of the package to save $$" The 47k could be replaced with a choke in the CRC filter, but the CRC does enough to reduce the 455 kHz ripple to such low levels that any remaining ripple at say -40 dB will not cause a bother in the audio amp. And the audio amp will have almost no gain at 455 kHz. This is something I will have to look into further. There are other detector circuit subtleties like this that may, or may not, be worth while. You need to use a soldering iron to find out about what I am promoting about AM detection. There is no other way. Actually I am fairly certain that there are several other ways, like me, you are simply prejudiced towards your own ways. I prefer my own methods, and I welcome anyone to post a schematic which works better. I think I finally understand what you are doing with the bias on your detector, and how it works. I believe I have misunderstood what your biased detector was all about and you haven't explained it. You have made statements like "This method means that detection of weak signal lower than the forward voltage of the Ge diode of 0.27v peak approx are not subject to the non linear turn on of the diode, ie, there is no clipping by the diode." This lead me to believe that you were using the bias to somehow "linearize" the diode, which I didn't understand. In a tubed diode detector used so often, the value of ripple voltage varies a lot between the bottom of a detected sine wave to the top of the crests of the wave, and its a distortion mechanism, worst when modulation % is high, which it is, on most transmitted AM signals these days; since efficient use of a carrier is wanted. In my biased detector, the variation in ripple voltage along all parts of the audio wave form is very nearly the same amplitude. The figure from the RDH4 which John Stewart posted finally made me realize that you were doing exactly the same thing with your bias as the RDH4 figure, except that you used a fixed bias and left out the tracking feature. My CF plus following Ge diode is totally different to anything in RDH4. The distortion reduction you claim makes sense in that context, because your receiver as described by the schematic you posted has an extremely poor AC/DC load ratio and I am sure the distortion is extreme without the bias. Your bias scheme presumably partially compensates for the poor AC/DC load ratio, rather than somehow improving the "non linear turn on of the diode" as I had erroneously assumed from what you have said. The diode detector schematic I did post does have a poor AC/DC load ratio, but still works fine to make a few volts without any wave clipping or added distortion because of the ratio. Even better results with capacity for a much higher undistorted output voltage is possible with the same first CF and Ge diode and CRC filter, but then directly coupled to a second CF, which is the other half of a twin triode, and behold, there is zero AC loading on the detector. If the load resistances of the CF are exchanged for transistor CCS the bulk of the small thd caused by the CF transfer character is also reduced to perhap 0.02% at 10v of audio output, depending on the triode used. 6DJ8 would be ideal... 12AU7 is OK, 6AQ8, ECC85, 6201, 12AT7 are probably better, since the gain reduction with 12AT7 is around say 35 times for a given load R, so if the open loop thd was 1%, its reduced to 1/35 % by the NFB effect of the follower. A CCS reduces the thd at least several times more. Now the only remaining question is, does the apparently pointless cathode follower driving the diode also compensate in some way for another unnoticed design flaw? I give up. If you cannot see the benefits of intelligent buffering, and you won't try an idea before roundly condemning it, then it becomes pointless for me to provide any more justification than I already have. Patrick Turner. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#13
|
|||
|
|||
|
John Byrns wrote:
In article , Patrick Turner wrote: John Byrns wrote: In article , Patrick Turner wrote: Better of course would be to have a CF tube to accept the IF envelope, and the low impedance output from the CF can then power a crystal diode, or a tube diode in a variety of ways I have previously explained in post on the matter. You still haven't explained how this added cathode follower, to drive the detector, helps matters? Many experts even make the claim that a finite source resistance can be beneficial in reducing distortion, especially high frequency distortion. IN my case there *is* a finite source resistance which is the 100k R across each IFT winding. Yes, but the diode is driven from the low source impedance of the cathode follower, not something on the order of 100k. Why not eliminate the cathode follower and choose the diode detector load so that it looks like 100k to the IFT at 455 kHz? The detector diode is a nonlinear load, where a resistor is. Not sure, but the non linearity might cause distortions in the IF transformer. In that the loading of the IF transformer's Q is varying, thus the bandwidth is varying. |
|
#14
|
|||
|
|||
|
In article , Patrick Turner
wrote: John Byrns wrote: Because a CF does it better. With 100k source impedance, one still gets distortion to the audio signal where the diodes conduct to charge the 100 pF C1 of thre CRC detector filter. I prefer to go the extra country mile at every part of the audio chain, and it all adds up to a good total outcome, rather than one which is medicocre along the way, which then adds up to a poor total outcome. I will NEVER build anything because " we only ever did it that way before ..." or any other equally gutless and stupid reason. But your reason is equally gutless and stupid, you are enamored with your circuit and refuse to look at how a diode envelope detector actually interfaces with the secondary of an IFT in the traditional approach, and whether it has any merit. Like others, yourself included, I am just prejudiced against some ideas, and a cathode follower between the IFT and detector is just something that I have little intention of trying. Try something different, and try abandoning your prejudice, just for an hour it takes to make something; maybe you hear something good. I may try the Selsted Smith detector given in the back of the RDH4. I'm not sure what hearing something good means, sometimes what sounds good is actually bad, or as some people call it euphonic. Measurements will confirm the improvement. Or they may only confirm that the cathode follower helps with your detector design for some as yet unexplained reason, but doesn't help in the general case, see my comments further along. Have you measured identical detectors with and without the cathode follower? I don't need to make the comparisons with measurements. Its so plain obvious with CRO experiments and dual trace tracing of the envelope shape against the detected signal. Now your prejudices are coming out, you say "I don't need to make the comparisons with measurements." I think I finally understand what you are doing with the bias on your detector, and how it works. I believe I have misunderstood what your biased detector was all about and you haven't explained it. You have made statements like "This method means that detection of weak signal lower than the forward voltage of the Ge diode of 0.27v peak approx are not subject to the non linear turn on of the diode, ie, there is no clipping by the diode." This lead me to believe that you were using the bias to somehow "linearize" the diode, which I didn't understand. In a tubed diode detector used so often, the value of ripple voltage varies a lot between the bottom of a detected sine wave to the top of the crests of the wave, and its a distortion mechanism, worst when modulation % is high, which it is, on most transmitted AM signals these days; since efficient use of a carrier is wanted. In my biased detector, the variation in ripple voltage along all parts of the audio wave form is very nearly the same amplitude. The figure from the RDH4 which John Stewart posted finally made me realize that you were doing exactly the same thing with your bias as the RDH4 figure, except that you used a fixed bias and left out the tracking feature. My CF plus following Ge diode is totally different to anything in RDH4. Only to the extent that you have added a cathode follower to the compensated diode circuit in the RDH4, but the cathode follower serves no useful purpose, and of course you have removed the tracking feature from the RDH4 circuit. The distortion reduction you claim makes sense in that context, because your receiver as described by the schematic you posted has an extremely poor AC/DC load ratio and I am sure the distortion is extreme without the bias. Your bias scheme presumably partially compensates for the poor AC/DC load ratio, rather than somehow improving the "non linear turn on of the diode" as I had erroneously assumed from what you have said. The diode detector schematic I did post does have a poor AC/DC load ratio, but still works fine to make a few volts without any wave clipping or added distortion because of the ratio. That is the bias at work compensating for the poor AC/DC load ratio, just as explained in the RDH4, it has nothing to do with the cathode follower. Even better results with capacity for a much higher undistorted output voltage is possible with the same first CF and Ge diode and CRC filter, but then directly coupled to a second CF, which is the other half of a twin triode, and behold, there is zero AC loading on the detector. Yes, I have said repeatedly that is where you should be using a cathode follower. Now the only remaining question is, does the apparently pointless cathode follower driving the diode also compensate in some way for another unnoticed design flaw? I give up. If you cannot see the benefits of intelligent buffering, and you won't try an idea before roundly condemning it, then it becomes pointless for me to provide any more justification than I already have. I can easily see the benefits of intelligent buffering, but a cathode follower between the IFT secondary and the diode is not "intelligent buffering". Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#15
|
|||
|
|||
|
In article , Robert Casey
wrote: John Byrns wrote: Yes, but the diode is driven from the low source impedance of the cathode follower, not something on the order of 100k. Why not eliminate the cathode follower and choose the diode detector load so that it looks like 100k to the IFT at 455 kHz? The detector diode is a nonlinear load, where a resistor is. Not sure, but the non linearity might cause distortions in the IF transformer. In that the loading of the IF transformer's Q is varying, thus the bandwidth is varying. Not a problem, the IF transformer acts like the electrical equivalent of a big flywheel providing the energy needed for those current pulses into the diode. Put a scope on the plate of the IF amplifier tube. If this didn't work radio would be in big trouble since many/most/all transmitters, or at least the old tube ones, depend on the same flywheel effect to work since the final only delivers short pulses of current into the output circuit. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#16
