Home |
Search |
Today's Posts |
|
#1
|
|||
|
|||
DELTA - Cap Analysis; Cap Biasing
In preparation for replacing or in some cases removing the electrolytics in my Soundcraft Delta, I've been cataloging those in the signal path. I've made offset measurements on both sides of many caps and found some with zero offset on either side. Good candidates for removal, I'd guess. (I did throw a 0.5 uf cap across the scope probe to filter the AC noise component, which was obscuring the small DC offsets.) I'm curious about two things. First, in a few cases where offset is present (a few mV) the voltage is opposite the labeled polarity on the cap. Obviously the cap hasn't blown apart from the small reversal, but does the distortion go up? Second item is why certain cap values were chosen by the designer. Most of the interstage couplers are 47 uf/25 V -- perhaps a large enough value so that after going through a number of them the LF roll off is still below 20 Hz. (Though if stuck using these, perhaps I should bump them up to 100 uf for better LF?) What I don't quite get, though, are the 2.2 uf/50V parts on the fader returns and outputs of most other rotary level controls. Is the idea to band-limit the LF so that the control action is clean(er) across a wide range of dirt in the part and aging? Or am I missing something? Next, the input to the L/R mix summing amps (and aux bus summers) is 1000 uf/6.3V. Why the huge value here? Why not 100 or 470? (Not saying they should be that way, just curious as to why.) CAP BIASING - where electrolytic caps must be used, this seems like an interesting idea to try, but I wonder how best to do this. How much bias V do we need? If I grab, say, the 17V positive rail through dropping resistors to provide a couple of volts, won't there be increased crosstalk by some signal getting back into the pos. rail and into other bias injection points? (Not to mention the op amps themselves.) Should I use a little regulator, say a 7803 (if there is such a beast, or a 7805 if not) local to each cap "pair" I'd want to bias? By pair I mean two electros replacing an original single cap, with the + sides tied back-to-back and the bias injected there. Would the output stages of such regulators impact the signal fidelity? Thanks again for the help and experiences in this area. Thanks to those who have responded to these Delta threads in the past, especially Mr. Williams, Mr. Dorsey, and Mr. Sank (sorry if I missed anybody). Frank Stearns Mobile Audio -- |
#2
|
|||
|
|||
On Fri, 22 Apr 2005 19:45:23 -0000, in rec.audio.pro Frank Stearns
wrote: snip Second item is why certain cap values were chosen by the designer. Most of the interstage couplers are 47 uf/25 V -- perhaps a large enough value so that after going through a number of them the LF roll off is still below 20 Hz. (Though if stuck using these, perhaps I should bump them up to 100 uf for better LF?) What I don't quite get, though, are the 2.2 uf/50V parts on the fader returns and outputs of most other rotary level controls. Is the idea to band-limit the LF so that the control action is clean(er) across a wide range of dirt in the part and aging? Or am I missing something? snip. Its quite interesting to look at the LF phase response, when simulated in LT spice, using voltage followers instead of all the gain stages. Just make 6 or so opamps with the usual AC coupling between stages and analyse the o/p wierd , probably says a lot about the "sound" of a console Thanks again for the help and experiences in this area. Thanks to those who have responded to these Delta threads in the past, especially Mr. Williams, Mr. Dorsey, and Mr. Sank (sorry if I missed anybody). Frank Stearns Mobile Audio martin After the first death, there is no other. (Dylan Thomas) |
#3
|
|||
|
|||
martin griffith wrote:
On Fri, 22 Apr 2005 19:45:23 -0000, in rec.audio.pro Frank Stearns wrote: snip Second item is why certain cap values were chosen by the designer. Most of the interstage couplers are 47 uf/25 V -- perhaps a large enough value so that after going through a number of them the LF roll off is still below 20 Hz. (Though if stuck using these, perhaps I should bump them up to 100 uf for better LF?) What I don't quite get, though, are the 2.2 uf/50V parts on the fader returns and outputs of most other rotary level controls. Is the idea to band-limit the LF so that the control action is clean(er) across a wide range of dirt in the part and aging? Or am I missing something? snip. Its quite interesting to look at the LF phase response, when simulated in LT spice, using voltage followers instead of all the gain stages. Just make 6 or so opamps with the usual AC coupling between stages and analyse the o/p wierd , probably says a lot about the "sound" of a console Unlikely. Kevin Aylward http://www.anasoft.co.uk SuperSpice, a very affordable Mixed-Mode Windows Simulator with Schematic Capture, Waveform Display, FFT's and Filter Design. |
#4
|
|||
|
|||
Frank Stearns wrote:
I'm curious about two things. First, in a few cases where offset is present (a few mV) the voltage is opposite the labeled polarity on the cap. Obviously the cap hasn't blown apart from the small reversal, but does the distortion go up? Yes. Second item is why certain cap values were chosen by the designer. Most of the interstage couplers are 47 uf/25 V -- perhaps a large enough value so that after going through a number of them the LF roll off is still below 20 Hz. (Though if stuck using these, perhaps I should bump them up to 100 uf for better LF?) For the most part, it's a matter of the low frequency corner vs. price and space. What I don't quite get, though, are the 2.2 uf/50V parts on the fader returns and outputs of most other rotary level controls. Is the idea to band-limit the LF so that the control action is clean(er) across a wide range of dirt in the part and aging? Or am I missing something? No, they are probably going into a thing with a high input impedance. Remember the low end corner is set by the input impedance and the series capacitance. Next, the input to the L/R mix summing amps (and aux bus summers) is 1000 uf/6.3V. Why the huge value here? Why not 100 or 470? (Not saying they should be that way, just curious as to why.) Because the input impedance on that amp is VERY low to keep the noise down. CAP BIASING - where electrolytic caps must be used, this seems like an interesting idea to try, but I wonder how best to do this. How much bias V do we need? If I grab, say, the 17V positive rail through dropping resistors to provide a couple of volts, won't there be increased crosstalk by some signal getting back into the pos. rail and into other bias injection points? (Not to mention the op amps themselves.) Enough that it never goes through the zero-crossing region ever. Should I use a little regulator, say a 7803 (if there is such a beast, or a 7805 if not) local to each cap "pair" I'd want to bias? By pair I mean two electros replacing an original single cap, with the + sides tied back-to-back and the bias injected there. Would the output stages of such regulators impact the signal fidelity? No, you can use resistors since you have a rail that is clean. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#5
