Home |
Search |
Today's Posts |
#41
|
|||
|
|||
On Sun, 05 Dec 2004 02:54:41 +1100, Patrick Turner
wrote: Stewart Pinkerton wrote: On Sat, 04 Dec 2004 01:18:26 +1100, Patrick Turner wrote: NFB is essnetial since the drain resistance of a mosfet is about 220 ohms and at the sec it will be 100 ohms, so enough loop FB is needed to reduce the Ro to about 2 ohms to make it the equivalant of the 300B amp with no loop FB. Hmmm, I was thinking of keeping it just to local degeneration, with no global NFB. Just to keep in the spirit of things SET. You mean source follower. Not really. That's cheating, see the headmaster immediately. The spirit of things means the NFB most not be greater than what's inside a 300B already. Well now, that's a number we could discuss for a *long* time! :-) The drain resistance does after all depend on the MOSFET(S) being used. Most output mosfets will have a finite drain resistance, and its far too high to ignore it. Even the Zen Pass amp with one active mosfet uses about 12 dB of NFB in a loop. Its reasonable to ask, is a Zen amp as good as a 300B? An excellent question, but first we must define 'good'............. One of the ruins that Henry knocked about a bit......... Henry V, or VIII? According to the song, VIII. My your wife getting on a bit. I'll pass on your message. Be afraid - be very afraid............. But if she's able to put up with you and still get dinner on the table, you must be doin OK in this terrible age of people spending so much time rubbishing the very idea of marriage. The first 30 years are the worst........ :-) Still haven't persuaded her that a Home Cinema is a *much* higher priority than a new car............ Did you tell her that the HT can't get run into or have its paintwork all scratched up in a car park? But boy, real HT needs a screen about the equivalant of 8 x 26 inch telies in size, and they are almost the price of a car. Exactly my point! 3-chip DLP, 10-foot Firehawk screen, 7 off Dynaudio 52SE, Pioneer 868 and 10, sure adds up! There ya go again makin out everyone has the same problems. Alas ye are mistaken.... Indeed, you seem to have a plethora all your own! :-) -- Stewart Pinkerton | Music is Art - Audio is Engineering |
#42
|
|||
|
|||
"Stewart Pinkerton" wrote in message ... : Did you tell her that the HT can't get run into or have its paintwork : all scratched up in a car park? But boy, real HT needs a screen : about the equivalant of 8 x 26 inch telies in size, and they are almost the : price of a car. : : Exactly my point! 3-chip DLP, 10-foot Firehawk screen, 7 off Dynaudio : 52SE, Pioneer 868 and 10, sure adds up! He, that might fit *into* a car .....well... the limo stretched ones Rudy : : Stewart Pinkerton | Music is Art - Audio is Engineering |
#43
|
|||
|
|||
: : Exactly my point! 3-chip DLP, 10-foot Firehawk screen, 7 off Dynaudio : 52SE, Pioneer 868 and 10, sure adds up! He, that might fit *into* a car .....well... the limo stretched ones Rudy : I knew someone who put a pair of top quality home stereo speakers in the back seats of his car. Sounded great! |
#44
|
|||
|
|||
Stewart Pinkerton wrote: On Sun, 05 Dec 2004 02:54:41 +1100, Patrick Turner wrote: Stewart Pinkerton wrote: On Sat, 04 Dec 2004 01:18:26 +1100, Patrick Turner wrote: NFB is essnetial since the drain resistance of a mosfet is about 220 ohms and at the sec it will be 100 ohms, so enough loop FB is needed to reduce the Ro to about 2 ohms to make it the equivalant of the 300B amp with no loop FB. Hmmm, I was thinking of keeping it just to local degeneration, with no global NFB. Just to keep in the spirit of things SET. You mean source follower. Not really. Oh well, I guess being sunday, its pointless discussing spiritual matters regarding "things SET" and as usual, I was easily caught out assuming apple source follower. ( Usually that's a nice drop of port ) The spiritual side of electronics is something youse believe in, or yer don't, and its unlikely The Pope is ever to convert to become a muslim, and so sunday is the wrong day to get you to change your beliefs either. In the name of the Father, Son, and God of Triodes, may your spirit rest in Peace. That's cheating, see the headmaster immediately. The spirit of things means the NFB most not be greater than what's inside a 300B already. Well now, that's a number we could discuss for a *long* time! :-) You'd need to have a good read of the ancient texts or Terman, Radio Engineering, 1937, then look up what Child has to say about triode "self regulation". But I am sure there is plenty elsewhere that most modern post deconstructionalist ppl who think they knows it all, but don't, should read, and its sunday, so its the day for spiritual contemplation, and cerebrally experiencing the wonderment of the universe if you ain't wandering out on a short walk up a mountain. And in fact since space is mainly a vaccum, there is plenty of room for spirits to lurk, although the nature of a spirit hasn't been defined yet by the scientists. They said it all began with the Big Karumba, and they get very offended when you ask them what happened before the BK. If they had the slightest idea, spirit definitions would be a doddle by comparison. Most are atheists, which don't give them much of a start in the "right attitude" stakes. The smart arses in white coats dunno how to have a strong suspicion they could be wrong, or that they grope in the dark, or are scratching at the surface of something far bigger than themsleves. The drain resistance does after all depend on the MOSFET(S) being used. Most output mosfets will have a finite drain resistance, and its far too high to ignore it. Even the Zen Pass amp with one active mosfet uses about 12 dB of NFB in a loop. Its reasonable to ask, is a Zen amp as good as a 300B? An excellent question, but first we must define 'good'............. Well if we say that a 300B can perform gooder as many other things, more or less, depending on circumstances, then I will re-fraze that question... Its reasonable to ask, is a Zen amp sonically similar to a 300B? One of the ruins that Henry knocked about a bit......... Henry V, or VIII? According to the song, VIII. My your wife getting on a bit. I'll pass on your message. Be afraid - be very afraid............. Don't worry, my gggggreat grandfather's suit of armour is a fine defense against the worst flame attacks the Net has yet to throw up at me. But ain't it cute the way we call a protection suit, a suit of 'armour'. Its not that different from amour, the expressed urge to woo some sheila. Only trouble is that to get a bonk these days, and maintaining the situation where a bonk is forthcoming at times deemed appropriate, ie, as often as one wishes, a warlike type of action has to be employed, albeit a more subtle methodology one than conducted in 1400. But it seem the origins of love and romance are embedded in the traditions of war, so we may as well use the same terminology. The ladies are more inclined to bonk if we persue them with riches, and to get riches we must do a bit of better warmongering than our mates, lest we be stuck with redheads and leftovers who are hopeless in bed, can't cook, fat, and argue about everything. And if in the arduous process we don't quite have things work out, war being somewhat a hit and miss affair, then time permitting after the battles are done, a little rape and pillage is all embraced by the terms armour, and ardour, and even "affair" can have connections with pools of blood. But just don't expect to woo a shiela with a box full of 300B, or a packet of mosfets. But if she's able to put up with you and still get dinner on the table, you must be doin OK in this terrible age of people spending so much time rubbishing the very idea of marriage. The first 30 years are the worst........ :-) And if the first 30 years don't occur, then a marriage started late would be utterly impossiblebe, assuming one starts these idiotic ventures early enough to have known there were better alternatives. Still haven't persuaded her that a Home Cinema is a *much* higher priority than a new car............ Did you tell her that the HT can't get run into or have its paintwork all scratched up in a car park? But boy, real HT needs a screen about the equivalant of 8 x 26 inch telies in size, and they are almost the price of a car. Exactly my point! 3-chip DLP, 10-foot Firehawk screen, 7 off Dynaudio 52SE, Pioneer 868 and 10, sure adds up! There ya go again makin out everyone has the same problems. Alas ye are mistaken.... Indeed, you seem to have a plethora all your own! :-) Even the cleanest of persons harbours all manner of bacteria. If you reckon yours don't smell you are kidding yourself, sundays teach us that our own ****e stinks like everyone eles's. Patrick Turner. -- Stewart Pinkerton | Music is Art - Audio is Engineering |
#45
|
|||
|
|||
Ruud Broens wrote: "Stewart Pinkerton" wrote in message ... : Did you tell her that the HT can't get run into or have its paintwork : all scratched up in a car park? But boy, real HT needs a screen : about the equivalant of 8 x 26 inch telies in size, and they are almost the : price of a car. : : Exactly my point! 3-chip DLP, 10-foot Firehawk screen, 7 off Dynaudio : 52SE, Pioneer 868 and 10, sure adds up! He, that might fit *into* a car .....well... the limo stretched ones Rudy But you'd have no room for the babes to lounge around..... Patrick Turner. : : Stewart Pinkerton | Music is Art - Audio is Engineering |
#46
|
|||
|
|||
On Sun, 28 Nov 2004 23:37:40 +0000 (UTC), Stewart Pinkerton
wrote: As noted, to optimise the design (however fundamentally crippled) is to avoid iron. Another POV is that the single-ended-triode designs' goals are to attack a different iron problem, low BH level "flat spots", and that proponents are attempting to optimize linearity near zero-crossing, if at some cost at higher level linearity. The older I get, the more I'm tending to agree with the premise. FWIW, nobody should jump to any blanket conclusions without first measuring his/her own personal normal monitoring SPL and cranking his/her monitoring speakers' sensitivity into a real-world number of needed peak amplifier output voltage. Now subtract 60 dB. This is the important power output number of your monitor amplifier. The range from -60 dB to peak is what we mostly hear. The range below there becomes interesting, and the range above there *isn't used*. Issues of the importance of fundamental (without feedback) linearity and distortion harmonic content skewing caused by feedback are still to be addressed properly. But real issues exist. Chris Hornbeck |
#47
|
|||
|
|||
On Sun, 05 Dec 2004 04:58:01 GMT, Chris Hornbeck
wrote: On Sun, 28 Nov 2004 23:37:40 +0000 (UTC), Stewart Pinkerton wrote: As noted, to optimise the design (however fundamentally crippled) is to avoid iron. Another POV is that the single-ended-triode designs' goals are to attack a different iron problem, low BH level "flat spots", and that proponents are attempting to optimize linearity near zero-crossing, if at some cost at higher level linearity. The older I get, the more I'm tending to agree with the premise. It's a fair point. Of course, if you use class A push-pull SS, you don't need to worry about this at all! :-) FWIW, nobody should jump to any blanket conclusions without first measuring his/her own personal normal monitoring SPL and cranking his/her monitoring speakers' sensitivity into a real-world number of needed peak amplifier output voltage. Now subtract 60 dB. This is the important power output number of your monitor amplifier. The range from -60 dB to peak is what we mostly hear. The range below there becomes interesting, and the range above there *isn't used*. Issues of the importance of fundamental (without feedback) linearity and distortion harmonic content skewing caused by feedback are still to be addressed properly. But real issues exist. Indeed they do, so make your amp as linear as is possible with local degeneration, and minimise (or eliminate) the global feedback. -- Stewart Pinkerton | Music is Art - Audio is Engineering |
#48
|
|||
|
|||
"Iain M Churches" wrote in message ... "Mike Gilmour" wrote in message ... "Iain M Churches" wrote in message ... It is clear from his posts on the KISS amp thread so far, that Andre has spent considerable time and effort in preparing what promises to be one of the most interesting articles on RAT for a very long time. [Clip] Why can't we extend to Andre the courtesy he deserves, and listen to what he has to tell us? There will, no doubt, be room for the discussion at the end of each chapter, and also the opportunity for questions and answers when the design process is complete. After that, perhaps Stewart will give us the benefit of his long experience as an engineer in SS analogue design, and begin a new thread to take us through the process of designing and building a lower power solid-state amp with a similar specification to Andre's valve amplifier. There are probably many of us who would be interested to build both designs, and measure and compare audibly the merits of each. Cordially, Iain Can we then expect to see two individual and hopefully unique designs posted on a binaries soon complete with the designers accompanying notes on their own design strategy and performance expectations. I look forward to this in anticipation of some really creative designs that reflect their indviduality and their respective long experiences in both valve and SS technologies. Mike Yes. Hopefully we can expect this. I for one would be interested to learn something from both parties, and build both their amplifiers. I am sure that each design will have its strengths and its weaknesses. If several members of this group build both amplifiers, this too will be a topic for a very interesting thread. Cordially, Iain To date I've found Andre's published circuit information to be very interesting. Are we to expect also some web input from Stewart, a schematic perhaps and description - before too long? In anticipation, Mike |
#49
|
|||
|
|||
On Mon, 6 Dec 2004 12:59:56 -0000, "Mike Gilmour"
wrote: "Iain M Churches" wrote in message ... "Mike Gilmour" wrote in message ... Can we then expect to see two individual and hopefully unique designs posted on a binaries soon complete with the designers accompanying notes on their own design strategy and performance expectations. I look forward to this in anticipation of some really creative designs that reflect their indviduality and their respective long experiences in both valve and SS technologies. Mike Yes. Hopefully we can expect this. I for one would be interested to learn something from both parties, and build both their amplifiers. I am sure that each design will have its strengths and its weaknesses. If several members of this group build both amplifiers, this too will be a topic for a very interesting thread. Cordially, Iain To date I've found Andre's published circuit information to be very interesting. Are we to expect also some web input from Stewart, a schematic perhaps and description - before too long? As it happens, I've started a design, targeted at 12 watts into 4 ohms, i.e. 10 v peak at 2.5 amps peak. This should have similar max output voltage to Andre's amp, but more current output for driving real-world speakers. I'm intending to base it on a very linear BJT loaded by a constant current source. I'm probably going to horrify everyone here by avoiding iron, and using a simple capacitor-coupled output. Currently (pun intended), the driver stage is taxing me somewhat, if simplicity is to be the keyword, but we are to keep below 1% THD at 10 watts into 4 ohms (another design target). I also intend the design to have no global feedback, but since it's not even complete on paper yet, never mind actually built and tested, that may change. Patience guys, Andre has been working on (and presumably refining) his KISS amp for God knows how long, I'm starting from a clean sheet. One thing, mine will be a lot cheaper to build, and a lot safer for a newcomer to DIY amps! -- Stewart Pinkerton | Music is Art - Audio is Engineering |
#50
|
|||
|
|||