|
|||
|
|||
|
Robert Casey wrote: John Byrns wrote: In article , Patrick Turner wrote: John Byrns wrote: In article , Patrick Turner wrote: Better of course would be to have a CF tube to accept the IF envelope, and the low impedance output from the CF can then power a crystal diode, or a tube diode in a variety of ways I have previously explained in post on the matter. You still haven't explained how this added cathode follower, to drive the detector, helps matters? Many experts even make the claim that a finite source resistance can be beneficial in reducing distortion, especially high frequency distortion. IN my case there *is* a finite source resistance which is the 100k R across each IFT winding. Yes, but the diode is driven from the low source impedance of the cathode follower, not something on the order of 100k. Why not eliminate the cathode follower and choose the diode detector load so that it looks like 100k to the IFT at 455 kHz? The detector diode is a nonlinear load, where a resistor is. Not sure, but the non linearity might cause distortions in the IF transformer. In that the loading of the IF transformer's Q is varying, thus the bandwidth is varying. The diode loading of the IFT does cause some non linearity in the IF amp, which is detected, and comes out as audio thd/imd. In my radio there is no AVC applied to the IF amp which is a sharp cut off type 6BX6, with an unbypassed Rk, and with the lower than usual IFT but linear R load only, there is a fair current change in the IF tube, so a fair amount of current fb from Rk, so the IF tube amplifies the enevelope with far better linearity than most other sets I have examined. Patrick Turner. |
|
#17
|
|||
|
|||
|
John Byrns wrote: In article , Patrick Turner wrote: John Byrns wrote: Because a CF does it better. With 100k source impedance, one still gets distortion to the audio signal where the diodes conduct to charge the 100 pF C1 of thre CRC detector filter. I prefer to go the extra country mile at every part of the audio chain, and it all adds up to a good total outcome, rather than one which is medicocre along the way, which then adds up to a poor total outcome. I will NEVER build anything because " we only ever did it that way before ..." or any other equally gutless and stupid reason. But your reason is equally gutless and stupid, you are enamored with your circuit and refuse to look at how a diode envelope detector actually interfaces with the secondary of an IFT in the traditional approach, and whether it has any merit. Well, I don't see any point in explaining things further. I suggest you try my ideas. I've tried all yours, and found them wanting. Like others, yourself included, I am just prejudiced against some ideas, and a cathode follower between the IFT and detector is just something that I have little intention of trying. Try something different, and try abandoning your prejudice, just for an hour it takes to make something; maybe you hear something good. I may try the Selsted Smith detector given in the back of the RDH4. I'm not sure what hearing something good means, sometimes what sounds good is actually bad, or as some people call it euphonic. RDH4 is a fine tome, but that doesn't mean you have to addopt every syllable as gospel. Measurements will confirm the improvement. Or they may only confirm that the cathode follower helps with your detector design for some as yet unexplained reason, but doesn't help in the general case, see my comments further along. Have you measured identical detectors with and without the cathode follower? I don't need to make the comparisons with measurements. Its so plain obvious with CRO experiments and dual trace tracing of the envelope shape against the detected signal. Now your prejudices are coming out, you say "I don't need to make the comparisons with measurements." By dual trace observation, I was making an objective measurement, but unquantified. I know all isn't perfect in the world of analog processes like a diode detector, but I do know a good one when I see one, and hear one. I think I finally understand what you are doing with the bias on your detector, and how it works. I believe I have misunderstood what your biased detector was all about and you haven't explained it. You have made statements like "This method means that detection of weak signal lower than the forward voltage of the Ge diode of 0.27v peak approx are not subject to the non linear turn on of the diode, ie, there is no clipping by the diode." This lead me to believe that you were using the bias to somehow "linearize" the diode, which I didn't understand. In a tubed diode detector used so often, the value of ripple voltage varies a lot between the bottom of a detected sine wave to the top of the crests of the wave, and its a distortion mechanism, worst when modulation % is high, which it is, on most transmitted AM signals these days; since efficient use of a carrier is wanted. In my biased detector, the variation in ripple voltage along all parts of the audio wave form is very nearly the same amplitude. The figure from the RDH4 which John Stewart posted finally made me realize that you were doing exactly the same thing with your bias as the RDH4 figure, except that you used a fixed bias and left out the tracking feature. My CF plus following Ge diode is totally different to anything in RDH4. Only to the extent that you have added a cathode follower to the compensated diode circuit in the RDH4, but the cathode follower serves no useful purpose, and of course you have removed the tracking feature from the RDH4 circuit. My circuit isn't anything like the RDH4 circuit, although is does use tubes, dides, wire, R&C bits, etc..., but the topology is quite different. The RDH4 does not attempt to isolate the last IFT coil against the effects of the diode; my circuit does, then it changes the source impedance of IF signal to a low impedance to feed to germanium diode + CRC. Its quite different to RDH4, and better, IMHO. The distortion reduction you claim makes sense in that context, because your receiver as described by the schematic you posted has an extremely poor AC/DC load ratio and I am sure the distortion is extreme without the bias. Your bias scheme presumably partially compensates for the poor AC/DC load ratio, rather than somehow improving the "non linear turn on of the diode" as I had erroneously assumed from what you have said. The diode detector schematic I did post does have a poor AC/DC load ratio, but still works fine to make a few volts without any wave clipping or added distortion because of the ratio. That is the bias at work compensating for the poor AC/DC load ratio, just as explained in the RDH4, it has nothing to do with the cathode follower. Even better results with capacity for a much higher undistorted output voltage is possible with the same first CF and Ge diode and CRC filter, but then directly coupled to a second CF, which is the other half of a twin triode, and behold, there is zero AC loading on the detector. Yes, I have said repeatedly that is where you should be using a cathode follower. Two cathode followers are better than one. Now the only remaining question is, does the apparently pointless cathode follower driving the diode also compensate in some way for another unnoticed design flaw? I give up. If you cannot see the benefits of intelligent buffering, and you won't try an idea before roundly condemning it, then it becomes pointless for me to provide any more justification than I already have. I can easily see the benefits of intelligent buffering, but a cathode follower between the IFT secondary and the diode is not "intelligent buffering". I leave you to your enjoyment of your ideas, and wish you well, but with all due respect I still think you are denying merit in a novel approach which I know works well, since I have tried and proven it at least to myself, and to a few customers, who were very surprised that AM could sound so good. Patrick Turner. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#18
|
|||
|
|||
|
In article , Patrick Turner
wrote: John Byrns wrote: In article , Patrick Turner wrote: Try something different, and try abandoning your prejudice, just for an hour it takes to make something; maybe you hear something good. I may try the Selsted Smith detector given in the back of the RDH4. I'm not sure what hearing something good means, sometimes what sounds good is actually bad, or as some people call it euphonic. RDH4 is a fine tome, but that doesn't mean you have to addopt every syllable as gospel. Haven't I made it clear that I don't worship at the altar of the RDH4? My interest in that circuit has little to do with the fact that the design is included in the RDH4. My CF plus following Ge diode is totally different to anything in RDH4. Only to the extent that you have added a cathode follower to the compensated diode circuit in the RDH4, but the cathode follower serves no useful purpose, and of course you have removed the tracking feature from the RDH4 circuit. My circuit isn't anything like the RDH4 circuit, although is does use tubes, dides, wire, R&C bits, etc..., but the topology is quite different. The RDH4 does not attempt to isolate the last IFT coil against the effects of the diode; my circuit does, then it changes the source impedance of IF signal to a low impedance to feed to germanium diode + CRC. Its quite different to RDH4, and better, IMHO. I did point out that you added the cathode follower at the input, and deleted the bias tracking function, but the underlying idea is the same in both cases. The distortion reduction you claim makes sense in that context, because your receiver as described by the schematic you posted has an extremely poor AC/DC load ratio and I am sure the distortion is extreme without the bias. Your bias scheme presumably partially compensates for the poor AC/DC load ratio, rather than somehow improving the "non linear turn on of the diode" as I had erroneously assumed from what you have said. The diode detector schematic I did post does have a poor AC/DC load ratio, but still works fine to make a few volts without any wave clipping or added distortion because of the ratio. That is the bias at work compensating for the poor AC/DC load ratio, just as explained in the RDH4, it has nothing to do with the cathode follower. Even better results with capacity for a much higher undistorted output voltage is possible with the same first CF and Ge diode and CRC filter, but then directly coupled to a second CF, which is the other half of a twin triode, and behold, there is zero AC loading on the detector. Yes, I have said repeatedly that is where you should be using a cathode follower. Two cathode followers are better than one. And three cathode followers would be better than two, that's not a joke, I actually have a circuit in mind that would use three cathode follower like circuits, although I would call it retched excess. I leave you to your enjoyment of your ideas, and wish you well, but with all due respect I still think you are denying merit in a novel approach which I know works well, since I have tried and proven it at least to myself, and to a few customers, who were very surprised that AM could sound so good. I don't believe I have ever said that your circuit doesn't sound very good, my only question is, does the cathode follower actually contribute to the good sound, or is it just a gimmick? Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#19