|
|||
|
|||
Scott Dorsey wrote:
Frank Stearns wrote: I'm curious about two things. First, in a few cases where offset is present (a few mV) the voltage is opposite the labeled polarity on the cap. Obviously the cap hasn't blown apart from the small reversal, but does the distortion go up? Yes. Not really. With only a few mv DC across the cap, distortion is not an issue. What matters is the ac level across it. This depends on the resistance loading the cap. Second item is why certain cap values were chosen by the designer. Most of the interstage couplers are 47 uf/25 V -- perhaps a large enough value so that after going through a number of them the LF roll off is still below 20 Hz. (Though if stuck using these, perhaps I should bump them up to 100 uf for better LF?) For the most part, it's a matter of the low frequency corner vs. price and space. And cost possibly. What I don't quite get, though, are the 2.2 uf/50V parts on the fader returns and outputs of most other rotary level controls. Is the idea to band-limit the LF so that the control action is clean(er) across a wide range of dirt in the part and aging? Or am I missing something? No, they are probably going into a thing with a high input impedance. Remember the low end corner is set by the input impedance and the series capacitance. Next, the input to the L/R mix summing amps (and aux bus summers) is 1000 uf/6.3V. Why the huge value here? Why not 100 or 470? (Not saying they should be that way, just curious as to why.) Because the input impedance on that amp is VERY low to keep the noise down. Actually the summing amp is essentially zero input resistance. The cap size is chosen based on the series resister value feeding the virtual Ov point, otherwise it would be several farads! CAP BIASING - where electrolytic caps must be used, this seems like an interesting idea to try, but I wonder how best to do this. How much bias V do we need? If I grab, say, the 17V positive rail through dropping resistors to provide a couple of volts, won't there be increased crosstalk by some signal getting back into the pos. The main issue here is noise form the power supply feeding into the input. rail and into other bias injection points? (Not to mention the op amps themselves.) Enough that it never goes through the zero-crossing region ever. What are you trying to say here? It dosnt make sense. Should I use a little regulator, say a 7803 (if there is such a beast, or a 7805 if not) local to each cap "pair" I'd want to bias? This is making a mountain out of a mole hill. If one really needs to have an unbiased electrolyte coupling cap, use a bi-polar one, which is two back to back ones. Kevin Aylward http://www.anasoft.co.uk SuperSpice, a very affordable Mixed-Mode Windows Simulator with Schematic Capture, Waveform Display, FFT's and Filter Design. |
#6
|
|||
|
|||
On Sat, 23 Apr 2005 17:07:47 GMT, "Kevin Aylward"
wrote: Enough that it never goes through the zero-crossing region ever. What are you trying to say here? It dosnt make sense. If the absolute voltage across the capacitor is reversed, or actually even too low, the capacitor is nonlinear. This is making a mountain out of a mole hill. If one really needs to have an unbiased electrolyte coupling cap, use a bi-polar one, which is two back to back ones. Yes, two unbiased ones. Chris Hornbeck "This has been an account for those who don't keep them" J-LG, _Tout Va Bien_ 1972 |
#7
|
|||
|
|||
Kevin Aylward wrote:
rail and into other bias injection points? (Not to mention the op amps themselves.) Enough that it never goes through the zero-crossing region ever. What are you trying to say here? It dosnt make sense. The problems with electrolytic linearity issues all have to do with cases when there is very little voltage across the cap. So the recommended design practice is to make sure there is sufficient DC bias across the cap so that at any reasonable signal level, the cap always has some voltage across it for any part of the waveform. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#8
|
|||
|
|||
Scott Dorsey wrote:
Kevin Aylward wrote: rail and into other bias injection points? (Not to mention the op amps themselves.) Enough that it never goes through the zero-crossing region ever. What are you trying to say here? It dosnt make sense. The problems with electrolytic linearity issues all have to do with cases when there is very little voltage across the cap. So the recommended design practice is to make sure there is sufficient DC bias across the cap so that at any reasonable signal level, the cap always has some voltage across it for any part of the waveform. --scott Ok, I understand what you are actually saying now with the term "zero crossing" in this context. This is a little unconventional way of talking about it. However, when there is litle signal accross a capacitor, it won't generate any significant distortion, so its simply not an issue. For example, its not uncommon to have a single electrolytic cap in the feedback network of a power amp putting out +/-70V. I can assure you that such a cap typically allows for the amp to have 0.001% thd. Kevin Aylward http://www.anasoft.co.uk SuperSpice, a very affordable Mixed-Mode Windows Simulator with Schematic Capture, Waveform Display, FFT's and Filter Design. |
#9
|
|||
|
|||
Kevin Aylward wrote:
For example, its not uncommon to have a single electrolytic cap in the feedback network of a power amp putting out +/-70V. I can assure you that such a cap typically allows for the amp to have 0.001% thd. Kevin, have you tried replacing that cap with something relevantly non-electrolytic and listened? Kevin Aylward Kind regards Peter Larsen -- ******************************************* * My site is at: http://www.muyiovatki.dk * ******************************************* |
#10
|
|||
|
|||
Scott Dorsey wrote:
Kevin Aylward wrote: rail and into other bias injection points? (Not to mention the op amps themselves.) Enough that it never goes through the zero-crossing region ever. What are you trying to say here? It dosnt make sense. The problems with electrolytic linearity issues all have to do with cases when there is very little voltage across the cap. So the recommended design practice is to make sure there is sufficient DC bias across the cap so that at any reasonable signal level, the cap always has some voltage across it for any part of the waveform. Shame about every console made for about the last 25+ yrs I guess ? Including the very top names. Truth is - zero biased electros are just fine - when used *intelligently*. Graham |