As it happens, I've started a design, targeted at 12 watts into 4 ohms, i.e. 10 v peak at 2.5 amps peak. This should have similar max output voltage to Andre's amp, but more current output for driving real-world speakers. I'm intending to base it on a very linear BJT loaded by a constant current source. I'm probably going to horrify everyone here by avoiding iron, and using a simple capacitor-coupled output. Currently (pun intended), the driver stage is taxing me somewhat, if simplicity is to be the keyword, but we are to keep below 1% THD at 10 watts into 4 ohms (another design target). I also intend the design to have no global feedback, but since it's not even complete on paper yet, never mind actually built and tested, that may change. Patience guys, Andre has been working on (and presumably refining) his KISS amp for God knows how long, I'm starting from a clean sheet. One thing, mine will be a lot cheaper to build, and a lot safer for a newcomer to DIY amps! Pinky, Are you familiar with IGBTs for output devices? (e.g. Toshiba GT20D101-Y, rated at 250V, 20A) Mark Alexander designed a PP amplifier using them (Analog Devices Application Note AN-211). I've got a stereo version of this amp maybe 70% complete. For you 'borgs, measured specs a Power Out: 70W RMS THD @ 1 kHz: 0.001% @ 50W THD @ 20 kHz: 0.009% @ 50W SMPTE IMD: 0.0004 @ 41.7W DIM-100: 0.0012% @ 50W Freq (-3 db): DC to 1 Mhz Slew Rate: 200V/µS Now, I wonder what it sounds like . . . Jon |
#51
|
|||
|
|||
In message , Jon Yaeger
writes As it happens, I've started a design, targeted at 12 watts into 4 ohms, i.e. 10 v peak at 2.5 amps peak. This should have similar max output voltage to Andre's amp, but more current output for driving real-world speakers. I'm intending to base it on a very linear BJT loaded by a constant current source. I'm probably going to horrify everyone here by avoiding iron, and using a simple capacitor-coupled output. Currently (pun intended), the driver stage is taxing me somewhat, if simplicity is to be the keyword, but we are to keep below 1% THD at 10 watts into 4 ohms (another design target). I also intend the design to have no global feedback, but since it's not even complete on paper yet, never mind actually built and tested, that may change. Patience guys, Andre has been working on (and presumably refining) his KISS amp for God knows how long, I'm starting from a clean sheet. One thing, mine will be a lot cheaper to build, and a lot safer for a newcomer to DIY amps! Pinky, Are you familiar with IGBTs for output devices? (e.g. Toshiba GT20D101-Y, rated at 250V, 20A) Mark Alexander designed a PP amplifier using them (Analog Devices Application Note AN-211). I've got a stereo version of this amp maybe 70% complete. For you 'borgs, measured specs a Power Out: 70W RMS THD @ 1 kHz: 0.001% @ 50W THD @ 20 kHz: 0.009% @ 50W SMPTE IMD: 0.0004 @ 41.7W DIM-100: 0.0012% @ 50W Freq (-3 db): DC to 1 Mhz Slew Rate: 200V/μS Now, I wonder what it sounds like . . . Jon Yes it did use IGBTs, but all the various versions of the 'Alexander Current Feedback Amplifier' I have seen have used standard MOSFETs as the output. His original design, which was also in an earlier app-note from ADI used Mosfets. I've measured and heard a MOSFET version, and it appears to be a very good design, though there are a lot of people who don't like using CFB in an audio amp. (BTW, a CFB amp really does need a dc servo, just in case you where wondering) -- Chris Morriss |
#52
|
|||
|
|||
Yes it did use IGBTs, but all the various versions of the 'Alexander Current Feedback Amplifier' I have seen have used standard MOSFETs as the output. His original design, which was also in an earlier app-note from ADI used Mosfets. I've measured and heard a MOSFET version, and it appears to be a very good design, though there are a lot of people who don't like using CFB in an audio amp. (BTW, a CFB amp really does need a dc servo, just in case you where wondering) *** Chris, Where did you see the MOSFET version? I've got a few Hafler DH500 heatsinks made for TO3 devices, and I've got plenty of Magnetek UST20 lateral Power MOSFETs, which should work fine. I'd much rather use the TO3 devices (what's holding me up from completion is that I'd need to make some PCBs to accommodate the IGBTs). It'd be great if you know of a source for a schematic. Offhand, which device do you think would be better for the app? TIA, Jon |
#53
|
|||
|
|||
On Mon, 06 Dec 2004 13:50:21 -0500, Jon Yaeger
wrote: As it happens, I've started a design, targeted at 12 watts into 4 ohms, i.e. 10 v peak at 2.5 amps peak. This should have similar max output voltage to Andre's amp, but more current output for driving real-world speakers. I'm intending to base it on a very linear BJT loaded by a constant current source. I'm probably going to horrify everyone here by avoiding iron, and using a simple capacitor-coupled output. Currently (pun intended), the driver stage is taxing me somewhat, if simplicity is to be the keyword, but we are to keep below 1% THD at 10 watts into 4 ohms (another design target). I also intend the design to have no global feedback, but since it's not even complete on paper yet, never mind actually built and tested, that may change. Patience guys, Andre has been working on (and presumably refining) his KISS amp for God knows how long, I'm starting from a clean sheet. One thing, mine will be a lot cheaper to build, and a lot safer for a newcomer to DIY amps! Pinky, Are you familiar with IGBTs for output devices? (e.g. Toshiba GT20D101-Y, rated at 250V, 20A) Yes, but I'm going with the Sanken 2SC2922. I'd rather like to show that a 'nasty' BJT can outlinear a 300B in an essentially similar circuit. Mark Alexander designed a PP amplifier using them (Analog Devices Application Note AN-211). I've got a stereo version of this amp maybe 70% complete. For you 'borgs, measured specs a Power Out: 70W RMS THD @ 1 kHz: 0.001% @ 50W THD @ 20 kHz: 0.009% @ 50W SMPTE IMD: 0.0004 @ 41.7W DIM-100: 0.0012% @ 50W Freq (-3 db): DC to 1 Mhz Slew Rate: 200V/µS Now, I wonder what it sounds like . . . Should sound very good (i.e. shouldn't 'sound' at all), but the SET gang will never accept it. My impossible mission (should I choose to accept it) is to show that a solid-state SE amplifier can produce results even better than is possible from tubes, while following the same 'minimalist' design philosophy and producing a similar spectral distribution of distortion artifacts (where observable at all!). I guess I'm trying to outZen the Zen and hoodwink the Linsley-Hood. The more I think about it, the more I'm up for the challenge! :-) Of course, I may create the world's loudest-barking dog of an amplifier, but hey, it's supposed to be a fun hobby, no? :-) -- Stewart Pinkerton | Music is Art - Audio is Engineering |
#54
|
|||
|
|||
Stewart Pinkerton wrote: On Mon, 6 Dec 2004 12:59:56 -0000, "Mike Gilmour" wrote: "Iain M Churches" wrote in message ... "Mike Gilmour" wrote in message ... Can we then expect to see two individual and hopefully unique designs posted on a binaries soon complete with the designers accompanying notes on their own design strategy and performance expectations. I look forward to this in anticipation of some really creative designs that reflect their indviduality and their respective long experiences in both valve and SS technologies. Mike Yes. Hopefully we can expect this. I for one would be interested to learn something from both parties, and build both their amplifiers. I am sure that each design will have its strengths and its weaknesses. If several members of this group build both amplifiers, this too will be a topic for a very interesting thread. Cordially, Iain To date I've found Andre's published circuit information to be very interesting. Are we to expect also some web input from Stewart, a schematic perhaps and description - before too long? As it happens, I've started a design, targeted at 12 watts into 4 ohms, i.e. 10 v peak at 2.5 amps peak. This should have similar max output voltage to Andre's amp, but more current output for driving real-world speakers. If you have 30 watts dissipated in just one BJT, say an MJL21193 or 94, then 12 watts into some load will be a doddle. 10V / 2.5A = 4 ohms. In the real world, 7 vrms into 4 ohms gives 12.2 watts. The rails could be +/- 15v, and idle current 2 amps. The active device will have 30 watts for its input power, anf the CCS will draw 30 watts of power at all times. I'm intending to base it on a very linear BJT loaded by a constant current source. Here is where you will have trouble. BJTs have splendid current linearity, ie the current in/current out is fairly constant. The voltages are not so linear...... Then you wil find the MJL21194 or 93 will have a current gain of about 30, so for 2 amps of load signal current you will need 66 mA of base drive current, and you'll find yourself being forced to use a darlington pair or triple if you want decently high input resistance. I'm probably going to horrify everyone here by avoiding iron, and using a simple capacitor-coupled output. You don't horrify me. But you are so wasteful. Iron allows twice the efficiency. But efficiency isn't audible, so be my guest and waste those watts. Sugden have been using caps to couple their class A amps since about 1969, and I have no problem with that. maybe use 10 x 1,000 uF caps.... Currently (pun intended), the driver stage is taxing me somewhat, if simplicity is to be the keyword, but we are to keep below 1% THD at 10 watts into 4 ohms (another design target). Well I did remind you that driving bjt outputs requires current drive, and since you will **not** be allowed to use emitter follower or global NFB, then the input Z to the output BJT will be a horror story. I doubt you would ever see 1% thd. Emitter follower mode is a severe case of loop series voltage negative feedback, and in the case of a lone class A BJT, amounts to far too much FB to comply with any KISS principle I also intend the design to have no global feedback, but since it's not even complete on paper yet, never mind actually built and tested, that may change. Nobody has yet been able to make a bjt amp with no loops of NFB and still get a low Ro. But look, I'll make a concession, and allow you just enough NFB so that your amp's Ro = 1 ohm, and no lower. That only gives you a DF of 4 with a 4 ohm load, but its the bare minimum, imho. Otherwise I won't come to the listening tests between the two amp genre. Why would I want to hear music via a current source with really high THD??? Patience guys, Andre has been working on (and presumably refining) his KISS amp for God knows how long, I'm starting from a clean sheet. One thing, mine will be a lot cheaper to build, and a lot safer for a newcomer to DIY amps! A Sziclai pair come to mind for you...... But I reckon pretty soon you'll be pulling all the hair out of your head, and screaming for someone to send you a few mosfets and j-fet drivers. It is in the realm of possibilities that your'e bald, and stubborn, and in that case, good luck, you'll need it! Patrick Turner.. -- Stewart Pinkerton | Music is Art - Audio is Engineering |
#55
|
|||
|
|||
Chris Morriss wrote: In message , Jon Yaeger writes As it happens, I've started a design, targeted at 12 watts into 4 ohms, i.e. 10 v peak at 2.5 amps peak. This should have similar max output voltage to Andre's amp, but more current output for driving real-world speakers. I'm intending to base it on a very linear BJT loaded by a constant current source. I'm probably going to horrify everyone here by avoiding iron, and using a simple capacitor-coupled output. Currently (pun intended), the driver stage is taxing me somewhat, if simplicity is to be the keyword, but we are to keep below 1% THD at 10 watts into 4 ohms (another design target). I also intend the design to have no global feedback, but since it's not even complete on paper yet, never mind actually built and tested, that may change. Patience guys, Andre has been working on (and presumably refining) his KISS amp for God knows how long, I'm starting from a clean sheet. One thing, mine will be a lot cheaper to build, and a lot safer for a newcomer to DIY amps! Pinky, Are you familiar with IGBTs for output devices? (e.g. Toshiba GT20D101-Y, rated at 250V, 20A) Mark Alexander designed a PP amplifier using them (Analog Devices Application Note AN-211). I've got a stereo version of this amp maybe 70% complete. For you 'borgs, measured specs a Power Out: 70W RMS THD @ 1 kHz: 0.001% @ 50W THD @ 20 kHz: 0.009% @ 50W SMPTE IMD: 0.0004 @ 41.7W DIM-100: 0.0012% @ 50W Freq (-3 db): DC to 1 Mhz Slew Rate: 200V/μS Now, I wonder what it sounds like . . . Jon Yes it did use IGBTs, but all the various versions of the 'Alexander Current Feedback Amplifier' I have seen have used standard MOSFETs as the output. His original design, which was also in an earlier app-note from ADI used Mosfets. I've measured and heard a MOSFET version, and it appears to be a very good design, though there are a lot of people who don't like using CFB in an audio amp. (BTW, a CFB amp really does need a dc servo, just in case you where wondering) Negative current feedback, CFB, raises Ro. Positive CFB lowers Ro, and even makes Ro negative, ie, the output voltage rises with a lower load value. NCFB reduces thd, and PCFB raises thd, and makes an amp more unstable. Keep it simple, don't use either!! Patrick Turner. -- Chris Morriss |
#56
|
|||
|
|||
Patrick Turner said:
Nobody has yet been able to make a bjt amp with no loops of NFB and still get a low Ro. Densen Beat and B10/20 models spring to mind. No loop NFB, just local. -- Sander de Waal " SOA of a KT88? Sufficient. " |
#57
|
|||
|
|||
On Tue, 07 Dec 2004 19:55:05 +1100, Patrick Turner
wrote: Stewart Pinkerton wrote: As it happens, I've started a design, targeted at 12 watts into 4 ohms, i.e. 10 v peak at 2.5 amps peak. This should have similar max output voltage to Andre's amp, but more current output for driving real-world speakers. If you have 30 watts dissipated in just one BJT, say an MJL21193 or 94, then 12 watts into some load will be a doddle. 10V / 2.5A = 4 ohms. In the real world, 7 vrms into 4 ohms gives 12.2 watts. But 10V peak gives 12.5 watts! Jeez, I thought just rounding to 12 was close enough for guvm'nt work! The rails could be +/- 15v, and idle current 2 amps. Er no, has to be 2.5 amps, and I'm going single-sided 30 volts. I'd rather use an OPT than add a DC servo. The active device will have 30 watts for its input power, anf the CCS will draw 30 watts of power at all times. Quite so (although I've got to dump a total of 75 watts among all the actives and passives), so a pair of good power trannies should be able to handle it easily, without heroic heatsinks or fan cooling. I thought of using the Sanken 2SC2922/2SA1216, the pnp being the current source. I'm intending to base it on a very linear BJT loaded by a constant current source. Here is where you will have trouble. BJTs have splendid current linearity, ie the current in/current out is fairly constant. The voltages are not so linear...... Yes, I'm aware of that, but it's worth trying, just to see what can be done. Then you wil find the MJL21194 or 93 will have a current gain of about 30, so for 2 amps of load signal current you will need 66 mA of base drive current, and you'll find yourself being forced to use a darlington pair or triple if you want decently high input resistance. I'm probably going to horrify everyone here by avoiding iron, and using a simple capacitor-coupled output. You don't horrify me. But you are so wasteful. Iron allows twice the efficiency. But efficiency isn't audible, so be my guest and waste those watts. Not a big deal with only 75 watts per channel, methinks. My trusty Krell sucks 300 watts out of the wall when idling! Sugden have been using caps to couple their class A amps since about 1969, and I have no problem with that. maybe use 10 x 1,000 uF caps.... Or one 5,000 with a film bypass. Currently (pun intended), the driver stage is taxing me somewhat, if simplicity is to be the keyword, but we are to keep below 1% THD at 10 watts into 4 ohms (another design target). Well I did remind you that driving bjt outputs requires current drive, and since you will **not** be allowed to use emitter follower or global NFB, then the input Z to the output BJT will be a horror story. What's wrong with using an emitter follower? I doubt you would ever see 1% thd. You may be right, but one must *start* with a reasonable target. Emitter follower mode is a severe case of loop series voltage negative feedback, and in the case of a lone class A BJT, amounts to far too much FB to comply with any KISS principle I find a transistor and one resistor to be quite simple, and triodes have loads of *internal* feedback. As you well know, there is no such thing as a 'zero feedback' SET amp. I also intend the design to have no global feedback, but since it's not even complete on paper yet, never mind actually built and tested, that may change. Nobody has yet been able to make a bjt amp with no loops of NFB and still get a low Ro. Possibly, nobody ever will! But look, I'll make a concession, and allow you just enough NFB so that your amp's Ro = 1 ohm, and no lower. That only gives you a DF of 4 with a 4 ohm load, but its the bare minimum, imho. Otherwise I won't come to the listening tests between the two amp genre. Why would I want to hear music via a current source with really high THD??? I'm not disagreeing with you, but they seem to sell pretty well, according to Cary, Jadis, Audio Note etc......... :-) Patience guys, Andre has been working on (and presumably refining) his KISS amp for God knows how long, I'm starting from a clean sheet. One thing, mine will be a lot cheaper to build, and a lot safer for a newcomer to DIY amps! A Sziclai pair come to mind for you...... Well, I've used Sziklai pairs in fully complementary amps before, and I'm still kicking possible topologies around on paper, trying to minimise the parts count.. But I reckon pretty soon you'll be pulling all the hair out of your head, and screaming for someone to send you a few mosfets and j-fet drivers. It is in the realm of possibilities that your'e bald, and stubborn, and in that case, good luck, you'll need it! I'm always open to good suggestions! However, I don't want to build a Zen clone, that's already been done. I agree with all that you say in theory, but I've never actually been dumb enough to build a simple SE BJT amp with no global NFB, and see what it actually sounds like, despite the horrible specs! :-) -- Stewart Pinkerton | Music is Art - Audio is Engineering |