|
|||
|
|||
|
Patrick Turner wrote:
John Byrns wrote: In article , Patrick Turner wrote: John Byrns wrote: In article , Patrick Turner wrote: Better of course would be to have a CF tube to accept the IF envelope, and the low impedance output from the CF can then power a crystal diode, or a tube diode in a variety of ways I have previously explained in post on the matter. You still haven't explained how this added cathode follower, to drive the detector, helps matters? Many experts even make the claim that a finite source resistance can be beneficial in reducing distortion, especially high frequency distortion. IN my case there *is* a finite source resistance which is the 100k R across each IFT winding. Yes, but the diode is driven from the low source impedance of the cathode follower, not something on the order of 100k. Why not eliminate the cathode follower and choose the diode detector load so that it looks like 100k to the IFT at 455 kHz? Because a CF does it better. With 100k source impedance, one still gets distortion to the audio signal where the diodes conduct to charge the 100 pF C1 of thre CRC detector filter. I prefer to go the extra country mile at every part of the audio chain, and it all adds up to a good total outcome, rather than one which is medicocre along the way, which then adds up to a poor total outcome. Pat, I just ran a simulation of the use of two cathode followers (I used 12AU7's) and a 6AL5 diode ddetector. I got very good results, using a 1KHz audio sine wave, modulated onto a 95% modulated 455KHz carrier, got 55dB down of the 2nd and 3rd harmonic. I also threw in a little bit of positive bias to the detector diode to partly get it up above the "Knee". About 300mV worth of bias. Of course this requires that the AM signal be at least 2Vp-p and that the modulation index be about 95% (leaving 5% carrier at the valleys). As I'm pushing a bit into the "other side" of the AM signal. IOW, I moved the zero crossing voltage a bit to make "fake exaulted carrier". See a screen capture of my simulation schematic and results in ABPR |
|
#20
|
|||
|
|||
|
John Byrns wrote: In article , Patrick Turner wrote: John Byrns wrote: In article , Patrick Turner wrote: Try something different, and try abandoning your prejudice, just for an hour it takes to make something; maybe you hear something good. I may try the Selsted Smith detector given in the back of the RDH4. I'm not sure what hearing something good means, sometimes what sounds good is actually bad, or as some people call it euphonic. RDH4 is a fine tome, but that doesn't mean you have to addopt every syllable as gospel. Haven't I made it clear that I don't worship at the altar of the RDH4? My interest in that circuit has little to do with the fact that the design is included in the RDH4. My CF plus following Ge diode is totally different to anything in RDH4. Only to the extent that you have added a cathode follower to the compensated diode circuit in the RDH4, but the cathode follower serves no useful purpose, and of course you have removed the tracking feature from the RDH4 circuit. My circuit isn't anything like the RDH4 circuit, although is does use tubes, dides, wire, R&C bits, etc..., but the topology is quite different. The RDH4 does not attempt to isolate the last IFT coil against the effects of the diode; my circuit does, then it changes the source impedance of IF signal to a low impedance to feed to germanium diode + CRC. Its quite different to RDH4, and better, IMHO. I did point out that you added the cathode follower at the input, and deleted the bias tracking function, but the underlying idea is the same in both cases. The distortion reduction you claim makes sense in that context, because your receiver as described by the schematic you posted has an extremely poor AC/DC load ratio and I am sure the distortion is extreme without the bias. Your bias scheme presumably partially compensates for the poor AC/DC load ratio, rather than somehow improving the "non linear turn on of the diode" as I had erroneously assumed from what you have said. The diode detector schematic I did post does have a poor AC/DC load ratio, but still works fine to make a few volts without any wave clipping or added distortion because of the ratio. That is the bias at work compensating for the poor AC/DC load ratio, just as explained in the RDH4, it has nothing to do with the cathode follower. Even better results with capacity for a much higher undistorted output voltage is possible with the same first CF and Ge diode and CRC filter, but then directly coupled to a second CF, which is the other half of a twin triode, and behold, there is zero AC loading on the detector. Yes, I have said repeatedly that is where you should be using a cathode follower. Two cathode followers are better than one. And three cathode followers would be better than two, that's not a joke, I actually have a circuit in mind that would use three cathode follower like circuits, although I would call it retched excess. I leave you to your enjoyment of your ideas, and wish you well, but with all due respect I still think you are denying merit in a novel approach which I know works well, since I have tried and proven it at least to myself, and to a few customers, who were very surprised that AM could sound so good. I don't believe I have ever said that your circuit doesn't sound very good, my only question is, does the cathode follower actually contribute to the good sound, or is it just a gimmick? Indeed the CF powered diode detector *does* contribute to the overall fidelity of any set with which it is used. For those without the room to fit a CF, an emitter follower is nearly as good using a well rated transistor emitter follower, since the input impedance is high, even during the current pulses at 455 kHz, but best is to use a pair of transistors connected as a darlington pair to make the follower buffer. I finalised the Kreisler radio with 100k added to each of the first 3 IFT coils, leaving the final coil loaded by the non CF detector. It has one end of the last IF coil taken to a pair of tube diodes in a 6N8 IF amp tube, then the other coil end goes to 100pF to 0V, then 47k to another 100pF to 0V, then 100k, then 220k, then 100k to 0V to make a total of 467k to 0V for the DC to flow from the 100 pF caps. The audio is taken from the top of the last 100k via a 0.27 uF feeding a 100k pot. There is an unbypassed Rk on the 6N8 to give a little negative current FB. The 220k value was found by testing for lowest thd with 100% modulation with enough signal at RF to generate -20v AVC voltage. Local strong stations generate less, so the R value was correct, and thd minimal at low level signals. There is 330pf across the 220k to slightly boost the recovered audio HF. The 100k before the 220k stops the 455 kHz ripple travelling further from the second 100pF via the 330 pF compensation cap. The bw increased from 2.7kHz to 6kHz. With treble boost in the power amp section, final bw of 8 kHz is possible. It sounds very nice indeed. The tuning is less critical, the AGC is still very OK for the locals which vary between 300 watts and 5,000+ watts. Then I stopped the cabinet from buzzing between metal speaker covers, and various bits of plywood, and I will complete by day's end, and give back a set to the world which is fit to listen to. The two 8" speakers have whizzer cones, so their response is at least to 8 kHz, although they probably have anything but a totally flat response. Patrick Turner. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#21
|
|||
|
|||
|
In article , Patrick Turner
wrote: The diode loading of the IFT does cause some non linearity in the IF amp, which is detected, and comes out as audio thd/imd. In my radio there is no AVC applied to the IF amp which is a sharp cut off type 6BX6, with an unbypassed Rk, and with the lower than usual IFT but linear R load only, there is a fair current change in the IF tube, so a fair amount of current fb from Rk, so the IF tube amplifies the enevelope with far better linearity than most other sets I have examined. Now this seems like a reasonable concern about the effect of the diode detector on the IF amplifier, unlike your concern about the current pulses into the diode which are rendered irrelevant by the filtering/flywheel action of the IFT. When the signal level is low, or the negative modulation high, the non linearity of the diode detector causes a lighter loading on the secondary of the IFT. This of course is going to create envelope distortion at the secondary of the IFT, and the plate of the IF amplifier tube, and many authorities claim this distortion is a good thing because it implies that given the finite source impedance of the IFT secondary, the drive voltage to the diode detector will increase at these points, partially offsetting the distortion of the envelope detector under these conditions. A cathode follower would negate this effect. One thing that I haven't seen mentioned is how the IFT plays into this since it has a sort of impedance inverting characteristic, as we discussed earlier, if the secondary is shorted, the impedance of the primary increases, just the opposite of what we would expect. I suppose this effect may not hold over the range of impedances we are actually talking about here, but if it does, it would seem that it could negate this claimed advantage of driving the diode directly from the IFT secondary without an intervening cathode follower. I am curious now and will have to crunch a few numbers, see what really happens with this, and see if the experts are really correct. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#22
|
|||
|
|||
|
In article , Robert Casey
wrote: Pat, I just ran a simulation of the use of two cathode followers (I used 12AU7's) and a 6AL5 diode ddetector. I got very good results, using a 1KHz audio sine wave, modulated onto a 95% modulated 455KHz carrier, got 55dB down of the 2nd and 3rd harmonic. I also threw in a little bit of positive bias to the detector diode to partly get it up above the "Knee". About 300mV worth of bias. Of course this requires that the AM signal be at least 2Vp-p and that the modulation index be about 95% (leaving 5% carrier at the valleys). As I'm pushing a bit into the "other side" of the AM signal. IOW, I moved the zero crossing voltage a bit to make "fake exaulted carrier". See a screen capture of my simulation schematic and results in ABPR Robert, Your simulated detector seems quite different than Patrick's, mainly because your bias network tends to turn the diode off, while Patrick's bias turns the diode on, compensating for his poor AC/DC load ratio, at least at one fixed carrier level. It took me a while to finally understand Patrick's diode bias scheme, I finally only realized what he was doing after John Stewart pointed out a similar circuit in the RDH4, perhaps I don't yet understand what your bias scheme is doing? Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#23