#11
|
|||
|
|||
Pooh Bear wrote:
Scott Dorsey wrote: The problems with electrolytic linearity issues all have to do with cases when there is very little voltage across the cap. So the recommended design practice is to make sure there is sufficient DC bias across the cap so that at any reasonable signal level, the cap always has some voltage across it for any part of the waveform. Shame about every console made for about the last 25+ yrs I guess ? Including the very top names. It's the case for a lot of them, but take a look at some of the Neve stuff for an example of how to do it right. Truth is - zero biased electros are just fine - when used *intelligently*. For a lot of things, that's perfectly true. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#12
|
|||
|
|||
Hi,
you mentioned you are looking at the DC bias with a scope. Good. I hope it is a dual trace scope. I have a suggestion. Set up the two channels of the scope by attaching the two probes to the same signal and adjust the gain and position of the scope channels so the two traces overlap. Use AC coupling on both channels of the scope. After you have done that, now connect the two probes to the cap in question, one probe on each side of the cap. Feed a tone through your circuit. You can now easily compare the signal in and out of the cap. At low audio frequencies you will probably see some phase shift. You can also see if the waveform is distorted by the cap. If you have a distortion analyzer or spectrum analyzer (RTA) (and a good generator) you can do it that way too. If you want to make these design changes, I suggest you try to measure the problem first so you can see how much improvement you made. After making the measurement you may decide you don't need to make the changes. Let us know what you find. Regarding your question about the cap values. The cap value needed is also a function of the impedance of the circuit. At 60 Ohms you need about 270 uF for -3 dB at 10 Hz. At 6,000 Ohms you only need 2.7 uF for -3 dB at 10 Hz. Mark |
#13
|
|||
|
|||
"Frank Stearns" I'm curious about two things. First, in a few cases where offset is present (a few mV) the voltage is opposite the labeled polarity on the cap. Obviously the cap hasn't blown apart from the small reversal, but does the distortion go up? ** No - not from just a few mV. Second item is why certain cap values were chosen by the designer. Most of the interstage couplers are 47 uf/25 V -- perhaps a large enough value so that after going through a number of them the LF roll off is still below 20 Hz. (Though if stuck using these, perhaps I should bump them up to 100 uf for better LF?) ** By using a large value, the AC signal voltage across the cap is kept small even at 20 Hz and full levels. What I don't quite get, though, are the 2.2 uf/50V parts on the fader returns and outputs of most other rotary level controls. Is the idea to band-limit the LF so that the control action is clean(er) across a wide range of dirt in the part and aging? Or am I missing something? ** The smaller value cap has lower DC leakage current - leakage would create noise when the fader is moved. Plus, the impedance of the circuit is higher here. Next, the input to the L/R mix summing amps (and aux bus summers) is 1000 uf/6.3V. Why the huge value here? Why not 100 or 470? (Not saying they should be that way, just curious as to why.) ** The summing inputs are "virtual earth" points - so no signal voltage appears at that point. This means that ( ideally) no signal crosstalk can find its way from one channel to another via the summing resistors. A small value cap at that point would allow low frequency crosstalk - so a big one is used instead. CAP BIASING - where electrolytic caps must be used, this seems like an interesting idea to try, but I wonder how best to do this. ** Forget it - to messy and troublesome. se bi-polar electos as these are *designed* to work with no bias and not distort. ............... Phil |
#14
|
|||
|
|||
Phil Allison wrote:
"Frank Stearns" I'm curious about two things. First, in a few cases where offset is present (a few mV) the voltage is opposite the labeled polarity on the cap. Obviously the cap hasn't blown apart from the small reversal, but does the distortion go up? ** No - not from just a few mV. Second item is why certain cap values were chosen by the designer. Most of the interstage couplers are 47 uf/25 V -- perhaps a large enough value so that after going through a number of them the LF roll off is still below 20 Hz. (Though if stuck using these, perhaps I should bump them up to 100 uf for better LF?) ** By using a large value, the AC signal voltage across the cap is kept small even at 20 Hz and full levels. What I don't quite get, though, are the 2.2 uf/50V parts on the fader returns and outputs of most other rotary level controls. Is the idea to band-limit the LF so that the control action is clean(er) across a wide range of dirt in the part and aging? Or am I missing something? ** The smaller value cap has lower DC leakage current - leakage would create noise when the fader is moved. In principle sure, in practise, irrelevant. The capacitor value at these points are not chosen for leakage considerations. Capacitors are usually chosen to be the smallest consistent with the desired frequency response, and possible current noise for the mic input. Plus, the impedance of the circuit is higher here. Next, the input to the L/R mix summing amps (and aux bus summers) is 1000 uf/6.3V. Why the huge value here? Why not 100 or 470? (Not saying they should be that way, just curious as to why.) ** The summing inputs are "virtual earth" points - so no signal voltage appears at that point. This means that ( ideally) no signal crosstalk can find its way from one channel to another via the summing resistors. A small value cap at that point would allow low frequency crosstalk - so a big one is used instead. This is not why a summer junction is used. Typically the op-amps drive the summer with a very low output resistance such that one signal feeding the other through its summing resisters will have its x-talk attenuated by this ideally zero output resistance, i.e. say 1/10k. The virtual earth summer is used to prevent *level* interaction between channels. Consider two channels sourced through say, 10k resisters with the load 10k. If one channel was now switched off by opening the resister, the gain of the other channel would change significantly due to the open channel not being in parallel with the load. The virtual summer keeps the resistance that each channel sees indepandant of the other channels state for gain reasons. This has nothing to do with x-talk. Kevin Aylward http://www.anasoft.co.uk SuperSpice, a very affordable Mixed-Mode Windows Simulator with Schematic Capture, Waveform Display, FFT's and Filter Design. |
