And rightly so.
By now, you should have realized that the forum you are talking about has a couple of people who:
1. Have learnt how to connectect two semiconductors without explosions, which makes then instant geniuses,
2. The name of the game is FET - if it ain't a FET, no way it can ever work properly,
3. Nelson Pass is God and nobody else matters, or even exists,
4, The owner and head honcho is a serviceman, who believes that a voltage regulator built around a MOSFET really can't "sound" good unless it is preceeded by a Darlington BD TO-220 device, etc.
Therefore, the initial mistake is yours, you assumed people would even bother to listen.
I understand that one of the few people who really knew their stuff there, Sipi, has also left the forum for more or less the same reasons you encountered. I was never really there, just dropped in a few times, but made sure I discussed nothing seriously, being a BJT guy I knew I'd end up being burned at the stake for blaspheming.
References?
I always took the approach that references should have stuff that I didn't have space to include in my main text, especially if it was stuff that was better put than any possible effort of mine.
ie references should be USEFUL .. far more important than pedantry.[*]
As an example, the evils of voltage driven Class B output stages (eg plain Miller etc) are comprehensively covered by Self in all his editions and I doubt if anyone can add more than a superficial gloss to the subject.
Of course, whether voltage drive of Class B output stages is a good idea is another matter entirely.
There are times when you quote something in the full knowledge that most people are unlikely to follow them up and that even fewer will be able to understand them.
My references to Bode fall in this category .. but not to do so is like not genuflecting in church.
What is particularly irritating are da pseudo gurus who quote obscure papers from Journal of Unobtainium Trans. just to show de ken reed en rite .. providing no useful info to us unwashed masses and only serving to obfuscate the subject with their semantic pedantic sh*t.
[*] Of course pedantry is important if you are proposing a new or novel technique; in which case, full attribution is appropriate .. whether useful or not.
In da old days wen I kud reed en rite en pri10dat to rite 4 da learned societies ..
I always took the approach that references should have stuff that I didn't have space to include in my main text, especially if it was stuff that was better put than any possible effort of mine.
ie references should be USEFUL .. far more important than pedantry.[*]
As an example, the evils of voltage driven Class B output stages (eg plain Miller etc) are comprehensively covered by Self in all his editions and I doubt if anyone can add more than a superficial gloss to the subject.
Of course, whether voltage drive of Class B output stages is a good idea is another matter entirely.
There are times when you quote something in the full knowledge that most people are unlikely to follow them up and that even fewer will be able to understand them.
My references to Bode fall in this category .. but not to do so is like not genuflecting in church.
What is particularly irritating are da pseudo gurus who quote obscure papers from Journal of Unobtainium Trans. just to show de ken reed en rite .. providing no useful info to us unwashed masses and only serving to obfuscate the subject with their semantic pedantic sh*t.
[*] Of course pedantry is important if you are proposing a new or novel technique; in which case, full attribution is appropriate .. whether useful or not.
Hi Waly,
I definitely hope to add more references in the next edition. Bear in mind that it is not just the number of references included, but just as importantly those that are missing that should be there. This is where I believe DS falls short.
I am always looking for input on references, just like any other input for my second edition. I especially will appreciate it if you or others here will bring to my attention any specific omissions in citations.
Cheers,
Bob
I plan to run the LTP pretty heavy, perhaps 5mA per transistor, and keep Re low, between 47 and 22 probably. Then use MIC to keep the slew rate up. Around 1nV/rt Hz looks achievable with complementary inputs and Toshiba low Rbb transistors. That will do.
Yes. The one sided view of some people provokes an opposite sided reaction.
People should find "what to not like" in both
I did think and do a bit of calculation before I posted. The above should have excellent noise and still fine THD. Current mirror noise can be made below concern. Source needs to be low, of course.
In all, the noise figures obtained in audio power amp designs using the 2 varients are very much the same.
BTW to lower the noise of CFA design is a simple matter of upping the current of the input configuration which is usually a diamond buffer. Another technique is to use two diamond buffers in tandem to lower noise, its more complex but then again it has other benefits as well.
Yes, more current and duplication works, and has some other benefits, in the LTP too, so that's why I plan to use it. But there are limits on how much the current can be usefully increased. What sort of results can the "CFB" achieve?
Best wishes
David
Supa LN power amp
Mr. Zan, in keeping with my naive efforts that tend to copy .. I mean improve other peoples stuff, may I recommend you look at making a 100W mike preamp.
The one to use is Graeme Cohen's as it is the basis of the only practical mike preamps with <1nV/rtHz Ein.