#58
|
|||
|
|||
In message , Jon Yaeger
writes Yes it did use IGBTs, but all the various versions of the 'Alexander Current Feedback Amplifier' I have seen have used standard MOSFETs as the output. His original design, which was also in an earlier app-note from ADI used Mosfets. I've measured and heard a MOSFET version, and it appears to be a very good design, though there are a lot of people who don't like using CFB in an audio amp. (BTW, a CFB amp really does need a dc servo, just in case you where wondering) *** Chris, Where did you see the MOSFET version? I've got a few Hafler DH500 heatsinks made for TO3 devices, and I've got plenty of Magnetek UST20 lateral Power MOSFETs, which should work fine. I'd much rather use the TO3 devices (what's holding me up from completion is that I'd need to make some PCBs to accommodate the IGBTs). It'd be great if you know of a source for a schematic. Offhand, which device do you think would be better for the app? TIA, Jon There's an extensive web site giving details of one version. I can't remember the site, but I'm sure a google for 'Alexander current feedback amplifier' will find it. The other is in my ADI app-note book. The Audio/video reference manual 1992. It's shown in the notes to the SSM-2131 audio op-amp. -- Chris Morriss |
#59
|
|||
|
|||
In message , Patrick Turner
writes Chris Morriss wrote: In message , Jon Yaeger writes As it happens, I've started a design, targeted at 12 watts into 4 ohms, i.e. 10 v peak at 2.5 amps peak. This should have similar max output voltage to Andre's amp, but more current output for driving real-world speakers. I'm intending to base it on a very linear BJT loaded by a constant current source. I'm probably going to horrify everyone here by avoiding iron, and using a simple capacitor-coupled output. Currently (pun intended), the driver stage is taxing me somewhat, if simplicity is to be the keyword, but we are to keep below 1% THD at 10 watts into 4 ohms (another design target). I also intend the design to have no global feedback, but since it's not even complete on paper yet, never mind actually built and tested, that may change. Patience guys, Andre has been working on (and presumably refining) his KISS amp for God knows how long, I'm starting from a clean sheet. One thing, mine will be a lot cheaper to build, and a lot safer for a newcomer to DIY amps! Pinky, Are you familiar with IGBTs for output devices? (e.g. Toshiba GT20D101-Y, rated at 250V, 20A) Mark Alexander designed a PP amplifier using them (Analog Devices Application Note AN-211). I've got a stereo version of this amp maybe 70% complete. For you 'borgs, measured specs a Power Out: 70W RMS THD @ 1 kHz: 0.001% @ 50W THD @ 20 kHz: 0.009% @ 50W SMPTE IMD: 0.0004 @ 41.7W DIM-100: 0.0012% @ 50W Freq (-3 db): DC to 1 Mhz Slew Rate: 200V/μS Now, I wonder what it sounds like . . . Jon Yes it did use IGBTs, but all the various versions of the 'Alexander Current Feedback Amplifier' I have seen have used standard MOSFETs as the output. His original design, which was also in an earlier app-note from ADI used Mosfets. I've measured and heard a MOSFET version, and it appears to be a very good design, though there are a lot of people who don't like using CFB in an audio amp. (BTW, a CFB amp really does need a dc servo, just in case you where wondering) Negative current feedback, CFB, raises Ro. Positive CFB lowers Ro, and even makes Ro negative, ie, the output voltage rises with a lower load value. NCFB reduces thd, and PCFB raises thd, and makes an amp more unstable. Keep it simple, don't use either!! Patrick Turner. -- Chris Morriss No I'm talking about CFB op-amps, nothing to do with using voltage feedback from a current-sensing resistor in the speaker lead. -- Chris Morriss |
#60
|
|||
|
|||
Sander deWaal wrote: Patrick Turner said: Nobody has yet been able to make a bjt amp with no loops of NFB and still get a low Ro. Densen Beat and B10/20 models spring to mind. No loop NFB, just local. It got FB, so it no get prize. If the Ro of a collector circuit of a power bjt was 3k, then 3,000 bjts in parallel with no FB would give you an Ro = 1 ohm. They'd all still make a non linear amp. And the input impedance would rather too low to be convivial. Patrick Turner. -- Sander de Waal " SOA of a KT88? Sufficient. " |
#61
|
|||
|
|||
Stewart Pinkerton wrote: On Tue, 07 Dec 2004 19:55:05 +1100, Patrick Turner wrote: Stewart Pinkerton wrote: As it happens, I've started a design, targeted at 12 watts into 4 ohms, i.e. 10 v peak at 2.5 amps peak. This should have similar max output voltage to Andre's amp, but more current output for driving real-world speakers. If you have 30 watts dissipated in just one BJT, say an MJL21193 or 94, then 12 watts into some load will be a doddle. 10V / 2.5A = 4 ohms. In the real world, 7 vrms into 4 ohms gives 12.2 watts. But 10V peak gives 12.5 watts! Jeez, I thought just rounding to 12 was close enough for guvm'nt work! Interdepartmental transfer of inordinately random co-opted projects always tend to have extrapolated variations in shared interest outcomes and result in small percentages of un-resolvable technical aberations due to the latest funding restrictions recommended by the Minister at the last conciliatory meeting to explore efficiency reforms concerning computer use and tea lady work contracts. The rails could be +/- 15v, and idle current 2 amps. Er no, has to be 2.5 amps, Gee, did that new junior officer advised me wrongly about the idle current?. It'd be better to have 3 amps, so that you could have +3 to -3 peak amp load change. With 15v rails, that's a dangerous 45 watts dissipation in a device? and I'm going single-sided 30 volts. I'drather use an OPT than add a DC servo. I figured from 30 watts dissipation in the gain device, you'd get a max of 15 watts if the amp was perfect, and about 12 in the real world. 12 watts into 4 ohms is 2.449 peak amps load current change, and you'd need rails of 9.8v in theory at an idle current of 2.449 in theory to get your 12 watts. Idle dissipation is then 2.449A x 9.8V = 24 watts, or about the same as an EL34 in class A. This is quite OK for one lonely lil ol bjt. But you still have to bank on getting only 40% efficiency from your bjt, so you'd want to have 2.6A x 11.5V rail operating conditions. This would be very nice in a car, since you already have a 12V battery. But you'd only be able to canoodle with Nicole K for 5 hrs before the amp hours mounted up, ( if nuthin else ) and the battery went flat. What's a few amps? a single car 100 W headlight might draw 8.3 amps... sue mucV 2 vto 10v3 in theoryp If you want 2.5 amps peak load current, you only need 1.25 amps idle. Then the load current varies from zero to +/- 2.5 amps The active device will have 30 watts for its input power, anf the CCS will draw 30 watts of power at all times. Quite so (although I've got to dump a total of 75 watts among all the actives and passives), so a pair of good power trannies should be able to handle it easily, without heroic heatsinks or fan cooling. I thought of using the Sanken 2SC2922/2SA1216, the pnp being the current source. I don't know these two. We have lots like that on our prison records, long numbers, long prison terms, bad crimes.... I will do a Google on them. The characteristics of the 2sc2922 are at http://www.ampslab.com/PDF/2sc2922.pdf I don't see any typical operating conditions for class A use in SE config with simple easy to understand collector resistance lines with Ec / Ic axes, like the Ea Ia axes for the plate resistance curves on a 300B. With the data for the 300B, you just get a ruler, and rule in a load line, and the gain, distortion, and power output info is all able to be calculated in 3 minutes on the back of an envelope. But of course the ppl at Sanken would think you quite barmy if you told them what you wanted one of their devices for, so they make sure those communist left wing lefty class Aists can't get hold of sufficient info to wreck the economy by sending up power consumption and threatening shortages of fuses with bjt meltdowns. You won't be able to keep it simple, the right wing pro status quo have made damn sure bjts have positive thermal coefficients, the hotter they get, the more current they draw, and this does wonders for sales of replacement parts...... So you have have an effective biasing circuit, probably more complex than the audio circuit.... I'm intending to base it on a very linear BJT loaded by a constant current source. Here is where you will have trouble. BJTs have splendid current linearity, ie the current in/current out is fairly constant. The voltages are not so linear...... Yes, I'm aware of that, but it's worth trying, just to see what can be done. I'll keep a bottle of anti-depressant pills handy. Then you wil find the MJL21194 or 93 will have a current gain of about 30, so for 2 amps of load signal current you will need 66 mA of base drive current, and you'll find yourself being forced to use a darlington pair or triple if you want decently high input resistance. I'm probably going to horrify everyone here by avoiding iron, and using a simple capacitor-coupled output. You don't horrify me. But you are so wasteful. Iron allows twice the efficiency. But efficiency isn't audible, so be my guest and waste those watts. Not a big deal with only 75 watts per channel, methinks. My trusty Krell sucks 300 watts out of the wall when idling! Sugden have been using caps to couple their class A amps since about 1969, and I have no problem with that. maybe use 10 x 1,000 uF caps.... Or one 5,000 with a film bypass. The electros made now are many and varied, and act as true capacitors better than their 1969 parent caps because of the the need for pure capacitance in SMPS, operating at HF, and to bypass devices with HF signals. Methinks size is of no consequence. Small is now beautiful, ( Martha says she doesn't agree ), but I figure more small paralleled caps will have the lowest esr, and greatest tendency to survive an attack by a roaming "inadvertant short circuit of a speaker cable" Currently (pun intended), the driver stage is taxing me somewhat, if simplicity is to be the keyword, but we are to keep below 1% THD at 10 watts into 4 ohms (another design target). Well I did remind you that driving bjt outputs requires current drive, and since you will **not** be allowed to use emitter follower or global NFB, then the input Z to the output BJT will be a horror story. What's wrong with using an emitter follower? You are using a great whacking ****e load of series voltage negative feedback. Ever wondered what the ***voltage*** gain of a power bjt is? Huge. The Gm of a power bjt is perhaps about 30 amps per volt, ( +/- 100%, but lets's settle on 30 ), so that means for a 2.5 amp load current change, only 0.083 volts of change occurs between the base and emitter. This means that if you have 4 ohms and 10v load change, the gain is 10 / 0.08 = 125. Transistors just start to turn on at about 0.4V between base and emitter, and are fully turned on at around 0.8V; there is not a large voltage swing required for then to work. They are all like this, biguns and smalluns. If the emitter voltage follows the base voltage which is 10.083V to get the emitter load voltage of 10V, then the gain reduction in EF mode is from 125 to slightly less than 1.0, so you have about 42 dB of applied series voltage NFB. If you had a mosfet, its Gm is only about 1A/V, so the "open loop gain" is only 4 into a 4 ohm load. Gain = load x Gm. So source follower = approximately 12 dB of NFB I doubt you would ever see 1% thd. You may be right, but one must *start* with a reasonable target. You would see considerably less that 1% with emitter follower, ( common collector mode ). If you tried to have the bjt in common emitter mode, with grounded emitter, load in the collector circuit, and with a small signal voltage applied to the base, I think the thd may be rather bad for what you intend. But the EF config reduces this thd by about 125 times, so if you had 10% open loop thd, expect about 0.08% with EF at the 12 watts. The use of EF has a technical benefit. Suppose the hfe of the power bjt was a healthy 50 at the idle current you wanna run yours at . Then for 2.5 amps load change, you'd need 0.05A of base current change. If the base-emitter voltage change was 0.08V, then Ri = 0.08/0.05 = 1.6 ohms, or less than the load you want to run. Most people use another midi sized bjt to power the base, and if its hfe is say 120, its base input current is 0.05/120 = 0.416 mA, and Ri = 0.08/ 0.0004166A = 192 ohms, or 120 x 1.6 ohms. Still, Ri is very low, but its doable for a drive amp because we only want 0.08 volts of input. But if we use EF, then the input voltage is 125 times greater, and the same current change occurs, so Ri becomes 125 times higher so it is then 200 ohms without the second midi driver bjt, or 24kohms if the midi driver is used in what is called the darlington connection. THEN, it should only be necessary to make a linear driver stage able to produce a substantially linear +/- 10volts. This is only possible if the driver stage has current fb in its emitter circuit. A single MJE 350 will have an enormous voltage gain, maybe 1,000 into 10k of RL, and attrocious open loop thd, but the current NFB from an emitter R can reduce the gain to only 12 as you want it, so 40 dB more NFB is to be used. Using bjts mean that there is always gain to burn, toooooooo muuuuuchhhh gain, and huge amounts of NFB are routinely used to control gain. Alternatively, one could use the MJE350 to drive the base of the output transistor, and then its voltage gain will be very small, since it is driving in current source mode, and the hfe of the MJE350 will convert the output transistor base input resistance to a more usable resistance. Applying series voltage loop NFB will then raise the input resistance. But a small amount of NFB ( say 12 dB ), still won't be anything like enough to raise Ri to be friendly to preamps. Emitter follower mode is a severe case of loop series voltage negative feedback, and in the case of a lone class A BJT, amounts to far too much FB to comply with any KISS principle I find a transistor and one resistor to be quite simple, and triodes have loads of *internal* feedback. As you well know, there is no such thing as a 'zero feedback' SET amp. True. How much applied NFB is there in a 300B? BTW, zero NFB in tube palance means no external loops of NFB. The NFB already inside the tube was put there by the God Of Triodes, and let's not ague with a God in charge of so many high voltages...... I also intend the design to have no global feedback, but since it's not even complete on paper yet, never mind actually built and tested, that may change. Nobody has yet been able to make a bjt amp with no loops of NFB and still get a low Ro. Possibly, nobody ever will! If you pull it off, the world will beat a path to your door.... But look, I'll make a concession, and allow you just enough NFB so that your amp's Ro = 1 ohm, and no lower. That only gives you a DF of 4 with a 4 ohm load, but its the bare