#15
|
|||
|
|||
Some pig arrogant pommy nut case: Phil Allison wrote: ** The smaller value cap has lower DC leakage current - leakage would create noise when the fader is moved. In principle sure, in practise, irrelevant. The capacitor value at these points are not chosen for leakage considerations. ** Arrant bull****. Next, the input to the L/R mix summing amps (and aux bus summers) is 1000 uf/6.3V. Why the huge value here? Why not 100 or 470? (Not saying they should be that way, just curious as to why.) ** The summing inputs are "virtual earth" points - so no signal voltage appears at that point. This means that ( ideally) no signal crosstalk can find its way from one channel to another via the summing resistors. A small value cap at that point would allow low frequency crosstalk - so a big one is used instead. This is not why a summer junction is used. ** Blatant straw man. The issue is the reason for the high cap value. Typically the op-amps drive the summer with a very low output resistance such that one signal feeding the other through its summing resisters will have its x-talk attenuated by this ideally zero output resistance, i.e. say 1/10k. ** Wrong. In a great many desks, Soundcraft and others, group and main buss summing resistors are driven from the wipers of the "pan" controls. .............. Phil |
#16
|
|||
|
|||
Phil Allison wrote:
Some pig arrogant pommy nut case: Phil Allison wrote: ** The smaller value cap has lower DC leakage current - leakage would create noise when the fader is moved. In principle sure, in practise, irrelevant. The capacitor value at these points are not chosen for leakage considerations. ** Arrant bull****. Here we go again... The dc leakage at the typical mv of offset between stages is completely insignificant. The capacitor is just there to ensure that as close to zero is on the input of each stage, so that dc offsets don't multiply up. In the ideal case, the caps dont do anything. Next, the input to the L/R mix summing amps (and aux bus summers) is 1000 uf/6.3V. Why the huge value here? Why not 100 or 470? (Not saying they should be that way, just curious as to why.) ** The summing inputs are "virtual earth" points - so no signal voltage appears at that point. This means that ( ideally) no signal crosstalk can find its way from one channel to another via the summing resistors. A small value cap at that point would allow low frequency crosstalk - so a big one is used instead. This is not why a summer junction is used. ** Blatant straw man. I am correcting your erroneous assertion. The issue is the reason for the high cap value. The fundamental reason for the high capacitor value is ensure that the gain settings of each channel do not interact with other. It is not there for x-talk reduction, although it might well do that as well. Typically the op-amps drive the summer with a very low output resistance such that one signal feeding the other through its summing resisters will have its x-talk attenuated by this ideally zero output resistance, i.e. say 1/10k. ** Wrong. In a great many desks, Soundcraft and others, group and main buss summing resistors are driven from the wipers of the "pan" controls. Oh dear...those cheaper ones...well, I'll give you some points for that one... Nothing changes mate. The fundamental reason, for the virtual earth summer, as I correctly stated, is to prevent selection of one signal effecting the level of another signal. Try adding voltages any other way, and you'll get the point, but then probably not. Kevin Aylward http://www.anasoft.co.uk SuperSpice, a very affordable Mixed-Mode Windows Simulator with Schematic Capture, Waveform Display, FFT's and Filter Design. |
#17
|
|||
|
|||
Some pig arrogant, ranting INSANE pommy arsehole: Phil Allison: ** The summing inputs are "virtual earth" points - so no signal voltage appears at that point. This means that ( ideally) no signal crosstalk can find its way from one channel to another via the summing resistors. A small value cap at that point would allow low frequency crosstalk - so a big one is used instead. This is not why a summer junction is used. ** Blatant straw man. I am correcting your erroneous assertion. ** No such assertion was ever posted by me. You are not sane, Kev. The issue is the reason for the high cap value. The fundamental reason for the high capacitor value is ensure that the gain settings of each channel do not interact with other. ** More arrant bull****. It is not there for x-talk reduction, although it might well do that as well. * Yet another total self contradiction. You are not sane, Kev. Typically the op-amps drive the summer with a very low output resistance such that one signal feeding the other through its summing resisters will have its x-talk attenuated by this ideally zero output resistance, i.e. say 1/10k. ** Wrong. In a great many desks, Soundcraft and others, group and main buss summing resistors are driven from the wipers of the "pan" controls. Oh dear... ** Is your missus dead yet ? those cheaper ones...well, I'll give you some points for that one... ** You are not sane, Kev. Nothing changes mate. The fundamental reason, for the virtual earth summer, as I correctly stated, is to prevent selection of one signal effecting the level of another signal. ** Same worthless old straw man. You are not sane, Kev. Try adding voltages any other way, and you'll get the point, but then probably not. ** You are not sane, Kev. **** the hell off. .................. Phil |
#18
|
|||
|
|||
Frank Stearns wrote:
In preparation for replacing or in some cases removing the electrolytics in my Soundcraft Delta, I've been cataloging those in the signal path. I've made offset measurements on both sides of many caps and found some with zero offset on either side. Good candidates for removal, I'd guess. (I did throw a 0.5 uf cap across the scope probe to filter the AC noise component, which was obscuring the small DC offsets.) Interesting. It seems like this information would be pretty easy to confirm from the schematics. I'm curious about two things. First, in a few cases where offset is present (a few mV) the voltage is opposite the labeled polarity on the cap. Obviously the cap hasn't blown apart from the small reversal, but does the distortion go up? In theory yes, but at this point actual performance is the most important thing. If you're so worried about distortion, why haven't you measured it? If you've got a PC with an audio interface, clever use of a free software package called the Audio Rightmark will give you a pretty complete report on noise, distortion, and frequency response in about a minute. Second item is why certain cap values were chosen by the designer. Most of the interstage couplers are 47 uf/25 V -- perhaps a large enough value so that after going through a number of them the LF roll off is still below 20 Hz. (Though if