|
|||
|
|||
|
John Byrns wrote:
In article , Robert Casey wrote: Pat, I just ran a simulation of the use of two cathode followers (I used 12AU7's) and a 6AL5 diode ddetector. I got very good results, using a 1KHz audio sine wave, modulated onto a 95% modulated 455KHz carrier, got 55dB down of the 2nd and 3rd harmonic. I also threw in a little bit of positive bias to the detector diode to partly get it up above the "Knee". About 300mV worth of bias. Of course this requires that the AM signal be at least 2Vp-p and that the modulation index be about 95% (leaving 5% carrier at the valleys). As I'm pushing a bit into the "other side" of the AM signal. IOW, I moved the zero crossing voltage a bit to make "fake exaulted carrier". See a screen capture of my simulation schematic and results in ABPR Robert, Your simulated detector seems quite different than Patrick's, mainly because your bias network tends to turn the diode off, while Patrick's bias turns the diode on, compensating for his poor AC/DC load ratio, at least at one fixed carrier level. The diagram is a bit confusing, but the simulation "power supply" which has its normally positive end tied to ground I made it a negative voltage. (I was lazy, easier to change the value and make it negative than to redraw the wires...). In any event, I'm biasing the diode a bit more "on". This to get up out of the curved portion of the vacuum tube diode characteristic. I still have to actually try this circuit (with or without bias) in hardware. I have a Heathkit AM tuner I could try it in. I'll use submini tubes with wire leads, smaller and no sockets to mount. |
|
#24
|
|||
|
|||
|
John Byrns wrote: In article , Patrick Turner wrote: The diode loading of the IFT does cause some non linearity in the IF amp, which is detected, and comes out as audio thd/imd. In my radio there is no AVC applied to the IF amp which is a sharp cut off type 6BX6, with an unbypassed Rk, and with the lower than usual IFT but linear R load only, there is a fair current change in the IF tube, so a fair amount of current fb from Rk, so the IF tube amplifies the enevelope with far better linearity than most other sets I have examined. Now this seems like a reasonable concern about the effect of the diode detector on the IF amplifier, unlike your concern about the current pulses into the diode which are rendered irrelevant by the filtering/flywheel action of the IFT. The load varies during thre audio cycle if the ripple voltage changes. So the IF amp gain changes, ie, there is intermodulation distortion caused by the detector. My concerns are not irrelevant. When the signal level is low, or the negative modulation high, the non linearity of the diode detector causes a lighter loading on the secondary of the IFT. This of course is going to create envelope distortion at the secondary of the IFT, and the plate of the IF amplifier tube, and many authorities claim this distortion is a good thing because it implies that given the finite source impedance of the IFT secondary, the drive voltage to the diode detector will increase at these points, partially offsetting the distortion of the envelope detector under these conditions. I don't know what authorities say that distortion is a "good thing" A cathode follower would negate this effect. You've been trying to tell me a CF isn't needed all along, but here you say its OK. One thing that I haven't seen mentioned is how the IFT plays into this since it has a sort of impedance inverting characteristic, as we discussed earlier, if the secondary is shorted, the impedance of the primary increases, just the opposite of what we would expect. I found that loading the secondary reduced the gain of the IF amp, ie, the load is reflected to the primary. I suppose this effect may not hold over the range of impedances we are actually talking about here, but if it does, it would seem that it could negate this claimed advantage of driving the diode directly from the IFT secondary without an intervening cathode follower. I am curious now and will have to crunch a few numbers, see what really happens with this, and see if the experts are really correct. Regards, John Byrns Happy number crunching, but I prefer late nights with a soldering iron & cro. Patrick Turner. Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#25
|
|||
|
|||
|
John Byrns wrote: In article , Robert Casey wrote: Pat, I just ran a simulation of the use of two cathode followers (I used 12AU7's) and a 6AL5 diode ddetector. I got very good results, using a 1KHz audio sine wave, modulated onto a 95% modulated 455KHz carrier, got 55dB down of the 2nd and 3rd harmonic. I also threw in a little bit of positive bias to the detector diode to partly get it up above the "Knee". About 300mV worth of bias. Of course this requires that the AM signal be at least 2Vp-p and that the modulation index be about 95% (leaving 5% carrier at the valleys). As I'm pushing a bit into the "other side" of the AM signal. IOW, I moved the zero crossing voltage a bit to make "fake exaulted carrier". See a screen capture of my simulation schematic and results in ABPR Robert, Your simulated detector seems quite different than Patrick's, mainly because your bias network tends to turn the diode off, while Patrick's bias turns the diode on, compensating for his poor AC/DC load ratio, at least at one fixed carrier level. It took me a while to finally understand Patrick's diode bias scheme, I finally only realized what he was doing after John Stewart pointed out a similar circuit in the RDH4, perhaps I don't yet understand what your bias scheme is doing? Regards, John Byrns The "poor" AC/DC" load ratio does not matter if the output audio voltage asked from the circuit is low. In a triode gain stage, if the DC carrying RL is large, and the cap coupled load is small, a similar effect occurs, and positive voltage swing is limited to a lot less than the B+. But the first few volts are still quite linear, and simply what one expects when load liane analysis is used. My own radio uses a good enough AC/DC load ratio to take the cut off effect fully into account. Patrick Turner. Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#26
|
|||
|
|||
|
Happy number crunching, but I prefer late nights with a soldering iron & cro. I number crunch (via simulations) first (to filter out the bad ideas that are doomed to perform poorly) and then plug the soldering iron in.... |
|
#27
|
|||
|
|||
|
Robert Casey wrote:
Patrick Turner wrote: Happy number crunching, but I prefer late nights with a soldering iron & cro. I number crunch (via simulations) first (to filter out the bad ideas that are doomed to perform poorly) and then plug the soldering iron in.... As a mechanical engineer with project management experience, we always first brought in the number crunchers for proposed designs before we went and built anything, big or small. Even though we recognized that simulation has its flaws, so one should not blindly accept the results, the models were sufficiently accurate that it at least put us in the right ballpark before we went out and built something. I'm glad bridges are designed and built with this process, and not Patrick's process. Now the question is how accurate are the simulations of circuits which include factors for real-world (e.g. non-linear) performance? Patrick keeps saying to get out the soldering gun and see how things turn out, which I find perplexing since I assume the circuit simulation codes that exist today are quite powerful. It would not surprise me if many commercial electronic circuits are first designed entirely by computer using simulation and optimization techniques, then prototypes built for final tweaking and testing. Jon Noring |
|
#28
|
|||
|
|||
|
Jon Noring wrote:
Now the question is how accurate are the simulations of circuits which include factors for real-world (e.g. non-linear) performance? Patrick keeps saying to get out the soldering gun and see how things turn out, which I find perplexing since I assume the circuit simulation codes that exist today are quite powerful. It would not surprise me if many commercial electronic circuits are first designed entirely by computer using simulation and optimization techniques, then prototypes built for final tweaking and testing. Jon Noring My opinion is that the simulated circuits generally do just fine unless there is one parameter that we might be looking for that isn't included. Quantifying or averaging human hearing is the classic example. Thats where you get into the fuzzy areas of things like which sounds better (or different)...11.5 kHz components at 6.5db down or 11.5 kc components at 12.3 db down. One guy may immediately note the difference, the next segment of the population wouldn't notice if it were 6.5db up. (I'm thinking of the 10kc wide AM radio here) Then the designer is faced with a real-world decision of whether of not this is an issue and what other implications might it have further down the chain with increased distortions, etc. The simulation can provide the data but with the 'wrong' criteria it can equally as efficiently spit out a worthless design. It would be nice to say "I want maximum bandwidth, flatness and minimum distortion" but at the end of the day a knowledgeable human being has to decide which of the results are what HE is looking for. Getting out the soldering iron quite often is the proof of the pudding. You may know all the parameters of the circuit down to three decimal points but does it sound the way you expect or want it to? (Again, the 10kc wide AM radio is a good example) And less importantly, will anyone else concur? -BM |
|
#29
|
|||
|
|||
|
Jon Noring said:
I'm glad bridges are designed and built with this process, and not Patrick's process. You rational types! What's the fun of that? It's like building an amplifier from the book. Now the question is how accurate are the simulations of circuits which include factors for real-world (e.g. non-linear) performance? Patrick keeps saying to get out the soldering gun and see how things turn out, which I find perplexing since I assume the circuit simulation codes that exist today are quite powerful. It would not surprise me if many commercial electronic circuits are first designed entirely by computer using simulation and optimization techniques, then prototypes built for final tweaking and testing. You said it, * commercial * electronic circuits. That's why the new stuff sounds like ****. :-) -- Sander deWaal Vacuum Audio Consultancy |
|
#30
|
|||
|
|||
|