minimum, imho. Otherwise I won't come to the listening tests between the two amp genre. Why would I want to hear music via a current source with really high THD??? I'm not disagreeing with you, but they seem to sell pretty well, according to Cary, Jadis, Audio Note etc......... :-) Er, the Ra of the triodes involved is usually less than 1/4 of RL, so such an item is considered to be a voltage source. If Ra was 4 times RL, sure, that would make a tube a current source. This is the case with many mantle radios using a lone beam tetrode like a 6V6. Ro = 20 ohms. Sound is fine up to 0.25 watts..... Patience guys, Andre has been working on (and presumably refining) his KISS amp for God knows how long, I'm starting from a clean sheet. One thing, mine will be a lot cheaper to build, and a lot safer for a newcomer to DIY amps! A Sziclai pair come to mind for you...... Well, I've used Sziklai pairs in fully complementary amps before, and I'm still kicking possible topologies around on paper, trying to minimise the parts count.. Betcha you use a shirtload more than Andre's 300B amp. But I reckon pretty soon you'll be pulling all the hair out of your head, and screaming for someone to send you a few mosfets and j-fet drivers. It is in the realm of possibilities that your'e bald, and stubborn, and in that case, good luck, you'll need it! I'm always open to good suggestions! However, I don't want to build a Zen clone, that's already been done. I agree with all that you say in theory, but I've never actually been dumb enough to build a simple SE BJT amp with no global NFB, and see what it actually sounds like, despite the horrible specs! :-) I'd have to say it may be impossible to design a bjt amp which has no more local or loop NFB than the 300B already has, and has only two gain devices. The input resistance issue was a pain to designers when bjts first started being used to replace tubes. Ah, at last, said engineers, we can get rid of the OPT. Then they sat down to design input transformers. They then tried to ban large currents from transistors, too bleeding unreliable. But then somebody replaced germanium with silicon in a special process, and that made matters simpler, but one still sees hellishly complex SS amps with huge numbers of bjts all strung out across vast pcbs, and its very hard to work out how these work, especially if there is a double sided board designed by a PC. Some are impressive, one Electronics World design had umpteen bjts, and puts out 300 watts, and has a slew rate of 250V/uS. Servicing of complex SS circuits is not supposed to be done. The circuits are supposed to be fully short circuit proof, and easy to fix. You pull out the plug in module with its 250 cheap parts, and bin it, and plug in the replacement. And pigs will fly when makers all addapt such ideas, and where they do, they charge a fortune for the replacement module; they'd rather you bought the latest amp, than repair the oldun, now 10 years "old". But a replacement of several blown output bjts and other collaterally caused failures is often far more than a set of new tubes.... Ppl keep sending me ancient Quad II amps, repairs are cheap, since 6L6 work OK in them. But they want upgrades anyhow, and like to see things already 50 years old be able to last them to their graves, maybe 30 years away. Will any solid state Sony, Marantz, Yamaha made in 1975 be serviceable in 2025? What of the gear made after 1985, with all that microprocessor control crap and switching behind the front panels? Who on earth will be able to decode the circuits if they have not got the exact workshop manual for the amp model they happen to be lumbered with? None of the special spare parts will be available. IMHO, a well re-engineered Quad II will hold its own with a 300B amp. Patrick Turner. -- Stewart Pinkerton | Music is Art - Audio is Engineering |
#62
|
|||
|
|||
Chris Morriss wrote: In message , Patrick Turner writes Chris Morriss wrote: In message , Jon Yaeger writes As it happens, I've started a design, targeted at 12 watts into 4 ohms, i.e. 10 v peak at 2.5 amps peak. This should have similar max output voltage to Andre's amp, but more current output for driving real-world speakers. I'm intending to base it on a very linear BJT loaded by a constant current source. I'm probably going to horrify everyone here by avoiding iron, and using a simple capacitor-coupled output. Currently (pun intended), the driver stage is taxing me somewhat, if simplicity is to be the keyword, but we are to keep below 1% THD at 10 watts into 4 ohms (another design target). I also intend the design to have no global feedback, but since it's not even complete on paper yet, never mind actually built and tested, that may change. Patience guys, Andre has been working on (and presumably refining) his KISS amp for God knows how long, I'm starting from a clean sheet. One thing, mine will be a lot cheaper to build, and a lot safer for a newcomer to DIY amps! Pinky, Are you familiar with IGBTs for output devices? (e.g. Toshiba GT20D101-Y, rated at 250V, 20A) Mark Alexander designed a PP amplifier using them (Analog Devices Application Note AN-211). I've got a stereo version of this amp maybe 70% complete. For you 'borgs, measured specs a Power Out: 70W RMS THD @ 1 kHz: 0.001% @ 50W THD @ 20 kHz: 0.009% @ 50W SMPTE IMD: 0.0004 @ 41.7W DIM-100: 0.0012% @ 50W Freq (-3 db): DC to 1 Mhz Slew Rate: 200V/μS Now, I wonder what it sounds like . . . Jon Yes it did use IGBTs, but all the various versions of the 'Alexander Current Feedback Amplifier' I have seen have used standard MOSFETs as the output. His original design, which was also in an earlier app-note from ADI used Mosfets. I've measured and heard a MOSFET version, and it appears to be a very good design, though there are a lot of people who don't like using CFB in an audio amp. (BTW, a CFB amp really does need a dc servo, just in case you where wondering) Negative current feedback, CFB, raises Ro. Positive CFB lowers Ro, and even makes Ro negative, ie, the output voltage rises with a lower load value. NCFB reduces thd, and PCFB raises thd, and makes an amp more unstable. Keep it simple, don't use either!! Patrick Turner. -- Chris Morriss No I'm talking about CFB op-amps, nothing to do with using voltage feedback from a current-sensing resistor in the speaker lead. Well if it works, then its usable, by someone, and they can claim they don't use negative feedback, whoopee! Maybe what they are really doing is a lot worse..... There are a zillion ways one can configure SS circuits, and many designers are looking for an edge.... Patrick Turner -- Chris Morriss |
#63
|
|||
|
|||
On Thu, 09 Dec 2004 00:04:21 +1100, Patrick Turner
wrote: Stewart Pinkerton wrote: On Tue, 07 Dec 2004 19:55:05 +1100, Patrick Turner wrote: Stewart Pinkerton wrote: As it happens, I've started a design, targeted at 12 watts into 4 ohms, i.e. 10 v peak at 2.5 amps peak. This should have similar max output voltage to Andre's amp, but more current output for driving real-world speakers. If you have 30 watts dissipated in just one BJT, say an MJL21193 or 94, then 12 watts into some load will be a doddle. 10V / 2.5A = 4 ohms. In the real world, 7 vrms into 4 ohms gives 12.2 watts. But 10V peak gives 12.5 watts! Jeez, I thought just rounding to 12 was close enough for guvm'nt work! Interdepartmental transfer of inordinately random co-opted projects always tend to have extrapolated variations in shared interest outcomes and result in small percentages of un-resolvable technical aberations due to the latest funding restrictions recommended by the Minister at the last conciliatory meeting to explore efficiency reforms concerning computer use and tea lady work contracts. I've been at the Bank too long, I understood that on the first readthrough! :-( The rails could be +/- 15v, and idle current 2 amps. Er no, has to be 2.5 amps, Gee, did that new junior officer advised me wrongly about the idle current?. It'd be better to have 3 amps, so that you could have +3 to -3 peak amp load change. With 15v rails, that's a dangerous 45 watts dissipation in a device? Nope, don't need 3 amps for 12.5 watts into 4 ohms, but do need 2.5 amps. and I'm going single-sided 30 volts. I'drather use an OPT than add a DC servo. I figured from 30 watts dissipation in the gain device, you'd get a max of 15 watts if the amp was perfect, and about 12 in the real world. Until it's all finalised, I can't be sure about the quiescent losses, but I'm going for 20V pk-pk output, and whatever supply voltage the losses demand above that. A 30 volt rail is just an edumacated guess at the moment. 12 watts into 4 ohms is 2.449 peak amps load current change, and you'd need rails of 9.8v in theory at an idle current of 2.449 in theory to get your 12 watts. Theoretically correct. Now we just need a theoretical speaker that remains at 4 ohms across the theoretical audio band! Idle dissipation is then 2.449A x 9.8V = 24 watts, or about the same as an EL34 in class A. This is quite OK for one lonely lil ol bjt. Indeed so, as is 30 watts. But you still have to bank on getting only 40% efficiency from your bjt, so you'd want to have 2.6A x 11.5V rail operating conditions. Oh, I'm banking on getting lot less than that from the circuit! :-) However, I'm not planning on having any idle current go mysteriously missing into what Andre might call the ether, so 2.5 amps through the output device should do nicely, and the various voltage drops can take their chances. This would be very nice in a car, since you already have a 12V battery. But you'd only be able to canoodle with Nicole K for 5 hrs before the amp hours mounted up, ( if nuthin else ) and the battery went flat. Yeah, and then the engine won't start, so I'll be stuck there all night with her. Another unexpected benefit of single-ended Class A! What's a few amps? a single car 100 W headlight might draw 8.3 amps... Indeed it might, but it won't sound that great. sue mucV 2 vto 10v3 in theoryp If you want 2.5 amps peak load current, you only need 1.25 amps idle. Not with single-ended operation. Then the load current varies from zero to +/- 2.5 amps With a P-P design, that's true. The active device will have 30 watts for its input power, anf the CCS will draw 30 watts of power at all times. Quite so (although I've got to dump a total of 75 watts among all the actives and passives), so a pair of good power trannies should be able to handle it easily, without heroic heatsinks or fan cooling. I thought of using the Sanken 2SC2922/2SA1216, the pnp being the current source. I don't know these two. We have lots like that on our prison records, long numbers, long prison terms, bad crimes.... I will do a Google on them. The characteristics of the 2sc2922 are at http://www.ampslab.com/PDF/2sc2922.pdf I don't see any typical operating conditions for class A use in SE config with simple easy to understand collector resistance lines with Ec / Ic axes, like the Ea Ia axes for the plate resistance curves on a 300B. With the data for the 300B, you just get a ruler, and rule in a load line, and the gain, distortion, and power output info is all able to be calculated in 3 minutes on the back of an envelope. But of course the ppl at Sanken would think you quite barmy if you told them what you wanted one of their devices for, so they make sure those communist left wing lefty class Aists can't get hold of sufficient info to wreck the economy by sending up power consumption and threatening shortages of fuses with bjt meltdowns. Well, they are at least designed for audio use, whereas most of the MOSFETs you find in audio amps are really switching devices. You won't be able to keep it simple, the right wing pro status quo have made damn sure bjts have positive thermal coefficients, the hotter they get, the more current they draw, and this does wonders for sales of replacement parts...... So you have have an effective biasing circuit, probably more complex than the audio circuit.... Well, I'm trying to figure cunning ways around that, without the support circuitry being ten times as complex as the signal path! In fact, that's the only reason I'm not using split rails. I'm intending to base it on a very linear BJT loaded by a constant current source. Here is where you will have trouble. BJTs have splendid current linearity, ie the current in/current out is fairly constant. The voltages are not so linear...... Yes, I'm aware of that, but it's worth trying, just to see what can be done. I'll keep a bottle of anti-depressant pills handy. Failure never depresses me, the next one can always be made better! Then you wil find the MJL21194 or 93 will have a current gain of about 30, so for 2 amps of load signal current you will need 66 mA of base drive current, and you'll find yourself being forced to use a darlington pair or triple if you want decently high input resistance. I'm probably going to horrify everyone here by avoiding iron, and using a simple capacitor-coupled output. You don't horrify me. But you are so wasteful. Iron allows twice the efficiency. But efficiency isn't audible, so be my guest and waste those watts. Not a big deal with only 75 watts per channel, methinks. My trusty Krell sucks 300 watts out of the wall when idling! Sugden have been using caps to couple their class A amps since about 1969, and I have no problem with that. maybe use 10 x 1,000 uF caps.... Or one 5,000 with a film bypass. The electros made now are many and varied, and act as true capacitors better than their 1969 parent caps because of the the need for pure capacitance in SMPS, operating at HF, and to bypass devices with HF signals. Ahh, that's true. If I use caps intended for SMPS, I shouldn't need the bypass - even simpler! Cheers! Methinks size is of no consequence. Small is now beautiful, ( Martha says she doesn't agree ), but I figure more small paralleled caps will have the lowest esr, and greatest tendency to survive an attack by a roaming "inadvertant short circuit of a speaker cable" True, but there's going to be a fuse in the power line for that. Currently (pun intended), the driver stage is taxing me somewhat, if simplicity is to be the keyword, but we are to keep below 1% THD at 10 watts into 4 ohms (another design target). Well I did remind you that driving bjt outputs requires current drive, and since you will **not** be allowed to use emitter follower or global NFB, then the input Z to the output BJT will be a horror story. What's wrong with using an emitter follower? You are using a great whacking ****e load of series voltage negative feedback. Yeah. So? So does a triode. Ever wondered what the ***voltage*** gain of a power bjt is? Huge. Alternatively zero depending how you look at it, because it doesn't respond to voltage per se. The Gm of a power bjt is perhaps about 30 amps per volt, ( +/- 100%, but lets's settle on 30 ), so that means for a 2.5 amp load current change, only 0.083 volts of change occurs between the base and emitter. This means that if you have 4 ohms and 10v load change, the gain is 10 / 0.08 = 125. Transistors just start to turn on at about 0.4V between base and emitter, and are fully turned on at around 0.8V; there is not a large voltage swing required for then to work. They are all like this, biguns and smalluns. No, really? :-) If the emitter voltage follows the base voltage which is 10.083V to get the emitter load voltage of 10V, then the gain reduction in EF mode is from 125 to slightly less than 1.0, so you have about 42 dB of applied series voltage NFB. If you had a mosfet, its Gm is only about 1A/V, so the "open loop gain" is only 4 into a 4 ohm load. Gain = load x Gm. So source follower = approximately 12 dB of NFB So what? BJT or FET, it's one device, using only local degeneration for feedback. And of course, the *current* gain of a FET is enormous, if you want to use that kind of logic. I doubt you would ever see 1% thd. You may be right, but one must *start* with a reasonable target. You would see considerably less that 1% with emitter follower, ( common collector mode ). If you tried to have the bjt in common emitter mode, with grounded emitter, load in the collector circuit, and with a small signal voltage applied to the base, I think the thd may be rather bad for what you intend. But the EF config reduces this thd by about 125 times, so if you had 10% open loop thd, expect about 0.08% with EF at the 12 watts. The use of EF has a technical benefit. Yes, but this is achieved by a higher level of feedback. I guess it all depends what kind of feedback the valvies will accept as 'good feedback', as opposed to global NFB. I'm basically trying to play