stuck using these, perhaps I should bump them up to 100 uf for better LF?) The size of the cap is only half the story - how big are the resistors that load the cap down? Tubed equipment usually has coupling caps on the order of 0.1 uF to 0.5 uF, but the load impedances are typically higher. What I don't quite get, though, are the 2.2 uf/50V parts on the fader returns and outputs of most other rotary level controls. Is the idea to band-limit the LF so that the control action is clean(er) across a wide range of dirt in the part and aging? Or am I missing something? You're missing 50% of the relevant information, which is the impedance that sets the time constant of an RC circuit, along with the size of the cap. Next, the input to the L/R mix summing amps (and aux bus summers) is 1000 uf/6.3V. Why the huge value here? Why not 100 or 470? (Not saying they should be that way, just curious as to why.) I guess that this circuit loads the caps down with lower impedances than the other circuits. CAP BIASING - where electrolytic caps must be used, this seems like an interesting idea to try, but I wonder how best to do this. How much bias V do we need? Ideally, the bias should be larger than the peak signal voltage. If I grab, say, the 17V positive rail through dropping resistors to provide a couple of volts, won't there be increased crosstalk by some signal getting back into the pos. rail and into other bias injection points? (Not to mention the op amps themselves.) The bias source should be heavily bypassed to ground to ensure that this doesn't happen. Using large biasing resistors would help. Should I use a little regulator, say a 7803 (if there is such a beast, or a 7805 if not) local to each cap "pair" I'd want to bias? By pair I mean two electros replacing an original single cap, with the + sides tied back-to-back and the bias injected there. Would the output stages of such regulators impact the signal fidelity? If you understand the purpose of bias, you will probably want to set the biasing voltage as high as practically possible. Obviously, you don't want to bias the caps beyond their working voltage. You don't want to bias the caps to the point where their leakage currents upset the operating points of the various circuits. Thanks again for the help and experiences in this area. Thanks to those who have responded to these Delta threads in the past, especially Mr. Williams, Mr. Dorsey, and Mr. Sank (sorry if I missed anybody). I generally don't respond to recap and reopamp posts because they are basically founded on the idea that some poorly-educated, inexperienced guy who doesn't even know how capacitors are sized is going to improve on a design by a well-experienced engineer who most likely does. ;-(. My ego is plenty inflated but not so inflated that I totally disrespect the engineering talents of the staff of the better-known, widely respected firms like Soundcraft. |
#19
|
|||
|
|||
Replace the 47 uf caps with Panasonic FM 220 uf/25 v. This removes low
end rolloffs and phase shift. If you're wiggy about reverse bias, use slow bipolar caps and bypass with a Wima MKP-2 .01 film cap. The fader amp sees a 100k impedance, that's why the 2.2 uf value. Replace with a Wima MKS-2 3.3 uf with a .01 MKP-2 across it. Use a Black Gate N or NG for the 1000 uf sum cap. Bypass with good films like MIT or InfiniCaps. Opamps with higher gain bandwidth like 5532/OPA2134 etc. require a redesign of the line input stage for stability. LT1358's will allow direct coupling due to low offsets. Replace the solo sum amp input cap to a 470 uf with a film bypass. Replace the 22k feedback resistor with a 10 k for unity gain. Replace the feedback cap with a 150 pf silver mica for stability. Reverse bias is common in these mixers, the EQ section will shift bias from positive to negative by simply rotating the frequency pot. I found it more important to have a quality bypass around these caps, it negates the slow response and allows the details through that normally would be converted to heat through the electrolytic capacitor. BTW, this info is offered as a help to the community, not to start any trouble with you folks... So don't blow a gasket, it's only audio ! Jim Williams Audio Upgrades |
#20
|
|||
|
|||
Frank,
if you decide to make all these changes, please make some before and after distortion and frequency response measurements and share the results with us. thanks Mark |
#21
|
|||
|
|||
"Mark" writes:
if you decide to make all these changes, please make some before and after distortion and frequency response measurements and share the results with us. I will indeed. Some excellent suggestions have been made on these various Delta threads and especially this one; the least I can do is report the results -- what worked; what didn't. Step one has been bringing the thing back to a starting point. Simply replacing all the worn/noisy switches, failing 1/4" normalling phone jacks, and the mother and monitor ribbon cables has made an obvious difference already. I have new power supplies awaiting mounting on a rack panel. I have at least two shorted caps; several others I suspect of losing value, as right now the bottom on this console just isn't what it ought to be. Unfortunately I don't have access to an AP, nor can I plunk down US$10K to get one. A friend has a Neutrik Minilyzer, but am not sure that it will have the resolution to do precise measurements but will investigate that, as well as PC-based analyzer options. I can still do some crude tests with the generator and scope. I'd like to have Jim Williams do the whole console, but alas funds are a bit tight and more importantly, I wouldn't learn nearly as much as I have by wrapping myself in the schematics and looking around with the scope. The basic design of each Delta module seems pretty good (and it *is* a modular console), but as is nearly always case these systems are designed to a price point -- at least with parts. Nothing wrong with that of course; meeting a stated design target on budget (and producing a good value for the market) is good engineering. But if in the process of replacing parts that need it anyway I can move that equivalent price point a little (or a lot), and not screw up the design, I should have a very sweet sounding console, and without getting into huge piles of cash. I hope. Thanks again to all for your input. More later. Frank Stearns Mobile Audio -- |
#22
|
|||
|
|||
Frank Stearns wrote:
