Uh-oh. I assume you are working in a lumber sales office, otherwise I can't explain the amount of nonsense you put together in a single post. You have only one possible excuse, if you confuse Miller compensation with lead compensation (which is indeed a big nono in CFAs).
So:
- You claim that CFA's can't be Miller compensated properly, neither TPC or TMC. I think Edmond may have something to say about, as he just posted an example of a TMC compensated (plus lead-lag) CFA.
- You claim that shunt compensation outperforms Miller, TPC (and TMC) compensation. That's something that Mr. Bode would not take easy.
The amount of CFAs (mostly Miller compensated) I am using in my projects on a daily basis exceeds your wildest imagination.
Only being posted on Bob Cordell site doesn't make a model correct. Those models for lateral mosfets are as rudimentary as possible. Spice has a problem with the default jfet/mosfet models, and more advanced models (EKV, etc...) are difficult to implement. If I remember correctly, andy_c implemented EKV models for the 2SK1530/2SJ201 vertical mosfets, and that's the only one I've seen in the public domain.
Ok, now we are down from 10MHz to 5MHz, that's half the GBW, and this is of course in a completely different league. I still can't find anything about the stability margins, though.
As others already mentioned, that amplifier has one quality: it's simple, otherwise it does not qualify as hifi by any known criteria. If you can't hear the TIM generated by the Ciss modulation but you can hear the 10MHZ (er... 5MHz) bandwidth effect, then you have a very selective hearing.
It wasn't Edmond that just posted that link
Currently looking at a fully populated CFA prototype that uses his scheme but with different values and split into a fully symmetric form. It sims fine and tomorrow shall be powered up. So we shall see...
Found that in sims this scheme has THD performance an order of magnitude better than using plain shunt compensation. Although I have sacrificed some PM but that happened in the pursuit of a good square wave.
Last edited:
Have you got a link to Andy_c's EKV models please? I've searched without success.
Bob promised us unwashed masses his EKV models but I think his day job is preventing him from 'hunt the needle in his hay drives'. Hint hint grovel grovel Mr. Cordell
Have you got a link to Andy_c's EKV models please? I've searched without success.
Bob promised us unwashed masses his EKV models but I think his day job is preventing him from 'hunt the needle in his hay drives'. Hint hint grovel grovel Mr. Cordell
Took me 30 seconds using the forum's search tool: 2SK1530, 2SJ201
That thread has a wealth of information and EKV models for other vertical mosfet devices (IRF, Fairchild).
Edmonds efforts are comendable and open possibilities indeed, the so called brutal shunt adds that its use is ideal with the topology although it is prudent to analyse it in detail first. I wonder where he finds the time for all this. Since youve simmed it youre the better judge when perhaps a lesser front end and outputstage is used.
Edmond's designs are very neat. I've spent many hours studying them. They're very logical and educational.
His DMTC is something else, you should try compensating one of your designs with it. Even a straight copy worked well for my design, 96 degrees PM and a quarter of the THD.
Wouldn't say necessarily lesser front end but definitely a lesser output stage
Sims are one thing though, real life is another.
Oh, please.
Again, and again, the subject was not this amp, but the comparison between one version of it in CFA and one in VFA, everithing around equal. Can be any amp. You always hit the man instead of the ball. Talking about the way Descarte was dressed when the subject is method....
And about Hifi, and your comment about an amp you have never measured or heard,, and other pure hypothetical assertions from you, please note that this amp was published in Elector december 1982 and had so much success near audiophile that elector published a little mod one year after. All PBs sold out.
Refer to google for real measurements that i do not publish to not make-you more ridiculous...
Stability margin ? This amp had sometimes problems, depending on the quality of the source resistors. Mine is perfectly stable with no overshoot on square waves, don't you read ?
Oh, no you don't.
Last regards, have a good life.
Edit: i cannot resist.
from Elector 2X140W in 8 Ohm, 2X180W in 4 Ohm, both channels at full power in the same time, <0.01% thd in all the audio band in both situations. (my measurements were better)
and the comment about my cfa version in this forum from a *real builder* of it:
"Thanks Christophe for a great update of the Crescendo. Results are very impressive and an exciting sound is produced.".
I suppose this guy is deaf, not able to hear TIM neither.
Last edited:
Manso, Can you post LTspice *.ASC of one of your simple CFA amps with good performance? Please
Dave, I was going to just post a sketch but was inspired by Esperado, Bonsai, Manso etc to investigate further.I asked for even a few numbers and have had no response yet, but I live in hope.