in their yard, not design a higjhly linear amp per se, otherwise I wouldn't start from here, so to speak. It's my understanding that an EF output stage is not liked, for some reason I can't quite fathom. Suppose the hfe of the power bjt was a healthy 50 at the idle current you wanna run yours at . Then for 2.5 amps load change, you'd need 0.05A of base current change. If the base-emitter voltage change was 0.08V, then Ri = 0.08/0.05 = 1.6 ohms, or less than the load you want to run. Most people use another midi sized bjt to power the base, and if its hfe is say 120, its base input current is 0.05/120 = 0.416 mA, and Ri = 0.08/ 0.0004166A = 192 ohms, or 120 x 1.6 ohms. Still, Ri is very low, but its doable for a drive amp because we only want 0.08 volts of input. But if we use EF, then the input voltage is 125 times greater, and the same current change occurs, so Ri becomes 125 times higher so it is then 200 ohms without the second midi driver bjt, or 24kohms if the midi driver is used in what is called the darlington connection. THEN, it should only be necessary to make a linear driver stage able to produce a substantially linear +/- 10volts. This is only possible if the driver stage has current fb in its emitter circuit. A single MJE 350 will have an enormous voltage gain, maybe 1,000 into 10k of RL, and attrocious open loop thd, but the current NFB from an emitter R can reduce the gain to only 12 as you want it, so 40 dB more NFB is to be used. Using bjts mean that there is always gain to burn, toooooooo muuuuuchhhh gain, and huge amounts of NFB are routinely used to control gain. Alternatively, one could use the MJE350 to drive the base of the output transistor, and then its voltage gain will be very small, since it is driving in current source mode, and the hfe of the MJE350 will convert the output transistor base input resistance to a more usable resistance. Applying series voltage loop NFB will then raise the input resistance. But a small amount of NFB ( say 12 dB ), still won't be anything like enough to raise Ri to be friendly to preamps. Emitter follower mode is a severe case of loop series voltage negative feedback, and in the case of a lone class A BJT, amounts to far too much FB to comply with any KISS principle I find a transistor and one resistor to be quite simple, and triodes have loads of *internal* feedback. As you well know, there is no such thing as a 'zero feedback' SET amp. True. How much applied NFB is there in a 300B? BTW, zero NFB in tube palance means no external loops of NFB. The NFB already inside the tube was put there by the God Of Triodes, and let's not ague with a God in charge of so many high voltages...... I also intend the design to have no global feedback, but since it's not even complete on paper yet, never mind actually built and tested, that may change. Nobody has yet been able to make a bjt amp with no loops of NFB and still get a low Ro. Possibly, nobody ever will! If you pull it off, the world will beat a path to your door.... But look, I'll make a concession, and allow you just enough NFB so that your amp's Ro = 1 ohm, and no lower. That only gives you a DF of 4 with a 4 ohm load, but its the bare minimum, imho. Otherwise I won't come to the listening tests between the two amp genre. Why would I want to hear music via a current source with really high THD??? I'm not disagreeing with you, but they seem to sell pretty well, according to Cary, Jadis, Audio Note etc......... :-) Er, the Ra of the triodes involved is usually less than 1/4 of RL, so such an item is considered to be a voltage source. If Ra was 4 times RL, sure, that would make a tube a current source. This is the case with many mantle radios using a lone beam tetrode like a 6V6. Ro = 20 ohms. Sound is fine up to 0.25 watts..... Patience guys, Andre has been working on (and presumably refining) his KISS amp for God knows how long, I'm starting from a clean sheet. One thing, mine will be a lot cheaper to build, and a lot safer for a newcomer to DIY amps! A Sziclai pair come to mind for you...... Well, I've used Sziklai pairs in fully complementary amps before, and I'm still kicking possible topologies around on paper, trying to minimise the parts count.. Betcha you use a shirtload more than Andre's 300B amp. Oh yes, but then I was trying to build a *good* amp, not just a minimally simple one! :-) Less is not necessarily more................. But I reckon pretty soon you'll be pulling all the hair out of your head, and screaming for someone to send you a few mosfets and j-fet drivers. It is in the realm of possibilities that your'e bald, and stubborn, and in that case, good luck, you'll need it! I'm always open to good suggestions! However, I don't want to build a Zen clone, that's already been done. I agree with all that you say in theory, but I've never actually been dumb enough to build a simple SE BJT amp with no global NFB, and see what it actually sounds like, despite the horrible specs! :-) I'd have to say it may be impossible to design a bjt amp which has no more local or loop NFB than the 300B already has, and has only two gain devices. That may be right, three devices (possibly two of them could be one darlington), seems more sensible - indeed, that's also common in 300B SETs. The input resistance issue was a pain to designers when bjts first started being used to replace tubes. Ah, at last, said engineers, we can get rid of the OPT. Then they sat down to design input transformers. They then tried to ban large currents from transistors, too bleeding unreliable. But then somebody replaced germanium with silicon in a special process, and that made matters simpler, but one still sees hellishly complex SS amps with huge numbers of bjts all strung out across vast pcbs, and its very hard to work out how these work, especially if there is a double sided board designed by a PC. Some are impressive, one Electronics World design had umpteen bjts, and puts out 300 watts, and has a slew rate of 250V/uS. Servicing of complex SS circuits is not supposed to be done. The circuits are supposed to be fully short circuit proof, and easy to fix. You pull out the plug in module with its 250 cheap parts, and bin it, and plug in the replacement. And pigs will fly when makers all addapt such ideas, and where they do, they charge a fortune for the replacement module; they'd rather you bought the latest amp, than repair the oldun, now 10 years "old". But a replacement of several blown output bjts and other collaterally caused failures is often far more than a set of new tubes.... Ppl keep sending me ancient Quad II amps, repairs are cheap, since 6L6 work OK in them. I'm not that bothered, since I've never had an amp failure. Oooohh, prob'ly shouldn't have said that! :-) But they want upgrades anyhow, and like to see things already 50 years old be able to last them to their graves, maybe 30 years away. Will any solid state Sony, Marantz, Yamaha made in 1975 be serviceable in 2025? What of the gear made after 1985, with all that microprocessor control crap and switching behind the front panels? Who cares? Just buy a new one. At 2025 labour rates, that'll have to be the cheap option! :-) Besides, they'll mostly be cheap and lightweight class Ds running direct off the mains by then (heh, that'll get the SET gang twitching!). Who on earth will be able to decode the circuits if they have not got the exact workshop manual for the amp model they happen to be lumbered with? None of the special spare parts will be available. IMHO, a well re-engineered Quad II will hold its own with a 300B amp. Hmmmm, I'd back a well-designed 300B P-P in class A against a Quad II any day, let alone an optimum 300B with 4 output valves. Of course, a 're-engineered' one isn't necessarily a Quad II any more, now is it? However, thanks for all the thought you've clearly put into this little enterprise. -- Stewart Pinkerton | Music is Art - Audio is Engineering |
#64
|
|||
|
|||
Stewart Pinkerton wrote: On Thu, 09 Dec 2004 00:04:21 +1100, Patrick Turner wrote: Stewart Pinkerton wrote: On Tue, 07 Dec 2004 19:55:05 +1100, Patrick Turner wrote: Stewart Pinkerton wrote: As it happens, I've started a design, targeted at 12 watts into 4 ohms, i.e. 10 v peak at 2.5 amps peak. This should have similar max output voltage to Andre's amp, but more current output for driving real-world speakers. If you have 30 watts dissipated in just one BJT, say an MJL21193 or 94, then 12 watts into some load will be a doddle. 10V / 2.5A = 4 ohms. In the real world, 7 vrms into 4 ohms gives 12.2 watts. But 10V peak gives 12.5 watts! Jeez, I thought just rounding to 12 was close enough for guvm'nt work! Interdepartmental transfer of inordinately random co-opted projects always tend to have extrapolated variations in shared interest outcomes and result in small percentages of un-resolvable technical aberations due to the latest funding restrictions recommended by the Minister at the last conciliatory meeting to explore efficiency reforms concerning computer use and tea lady work contracts. I've been at the Bank too long, I understood that on the first readthrough! :-( Bankers are pretty dry sort of people. I see why you mind is at ease with 0.001% amplifiers. You worry about the cents! The rails could be +/- 15v, and idle current 2 amps. Er no, has to be 2.5 amps, Gee, did that new junior officer advised me wrongly about the idle current?. It'd be better to have 3 amps, so that you could have +3 to -3 peak amp load change. With 15v rails, that's a dangerous 45 watts dissipation in a device? Nope, don't need 3 amps for 12.5 watts into 4 ohms, but do need 2.5 amps. For goodness sake, let us never build in any slight capacity to power a lower load than 4 ohms. In actual fact, what you suggest will do because the amp should never be used where it would clip, which means an average low level of a watt would be all you'd be using. Well, that's ok for rockn'roll. Classical might catch you out, since we turn it up to hear the lone violin at the beginning, not realising that 5 minutes later 75 other artistes will join in loudly..... and I'm going single-sided 30 volts. I'drather use an OPT than add a DC servo. I figured from 30 watts dissipation in the gain device, you'd get a max of 15 watts if the amp was perfect, and about 12 in the real world. Until it's all finalised, I can't be sure about the quiescent losses, but I'm going for 20V pk-pk output, and whatever supply voltage the losses demand above that. A 30 volt rail is just an edumacated guess at the moment. I leave you to it. 12 watts into 4 ohms is 2.449 peak amps load current change, and you'd need rails of 9.8v in theory at an idle current of 2.449 in theory to get your 12 watts. Theoretically correct. Now we just need a theoretical speaker that remains at 4 ohms across the theoretical audio band! Idle dissipation is then 2.449A x 9.8V = 24 watts, or about the same as an EL34 in class A. This is quite OK for one lonely lil ol bjt. Indeed so, as is 30 watts. But you still have to bank on getting only 40% efficiency from your bjt, so you'd want to have 2.6A x 11.5V rail operating conditions. Oh, I'm banking on getting lot less than that from the circuit! :-) However, I'm not planning on having any idle current go mysteriously missing into what Andre might call the ether, so 2.5 amps through the output device should do nicely, and the various voltage drops can take their chances. This would be very nice in a car, since you already have a 12V battery. But you'd only be able to canoodle with Nicole K for 5 hrs before the amp hours mounted up, ( if nuthin else ) and the battery went flat. Yeah, and then the engine won't start, so I'll be stuck there all night with her. Another unexpected benefit of single-ended Class A! Well ya better walk her to the nearest railway station, she's due here soon, and I want her in fine condition. I have a lot of mowing and hedge clipping needing doing. What's a few amps? a single car 100 W headlight might draw 8.3 amps... Indeed it might, but it won't sound that great. sue mucV 2 vto 10v3 in theoryp If you want 2.5 amps peak load current, you only need 1.25 amps idle. Not with single-ended operation. Then the load current varies from zero to +/- 2.5 amps With a P-P design, that's true. The active device will have 30 watts for its input power, anf the CCS will draw 30 watts of power at all times. Quite so (although I've got to dump a total of 75 watts among all the actives and passives), so a pair of good power trannies should be able to handle it easily, without heroic heatsinks or fan cooling. I thought of using the Sanken 2SC2922/2SA1216, the pnp being the current source. I don't know these two. We have lots like that on our prison records, long numbers, long prison terms, bad crimes.... I will do a Google on them. The characteristics of the 2sc2922 are at http://www.ampslab.com/PDF/2sc2922.pdf I don't see any typical operating conditions for class A use in SE config with simple easy to understand collector resistance lines with Ec / Ic axes, like the Ea Ia axes for the plate resistance curves on a 300B. With the data for the 300B, you just get a ruler, and rule in a load line, and the gain, distortion, and power output info is all able to be calculated in 3 minutes on the back of an envelope. But of course the ppl at Sanken would think you quite barmy if you told them what you wanted one of their devices for, so they make sure those communist left wing lefty class Aists can't get hold of sufficient info to wreck the economy by sending up power consumption and threatening shortages of fuses with bjt meltdowns. Well, they are at least designed for audio use, whereas most of the MOSFETs you find in audio amps are really switching devices. So why do they design bjt amps which spend all their darn life switching on and bloody off? Give me a mosfet any day, and I can give you music.... You won't be able to keep it simple, the right wing pro status quo have made damn sure bjts have positive thermal coefficients, the hotter they get, the more current they draw, and this does wonders for sales of replacement parts...... So you have have an effective biasing circuit, probably more complex than the audio circuit.... Well, I'm trying to figure cunning ways around that, without the support circuitry being ten times as complex as the signal path! In fact, that's the only reason I'm not using split rails. With cap coupled loads, no need for split rails. Split rails became elegant when the fashions changed to using direct coupled pnp and npn complementary outputs, and it made the differential input pair easy to bias and arrange. John Linsley Hood had quite a few class A designs in Wireless World. There is a wealth of info in those magazines and I suggest that if you were serious, try spending a sunday at the local university library which may still have them tucked away in an archive basement, that is if they have not cleared them out, and convereted them to gyms for $10 entry per hour. I'm intending to base it on a very linear BJT loaded by a constant current source. Here is where you will have trouble. BJTs have splendid current linearity, ie the current in/current out is fairly constant. The voltages are not so linear...... Yes, I'm aware of that, but it's worth trying, just to see what can be done. I'll keep a bottle of anti-depressant pills handy. Failure never depresses me, the next one can always be made better! Like a weeks of flooding rains is OK, when the sun does make an appearance, whoopee! Then you wil find the MJL21194 or 93 will have a current gain of about 30, so for 2 amps of load signal current you will need 66 mA of base drive current, and you'll find yourself being forced to use a darlington pair or triple if you want decently high input resistance. I'm probably going to horrify everyone here by avoiding iron, and using a simple capacitor-coupled output. You don't horrify me. But you are so wasteful. Iron allows twice the efficiency. But efficiency isn't audible, so be my guest and waste those watts. Not a big deal with only 75 watts per channel, methinks. My trusty Krell sucks 300 watts out of the wall when idling! Sugden have been using caps to couple their class A amps since about 1969, and I have no problem with that. maybe use 10 x 1,000 uF caps.... Or one 5,000 with a film bypass. The electros made now are many and varied, and act as true capacitors better than their 1969 parent caps because of the the need for pure capacitance in SMPS, operating at HF, and to bypass devices with HF signals. Ahh, that's true. If I use caps intended for SMPS, I shouldn't need the bypass - even simpler! Cheers! Methinks size is of no consequence. Small is now beautiful, ( Martha says she doesn't agree ), but I figure more small paralleled caps will have the lowest esr, and greatest tendency to survive an attack by a roaming "inadvertant short circuit of a speaker cable" True, but there's going to be a fuse in the power line for that. Yeah, but the fuse has to be say 2 amps, and then if a frayed speaker cable just has one fine strand causing a short, the load current is quite high, but not high enough to blow a fuse, and before you notice what's wrong with one channel, the bjt blows instead of the fuse, due to excessive dissipation. Only unlike a fuse, the bjt goes to a short circuit itself, and then something else fuses, a rail fuse, or the supply diodes..... I get clouds of amps which have blown up like this every year. Almost no amps have circuits which work out if the load value is below 2 ohms, and then trigger the protection circuit, ot just turn the amp off, which means you need an auxiliary PS for the "protection and management regime". Currently (pun intended), the driver stage is taxing me somewhat, if simplicity is to be the keyword, but we are to keep below 1% THD at 10 watts into 4 ohms (another design target). Well I did remind you that driving bjt outputs requires current drive, and since you will **not** be allowed to use emitter follower or global NFB, then the input Z to the output BJT will be a horror story. What's wrong with using an emitter follower? You are using a great whacking ****e load of series voltage negative feedback. Yeah. So? So does a triode. I think you may find that EF bjt connection involves far more applied NFB than the feedback in a 300B. Ever wondered what the ***voltage*** gain of a power bjt is? Huge. Alternatively zero depending how you look at it, because it doesn't respond to voltage per se. Oh but they do. In addition to a voltage signal input, you also need a current. Without the voltage, the current won't flow. The Gm of a power bjt is perhaps about 30 amps per volt, ( +/- 100%, but lets's settle on 30 ), so that means for a 2.5 amp load current change, only 0.083 volts of change occurs between the base and emitter. This means that if you have 4 ohms and 10v load change, the gain is 10 / 0.08 = 125. Transistors just start to turn on at about 0.4V between base and emitter, and are fully turned on at around 0.8V; there is not a large voltage swing required for then to work. They are all like this, biguns and smalluns. No, really? :-) If the emitter voltage follows the base voltage which is 10.083V to get the emitter load voltage of 10V, then the gain reduction in EF mode is from 125 to slightly less than 1.0, so you have about 42 dB of applied series voltage NFB. If you had a mosfet, its Gm is only about 1A/V, so the "open loop gain" is only 4 into a 4 ohm load. Gain = load x Gm. So source follower = approximately 12 dB of NFB So what? BJT or FET, it's one device, using only local degeneration for feedback. And of course, the *current* gain of a FET is enormous, if you want to use that kind of logic. But the *amount* of series volage feedback in bjt EF is far greater than with SF in a mosfet. The distortion from a pair of npn and pnp average power mosfets in class A complementary pair SF is about the same as a pair of triodes in PP class A, about 1%. The EF bjt pair in class A should have a lot less thd, since far more NFB is involved, since there is far greater gain reduction.... I doubt you would ever see 1% thd. You may be right, but one must *start* with a reasonable target. You would see considerably less that 1% with emitter follower, ( common collector mode ). If you tried to have the bjt in common emitter mode, with grounded emitter, load in the collector circuit, and with a small signal voltage applied to the base, I think the thd may be rather bad for what you intend. But the EF config reduces this thd by about 125 times, so if you had 10% open loop thd, expect about 0.08% with EF at the 12 watts. The use of EF has a technical benefit. Yes, but this is achieved by a higher level of feedback. I guess it all depends what kind of feedback the valvies will accept as 'good feedback', as opposed to global NFB. I'm basically trying to play in their yard, not design a higjhly linear amp per se, otherwise I wouldn't start from here, so to speak. It's my understanding that an EF output stage is not liked, for some reason I can't quite fathom. Well then to be fair, don't use any more FB than is used in a 300B. Therfore EF is not an allowable solution. The output bjt should be set up in common collector, as a gain device, load cap coupled off its collector, not its emitter, and the base operated by another single bjt, and I'll allow you only enough loop or other NFB to give the same Ro as achieved by the 300B I doubt anyone can do this, because of the base input resistances involved. Nelson Pass managed the simplest SS amp of all, one mosfet, one CCS, and shunt NFB in the input-gate-collector circuit. But it needed 4.5 volts of drive, and still rather low input impedance, because that's what you get with shunt feedback input circuits. He would have done better to excomunicate the CCS, so one device out and gone, and a big hot lazy ******* it is too. Then have an OPT, and a single drive j-fet, with some normal NFB, still only 12 dB required, and hed have ahd himself a better simple amp imho. Or he could have had a simple choke feed to the output fet, and cap couple to the load, so no hassles with OPTs, but a cap coupled OPT using a toroidal transformer works very well. What we are doing here is true simplicity, tiny amount of gain, very few devices, small quantity of NFB, low noise, lowish thd, wide bw, therefore usable. OK, the average diyer gets horified by having to use hardware instead of multiple devices which are cheap along with lots of NFB, also easy but only if you know what you are doing, because anyone **designing** and building an amp can easily end up with an oscillator instaed of an amp. Well, I've used Sziklai pairs in fully complementary amps before, and I'm still kicking possible topologies around on paper, trying to minimise the parts count.. Betcha you use a shirtload more than Andre's 300B amp. Oh yes, but then I was trying to build a *good* amp, not just a minimally simple one! :-) Less is not necessarily more................. Goodness and badness are never agreed upon. But I reckon pretty soon you'll be pulling all the hair out of your head, and screaming for someone to send you a few mosfets and j-fet drivers. It is in the realm of possibilities that your'e bald, and stubborn, and in that case, good luck, you'll need it! I'm always open to good suggestions! However, I don't want to build a Zen clone, that's already been done. I agree with all that you say in theory, but I've never actually been dumb enough to build a simple SE BJT amp with no global NFB, and see what it actually sounds like, despite the horrible specs! :-) I'd have to say it may be impossible to design a bjt amp which has no more local or loop NFB than the 300B already has, and has only two gain devices. That may be right, three devices (possibly two of them could be one darlington), seems more sensible - indeed, that's also common in 300B SETs. I am a "3 device" SE amp maker. None of the devices need much gain. I also use some loop NFB, rather than none. Then the requirements for the preamp are no more than a gain pot, or a preamp which will operate at such low levels, the thd is usually less than 0.01% at all times, and totally blameless, imho. I know, I just measured a preamp I built for a client. And pigs will fly when makers all addapt such ideas, and where they do, they charge a fortune for the replacement module; they'd rather you bought the latest amp, than repair the oldun, now 10 years "old". But a replacement of several blown output bjts and other collaterally caused failures is often far more than a set of new tubes.... Ppl keep sending me ancient Quad II amps, repairs are cheap, since 6L6 work OK in them. I'm not that bothered, since I've never had an amp failure. Oooohh, prob'ly shouldn't have said that! :-) I have seen Sugden class A amps with major problems with stuffed coupling caps and other faults from shorted speaker leads. But at least the speakers didn't fry with DC. Hasn't your Krell got some protection? But they want upgrades anyhow, and like to see things already 50 years old be able to last them to their graves, maybe 30 years away. Will any solid state Sony, Marantz, Yamaha made in 1975 be serviceable in 2025? What of the gear made after 1985, with all that microprocessor control crap and switching behind the front panels? Who cares? Just buy a new one. At 2025 labour rates, that'll have to be the cheap option! :-) Besides, they'll mostly be cheap and lightweight class Ds running direct off the mains by then (heh, that'll get the SET gang twitching!). I forsee extended wars of words between digital and analog camps. Who on earth will be able to decode the circuits if they have not got the exact workshop manual for the amp model they happen to be lumbered with? None of the special spare parts will be available. IMHO, a well re-engineered Quad II will hold its own with a 300B amp. Hmmmm, I'd back a well-designed 300B P-P in class A against a Quad II any day, let alone an optimum 300B with 4 output valves. Of course, a 're-engineered' one isn't necessarily a Quad II any more, now is it? Well there isn't too much wrong with the Quad II when used with Quad speakers. The ESL57 isn't a bad load for them. But I get folks wanting to use modern dynamics, maybe only 4 ohms, and that reduces their charms immediately, because the circuit luches into class AB above a low threshold, and the sound falls apart, mainly because of their appalling power supply, designed before silicon diodes and decent small electros. The simplest improvements to a Quad II are 1, Fit individual RC bypassing to each output tube cathode. 2, At least have a GZ34 as the rectifier. 3, Remove the bulky low value boxed electros, 4, Install a CLC filter, 47uF, 2H, 100uF instead of the large cap for the reduction of hum appearing at the CT of the OPT. 5, Replace all the old Hunts caps, and resistors with modern caps and metal films. 6, Make sure the tubes are in good operating order. 7, Set up the OPT settings for 8 ohms, and never use the 16 ohm setting, 8, Add stabilisation network to prevent oscilations into C loads with the 8 ohm setting, 9, Measure the thd and try swapping positions of the EF86, and the KT66 to get the lowest thd at 3 watts into 8 ohms. The Quad 22 control unit is another story, another time. However, thanks for all the thought you've clearly put into this little enterprise. The group may want to know about the simplicity or complexity issues regarding the amps they might use, or might build. Patrick Turner. -- Stewart Pinkerton | Music is Art - Audio is Engineering |
#65
|
|||
|
|||
On Fri, 10 Dec 2004 01:37:03 +1100, Patrick Turner
wrote: Stewart Pinkerton wrote: On Thu, 09 Dec 2004 00:04:21 +1100, Patrick Turner wrote: Stewart Pinkerton wrote: On Tue, 07 Dec 2004 19:55:05 +1100, Patrick Turner wrote: Stewart Pinkerton wrote: As it happens, I've started a design, targeted at 12 watts into 4 ohms, i.e. 10 v peak at 2.5 amps peak. This should have similar max output voltage to Andre's amp, but more current output for driving real-world speakers. If you have 30 watts dissipated in just one BJT, say an MJL21193 or 94, then 12 watts into some load will be a doddle. 10V / 2.5A = 4 ohms. In the real world, 7 vrms into 4 ohms gives 12.2 watts. But 10V peak gives 12.5 watts! Jeez, I thought just rounding to 12 was close enough for guvm'nt work! Interdepartmental transfer of inordinately random co-opted projects always tend to have extrapolated variations in shared interest outcomes and result in small percentages of un-resolvable technical aberations due to the latest funding restrictions recommended by the Minister at the last conciliatory meeting to explore efficiency reforms concerning computer use and tea lady work contracts. I've been at the Bank too long, I understood that on the first readthrough! :-( Bankers are pretty dry sort of people. I see why you mind is at ease with 0.001% amplifiers. You worry about the cents! The rails could be +/- 15v, and idle current 2 amps. Er no, has to be 2.5 amps, Gee, did that new junior officer advised me wrongly about the idle current?. It'd be better to have 3 amps, so that you could have +3 to -3 peak amp load change. With 15v rails, that's a dangerous 45 watts dissipation in a device? Nope, don't need 3 amps for 12.5 watts into 4 ohms, but do need 2.5 amps. For goodness sake, let us never build in any slight capacity to power a lower load than 4 ohms. In actual fact, what you suggest will do because the amp should never be used where it would clip, which means an average low level of a watt would be all you'd be using. Well, that's ok for rockn'roll. Classical might catch you out, since we turn it up to hear the lone violin at the beginning, not realising that 5 minutes later 75 other artistes will join in loudly..... and I'm going single-sided 30 volts. I'drather use an OPT than add a DC servo. I figured from 30 watts dissipation in the gain device, you'd get a max of 15 watts if the amp was perfect, and about 12 in the real world. Until it's all finalised, I can't be sure about the quiescent losses, but I'm going for 20V pk-pk output, and whatever supply voltage the losses demand above that. A 30 volt rail is just an edumacated guess at the moment. I leave you to it. 12 watts into 4 ohms is 2.449 peak amps load current change, and you'd need rails of 9.8v in theory at an idle current of 2.449 in theory to get your 12 watts. Theoretically correct. Now we just need a theoretical speaker that remains at 4 ohms across the theoretical audio band! Idle dissipation is then 2.449A x 9.8V = 24 watts, or about the same as an EL34 in class A. This is quite OK for one lonely lil ol bjt. Indeed so, as is 30 watts. But you still have to bank on getting only 40% efficiency from your bjt, so you'd want to have 2.6A x 11.5V rail operating conditions. Oh, I'm banking on getting lot less than that from the circuit! :-) However, I'm not planning on having any idle current go mysteriously missing into what Andre might call the ether, so 2.5 amps through the output device should do nicely, and the various voltage drops can take their chances. This would be very nice in a car, since you already have a 12V battery. But you'd only be able to canoodle with Nicole K for 5 hrs before the amp hours mounted up, ( if nuthin else ) and the battery went flat. Yeah, and then the engine won't start, so I'll be stuck there all night with her. Another unexpected benefit of single-ended Class A! Well ya better walk her to the nearest railway station, she's due here soon, and I want her in fine condition. I have a lot of mowing and hedge clipping needing doing. What's a few amps? a single car 100 W headlight might draw 8.3 amps... Indeed it might, but it won't sound that great. sue mucV 2 vto 10v3 in theoryp If you want 2.5 amps peak load current, you only need 1.25 amps idle. Not with single-ended operation. Then the load current varies from zero to +/- 2.5 amps With a P-P design, that's true. The active device will have 30 watts for its input power, anf the CCS will draw 30 watts of power at all times. Quite so (although I've got to dump a total of 75 watts among all the actives and passives), so a pair of good power trannies should be able to handle it easily, without heroic heatsinks or fan cooling. I thought of using the Sanken 2SC2922/2SA1216, the pnp being the current source. I don't know these two. We have lots like that on our prison records, long numbers, long prison terms, bad crimes.... I will do a Google on them. The characteristics of the 2sc2922 are at http://www.ampslab.com/PDF/2sc2922.pdf I don't see any typical operating conditions for class A use in SE config with simple easy to understand collector resistance lines with Ec / Ic axes, like the Ea Ia axes for the plate resistance curves on a 300B. With the data for the 300B, you just get a ruler, and rule in a load line, and the gain, distortion, and power output info is all able to be calculated in 3 minutes on the back of an envelope. But of course the ppl at Sanken would think you quite barmy if you told them what you wanted one of their devices for, so they make sure those communist left wing lefty class Aists can't get hold of sufficient info to wreck the economy