I have at least two shorted caps; several others I suspect of losing value, as right now the bottom on this console just isn't what it ought to be. Unfortunately I don't have access to an AP, nor can I plunk down US$10K to get one. A friend has a Neutrik Minilyzer, but am not sure that it will have the resolution to do precise measurements but will investigate that, as well as PC-based analyzer options. I can still do some crude tests with the generator and scope. You can do some incredibly powerful tests with the generator and scope. The handiest one is just running a 1 KHz square wave through the thing with the controls flat. If the top of the squarewave slopes down, you have a low frequency response problem. If it slopes up, you have a high frequency response problem. If the leading edge rings, you have an impulse problem of some sort. It is a very quick and easy way to get a good notion of the average health of the system. And if you have a problem, you can use the scope and go from stage to stage until you find the offender. But if in the process of replacing parts that need it anyway I can move that equivalent price point a little (or a lot), and not screw up the design, I should have a very sweet sounding console, and without getting into huge piles of cash. I hope. Replace all the electrolytics first.... switches and pots might be fixable by cleaning and a little Cailube, but until you get the leaky caps out you don't know if there's a real problem or not. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#23
|
|||
|
|||
On Mon, 25 Apr 2005 01:20:54 -0000, Frank Stearns
wrote: "Mark" writes: if you decide to make all these changes, please make some before and after distortion and frequency response measurements and share the results with us. I will indeed. Some excellent suggestions have been made on these various ... be. Unfortunately I don't have access to an AP, nor can I plunk down US$10K to get one. A friend has a Neutrik Minilyzer, but am not sure that it will have the resolution to do precise measurements but will investigate that, as well as PC-based analyzer options. I can still do some crude tests with the generator and scope. Try this: http://rightmark.org/ Frank Stearns Mobile Audio ----- http://mindspring.com/~benbradley |
#25
|
|||
|
|||
Sounds like someone has blown a gasket. If you react so negativly to
freely offered information, I suspect you may have a heart attack if anyone dared charge you for anything. I don't think you own a Delta console so I wonder why you are even here... BTW, I don't need to advertise for work, I have enough already, feel free to not ever ask me for anything. Go take a walk in the garden, find a girlfriend, go kill a cat and switch to de-caf! One of these should work for you. Jim Williams Audio Upgrades |
#26
|
|||
|
|||
wrote in message ups.com... Sounds like someone has blown a gasket. If you react so negativly to freely offered information, I suspect you may have a heart attack if anyone dared charge you for anything. I don't think you own a Delta console so I wonder why you are even here... To spoil the party for a potentially large number of people around the world who silently follow this thread with keen interest in the subject - which happens to be related to rather popular Soundcraft Delta consoles and not to some frustrations of personal nature - the people who stand a chance of actually benefitting from your generous sharing of what could be considered proprietary information. That's how it looks from where I stand. Predrag |
#27
|
|||
|
|||
"Predrag Trpkov" writes:
wrote in message oups.com... Sounds like someone has blown a gasket. If you react so negativly to freely offered information, I suspect you may have a heart attack if anyone dared charge you for anything. I don't think you own a Delta console so I wonder why you are even here... To spoil the party for a potentially large number of people around the world who silently follow this thread with keen interest in the subject - which happens to be related to rather popular Soundcraft Delta consoles and not to some frustrations of personal nature - the people who stand a chance of actually benefitting from your generous sharing of what could be considered proprietary information. That's how it looks from where I stand. Indeed, Mr. Williams has been very generous with his time and help. Years ago I was in the "if the numbers don't show it, it's not real" camp. Then I got into a very good room with decent monitoring and a whole new world opened up. Differences brought on by changing parts that spec'd the "same" were no longer subtle or ambiguous, they were consistently (and sometime painfully) obvious. Numbers can tell the story -- problem is that after all these years we *still* don't commonly use enough descriptive numbers to tell the whole story. Moreover, marketing departments would have a hard time trying to put a good spin on these numbers and not confuse the average user. In the late 1940s engineers fought over whether AM radio would benefit by extending the bandwidth from 5 KHz to 15 KHz. The advocates were considered silly and ill-informed -- why, anyone knew that a 15KHz bandwidth sounded terrible so why waste all that money upgrading the systems? If I remember the story correctly, Avery Fisher set up a *live* orchestra behind an acoustically transparent (but visually opaque) curtain that in turn had tunable acoustic elements behind it. The live audience would hear either a 5KHz or 15KHz acoustical bandwidth. A nice ABX test. Of course, the full-bandwidth sound won easily. Follow-on research revealed that the original electrical systems had terrible distortion in the high bands, something that in the day many opposing engineers had not considered. It was outside their current orthodoxy, so they ignored or belittled the notion of 15KHz bandwidth. (We saw similar attitudes among some engineers early on with CDs. CDs measured "perfect" therefore were perfect, until someone described the notion of jitter -- with numbers -- and how it could affect the perceived sound.) And so it is with the modifications that Mr. Williams and others do -- there's nothing voodoo about them. (And many design engineers probably know exactly the same tricks but are limited by management to a specific design/build budget.) One of my contentions is that some of these mods really show their true worth at the very *end* of a signal chain -- the consumer's playback system. The cleaner you start back at the very beginning, the more "room" you have to accumulate various inevitable distortions before bumping into the corresponding thresholds of human perception. And this accumulation seems to be of a logrithmic nature, which reinforces the notion of keeping that signal as pure as you can for as long as you can before those accumulative step-ups start getting large. Happy recording. Frank Stearns Mobile Audio -- |
#28
|
|||
|
|||
wrote:
Sounds like someone has blown a gasket. If you react so negativly to freely offered information, I suspect you may have a heart attack if anyone dared charge you for anything. I have an aversion to people ripping of the general layman public with snakeoilsmanship. I suppose you also sell those $200 mains cables that add dynamic transparency to the overall sonic enjoyment. I don't think you own a Delta console so I wonder why you are even here... Why on earth would I want to own one? Kevin Aylward http://www.anasoft.co.uk/EE/index.html http://www.anasoft.co.uk SuperSpice, a very affordable Mixed-Mode Windows Simulator with Schematic Capture, Waveform Display, FFT's and Filter Design. |