Jon Noring wrote: Robert Casey wrote: Patrick Turner wrote: Happy number crunching, but I prefer late nights with a soldering iron & cro. I number crunch (via simulations) first (to filter out the bad ideas that are doomed to perform poorly) and then plug the soldering iron in.... As a mechanical engineer with project management experience, we always first brought in the number crunchers for proposed designs before we went and built anything, big or small. Even though we recognized that simulation has its flaws, so one should not blindly accept the results, the models were sufficiently accurate that it at least put us in the right ballpark before we went out and built something. I'm glad bridges are designed and built with this process, and not Patrick's process. Unfortunately, you have guessed wildly at what my methods really are. I crunch pages full of numbers before I actually solder a single joint. I would have 200 exercize books full of sketches and drawings of all sorts of things, and pages of calculations, and my folder which contains my summary of useful formulas is at least twenty handwritten pages, taken from many sources. I should throw all these out, to save room in a cupboard, but sometimes I return to look, and I don't have a nagging woman to tell me what I must throw out.... Before I took up electronics seriously, I trained as a carpenter who did a Building Certificate course over 5 years at night school. I spent 25 years constructing buildings, and then had another 15 years as a design and construction contractor. At night school, when I didn't fall asleep in class, they slowly taught me how to design simple bridges and retaining walls, and dams, and multi-story buildings, and had I had the job of designing the trade centres in NY, they might be still standing, with repair work about completed by now, because I don't like flimsy design, or ideas. So if you ever buy a house which I'd worked on, have no fear, it won't collapse. My website is full of formulas. Now the question is how accurate are the simulations of circuits which include factors for real-world (e.g. non-linear) performance? Well judging by the success of the space mission to Saturn and its 31 known moons, there is a lot to be said for simulation, and accurate calculations. I wonder about God though. They say he made a sheila, and left a little hole, Then he made a bloke, and gave him a little pole. Somehow these creatures of the mud got together, and each version have been trying to get the hang of life ever since, and in doing so are rooting up this planet in a way that's worse than when one of God's tradgectory simulations went awry, and an asteroid hit the earth, to give the huge lizards a rotten surprize about the weather. We are left wondering about God, who the scientists say couldn't have existed before the Big Bang, but I ask these morons "but what was there before the BB?" and they haven't the feintest idea. Now you should remember the COWPAT formula. COWPAT = chance of working perfectly any time = 1 divided by N squared, where N is the number of things you have to have taken into consideration for something to turn out right. So if you have 3 basic numbers wrong, picked the wrong tube to plug in, had a blue wiith the missus the night before, and got drunk, and your soldering iron fused this morning, you have 5 crook things, and so the chance of finishing the latest IF amp stage by midnight is 1/25. God had rather a lot of things to consider, so COWPAT virtually = 1 / infinity, or 0. Patrick keeps saying to get out the soldering gun and see how things turn out, which I find perplexing since I assume the circuit simulation codes that exist today are quite powerful. But I don't need to spend a month of sundays to learn to drive simulators. I can simulate well enough in my brain for a simpe thing like a detector. It would not surprise me if many commercial electronic circuits are first designed entirely by computer using simulation and optimization techniques, then prototypes built for final tweaking and testing. Nearly all things are now designed by simulation. A computer works it all out. But there comes a time in a man's life where you develop a feel for an idea, knowing the capabilities of the elements at hand, and presto, you make something half decent without *too much* calculation. Its a bit like playing chess. Deep Fritz, the computor, can beat Kasparov, but DF has to make perhaps billions of calculations and simulations before he can decide where to move a pawn. Kasparov, on the other hand takes a look at the board, thinks for awhile, and makes a move. Its only in the last few years that puters developed enough power and speed to beat the best men, and both are essential with chess because its a timed game. Kasparov will never be able to think like DF, and calculate and plot the same way, he'd take 20 years to do one move. Picasso had this brilliance with a canvas, and probably was hopeless with calculations and simulations. He just "had it". Alas, my friend here who has been our town chess champion a few times will beat me at the game when he gives me 10 minutes on the clock and he has set his clock for 1 minute. He is utterly helpless with any sort of tool used to do anything real. His ex wife and students gave him a nervous breakdown, and he had to give up his career as a math teacher. But at chess he is something else, he soars above the rest of us at the cafe. I don't like playing him much, its like being beaten up by a thug. When you browse the pages of the last 50 years of magazines like Wireless World which later evolved into Electronics World, you see the evolution of some very gifted guys who are far more able than I to contemplate ideas, and bring electronics from what it was in 1954 to what it is now, bleeding incomprehensible for us poor mortals. Did the fathers of the PC ever decide they had to make a detector better before moving on?. Nope, they probably were bored ****eless by radio, since it wasn't interactive enough to warrant their attention. Even bright geeks are emotional! Simulation saves an enormous amount of calculation time with modern circuits. Computers design computers, we just give them the basic functions they have to perform, and leave them to it, and we get other computers to iron out any bugs. Some of the boffins in charge are bright enough to have some alarmingly clever insights as to why a persistant bug exists in a given prototype, and the human's way of thinking intuitively combined with the machine way of thought is going to build a future existance for our species which is utterly unimaginable at present. Clashes between men, and computers like HAL, in 2001, are unlikely. And it is cheaper and probably more effective *not* to send men on a Saturn mission, because men are complex, and have to eat, and poo, and get along, all problematic on long space flights. A man is too complex, like an output transformer, to ever simulate correctly. So there are no output trannies, or men, on the Saturn jaunt. But if there was a good supply of oil out there, maybe the US would send the troops to secure it. They might find a billion times what Earth already has, and that would spell catastophe for us, because we can't afford to fit our sky with all that smoke. Jon Noring Patrick Turner. |
|
#31
|
|||
|
|||
|
".Bill" wrote: Jon Noring wrote: Now the question is how accurate are the simulations of circuits which include factors for real-world (e.g. non-linear) performance? Patrick keeps saying to get out the soldering gun and see how things turn out, which I find perplexing since I assume the circuit simulation codes that exist today are quite powerful. It would not surprise me if many commercial electronic circuits are first designed entirely by computer using simulation and optimization techniques, then prototypes built for final tweaking and testing. Jon Noring My opinion is that the simulated circuits generally do just fine unless there is one parameter that we might be looking for that isn't included. Quantifying or averaging human hearing is the classic example. Thats where you get into the fuzzy areas of things like which sounds better (or different)...11.5 kHz components at 6.5db down or 11.5 kc components at 12.3 db down. One guy may immediately note the difference, the next segment of the population wouldn't notice if it were 6.5db up. (I'm thinking of the 10kc wide AM radio here) But FM radio surely sounds better than AM radio. So we should aim for wide audio bandwidth. Then the designer is faced with a real-world decision of whether of not this is an issue and what other implications might it have further down the chain with increased distortions, etc. The simulation can provide the data but with the 'wrong' criteria it can equally as efficiently spit out a worthless design. It would be nice to say "I want maximum bandwidth, flatness and minimum distortion" but at the end of the day a knowledgeable human being has to decide which of the results are what HE is looking for. Getting out the soldering iron quite often is the proof of the pudding. You may know all the parameters of the circuit down to three decimal points but does it sound the way you expect or want it to? (Again, the 10kc wide AM radio is a good example) And less importantly, will anyone else concur? Well quite a few would, and there have been quite a few designs for a least technically excellent AM radio receivers over the last 30 years which make what preceeded them look like a POS, if the Kreisler type of radiogram is anything to go by. 99% of the public bought what worked, and was cheap, and that excluded the finer gear made by so very few makers. The modern radios didn't look very nice though, they were very bland, and the real charm was in the radios of the time when radio was all there was. Huge resonant wooden enclosures were used, and then elaborate dial glasses, neither of which improved what was heard one bit. More value was in the pretentious appearence than in the electronics. Part of putting on the agony and putting on the style. People were very fussy and even snobbish about their prized radios they owned, and those who had their youth in that time are now very nostalgic about the symbols of that time. I doubt I will ever be nostagic about my CD player. Its black, and reminds me of a funeral, and I need a torch to read the many horid buttons, and sounds worse than vinyl when vinyl is good, and if something goes wrong, I damn well can't fix it. Modern junk. Patrick Turner. -BM |
|
#32
|
|||
|
|||
|
In article , Patrick Turner