And you haven't replied to my enquiry about the "low noise CFB" that you mentioned.
It appears JLH actually had several MC preamps & other stuff based on what I had in mind http://www.angelfire.com/sd/paulkemble/sound3.html
I don't think his MC amp is that supa and unfortunately, his topology has DC offset & other issues for a general purpose power amplifier. I got sorta carried away before admitting it needed a lot more work than I have time for at the present.
The inevitable caps in a CFA also means I can't use my beloved THD analyser without a LOT of extra work. Sorry about all dis SPICE world sh*t but this beach bum really can't afford to play 'real life' this millenium.
My interest is in the LN potential and also Bonsai said he had very good DC offsets. Alas, this appears to need evil diamond inputs which negates the stuff which appeals to me .. LN & simplicity.
______________
For your application, may I recommend you have at least a high quality 10u film cap in series with your horns. I've done some very complex commercial systems with electronic xovers but IMHO, you'll thank me for suggesting you combine some simple passive xovers with the fancy electronic stuff.
______________
I'll repeat that yus CFA fans (I'm on the fence) should by trying to persuade Bob to
- burn solder on CFAs
- pontificate on something he doesn't believe in.
For pontificating, I'd rather have it from people who believe in it than someone who doesn't. So hands up those of you who are going to collaborate on 'Supa CFA amps for Dummies'.
Please make sure your sims, 'real life' measurements, Double Blind Listening Tests bla bla are at least up to Cordell/Self standards .. and don't forget comparisons with equivalent VFAs too.
Attachments
And here we were thinking you clean toilets for a living. Just maybe we both guessed wrong.
You missed the part where Edmond s design does in fact use shunt together with another scheme but it doesnt surprise me that you did.
You also resort now to leaving fine details in statements I see, its childish and irritating, my little daughter does this so Im quite familiar with it.
I made no claim that shunt compensation outperforms MIller TPC and TMC.
I did claim so when it is used with the diamond input circuit in a CFA amp.
Does this need to be made any clearer.
Dont fear I dont confuse miller for lead compensation. However youre again missing some basic principles about CFAs operation that would allow the use of lead compensation together with shunt compensation. Did you pay any attention to the Cyrus 3 schematic ?? I was 70 % sure you havent designed much in the way of discrete audio CFAs, Im even more convinced now.
Dont bother answering I will focus my attention from now on those that actually can bring something new to the table.
Credit usually goes to who either publishes first or patents first. Otherwise, we have no way of know what is in anyone's mind or who thought of something first. The diamond buffer circuit is pretty old and applying it to the front end or output is a later refinement as are many other such refinemnts and patents since.
At the time, I wanted a direct coupled design as well. Because I knew the caps of the era were pretty bad and introduced harmonics - though that was much less understood than it is now. The use of no caps was a factor as well as very linear at all freqs. I introduced my concept of dc servo to W.jung for his designs. [Only J.Curl knew this until recently it was thought WJ was the insperation for the dc servo concept.] That meant a complimentary push pull and not too noisy. So only low Z values was decided... no values more than 1K were used. And, then I used supplies that were dual 24vdc. Local distributed voltage regulation At the circuit. etc etc. All the things that have since been picked up and become standard high performance circuitry was all described in those breaking articles.
Note too the use of low z parts in the topology meant the circuit tray C's and device C's would produce time constants that are very low... also leading to wide bandwidths. It was all folded into one design. At that time of publication, john curl called the publisher and got my telephone number and that started a long relationship... he in advanced audio design and me in government R&D at Lawrence Livermore National Laboratory (25 years and retired from there). I got permission from the labs legal department to publish as i couldnt give them an application for whih it could be patented that the goverment would be interested (audio was of no interest to them). And, if i wanted to pusue a patent it was OK. i didnt want to so i published.... beating the IC makers and thier patents/changes/mods/ to make it patentable over my publication.
There are some fundemantal differences that are not appreciated --- the input of a diff stage is compressive while the CFA is expansive. The use of low z means you can also have lower voltage swings which leads to lower device distortion. [I know im glossing over this for later talk].
Anyway, it has found a unique place in amp designer's bag of tricks and is very refined. Today, i have turned my attention to distortion cancellation techniques rather than purely neg feedback approach for the freedom it gives in optimization of parameters deemed important and lower parts count in many cases while being very stable. That's where the cutting edge is IMO.
Thx-RNMarsh
Last edited:
I don't believe in sims, as models are not able to follow all the evils of real devices (temperature distortions, variation of capacitance with levels etc.)