#29
|
|||
|
|||
wrote in message
ups.com... Sounds like someone has blown a gasket. If you react so negativly to freely offered information, I suspect you may have a heart attack if anyone dared charge you for anything. I don't think you own a Delta console so I wonder why you are even here... BTW, I don't need to advertise for work, I have enough already, feel free to not ever ask me for anything. Go take a walk in the garden, find a girlfriend, go kill a cat and switch to de-caf! One of these should work for you. Jim Williams Audio Upgrades FWIW Jim, I smacked lips with delight over your post. First, because I own a couple Deltas and may actually put that info to use someday. Knowing your track record with Deltas, I know better than to second-guess your suggestions about that particular device! g Second, and more importantly, because it made me feel good to see that kind of expertise sharing going on here. Gabe would be proud! Don't let someone who sees things differently discourage you. You done good. -- "It CAN'T be too loud... some of the red lights aren't even on yet!" - Lorin David Schultz in the control room making even bad news sound good (Remove spamblock to reply) |
#31
|
|||
|
|||
wrote:
Have you actually _listened_ to said preamp? Why? It will sound identical to any other preamp with 1/10 the spec, and 1/100 the price. Kevin Aylward http://www.anasoft.co.uk/EE/index.html http://www.anasoft.co.uk SuperSpice, a very affordable Mixed-Mode Windows Simulator with Schematic Capture, Waveform Display, FFT's and Filter Design. |
#32
|
|||
|
|||
Kevin Aylward wrote: wrote: Have you actually _listened_ to said preamp? Why? It will sound identical to any other preamp with 1/10 the spec, and 1/100 the price. I like this man. :-) Bob -- "Things should be described as simply as possible, but no simpler." A. Einstein |
#33
|
|||
|
|||
Kevin Aylward wrote:
wrote: Have you actually _listened_ to said preamp? Why? It will sound identical to any other preamp with 1/10 the spec, and 1/100 the price. So Jim's preamp sounds the same as a two channel preamp that costs $15? |
#34
|
|||
|
|||
"Kevin Aylward" wrote in message
... wrote: Have you actually _listened_ to said preamp? Why? It will sound identical to any other preamp with 1/10 the spec, and 1/100 the price. Dear Lord, Please please please please please please please adjust the universe to make this true. I *SO* want Kevin to be right! Oh, and deliver us from trolls. Amen. -- "It CAN'T be too loud... some of the red lights aren't even on yet!" - Lorin David Schultz in the control room making even bad news sound good (Remove spamblock to reply) |
#35
|
|||
|
|||
Perhaps a career in video would suit your talents better. If you can't
hear any of this stuff, maybe you can see some of these things. I use the Audio Precision analyzer in the design and testing of the High Speed Mic Preamp as well as mods done to other manufacturers equipment. Yes, it does show subtle test variations. It also will show measurements that may be nearly identical, but sound different. It, like any test rig has severe limitations due to the fact it uses steady state stimulus rather than the violent waveforms of actual music. Folks that design gear based on test results will miss the boat every time. Test gear will tell you if something is wrong, only your ears will tell you when something is right. Yes, once the bugs are worked out on a design, there comes a point when you turn off the analyzer and turn on the ears. Jim Williams Audio Upgrades BTW, replace the 220 uf or 47 uf power supply filters on each Delta module with a Panasonic 470 uf 25v FM series. Replace the 10 ohm 1/4 watt fusing resistors wth 10 ohm 1/2 watt. Better low end punch. |
#36
|
|||
|
|||
wrote:
Have you actually _listened_ to said preamp? Why? It will sound identical to any other preamp with 1/10 the spec, and 1/100 the price. Kevin Aylward "Peoples perceptions are based on their belief of the facts, not the facts"=20 You're not immune either you know. |
#37
|
|||
|
|||
Frank Stearns wrote: In preparation for replacing or in some cases removing the electrolytics in my Soundcraft Delta, I've been cataloging those in the signal path. I've made offset measurements on both sides of many caps and found some with zero offset on either side. Good candidates for removal, I'd guess. Not really. The voltages you measure are only valid at the time they're made. Audio circuits aren't precision DC amplifiers ! The voltages will drift with time and temperature. (I did throw a 0.5 uf cap across the scope probe to filter the AC noise component, which was obscuring the small DC offsets.) I'm curious about two things. First, in a few cases where offset is present (a few mV) the voltage is opposite the labeled polarity on the cap. Obviously the cap hasn't blown apart from the small reversal, You'll need rather more than millivolts to do that ! but does the distortion go up? No. For the same reason that polarised electrolytics work perfectly as coupling caps under zero bias conditions when used intelligently. You'll typically need to see a few hundred mV reverse bias before there's any consequence. Second item is why certain cap values were chosen by the designer. Most of the interstage couplers are 47 uf/25 V -- perhaps a large enough value so that after going through a number of them the LF roll off is still below 20 Hz. (Though if stuck using these, perhaps I should bump them up to 100 uf for better LF?) I use 100uF/16V for output coupling caps mainly. Always been fine. Works for Neve too ! What I don't quite get, though, are the 2.2 uf/50V parts on the fader returns and outputs of most other rotary level controls. Is the idea to band-limit the LF Uh ! ????? so that the control action is clean(er) across a wide range of dirt in the part and aging? Or am I missing something? I think you're missing something. Nevertheless I use 10uF/16V parts typically in the instances you refer to. Next, the input to the L/R mix summing amps (and aux bus summers) is 1000 uf/6.3V. Why the huge value here? Why not 100 or 470? (Not saying they should be that way, just curious as to why.) Why ? Because the source resistance is the parallel value of *all* the mix bus Rs - i.e. *low* ! CAP BIASING - where electrolytic caps must be used, this seems like an interesting idea to try, but I wonder how best to do this. How much bias V do we need? If I grab, say, the 17V positive rail through dropping resistors to provide a couple of volts, won't there be increased crosstalk by some signal getting back into the pos. rail