wrote: John Byrns wrote: Now this seems like a reasonable concern about the effect of the diode detector on the IF amplifier, unlike your concern about the current pulses into the diode which are rendered irrelevant by the filtering/flywheel action of the IFT. The load varies during thre audio cycle Yes, that is correct, the reason is that the diode isn't "perfect". if the ripple voltage changes. No, that is wrong, even with a "perfect" diode which would present the driving circuit with a constant load over the complete audio cycle if such a diode existed, the ripple voltage will vary, being very low at the negative modulation peaks, and high at the positive modulation peaks, the amplitude of the ripple tracks the envelope, it is just the natural way a diode peak detector works, there is nothing to be done about it that I know of. You have made conflicting statements about the 455 kHz ripple voltage out of your detectors, first saying that your detector had a constant ripple voltage over the audio cycle, then saying that the ripple voltage varied over the audio cycle, and most recently you seem to be back to your detector somehow having constant ripple voltage. This doesn't add up and there is something fishy here. So the IF amp gain changes, ie, there is intermodulation distortion caused by the detector. Obviously the IF amplifier gain changes, I don't think anyone claimed otherwise. Interestingly, one of the things I have learned as a result of this discussion is that when a double tuned IFT is used to drive a real diode the gain does not change in the direction you would expect due to the impedance inverting properties of double tuned IFTs. My concerns are not irrelevant. I said your concerns about the diode RF current pulses are irrelevant in my opinion, I didn't say that the varying load presented by the diode over the modulation cycle is irrelevant. Interestingly the issue of the varying load seems to be much more complex than I have seen mentioned in any text book, I have never seen a text book which mentioned the impedance inverting properties of double tuned IFTs in connection with the varying diode load. When the signal level is low, or the negative modulation high, the non linearity of the diode detector causes a lighter loading on the secondary of the IFT. This of course is going to create envelope distortion at the secondary of the IFT, and the plate of the IF amplifier tube, and many authorities claim this distortion is a good thing because it implies that given the finite source impedance of the IFT secondary, the drive voltage to the diode detector will increase at these points, partially offsetting the distortion of the envelope detector under these conditions. I don't know what authorities say that distortion is a "good thing" I didn't say that they said distortion was a good thing, I said that some authorities point out that a finite non zero source impedance will work to minimize the distortion caused by the diode peak detector at high modulation levels, the fact that the signal wave form at the anode of the IF amplifier tube becomes distorted as a result is irrelevant. Think of it as a distortion cancellation mechanism if you like, I know you have done work in that area yourself. The text books speak of this mechanism as if the diode were driven from a tube and a simple resistance, and it seems rather obvious that the idea works in that case, but when you put a double tuned IFT between the IF amplifier tube and the diode, then things change and it is no longer obvious exactly what happens, and what the final result is. I have crunched a few numbers and found that my hypothesized impedance inverting effect does indeed occur, causing the gain of the IF amplifier to decrease rather than increase as you might expect, but so far it appears that the increase in voltage driving the diode due to the higher load resistance presented by the diode more than compensates for the lowered gain of the IF amplifier tube. One unfortunate thing I have discovered is that the impedance inverting effect is most pronounced at the carrier frequency, and less so at the sideband frequencies, which means that this effect increases the modulation percentage. Fortunately in a narrow band IF this increase in modulation percentage is probably completely swamped by the bandpass effects of the IFT on the sidebands, and this might partly account for John Doty's comments that wideband AM IFs often sound more distorted. There are a lot of numbers to be crunched with respect to the effects of the double tuned IFT in this regard, it might make a good subject for someone's masters degree thesis, that is if they can find a professor interested in something like this, that has little relevance to current technology. A cathode follower would negate this effect. You've been trying to tell me a CF isn't needed all along, but here you say its OK. No, I didn't say it was OK, or for that matter that it wasn't, what I was saying is that if driving a diode envelope detector with a finite source resistance does reduce distortion on weak or heavily modulated signals, then using a cathode follower would negate the distortion reducing effect of the finite source resistance, I didn't mean that a cathode follower would negate the distortion! One thing that I haven't seen mentioned is how the IFT plays into this since it has a sort of impedance inverting characteristic, as we discussed earlier, if the secondary is shorted, the impedance of the primary increases, just the opposite of what we would expect. I found that loading the secondary reduced the gain of the IF amp, ie, the load is reflected to the primary. When a double tuned IFT is used, the load reflected in the primary is opposite to the actual load on the secondary, that is when the load resistance presented to the secondary of the IFT by the diode goes up, the tube sees a load resistance from the primary that is going down, just the opposite of what you might expect. If by loading the secondary you meant decreasing the load resistance, then you should have found that the power gain actually went up, at least from the perspective of the primary. See my discussion above for more detail. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#33
|
|||
|
|||
|
In article , Patrick Turner
wrote: The "poor" AC/DC" load ratio does not matter if the output audio voltage asked from the circuit is low. But the "audio voltage asked from the circuit" is not low when the negative modulation percentage is high, it is actually quite high relatively speaking. I think you have to do some more thinking about how diode envelope detectors actually work. In a triode gain stage, if the DC carrying RL is large, and the cap coupled load is small, a similar effect occurs, and positive voltage swing is limited to a lot less than the B+. But the first few volts are still quite linear, and simply what one expects when load liane analysis is used. A triode gain stage is not a good analogy, as the signal level can be independently adjusted for a reasonable amount of distortion. In a detector it is the transmitter that controls the modulation, and there is not a lot you can do about it short of somehow adding in more carrier, or filtering out some of the sidebands. My own radio uses a good enough AC/DC load ratio to take the cut off effect fully into account. That may well be, but it certainly wasn't the case with the AM tuner schematic that you posted. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#34
|
|||
|
|||
|
"John Byrns" wrote in message ... In article , Patrick Turner wrote: John Byrns wrote: Now this seems like a reasonable concern about the effect of the diode detector on the IF amplifier, unlike your concern about the current pulses into the diode which are rendered irrelevant by the filtering/flywheel action of the IFT. The load varies during thre audio cycle Yes, that is correct, the reason is that the diode isn't "perfect". if the ripple voltage changes. No, that is wrong, even with a "perfect" diode which would present the driving circuit with a constant load over the complete audio cycle if such a diode existed, the ripple voltage will vary, being very low at the negative modulation peaks, and high at the positive modulation peaks, the amplitude of the ripple tracks the envelope, it is just the natural way a diode peak detector works, there is nothing to be done about it that I know of. You have made conflicting statements about the 455 kHz ripple voltage out of your detectors, first saying that your detector had a constant ripple voltage over the audio cycle, then saying that the ripple voltage varied over the audio cycle, and most recently you seem to be back to your detector somehow having constant ripple voltage. This doesn't add up and there is something fishy here. John, If I follow what you are saying, you are saying that the sawtooth decay of the capacitor changes and differs significantly between negative modulation peaks and positive modulation peaks. If so, then that ripple is caused by the RC circuit connected to the diode where the resistor is discharging the capacitor. If you replace the resistor with a constant current source (to discharge the capacitor), the ripple will be much more uniform. Or instead of tying the resistor to ground, use a much larger resistor connected to a negative supply. If the discharge current is constant (does not change due to the voltage on the capacitor), then this approach leads to ripple that does not vary with the voltage on the capacitor. I would think that this improves the linearity of the detector and this improve distortion, however, I don't know how much of an improvement it will make. I also believe it is a lot easier to do this in the solid state world than in the tube world. craigm |
|
#35
|
|||
|
|||
|
John Byrns wrote: In article , Patrick Turner wrote: John Byrns wrote: Now this seems like a reasonable concern about the effect of the diode detector on the IF amplifier, unlike your concern about the current pulses into the diode which are rendered irrelevant by the filtering/flywheel action of the IFT. The load varies during thre audio cycle Yes, that is correct, the reason is that the diode isn't "perfect". if the ripple voltage changes. No, that is wrong, even with a "perfect" diode which would present the driving circuit with a constant load over the complete audio cycle if such a diode existed, the ripple voltage will vary, being very low at the negative modulation peaks, and high at the positive modulation peaks, the amplitude of the ripple tracks the envelope, it is just the natural way a diode peak detector works, there is nothing to be done about it that I know of. There is something to be done about it. Use my design of detector. You have made conflicting statements about the 455 kHz ripple voltage out of your detectors, first saying that your detector had a constant ripple voltage over the audio cycle, then saying that the ripple voltage varied over the audio cycle, and most recently you seem to be back to your detector somehow having constant ripple voltage. This doesn't add up and there is something fishy here. Try my detector, and all will be revealed. The variation of ripple voltage is much lower than with most other types of detector. So the IF amp gain changes, ie, there is intermodulation distortion caused by the detector. Obviously the IF amplifier gain changes, I don't think anyone claimed otherwise. Interestingly, one of the things I have learned as a result of this discussion is that when a double tuned IFT is used to drive a real diode the gain does not change in the direction you would expect due to the impedance inverting properties of double tuned IFTs. My concerns are not irrelevant. I said your concerns about the diode RF current pulses are irrelevant in my opinion, I didn't say that the varying load presented by the diode over the modulation cycle is irrelevant. Interestingly the issue of the varying load seems to be much more complex than I have seen mentioned in any text book, I have never seen a text book which mentioned the impedance inverting properties of double tuned IFTs in connection with the varying diode load. Using my detector, what becomes irelevant is the loading effect by the detector circuit on the IFT and amp. My circuit removes such loading entirely. When the signal level is low, or the