The Millenia Media (1nV/rtHz) uses this circuit and almost certainly, the Earthworks too. The Earthworks is the quietest mike preamp you can buy but it doesn't quite meet Cohen's own performance. There are many DIY Cohen copies that don't come anywhere near the potential performance.
You'll note it is a 3 stage design but the final amp could well be a 100+W OPA. This ties in with my philosophy of favouring 'tried & tested' above 'new fangled' as long as T&T will do da job.
You'll have your work cut out making your power amp a good differential amp as is required for the Cohen output stage.
I'd recommend you arrange for quite high gain for the final amp stage in your system (difficult I know with your horns) as its easy to do 60dB LN amps and 0dB LN amps but very difficult to do 1dB or 20dB LN amps.
_______________
If you decide to stick with a 'single PA stage' you might like to look at making it as simple as possible, eg no enhanced VAS or triples, relying on the low Closed Loop gain to achieve low THD. Then you might get away with no evil degeneration on the IPS which is death for LN.
Oh. A single device IPS (no LTP) helps a lot too.
I refuse to call it a CFA but these tend to parallel the input devices which helps too.
How much power are you after and into what Z?
Mr. Zan, in keeping with my naive efforts that tend to copy .. I mean improve other peoples stuff, may I recommend you look at making a 100W mike preamp.
The one to use is Graeme Cohen's as it is the basis of the only practical mike preamps with <1nV/rtHz Ein.
The Millenia Media (1nV/rtHz) uses this circuit and almost certainly, the Earthworks too. The Earthworks is the quietest mike preamp you can buy but it doesn't quite meet Cohen's own performance. There are many DIY Cohen copies that don't come anywhere near the potential performance.
You'll note it is a 3 stage design but the final amp could well be a 100+W OPA. This ties in with my philosophy of favouring 'tried & tested' above 'new fangled' as long as T&T will do da job.
You'll have your work cut out making your power amp a good differential amp as is required for the Cohen output stage.
I'd recommend you arrange for quite high gain for the final amp stage in your system (difficult I know with your horns) as its easy to do 60dB LN amps and 0dB LN amps but very difficult to do 1dB or 20dB LN amps.
_______________
If you decide to stick with a 'single PA stage' you might like to look at making it as simple as possible, eg no enhanced VAS or triples, relying on the low Closed Loop gain to achieve low THD. Then you might get away with no evil degeneration on the IPS which is death for LN.
Oh. A single device IPS (no LTP) helps a lot too.
I refuse to call it a CFA but these tend to parallel the input devices which helps too.
How much power are you after and into what Z?
Recent Pioneer amps have Sziklai outputs but with BJT driver and Hex fet output transisitor. They sound good to me. I have not seen Sziklai combination with FET driver and BJT output.
Thanks for the comments.
The idea is actually to develop a modular circuit concept for a tri-amped active system.
600W into 4 ohms woofer.
150W into 8 ohms compression drivers.
40W into 8-16 ohms horn tweeters.
and
300W into 8 ohms subwoofer that is essentially half a woofer amp.
A bit overkill, but I have all these JBL drivers that aren't "home" cinema, they're just cinema
.I aim to use tried and true too rather than an excessive emphasis on one aspect.
Noise is important but I will stick with an LTP for more "balanced" performance, in both senses.
I plan to keep the Re low but non zero. After a while the feedback resistor dominates anyway and it becomes a matter of how much power one is prepared to burn in the the feedback network to keep its impedance low.
A complementary input layout is effectively paralleled devices, and it mostly cancels input bias currents as a free benefit.
Around 1 nV/rt Hz looks achievable without serious compromise to other parameters and would be lower than practically any commercial product I know.
And I value simplicity too
Partly because I have to build a whole set of them.OnSemi 4xxx series ThermalTrak outputs.
Distributed Driver 2SC3503F + complements
Toshiba Low noise Inputs and dual monolithic low noise current mirrors.
Elevated, clean, front end rails up to the drivers.
Bode optimised output inclusive compensation with RHP zero cancellation.
Some form of MIC for the inner loop, probably.
No cascodes on the LTP, probably.
A VAS with no EF added but with optimised cascode looks attractive at the moment but I used to think EF with no cascode was better, so I still don't have this clear.
Between the VAS and the drivers, and the bias for the ThermalTraks, is still not clear.
Best wishes
David
Nice reference. Maybe I have to re-think, once more.
Last edited:
Dave Zan: " . . . What sort of results can the "CFB" achieve?"
Very good.