by sending up power consumption and threatening shortages of fuses with bjt meltdowns. Well, they are at least designed for audio use, whereas most of the MOSFETs you find in audio amps are really switching devices. So why do they design bjt amps which spend all their darn life switching on and bloody off? Give me a mosfet any day, and I can give you music.... You won't be able to keep it simple, the right wing pro status quo have made damn sure bjts have positive thermal coefficients, the hotter they get, the more current they draw, and this does wonders for sales of replacement parts...... So you have have an effective biasing circuit, probably more complex than the audio circuit.... Well, I'm trying to figure cunning ways around that, without the support circuitry being ten times as complex as the signal path! In fact, that's the only reason I'm not using split rails. With cap coupled loads, no need for split rails. Split rails became elegant when the fashions changed to using direct coupled pnp and npn complementary outputs, and it made the differential input pair easy to bias and arrange. John Linsley Hood had quite a few class A designs in Wireless World. There is a wealth of info in those magazines and I suggest that if you were serious, try spending a sunday at the local university library which may still have them tucked away in an archive basement, that is if they have not cleared them out, and convereted them to gyms for $10 entry per hour. I already have copies of all his articles, indeed one of my earliest audio construction projects was his original Class A circuit, built within weeks of the article appearing in WW. However, I can't use his excellent circuit for two reasons: it uses a lot of global NFB and it's arguably not single-ended. I'm intending to base it on a very linear BJT loaded by a constant current source. Here is where you will have trouble. BJTs have splendid current linearity, ie the current in/current out is fairly constant. The voltages are not so linear...... Yes, I'm aware of that, but it's worth trying, just to see what can be done. I'll keep a bottle of anti-depressant pills handy. Failure never depresses me, the next one can always be made better! Like a weeks of flooding rains is OK, when the sun does make an appearance, whoopee! Yup, the barbie is still in the garden, just in case! :-) Then you wil find the MJL21194 or 93 will have a current gain of about 30, so for 2 amps of load signal current you will need 66 mA of base drive current, and you'll find yourself being forced to use a darlington pair or triple if you want decently high input resistance. I'm probably going to horrify everyone here by avoiding iron, and using a simple capacitor-coupled output. You don't horrify me. But you are so wasteful. Iron allows twice the efficiency. But efficiency isn't audible, so be my guest and waste those watts. Not a big deal with only 75 watts per channel, methinks. My trusty Krell sucks 300 watts out of the wall when idling! Sugden have been using caps to couple their class A amps since about 1969, and I have no problem with that. maybe use 10 x 1,000 uF caps.... Or one 5,000 with a film bypass. The electros made now are many and varied, and act as true capacitors better than their 1969 parent caps because of the the need for pure capacitance in SMPS, operating at HF, and to bypass devices with HF signals. Ahh, that's true. If I use caps intended for SMPS, I shouldn't need the bypass - even simpler! Cheers! Methinks size is of no consequence. Small is now beautiful, ( Martha says she doesn't agree ), but I figure more small paralleled caps will have the lowest esr, and greatest tendency to survive an attack by a roaming "inadvertant short circuit of a speaker cable" True, but there's going to be a fuse in the power line for that. Yeah, but the fuse has to be say 2 amps, and then if a frayed speaker cable just has one fine strand causing a short, the load current is quite high, but not high enough to blow a fuse, and before you notice what's wrong with one channel, the bjt blows instead of the fuse, due to excessive dissipation. Only unlike a fuse, the bjt goes to a short circuit itself, and then something else fuses, a rail fuse, or the supply diodes..... Ah, your glass is always half-empty, is it? :-) I get clouds of amps which have blown up like this every year. Almost no amps have circuits which work out if the load value is below 2 ohms, and then trigger the protection circuit, ot just turn the amp off, which means you need an auxiliary PS for the "protection and management regime". I might end up with another much higher voltage low-current rail, for quite different reasons. Currently (pun intended), the driver stage is taxing me somewhat, if simplicity is to be the keyword, but we are to keep below 1% THD at 10 watts into 4 ohms (another design target). Well I did remind you that driving bjt outputs requires current drive, and since you will **not** be allowed to use emitter follower or global NFB, then the input Z to the output BJT will be a horror story. What's wrong with using an emitter follower? You are using a great whacking ****e load of series voltage negative feedback. Yeah. So? So does a triode. I think you may find that EF bjt connection involves far more applied NFB than the feedback in a 300B. Yeah, so? Where's the harm? Ever wondered what the ***voltage*** gain of a power bjt is? Huge. Alternatively zero depending how you look at it, because it doesn't respond to voltage per se. Oh but they do. In addition to a voltage signal input, you also need a current. Without the voltage, the current won't flow. The Gm of a power bjt is perhaps about 30 amps per volt, ( +/- 100%, but lets's settle on 30 ), so that means for a 2.5 amp load current change, only 0.083 volts of change occurs between the base and emitter. This means that if you have 4 ohms and 10v load change, the gain is 10 / 0.08 = 125. Transistors just start to turn on at about 0.4V between base and emitter, and are fully turned on at around 0.8V; there is not a large voltage swing required for then to work. They are all like this, biguns and smalluns. No, really? :-) If the emitter voltage follows the base voltage which is 10.083V to get the emitter load voltage of 10V, then the gain reduction in EF mode is from 125 to slightly less than 1.0, so you have about 42 dB of applied series voltage NFB. If you had a mosfet, its Gm is only about 1A/V, so the "open loop gain" is only 4 into a 4 ohm load. Gain = load x Gm. So source follower = approximately 12 dB of NFB So what? BJT or FET, it's one device, using only local degeneration for feedback. And of course, the *current* gain of a FET is enormous, if you want to use that kind of logic. But the *amount* of series volage feedback in bjt EF is far greater than with SF in a mosfet. Yeah, so? The distortion from a pair of npn and pnp average power mosfets in class A complementary pair SF is about the same as a pair of triodes in PP class A, about 1%. The EF bjt pair in class A should have a lot less thd, since far more NFB is involved, since there is far greater gain reduction.... And this is a bad thing, exactly how? I doubt you would ever see 1% thd. You may be right, but one must *start* with a reasonable target. You would see considerably less that 1% with emitter follower, ( common collector mode ). If you tried to have the bjt in common emitter mode, with grounded emitter, load in the collector circuit, and with a small signal voltage applied to the base, I think the thd may be rather bad for what you intend. But the EF config reduces this thd by about 125 times, so if you had 10% open loop thd, expect about 0.08% with EF at the 12 watts. The use of EF has a technical benefit. Yes, but this is achieved by a higher level of feedback. I guess it all depends what kind of feedback the valvies will accept as 'good feedback', as opposed to global NFB. I'm basically trying to play in their yard, not design a higjhly linear amp per se, otherwise I wouldn't start from here, so to speak. It's my understanding that an EF output stage is not liked, for some reason I can't quite fathom. Well then to be fair, don't use any more FB than is used in a 300B. Therfore EF is not an allowable solution. ********, the point is the equivalent simplicity of topolgy, combined with an absence of global feedback. I'm not trying to *simulate* the 300B amp, just come up with a notionally similar KISASS amp. That's uh, Keep It Simple And Solid State! :-) The output bjt should be set up in common collector, as a gain device, load cap coupled off its collector, not its emitter, and the base operated by another single bjt, and I'll allow you only enough loop or other NFB to give the same Ro as achieved by the 300B I doubt anyone can do this, because of the base input resistances involved. Nelson Pass managed the simplest SS amp of all, one mosfet, one CCS, and shunt NFB in the input-gate-collector circuit. But it needed 4.5 volts of drive, and still rather low input impedance, because that's what you get with shunt feedback input circuits. Indeed so, and there's no point in making a Zen clone anyway, since that game's already been played, and we know the result. He would have done better to excomunicate the CCS, so one device out and gone, and a big hot lazy ******* it is too. Then have an OPT, and a single drive j-fet, with some normal NFB, still only 12 dB required, and hed have ahd himself a better simple amp imho. Depends how you look at it - the CCS has to be cheaper than the OPT, so why use an OPT just for KISS brownie points? Or he could have had a simple choke feed to the output fet, and cap couple to the load, so no hassles with OPTs, but a cap coupled OPT using a toroidal transformer works very well. What we are doing here is true simplicity, tiny amount of gain, very few devices, small quantity of NFB, low noise, lowish thd, wide bw, therefore usable. OK, the average diyer gets horified by having to use hardware instead of multiple devices which are cheap along with lots of NFB, also easy but only if you know what you are doing, because anyone **designing** and building an amp can easily end up with an oscillator instaed of an amp. Indeed so, in fact my old mentor used to say that the best way to design an oscillator was to *think* that you were designing an amplifier! :-) Well, I've used Sziklai pairs in fully complementary amps before, and I'm still kicking possible topologies around on paper, trying to minimise the parts count.. Betcha you use a shirtload more than Andre's 300B amp. Oh yes, but then I was trying to build a *good* amp, not just a minimally simple one! :-) Less is not necessarily more................. Goodness and badness are never agreed upon. Not around here, that's for sure! But I reckon pretty soon you'll be pulling all the hair out of your head, and screaming for someone to send you a few mosfets and j-fet drivers. It is in the realm of possibilities that your'e bald, and stubborn, and in that case, good luck, you'll need it! I'm always open to good suggestions! However, I don't want to build a Zen clone, that's already been done. I agree with all that you say in theory, but I've never actually been dumb enough to build a simple SE BJT amp with no global NFB, and see what it actually sounds like, despite the horrible specs! :-) I'd have to say it may be impossible to design a bjt amp which has no more local or loop NFB than the 300B already has, and has only two gain devices. That may be right, three devices (possibly two of them could be one darlington), seems more sensible - indeed, that's also common in 300B SETs. I am a "3 device" SE amp maker. None of the devices need much gain. I also use some loop NFB, rather than none. Then the requirements for the preamp are no more than a gain pot, or a preamp which will operate at such low levels, the thd is usually less than 0.01% at all times, and totally blameless, imho. I know, I just measured a preamp I built for a client. And pigs will fly when makers all addapt such ideas, and where they do, they charge a fortune for the replacement module; they'd rather you bought the latest amp, than repair the oldun, now 10 years "old". But a replacement of several blown output bjts and other collaterally caused failures is often far more than a set of new tubes.... Ppl keep sending me ancient Quad II amps, repairs are cheap, since 6L6 work OK in them. I'm not that bothered, since I've never had an amp failure. Oooohh, prob'ly shouldn't have said that! :-) I have seen Sugden class A amps with major problems with stuffed coupling caps and other faults from shorted speaker leads. But at least the speakers didn't fry with DC. Hasn't your Krell got some protection? Yes, but not being a total klutz, I don't short my speaker leads. Of course, if I had warranty returns to consider, I'd use a sophisticated protection circuit! But they want upgrades anyhow, and like to see things already 50 years old be able to last them to their graves, maybe 30 years away. Will any solid state Sony, Marantz, Yamaha made in 1975 be serviceable in 2025? What of the gear made after 1985, with all that microprocessor control crap and switching behind the front panels? Who cares? Just buy a new one. At 2025 labour rates, that'll have to be the cheap option! :-) Besides, they'll mostly be cheap and lightweight class Ds running direct off the mains by then (heh, that'll get the SET gang twitching!). I forsee extended wars of words between digital and analog camps. No, really? :-) Who on earth will be able to decode the circuits if they have not got the exact workshop manual for the amp model they happen to be lumbered with? None of the special spare parts will be available. IMHO, a well re-engineered Quad II will hold its own with a 300B amp. Hmmmm, I'd back a well-designed 300B P-P in class A against a Quad II any day, let alone an optimum 300B with 4 output valves. Of course, a 're-engineered' one isn't necessarily a Quad II any more, now is it? Well there isn't too much wrong with the Quad II when used with Quad speakers. The ESL57 isn't a bad load for them. But I get folks wanting to use modern dynamics, maybe only 4 ohms, and that reduces their charms immediately, because the circuit luches into class AB above a low threshold, and the sound falls apart, mainly because of their appalling power supply, designed before silicon diodes and decent small electros. Quite so - it's a classic design, and should be used only with classic partners, and an expectation of vintage sound. The simplest improvements to a Quad II are 1, Fit individual RC bypassing to each output tube cathode. 2, At least have a GZ34 as the rectifier. 3, Remove the bulky low value boxed electros, 4, Install a CLC filter, 47uF, 2H, 100uF instead of the large cap for the reduction of hum appearing at the CT of the OPT. 5, Replace all the old Hunts caps, and resistors with modern caps and metal films. 6, Make sure the tubes are in good operating order. 7, Set up the OPT settings for 8 ohms, and never use the 16 ohm setting, 8, Add stabilisation network to prevent oscilations into C loads with the 8 ohm setting, 9, Measure the thd and try swapping positions of the EF86, and the KT66 to get the lowest thd at 3 watts into 8 ohms. So now you have a nicer amp - but it's not a Quad II! The Quad 22 control unit is another story, another time. However, thanks for all the thought you've clearly put into this little enterprise. The group may want to know about the simplicity or complexity issues regarding the amps they might use, or might build. Indeed, and that's a vexed question when it comes to such things as CCS additions not traditionally seen as being 'in the signal path'. Some seem not to count those, others feel that *any* active device goes against KISS. Personally, I must say that this 'signal path' thing is ******** for me, because the real signal path starts at the power supply. After all, why are they called valves if not because they control the energy flow from the PS? -- Stewart Pinkerton | Music is Art - Audio is Engineering |
#66
|
|||
|
|||
http://www.tcaas.btinternet.co.uk/
about JLH also www.diyaudio.com look for solid state forum -- -- .................................................. ........................ Choky Prodanovic Aleksandar YU "don't use force, "don't use force, use a larger hammer" use a larger tube - Choky and IST" - ZM .................................................. ........................... "Patrick Turner" wrote in message ... |
#67
|
|||
|
|||
Gee its got to be a large post, pardon me snipping what needs to be snipped....