negative modulation high, the non linearity of the diode detector causes a lighter loading on the secondary of the IFT. This of course is going to create envelope distortion at the secondary of the IFT, and the plate of the IF amplifier tube, and many authorities claim this distortion is a good thing because it implies that given the finite source impedance of the IFT secondary, the drive voltage to the diode detector will increase at these points, partially offsetting the distortion of the envelope detector under these conditions. I don't know what authorities say that distortion is a "good thing" I didn't say that they said distortion was a good thing, I said that some authorities point out that a finite non zero source impedance will work to minimize the distortion caused by the diode peak detector at high modulation levels, the fact that the signal wave form at the anode of the IF amplifier tube becomes distorted as a result is irrelevant. Think of it as a distortion cancellation mechanism if you like, I know you have done work in that area yourself. There is an optimal load range for an IF amp pentode, too high, or too low results in greater distortion. Its important to have the IF tube operate with a load within the range for a fairly linear outcome, and at a not too high an output voltage, since there is no negative FB that can be applied around the device to linearize it, except perhaps by negative current FB by usung an unbypassed Rk. Pentodes have terrible distortion when the load is really high or low. The text books speak of this mechanism as if the diode were driven from a tube and a simple resistance, and it seems rather obvious that the idea works in that case, but when you put a double tuned IFT between the IF amplifier tube and the diode, then things change and it is no longer obvious exactly what happens, and what the final result is. I have crunched a few numbers and found that my hypothesized impedance inverting effect does indeed occur, causing the gain of the IF amplifier to decrease rather than increase as you might expect, but so far it appears that the increase in voltage driving the diode due to the higher load resistance presented by the diode more than compensates for the lowered gain of the IF amplifier tube. One unfortunate thing I have discovered is that the impedance inverting effect is most pronounced at the carrier frequency, and less so at the sideband frequencies, which means that this effect increases the modulation percentage. Fortunately in a narrow band IF this increase in modulation percentage is probably completely swamped by the bandpass effects of the IFT on the sidebands, and this might partly account for John Doty's comments that wideband AM IFs often sound more distorted. There are a lot of numbers to be crunched with respect to the effects of the double tuned IFT in this regard, it might make a good subject for someone's masters degree thesis, that is if they can find a professor interested in something like this, that has little relevance to current technology. Maybe only about 3 ppl in the world care about this whole issue at this time. I decided it'd be better to remove diode loading effects, period. No more vague notions to spoil the music. A cathode follower would negate this effect. You've been trying to tell me a CF isn't needed all along, but here you say its OK. No, I didn't say it was OK, or for that matter that it wasn't, ??? what I was saying is that if driving a diode envelope detector with a finite source resistance does reduce distortion on weak or heavily modulated signals, then using a cathode follower would negate the distortion reducing effect of the finite source resistance, I didn't mean that a cathode follower would negate the distortion! Depends how the diode and resistances are set up. But I would say the CF definately does negate, reduce, banish, expunge the distortion of the usual "conventional" detection circuit, so ppl should try it, if they can live with an extra twin triode in their set, but if not, use a couple of emitter follower buffers, using darlington pair connected signal transistors. These display very high input impedance, and much lower output impedance than any tube, and lower thd. One thing that I haven't seen mentioned is how the IFT plays into this since it has a sort of impedance inverting characteristic, as we discussed earlier, if the secondary is shorted, the impedance of the primary increases, just the opposite of what we would expect. I found that loading the secondary reduced the gain of the IF amp, ie, the load is reflected to the primary. When a double tuned IFT is used, the load reflected in the primary is opposite to the actual load on the secondary, that is when the load resistance presented to the secondary of the IFT by the diode goes up, the tube sees a load resistance from the primary that is going down, just the opposite of what you might expect. Nope, when I placed 100k resistors on the IFT secs, the gain went down. If by loading the secondary you meant decreasing the load resistance, then you should have found that the power gain actually went up, at least from the perspective of the primary. See my discussion above for more detail. If a 100k resistance is connected across the secondary of the last IFT LC, then the the voltage ratio of the IFT increases. Less output IF voltage comes out, so less AVC voltage, to the IF amp is biased more positively, so there is more voltage produced at the primary or anode connection of the IFT. That's the only mechanism I know by which gain will increase on the IF amp tube, if R loading is applied to the IFT sec. If the AVC voltage is removed and replaced with a fixed bias then its a different story, and loading the sec of an IFT presents a lower load at the pri, and the gain falls. That's what I have found, and there are limits about how much loading to apply, or the gain reduces to too low a value. 100k is about right for most IFT LC circuits to widen bw. 47k is a bit too much, but might be OK if you had two IF amps. If two IF amps are used, perhaps 22k could be used across the second IFT, since the amp need only make a small amount of gain and a small output voltage ahead of the second IF amp. The 3 IFTs would give better skirt selectivity. Patrick Turner. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#36
|
|||
|
|||
|
John Byrns wrote: In article , Patrick Turner wrote: The "poor" AC/DC" load ratio does not matter if the output audio voltage asked from the circuit is low. But the "audio voltage asked from the circuit" is not low when the negative modulation percentage is high, it is actually quite high relatively speaking. I think you have to do some more thinking about how diode envelope detectors actually work. I suggest you try my circuit, before you say I dunno how enevlope detectors work. In a triode gain stage, if the DC carrying RL is large, and the cap coupled load is small, a similar effect occurs, and positive voltage swing is limited to a lot less than the B+. But the first few volts are still quite linear, and simply what one expects when load liane analysis is used. A triode gain stage is not a good analogy, as the signal level can be independently adjusted for a reasonable amount of distortion. In a detector it is the transmitter that controls the modulation, and there is not a lot you can do about it short of somehow adding in more carrier, or filtering out some of the sidebands. Build a better detector, for a start. My own radio uses a good enough AC/DC load ratio to take the cut off effect fully into account. That may well be, but it certainly wasn't the case with the AM tuner schematic that you posted. All the circuits I have posted work fine. Better than most commercial bean counter driven designs. If you don't try my circuits, maybe you'll never know how good they are. Make the measurement comparisons between what I have posted, and what exists in a typical junky radio laying around. I am not afraid of your test results. Patrick Turner. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#37
|
|||
|
|||
|
craigm wrote: "John Byrns" wrote in message ... In article , Patrick Turner wrote: John Byrns wrote: Now this seems like a reasonable concern about the effect of the diode detector on the IF amplifier, unlike your concern about the current pulses into the diode which are rendered irrelevant by the filtering/flywheel action of the IFT. The load varies during thre audio cycle Yes, that is correct, the reason is that the diode isn't "perfect". if the ripple voltage changes. No, that is wrong, even with a "perfect" diode which would present the driving circuit with a constant load over the complete audio cycle if such a diode existed, the ripple voltage will vary, being very low at the negative modulation peaks, and high at the positive modulation peaks, the amplitude of the ripple tracks the envelope, it is just the natural way a diode peak detector works, there is nothing to be done about it that I know of. You have made conflicting statements about the 455 kHz ripple voltage out of your detectors, first saying that your detector had a constant ripple voltage over the audio cycle, then saying that the ripple voltage varied over the audio cycle, and most recently you seem to be back to your detector somehow having constant ripple voltage. This doesn't add up and there is something fishy here. John, If I follow what you are saying, you are saying that the sawtooth decay of the capacitor changes and differs significantly between negative modulation peaks and positive modulation peaks. If so, then that ripple is caused by the RC circuit connected to the diode where the resistor is discharging the capacitor. If you replace the resistor with a constant current source (to discharge the capacitor), the ripple will be much more uniform. Or instead of tying the resistor to ground, use a much larger resistor connected to a negative supply. In my detector with Ge diode after a CF the 1M resistor with 55 volts across it gives discharge current of 0.055 mA from the 270 pF used for the charge C off the diode, and this current stays fairly constant, regardless of signal level. But were the R grounded so that the current was negligible at low IF sig levels, then huge variations in ripple voltage occur, and well as audio distortions at high audio voltages, especially if the audio F is 10 khz. Try building a detector or two, and use your CRO to see what happens. Make sure the probe capoacitance does not affect what you are viewing. If the discharge current is constant (does not change due to the voltage on the capacitor), then this approach leads to ripple that does not vary with the voltage on the capacitor. I would think that this improves the linearity of the detector and this improve distortion, however, I don't know how much of an improvement it will make. I also believe it is a lot easier to do this in the solid state world than in the tube world. No it isn't, and the circuit using two CF tubes and a couple of Ge diodes etc prove this, and the dynamic range of my detector circuit is way above most SS circuits. Its no use only discussing this for 100 years to understand; you must away from the PC, and build and measure something. Observe the waveforms. Patrick Turner. craigm |