and into other bias injection points? (Not to mention the op amps themselves.) They don't need biasing. Why ? Because the AC component of signal across them is so tiny that they don't *need* to be polarised. The AC signal appearing across the cap is so small that this is why electrolytics work so well as coupling caps - it introduces no issues of significance. This is actually a *long story* - suffice to say I've measured electrolytics as coupling caps - and while such caps aren't *perfect* - their contribution to any signal degradation is miniscule. To see the effect you have to measure *across* the coupling cap itself ( where there might be a mV or so of signal perhaps - compared to Volts in the circuit under test ) . Measuring at the load will never reveal any imperfections. Should I use a little regulator, say a 7803 (if there is such a beast, or a 7805 if not) local to each cap "pair" I'd want to bias? By pair I mean two electros replacing an original single cap, with the + sides tied back-to-back and the bias injected there. Would the output stages of such regulators impact the signal fidelity? Where there is a potentially large DC offset in the signal - I use the back to back method and bias the centre point from the op-amps' regulated supply via a very large value R - say 1M. And you needn't worry about the regulators - thery're injecting any noise via 1 Meg into a few ohms ( the signal path impedance ) - and the noise on a decent regulated rail should be in the hundreds of uV max anyway. I.e. you're injecting femto Volts of noise ( if that ) into the signal chain most likely ! You could also used non-polarised electrolytics Graham |
#38
|
|||
|
|||
Graham,
I'll have to second everything you said. Reading this crazy thread brought my curiosity to the point where I went to the lab today and hooked up a 10 uF 16 Volt cheap electrolytic cap to a generator and scope. The generator source z is 50 Ohms and I set the scope for 50 Ohm input so the cap is seeing 100 Ohms. I had 0 volts DC bias on the cap. At about 20 Hz the cap was attenuating the signal about 6 dB, ie 1/2 the gen output was across the cap and 1/2 across the load. I had to raise the signal swing to over 5 volts p-p before I could see some effect of the signal reverse biasing the cap. In most applications the purpose of the cap is to couple the signal so there is very very little signal voltage ACROSS the cap so it has little opportunity to create any distortion. And the signal level in line and mic circuits are pretty low anyway. The only case I could think of where there might be an issue is in a high level high pass filter (speaker crossover?) where the cap might see a large portion of the signal swing across it. I completly agree with you, if the signal swing ACROSS the cap is less then 1 Volt or so, the cap does not need to be biased. Note the signal passing THROUGH the cap can be much larger than the swing ACROSS the cap except in the case of high pass filters. But there will still be folks that swear they can hear something... Whatever... Maybe the other guys have the right idea...anybody want to buy a set of gold plated matched coupling caps, cheap , only $50 a pair. Mark |
#39
|
|||
|
|||
Mark wrote:
Graham, I'll have to second everything you said. Reading this crazy thread brought my curiosity to the point where I went to the lab today and hooked up a 10 uF 16 Volt cheap electrolytic cap to a generator and scope. The generator source z is 50 Ohms and I set the scope for 50 Ohm input so the cap is seeing 100 Ohms. I had 0 volts DC bias on the cap. At about 20 Hz the cap was attenuating the signal about 6 dB, ie 1/2 the gen output was across the cap and 1/2 across the load. I had to raise the signal swing to over 5 volts p-p before I could see some effect of the signal reverse biasing the cap. In most applications the purpose of the cap is to couple the signal so there is very very little signal voltage ACROSS the cap so it has little opportunity to create any distortion. And the signal level in line and mic circuits are pretty low anyway. The only case I could think of where there might be an issue is in a high level high pass filter (speaker crossover?) where the cap might see a large portion of the signal swing across it. I completly agree with you, if the signal swing ACROSS the cap is less then 1 Volt or so, the cap does not need to be biased. Note the signal passing THROUGH the cap can be much larger than the swing ACROSS the cap except in the case of high pass filters. But there will still be folks that swear they can hear something... Whatever... Maybe the other guys have the right idea...anybody want to buy a set of gold plated matched coupling caps, cheap , only $50 a pair. LOL ! Some months ago, stimulated by a similar thread, I actually made some measurements on non-polarised electrolytics as coupling caps. I used an Audio Precision generator / analyser ( 40 ohms output Z ) and hooked up my favourite 100uF/16V output coupling caps driving a choice of 600 ohm ( worst case ) or 10k ohm ( more typical ) load. In order to get some volts across the cap I ran the tets with a 10V input signal. Naturally, the volts across the cap are larger at low frequencies - and I was *just* able to measure a distortion degradion with a 10V signal @ 20 Hz with the 600 ohm load. It was undetectable with the 10k ohm load. That's a pretty tough and untypical test ( 10V @ 20Hz into a real 600 ohm load ) and still the distortion was barely affected much beyond oscillator residual which is around -100 dB ! @ 3V there was no degradation under the same conditions btw. Measuring *across* the cap was interesting though ! There is indeed some non-linearity but not necessarily where you'd expect it. I saw some funny waveforms @ 2 kHz actually - but then there was only about 1mV across the cap at that frequency, -80dB relative to the test signal - so 3% THD across the cap equates to -110dB SINAD ref the signal. ( figures quoted are off the top of my head from memory ) Graham |
#40
|
|||
|
|||
The bottom line for me is: there's much hard data to utilize in these
issues but there is also judgement used in how this data is applied in real world applications. And, sorry, I can't separate the judgement of someone whose lead off promo picture of himself is standing in the kitchen doorway in his socks with a banana headstock Explorer. And somehow correcting the timing of web midi files and reposting them seems kinda far from how most of the propellorheads I know spend their time. "shrug" |
Reply |
|
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
The Art of Cultural Analysis: An Analagy | High End Audio | |||
What are they Teaching | Audio Opinions | |||
Doppler Distortion - Fact or Fiction | Pro Audio | |||
Synching Muyltiple M-Audio Delta Cards | Pro Audio | |||
Use of 2 M-Audio Delta boards in parallel ? | Pro Audio |