If its low distortion you are after, it seems low single digit ppm at 20k and high power levels is imminently doable. At the usual 1W reference out, ppb levels at 20k
High SR's come with the topology - so equaling or bettering MIC type performance in VFA's
PSRR is cited as an issue, but a simple cap multiplier and filtering of the front end stage supply (which is normally regulated in a CFA analogous to the LTP current source) and very respectable > 100+ dB figures are attainable. And its symmetrical
Noise - CFA wont match VFA designs in general, but as I mentioned on a previous post, ultra low noise is not a requirement for a power amp
Very good.
If its low distortion you are after, it seems low single digit ppm at 20k and high power levels is imminently doable. At the usual 1W reference out, ppb levels at 20k
High SR's come with the topology - so equaling or bettering MIC type performance in VFA's
PSRR is cited as an issue, but a simple cap multiplier and filtering of the front end stage supply (which is normally regulated in a CFA analogous to the LTP current source) and very respectable > 100+ dB figures are attainable. And its symmetrical
Noise - CFA wont match VFA designs in general, but as I mentioned on a previous post, ultra low noise is not a requirement for a power amp
I find CFA have a too eratic behaviour. It's all talk about how a transistor is removed from the feedback loop, but in reality, it can't work without another symmetrical half to generate the push-pull feedback. So we have two feedback paths in CFA that have to be well balanced with their associated dead-zone. I dislike multiple signalpaths (be it forward or feedback) because of (compensation) interaction that comes with it. CFA is fast, but it isn't 'firm', which is often seen by the ringing remnants on square waves. All my tests on CFAs show inferior THD compared to VFA. The only thing CFA has going for it is the slewrate. These days it's not hard to get multi hundred volts SR in VFA either.
1. Theres only 1 feedback path in a CFA
2. Theres nothing erratic about the topology - 100's of millions of them are manufactured in IC form every year for everything from audio, to video to telecoms applications etc. I am quite sure if they were erratic, they would have been abandoned long ago!
3. There is no ringing on a properly compensated CFA - just as there should be no ringing on a properly compensated VFA
4. THD is definitely not inferior - in fact some one on the forum showed that CFA linarity was easy the equal or better of VFA's (Edmond, I think it was you)
5. I am very pleased with my sx and nx-Amps - please take a read because they suffer from none of the problems you list above.
1) There's a resistor going from VAS out to the emitter of an input of one half, and a resistor going from VAS out to the emitter of the input of the other half. It's a complementary input that generates two feedback nodes (the two emitters).
2) Erraticness is what I found in personal experiences. You can remove erraticness with compensation, but then you're back to square (pun) one.
3) See 2).
5) I simmed a number of CFAs from the forums, none impressed me THD wise
6) I'm sure your amps will be great; I'm not the one to Judge whether they're pleasing or not
I'm just not sure if they would be a convincing case to go with CFA over VFA.The name of the forum is diyaudio. Much too often people tend to place subjective impression too low on the list of priorities and give completely irrelevant arguments. Which lab instrument has ears? Is the human perception the least important aspect of diyaudio?
CFB amp does not have to be symmetrical, and symmetrical does not have to use two feedback paths.
Maybe take some of your own advice? I simply stated an opinion about CFA and didn't dulge into stating my opinion in a subjective "listening" manner deliberately. This shouldn't be mistaken that listening isn't used to judge sound quality. The anti techies harp too much on this aspect and as such try to keep their voodoo magic in place. Sooo.. yeah, your comment is totally irrelevant to my opinion on CFA.
Pretty much every CFA in this forum is of the symmetrical, dual feedback type. Why would that be you think?
Magic Box wrote: "Pretty much every CFA in this forum is of the symmetrical, dual feedback type. Why would that be you think? "
Those are variations on the CFA theme - examples being the Mosquito + derivatives and the Musical Fidelity A1. Dual feedback paths are primarily used to set the DC operating point, and in the case of the MF A1, used to set the output standing current.
Those are classic designs but they are derivatives of the standard CFA topology (I think they do a good job of getting good sound with very few parts by the way).
The classic CFA is per the attached diagram (although IC types typically load the CB stage with mirrors).
There is only one feedback resistor.
Those are variations on the CFA theme - examples being the Mosquito + derivatives and the Musical Fidelity A1. Dual feedback paths are primarily used to set the DC operating point, and in the case of the MF A1, used to set the output standing current.
Those are classic designs but they are derivatives of the standard CFA topology (I think they do a good job of getting good sound with very few parts by the way).
The classic CFA is per the attached diagram (although IC types typically load the CB stage with mirrors).
There is only one feedback resistor.
Attachments
Last edited:
That I couldn't count and skipped his 4) bulletmark doesn't mean I completely ignored his objective arguments. I addressed all of them with my experience and opinions. Let's leave the sudden grasp for listening/listening tests in the listening room. Here we're talking about technical aspects of a topology.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.