|
#38
|
|||
|
|||
|
In article , "craigm"
wrote: John, If I follow what you are saying, you are saying that the sawtooth decay of the capacitor changes and differs significantly between negative modulation peaks and positive modulation peaks. If so, then that ripple is caused by the RC circuit connected to the diode where the resistor is discharging the capacitor. If you replace the resistor with a constant current source (to discharge the capacitor), the ripple will be much more uniform. Or instead of tying the resistor to ground, use a much larger resistor connected to a negative supply. If the discharge current is constant (does not change due to the voltage on the capacitor), then this approach leads to ripple that does not vary with the voltage on the capacitor. With one exception, see below, the ripple should be more or less constant with this scheme. At one time I was enamored with this scheme, but gave it up for reasons I no longer remember. A reason that comes to mind now, but was not the one I can't remember, is that while the ripple may be relatively constant with modulation, the frequency at which tangential clipping sets in will be dependent on the carrier amplitude, since the slope of the discharge is independent of carrier level. I would think that this improves the linearity of the detector and this improve distortion, however, I don't know how much of an improvement it will make. I'm not sure, but I would think that to a first approximation the variable ripple voltage of the normal envelope detector doesn't add distortion or nonlinearity, but with the current source approach, there is going to be distortion on negative modulation peaks when the amplitude of the ripple wave form is clamped by the diode before the next carrier cycle starts. For example at 100% negative modulation the ripple voltage must go to zero even with a current source, and this is a nonlinear effect, since the ripple voltage is constant up to a certain modulation depth, at which point it abruptly starts reducing towards zero on 100 % negative modulation peaks. I also believe it is a lot easier to do this in the solid state world than in the tube world. There are high voltages available in vacuum tube circuits which in combination with a large resistor would make a pretty good approximation to a current source. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#39
|
|||
|
|||
|
In article , Patrick Turner
wrote: John Byrns wrote: In article , Patrick Turner wrote: John Byrns wrote: One thing that I haven't seen mentioned is how the IFT plays into this since it has a sort of impedance inverting characteristic, as we discussed earlier, if the secondary is shorted, the impedance of the primary increases, just the opposite of what we would expect. I found that loading the secondary reduced the gain of the IF amp, ie, the load is reflected to the primary. When a double tuned IFT is used, the load reflected in the primary is opposite to the actual load on the secondary, that is when the load resistance presented to the secondary of the IFT by the diode goes up, the tube sees a load resistance from the primary that is going down, just the opposite of what you might expect. Nope, when I placed 100k resistors on the IFT secs, the gain went down. Well we have to be careful how we are defining "gain", and where we are measuring it. As I said this could be a subject for a Masters Thesis, and I have only crunched the first few of many numbers that would have to be crunched if I were writing the complete thesis. In any case when you add a 100k resistor across the secondary of a double tuned IFT you will see an increase in primary voltage if it is being driven from a constant current source like a pentode valve. The increase in primary voltage directly implies an increase in the voltage gain at that point, and since the voltage increased, and the current is constant, the power delivered to the IFT must also have increased and hence the power gain at the primary also increased. Now as far as what happens at the secondary, that is affected by many factors, among them where the losses in the circuit are, and what the value of "k" is. That is a major part of the thesis and one that I have only taken a small peak at. In any case I think it is probably safe to say that the voltage gain to the secondary, that is including the IFT, goes down when you add the 100k resistor. The inescapable fact is that when you put a resistor across the secondary of an IFT, the gain as measured at the plate of the IF amplifier tube goes up, and at least the voltage gain at the secondary of the IFT goes down. If by loading the secondary you meant decreasing the load resistance, then you should have found that the power gain actually went up, at least from the perspective of the primary. See my discussion above for more detail. If a 100k resistance is connected across the secondary of the last IFT LC, then the the voltage ratio of the IFT increases. Less output IF voltage comes out, so less AVC voltage, to the IF amp is biased more positively, so there is more voltage produced at the primary or anode connection of the IFT. Let's leave the AGC voltage out of this and make it constant, AGC is a separate issue and should be a separate discussion. That's the only mechanism I know by which gain will increase on the IF amp tube, if R loading is applied to the IFT sec. This simply indicates to me that you haven't done the lab work I assigned you earlier, and this is the second time I have had to remind you about this. I will remind you of the assignment again, it is to take a properly tuned IFT and measure the primary impedance at resonance, and then to short the secondary and again measure the primary impedance. Extra credit will be given for also measuring the primary impedance with a 100 k resistor connected across the secondary, and for repeating all the above measurements with several different values of "k". Once you have all this data logged in your notebook, you can sit down in your lounge and contemplate what it all means. That the world isn't as simple as it first seems is one conclusion, and as an old associate often said, you may reach the next level of consciousness. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
|
#40
|
|||
|
|||
|
John Byrns wrote: In article , "craigm" wrote: John, If I follow what you are saying, you are saying that the sawtooth decay of the capacitor changes and differs significantly between negative modulation peaks and positive modulation peaks. If so, then that ripple is caused by the RC circuit connected to the diode where the resistor is discharging the capacitor. If you replace the resistor with a constant current source (to discharge the capacitor), the ripple will be much more uniform. Or instead of tying the resistor to ground, use a much larger resistor connected to a negative supply. If the discharge current is constant (does not change due to the voltage on the capacitor), then this approach leads to ripple that does not vary with the voltage on the capacitor. With one exception, see below, the ripple should be more or less constant with this scheme. At one time I was enamored with this scheme, but gave it up for reasons I no longer remember. A reason that comes to mind now, but was not the one I can't remember, is that while the ripple may be relatively constant with modulation, the frequency at which tangential clipping sets in will be dependent on the carrier amplitude, since the slope of the discharge is independent of carrier level. The "tangential clipping" also becomes more likely if the audio modulation F is higher and the audio voltage is high. That's because of the discharge rate of the cap charged by the diode. Its worse with a detector which does not have a biased diode and constant current source. I would think that this improves the linearity of the detector and this improve distortion, however, I don't know how much of an improvement it will make. I'm not sure, but I would think that to a first approximation the variable ripple voltage of the normal envelope detector doesn't add distortion or nonlinearity, but with the current source approach, there is going to be distortion on negative modulation peaks when the amplitude of the ripple wave form is clamped by the diode before the next carrier cycle starts. For example at 100% negative modulation the ripple voltage must go to zero even with a current source, and this is a nonlinear effect, since the ripple voltage is constant up to a certain modulation depth, at which point it abruptly starts reducing towards zero on 100 % negative modulation peaks. In my detector, the output voltage cannot ever go to to zero, or 0V. Its at 55 volts even with no RF input, and then when a carrier of 10v p-p appears, you get a rise in the direct voltage voltage across the 270pF of 5 volts to +60v. Then if there was 100% modulation, you get a varying voltage at the 270 pF between +55v and +65v. For low frequency audio, there is almost no change in the ripple voltage at any part of the audio wave form. But at 20 kHz modulation F, the ripple voltage is less on the negative going slopes of the sine wave. But the recovered audio wave is very close to the shape of the envelope. The process introduces a slight phase lag in HF audio modulation. The discharge rate of the 270 pF in my detector and the 1M form a time constant of 270 uS. But when you draw the curve for a 270 uS cap discharge, the first 10v discharge is a nearly a straight line discharge rate of 55v per 135 uS, or 2.45 V/uS, or 5V/12.2 uS, and this allows an undistorted sine wave of 2.5v peak v and at 20.49 kHz. Or 10 peak volts at 5 kHz. The undistorted sine wave maximum rises as F reduces. But the HF content of music reduces as F rises, so this is not a problem. Its possible to increase the current discharge from the 270 pF cap by using 470k instead of 1M but all this would do is raise the undistorted threshold of the sine wave by twice, when it isn't really needed if you arrange the detector to produce no more than around 3 vrms of audio at 100% modulation. This sort of audio voltage would mean the IF amp isn't working too hard, and into its high distortion region. I also believe it is a lot easier to do this in the solid state world than in the tube world. There are high voltages available in vacuum tube circuits which in combination with a large resistor would make a pretty good approximation to a current source. If you use my detector circuit, you could easily establish a -200v voltage source, and take a 4.7M resistor from the 270 pF, to give a closer approximation to a CCS. Patrick Turner. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
| Reply |
| Thread Tools | |
| Display Modes | |
|
|
Similar Threads
|
||||
| Thread | Forum | |||
| Infrasound detector | Tech | |||
| An Improved Method of Noise Cancellation | Tech | |||
| Norman Koren-Spiced Improved Preamp | Vacuum Tubes | |||
| Radar Detector Advice | Car Audio | |||
| The other triode of my infinite impedance AM detector had distortionfrom stray RF | Vacuum Tubes | |||
Linear Mode