I don't believe in sims, as the evils of components, you know, resistances and capacitance distortion etc.
I don't believe in sims because the evils of printed boards, you know, parasitic capacitances and inductances of tracks etc.
But i believe that, if some mod give better results in simulation, it has a chance it will give better results in real life too. In other term, only relative values are really interesting. It is a tool to replace a calculator, nothing else.
Having a judgment on any audio device via simulators is like having a judgment on poetry via a Word orthographic corrector report or the beauty of a photography via a luminance curve from Photoshop.
To Richard, thanks for all your clever words. Your understanding of audio is not confused and fulfill this: "what is well conceived is stated clearly.
I don't believe in sims, as the evils of components, you know, resistances and capacitance distortion etc.
I don't believe in sims because the evils of printed boards, you know, parasitic capacitances and inductances of tracks etc.
But i believe that, if some mod give better results in simulation, it has a chance it will give better results in real life too. In other term, only relative values are really interesting. It is a tool to replace a calculator, nothing else.
Having a judgment on any audio device via simulators is like having a judgment on poetry via a Word orthographic corrector report or the beauty of a photography via a luminance curve from Photoshop.
To Richard, thanks for all your clever words. Your understanding of audio is not confused and fulfill this: "what is well conceived is stated clearly.
Last edited:
LN and simplicity appeal to me too, of course, so I was hopeful, but the EF inputs don't help the noise.
Yes, I had already decided to do this. In fact it was one of my earliest decisions. If one has a capacitor for fail safe then may as well make it useful as part of the x-over too.
I read all these claims too and what I really want to ask is - "How did you make it unbiased?"
Not to even mention "What experimental procedure, What sample size? What level of statistical confidence"
What a waste of time.
So thanks for the schematic, an actually useful contribution.
This is your distillation of JLH?
Best wishes
David
Last edited:
Ricardo dont fear, I will get around to posting a stripped down design that show good figures. Still learning all the ins and outs of LTspice as Im not a regular user. I believe Bonsai has a design in the pipeline as well.
The one youre toying with is ancient, but capable of high performance when tweaked.
Maybe this should be done in another thread though, its way off topic here unless the aim is to try persuade Self to write something on it.
What publication was this? A reference is helpful but a Net readable version would be even better.
Best wishes
David
Proper Double Blind Listening Tests bla bla are a huge topic and not easy to carry out properly. It also costs a lot of money. I was a real Double Blind Listening Test bla bla guru in my previous life.
I might write some of this down before the end of the Millenium but don't hold your breath waiting
Some useful hints (which aren't covered by the Lipsh*tz & Vanderkooy papers).
- Do ABC tests rather than ABX tests. You get statistical significance MUCH faster. For something like CFA vs VFA, 2 presentations would be of one of these BUT YOU NEVER TELL THE VICTIM THIS OR EVEN WHAT HE IS SUPPOSED TO BE LISTENING TO. He must be aware that all three presentations might be different (or even all three the same. You might do this evil thing to weed out the pseudo Golden Pinnae.)
- Test each victim individually. NOBODY else in the same room
- Let the victim choose the music and levels. But he needs to listen to AT LEAST 5 pieces of music and take at least 30min. He can take up to 1/2 day.
- The 5 pieces must include simple / complex / instrumental / vocal / mixed but they can be eg all heavy metal.
- Let the victim make his OWN comments. Often this is the most useful info you get out
- Be prepared to find most (all?) self-styled Golden Pinnae are deaf including HiFi reviewers
- Be prepared to find the true golden pinnae are your girlfriend/wife who disclaim any interest in HiFi etc
- Be prepared to dump several days of work and tests cos you discover a flaw that allows bias in your procedure
- But do the tests!
Hardly a distillation of JLH or even a working model .. more rediscovering how flaky his designs were
Thanks for this Waly. Alas it also has lotsa liquid BS but I'm ploughing through it now to find the pearls.
I trained in statistics, both the mathematical foundations and practical applications, before I moved into other studies.
What was a real education was to see the cluelessness of many people who had to do a practical statistics course as a prerequisite for another subject.
I also plan to do an article on the subject but would be happy if you saved me the effort.
It's more because I think it should be explained rather than anyone seems to want to learn and have to think.
I could be incorrect, perhaps people will say "yes, I checked my impressions with a blind test to remove placebo effect and I found..."
Best wishes
David
Why do you use the dual feedback network? You are creating unecessary complexity. I can assure from a practical point of view, that the single network works exceedingly well. The sx and nx-Amps are 100% DC coupled.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.