John Linsley Hood had quite a few class A designs in Wireless World. There is a wealth of info in those magazines and I suggest that if you were serious, try spending a sunday at the local university library which may still have them tucked away in an archive basement, that is if they have not cleared them out, and convereted them to gyms for $10 entry per hour. I already have copies of all his articles, indeed one of my earliest audio construction projects was his original Class A circuit, built within weeks of the article appearing in WW. However, I can't use his excellent circuit for two reasons: it uses a lot of global NFB and it's arguably not single-ended. That just leaves the minor detail of designing something simpler..... I'm intending to base it on a very linear BJT loaded by a constant current source. Here is where you will have trouble. BJTs have splendid current linearity, ie the current in/current out is fairly constant. The voltages are not so linear...... Yes, I'm aware of that, but it's worth trying, just to see what can be done. I'll keep a bottle of anti-depressant pills handy. Failure never depresses me, the next one can always be made better! Like a weeks of flooding rains is OK, when the sun does make an appearance, whoopee! Yup, the barbie is still in the garden, just in case! :-) There was a lecturer in "human relations" at the local uni who tried to say to his students that sex makes people happy, and the more they have, the happier they'll be. Hands up who had sex 3 times this week?, he asked. About 1/2 the smiling youngsters raised their hands.. Who had sex only once in the last month? About 10 people raised their hands, and they didn't look too happy. Who has sex once a year? One guy up the back lept to his feet, both hands raised, big happy face... Well Sir, you seem to confound my theory, why are you so happy if you have sex so infrequently?? Tonight's the night, he replied..... What's wrong with using an emitter follower? You are using a great whacking ****e load of series voltage negative feedback. Yeah. So? So does a triode. I think you may find that EF bjt connection involves far more applied NFB than the feedback in a 300B. Yeah, so? Where's the harm? Its cheating in this case. If you insist on so much NFB, then to make the playing field level, we should allow the guy with the 300B amp to add 50 dB more voltage gain, and then apply 50 dB of NFB, and that would reduce measure thd levels to the same approximate level. But I have spoken to the other contestant, and he has no intention of agreeing to any more complexity or NFB, and intends on fully presenting his amplifier as is regardless of what you do with yours. But the *amount* of series volage feedback in bjt EF is far greater than with SF in a mosfet. Yeah, so? The amount of voltage NFB with a mosfet SF amp output stage is less than in the NFB in an EF bjt stage. A single fet driving 4 ohms in class A has an open loop gain of about 4 x Gm = about 3.2, so when used as SF, the closed loop gain is about 3.2 / ( 1 + 3.2 ) = 0.76, so a gain reduction of about 13 dB has occured. But the amount of this FB also yields an output stage with Ro = Rd / [ 1 + ( u x B) ] = approx 220 / [ 1 + ( 180 x 1.0 ) ] = 1.2 ohms, and not that much different to a 300B amp when measured at its OPT secondary. This don't mean the 300B has 13 dB of internal FB applied. Or more simply, the Ro of a source follower, cathode follower, or emitter follower is = 1 / Gm, and if Gm = 0.82 A/V, Ro = 1.2 ohms. If Gm of the bjt is 30A/V, the EF Ro = 0.033 ohms. I am still waiting from the boffins out there to place a figure in Db for the FB in a 300B. The distortion from a pair of npn and pnp average power mosfets in class A complementary pair SF is about the same as a pair of triodes in PP class A, about 1%. The EF bjt pair in class A should have a lot less thd, since far more NFB is involved, since there is far greater gain reduction.... And this is a bad thing, exactly how? I am not into goodness or badness here; I ain't a moral guardian. I am just talking the facts about the basic engineering. I thought your original stated intention was to KISS, and to not use any NFB. Now your'e squirming around trying to escape the fact that EF is a ****eload of NFB. Even if you do have an SE follower bjt output, you still have to get the input impedance to be high, and try not to use any more than two devices. I have never seen it done. Doable with mosfets, since their Rin at their gates is high, like a 300B, until the capacitance becomes a problem, but only above 20 kHz. But the EF config reduces this thd by about 125 times, so if you had 10% open loop thd, expect about 0.08% with EF at the 12 watts. The use of EF has a technical benefit. Yes, but this is achieved by a higher level of feedback. I guess it all depends what kind of feedback the valvies will accept as 'good feedback', as opposed to global NFB. I'm basically trying to play in their yard, not design a higjhly linear amp per se, otherwise I wouldn't start from here, so to speak. It's my understanding that an EF output stage is not liked, for some reason I can't quite fathom. Well then to be fair, don't use any more FB than is used in a 300B. Therfore EF is not an allowable solution. ********, the point is the equivalent simplicity of topolgy, combined with an absence of global feedback. I'm not trying to *simulate* the 300B amp, just come up with a notionally similar KISASS amp. That's uh, Keep It Simple And Solid State! :-) Well then we await your schematic, ******** etc permitting.... The output bjt should be set up in common collector, as a gain device, load cap coupled off its collector, not its emitter, and the base operated by another single bjt, and I'll allow you only enough loop or other NFB to give the same Ro as achieved by the 300B I doubt anyone can do this, because of the base input resistances involved. Nelson Pass managed the simplest SS amp of all, one mosfet, one CCS, and shunt NFB in the input-gate-collector circuit. But it needed 4.5 volts of drive, and still rather low input impedance, because that's what you get with shunt feedback input circuits. Indeed so, and there's no point in making a Zen clone anyway, since that game's already been played, and we know the result. Nelson's idea could be extended to use say 6 paralled gain devices, and still keep all the rest of the circuit identical, along with the PS and its currents and voltages. Then the 6 mosfets would each see 24 ohms as their load, or in effect, Gm = say 5A/V, and the Ro would then be aproximately 6 times lower, which means less NFB could be used, certainly less than the parsimonious 12 dB that he uses. Perhaps if you use just one bjt, with its load in the collector, and then have a mosfet driver to power the bjt base, and the fet gives the high imput resistance. You would find that the amount of loop FB needed to get Ro = approx 1 ohm would be rather low.... He would have done better to excomunicate the CCS, so one device out and gone, and a big hot lazy ******* it is too. Then have an OPT, and a single drive j-fet, with some normal NFB, still only 12 dB required, and hed have ahd himself a better simple amp imho. Depends how you look at it - the CCS has to be cheaper than the OPT, so why use an OPT just for KISS brownie points? Expense doesn't come into this hypothetical amp designing competition. We are expected by our jeering public to be purist godammit. An OPT is One Transformer. A CCS has a gain device, biasing network, voltage reference, etc, etc, etc. The OPT replaces One Output Capacitor. Or he could have had a simple choke feed to the output fet, and cap couple to the load, so no hassles with OPTs, but a cap coupled OPT using a toroidal transformer works very well. What we are doing here is true simplicity, tiny amount of gain, very few devices, small quantity of NFB, low noise, lowish thd, wide bw, therefore usable. OK, the average diyer gets horified by having to use hardware instead of multiple devices which are cheap along with lots of NFB, also easy but only if you know what you are doing, because anyone **designing** and building an amp can easily end up with an oscillator instaed of an amp. Indeed so, in fact my old mentor used to say that the best way to design an oscillator was to *think* that you were designing an amplifier! :-) An oscillator is an amplifier that oscillates due to positive voltage FB. Well, I've used Sziklai pairs in fully complementary amps before, and I'm still kicking possible topologies around on paper, trying to minimise the parts count.. Betcha you use a shirtload more than Andre's 300B amp. Oh yes, but then I was trying to build a *good* amp, not just a minimally simple one! :-) Less is not necessarily more................. Goodness and badness are never agreed upon. Not around here, that's for sure! Not anywhere on earth and not on any issue. I forsee extended wars of words between digital and analog camps. No, really? :-) ; Well there isn't too much wrong with the Quad II when used with Quad speakers. The ESL57 isn't a bad load for them. But I get folks wanting to use modern dynamics, maybe only 4 ohms, and that reduces their charms immediately, because the circuit luches into class AB above a low threshold, and the sound falls apart, mainly because of their appalling power supply, designed before silicon diodes and decent small electros. Quite so - it's a classic design, and should be used only with classic partners, and an expectation of vintage sound. The simplest improvements to a Quad II are 1, Fit individual RC bypassing to each output tube cathode. 2, At least have a GZ34 as the rectifier. 3, Remove the bulky low value boxed electros, 4, Install a CLC filter, 47uF, 2H, 100uF instead of the large cap for the reduction of hum appearing at the CT of the OPT. 5, Replace all the old Hunts caps, and resistors with modern caps and metal films. 6, Make sure the tubes are in good operating order. 7, Set up the OPT settings for 8 ohms, and never use the 16 ohm setting, 8, Add stabilisation network to prevent oscilations into C loads with the 8 ohm setting, 9, Measure the thd and try swapping positions of the EF86, and the KT66 to get the lowest thd at 3 watts into 8 ohms. So now you have a nicer amp - but it's not a Quad II! Its a Quadner, or Turquad, take your pick. The glory of the original simple design is allowed to shine without the unecessary intermodulation distortion from the power supply which equals the thd of the tubes et all even when its in class A. The Quad 22 control unit is another story, another time. However, thanks for all the thought you've clearly put into this little enterprise. The group may want to know about the simplicity or complexity issues regarding the amps they might use, or might build. Indeed, and that's a vexed question when it comes to such things as CCS additions not traditionally seen as being 'in the signal path'. Some seem not to count those, others feel that *any* active device goes against KISS. Personally, I must say that this 'signal path' thing is ******** for me, because the real signal path starts at the power supply. After all, why are they called valves if not because they control the energy flow from the PS? I like hum free power supplies. CCS are the antithesis of what could be detrimental to a signal by Straus waltsing its way along the signal path. If we replaced the input triode driver proposed by the 300B camp with a bjt as an SE driver device, we may well find the distortion and thd spectra rather dismally high, and complex, and its not what I would ever recommend. It'd be as bad at least as using a pentode as an input driver device for the 300B. However, should we have the cheek and temerity to use a transistor CCS instead og the triode driver's resistance load to convey a DC flow to the tube, we would have the opposite effect on the thd of this crucial stage for the 300B amp. THD would be found to have reduced from whatever it is with an R load, by perhaps 10 dB, because the load seen by the triode becomes a larger ohmic load. Triodes have their most naturally linear performance when there is **no change** to the anode current, anf there is just a pure electrostatic voltage change. This is hard for worshippers of current activated devices to comprehend, but tube's ain't called voltage devices for nothing. The triode linearity with no current change is due to the maximal effect of the anode's **voltage** change on the flow of electrons from the cathode. Pentodes and beam tetrodes have a helical screen grid between the control grid and anode, and the voltage feedback effect on the current from the cathode is screened off, since the screen is held at a fixed potential. Then the linearity is little better than a bjt, mosfet, etc, and in fact the worst voltage thd occurs when the load on a pentode is a CCS, and gain is maybe several thousand times, and many more times than a triode.. bjt gain with a CCS load is also very high and thd pretty dreadful. In these cases where CCS are cited as being mischeivous sonic culprits, there is usually a lack of understanding of the nature of the CCS. A typical CCS using a lone MJE350 can have the finite CCS value of 30Mohms. This is right acoss the audio range, and at any voltage. Compound CCS using two devices have an unmeasurable and high equivalent resistance. This allows a triode to do more good than harm, ie, perform without having to produce audio power in its DC supply resistor, and hence produce the associated thd. In all the circuits where I have used CCS loadings for triode gain stages, people have told me they like the sound. Then there is the mu-follower, or "bootstrapped follower" gain stage which can be done using all bjts, but for the fact that there would be far too much gain, and there has to be some loop FB to at least control gain. I take the signal path as seriously as I think I should; its not a bollocky issue. NFB also reduces PS and other noise in amplifiers, in general. Since large amounts of NFB are never used by any tube amps with comparison of the huge total amounts used in most bjt amps, then the amount of noise reduction and PS caused IMD reductions due to FB in tube amps is quite limited, and so we need to give the tube amp some quiet PS to work from, and thus depend less on NFB for the sonic outcome. If you design a blameless KIS SE bjt amp, the same economy of NFB with compensating emphasis on passively reducing PS noise will apply, and you will perhaps need CLC filtered rail voltages. Patrick Turner. |
#68
|
|||
|
|||
Choky wrote: http://www.tcaas.btinternet.co.uk/ about JLH also www.diyaudio.com look for solid state forum There is for example a transformer coupled SS amp which uses a pair of npn mosfets in source follower at http://www.diyaudio.com/forums/showt...threadid=42259 The "amp" is a source follower, so the input voltage has to be larger than the output voltage. The title of the post is "zero FB impedance amplifiers" This unfortunately is misleading to me, since there definately *is* NFB usage, because of the source follower connection, which is an application here of about 20 dB of series voltage NFB, and I guess that the words "impedance amplifier" means that there is a change between the input and output impedance. Amplifying impedance usually means increasing impedance, not drastically reducing it, as is the case with the amp in the schematic. It is really what is known as a source follower buffer amp. If I drove down into that forum, I might find myself bogged to the axels explaining the basics. Patrick Turner. -- -- .................................................. ....................... Choky Prodanovic Aleksandar YU "don't use force, "don't use force, use a larger hammer" use a larger tube - Choky and IST" - ZM .................................................. .......................... "Patrick Turner" wrote in message ... |
Reply |
|
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
Lionel's Demonstration of His Insanity = His Delusional Attack Threads | Audio Opinions | |||
KISS 114 by Andre Jute | Vacuum Tubes | |||
KISS 113 by Andre Jute | Vacuum Tubes | |||
KISS 111 by Andre Jute | Vacuum Tubes | |||
KISS 102 by Andre Jute | Vacuum Tubes |