You are very affirmative, man.
For PSSR, as I said, it is on the way. Will try this, but I'm unsure; C1 is here for the current source and input stage.
For compensation, you are TOTALY wrong. Yes it is a little tricky, but reduce distortion by a 10 or more factor. Just try-it. I do-it since 1980. And there is two reasons for this. First, it add feedback at very HF to reduce the OPS crossover distortion, second, it do not add a load to the VAS.
About stability, Don't you think margins are good enough ?
Gate resistor ? I removed them on purpose. The feedback loop is pretty complex, considering poles and phase turns. And this added both margins and
diminished distortions.
Last edited:
Not my experience neither. PSSR of CFA's are worse than VFA's at LF. While a little better at HF. And i have not found a better way to improve-them till now.
Well, let's talk about this. As Mark Johnson noticed, we are at the limit of the Power safety of the output devices. Reason of the actual Rails voltages to limit the power. I don't really care to add some watts on the paper for a 1dB benefit, see what I mean ?
But if you know a better way to improve the PSSR, that is the weak point of CFAs, you are welcomed.
If you really need more power it is possible to add some voltage on the input+VAS rails using the 12V transfo that is needed for the Servo and future protection. We need less than 100mA. We could use a double output coil one, one for the +-12V of the servo, an other to add voltage to those rails. But we will need more output HEX(or Moss) FETs. And, to go to 75V with 2sk135 and 2sj50.
To be tried. But, really, folks, don't you think distortion numbers are not good enough ? If something has to be improved, it is the Slew rate.
A friend of mine likes to install every possible stunt, contrivance, trick, and embellishment into the revision-A circuit board. Then, in the lab, he tests the features individually. Measure the amp with gimmick-1 installed, and measure again with gimmick-1 removed. Did gimmick-1 provide any benefit at all? Write it down in the lab notebook. Reinstall gimmick-1 and now measure the amp with, and without, gimmick-2 installed. Did gimmick-2 provide any benefit at all? Write it down. Proceed until all gimmicks have been individually evaluated.
Now build-up a second copy of the PCB with only the beneficial gimmicks installed and all non-beneficial gimmicks removed. Measure the amp. Does it behave as expected from the first round of experiments? If so, put it into revision-B of the circuit board, which is now the final amplifier.
Now the answer to the question "what about feature X?" becomes "I tried X and found, in the lab, that it did not help. So I omitted it."
Now build-up a second copy of the PCB with only the beneficial gimmicks installed and all non-beneficial gimmicks removed. Measure the amp. Does it behave as expected from the first round of experiments? If so, put it into revision-B of the circuit board, which is now the final amplifier.
Now the answer to the question "what about feature X?" becomes "I tried X and found, in the lab, that it did not help. So I omitted it."
Thanks, ASKA. You are a gentleman, and this is a refreshing change from the usual atmosphère of this place. What they believe ? That those these performances with such a simple design have fallen from the sky ?
As you noticed, the input stage add some gain. outside from the loop. It is very tricky to get the minimal distortion by an optimal balance between this gain and the VAS ones. Something to be tried.
Once I'll found some time to finish to clean my ASC's, i will publish hem: at this moment, you will try on your side and publish your results. As I said, this is supposed to be a collaborative work in progress.
My purpose is to offer the ASC, and people propose REAL improvements with their own asc.
The idea is a simple circuit, with minimal components parts. Apart SR that is not outstanding, I wonder why someone would like to reduce distortion numbers (on sims) that are yet VERY good, don't you think ?
The most interesting CFA I have found here is the Dadod 200W one. I noticed he got, like-me, very low distortions (better at HF) and has too a limited SR of 200V/µs. That is the first point to be improved.
Mark, let's dissipate some misunderstandings between us.
It is the only way to do things, on my opinion.
And, most of all to listen at each change.
I do-it since 1970, when I first entered in a R&D department of an Hifi company. It was my first job, pationating. I was a self made man at this time, just pationated by electronic since the age of 13., building kits, trying to understand... And, because I was not academic, always trying things from an new point of view, I was the one that had deposed the most patents for them at this time. See what I mean ?
I'm, not comfortable at all with Sims. It is relatively new for me. So I don't believe so much in it. I consider-it as a good calculation tool.
I expect to learn from all of you all.
Neither had read any of those books that seems like a Bible, here: Bob Cordell or Douglas Self ?
Not my native Language, lack of time etc.
What I know, I have learned it or discovered on the bench. By myself, long before someone even put a name on things. It explains why I'm not academic at all. But what I know is for real and I had the luck to design some well regarded gears during my life. Now, with the time offered by my retirement, I get back to my original passion. And have a lot to learn of what had changed while I was in others activities (Not so far ;-).
Hoping that we, all together, will be able to really improve this project, and end with something both consistent and real ... not showing our g... ;-)
Last edited:
Both will be one trim pot, to adjust the output offset before to connect the DC servo.
In this regard, it would not be a good idea to inject this current by taking the voltage of the bridge from +- 12 V of the DC servo rather than rails of the amp ?
Reminds me that at this time I designed 5kW LatFETamps for the German Navy that worked successfully over the years.
I don't reacted to this. Are-you kidding ? Between both emitters (or the 12V protection zeners) you'll see 0V at clipping and less than 1V when no signal at 100 mA of quiescent current: It behaves the opposite as you pretend.
If ever you go to 6V, the quiescent will be so high that your output devices will be dead long before.
Man about this compensation scheme, It seems I'm not the only one to use-it now., it is no more esoteric. Make a quick search for the various "inclusive output compensation".Reminds me that at this time I designed 5kW LatFETamps for the German Navy that worked successfully over the years.
Anyway, you have the SIM file available for now: easy to see the difference and discover something.
What was exactly the purpose of your input ? Stability margins are not good enough on your point of view ? Can-you provide a way to reduce further so low distortions? If yes, please, provide your solution with better simulated results.
I think, for the benefit of all, that we shall progress in a practical way: Every suggested change to come with simulation results that will prove a real improvement.
A collaborative WORK, not useless discussions on the angel's sex.
Last edited:
Your design includes some flaws that you do not want to know.
Post your asc-file and I will be back again here.
Sim File is just in my few posts before.
My friend Nyquist and me were very surprised by you care about phases behaviors at less than -10dB under and after the 0dB open loop gain.
I thank-you for your "flaw". Now, you just have to come back with proof of your strange assertions with better results. (stability margins+distortion numbers, good luck)
Sorry to say this, but if somebody between us don't want "to know", it is you, not me. The traditional compensation between base and collector of the VAS is known and experienced since 60 years by everybody, me included. An elementary basis.
End about this subject for me. So strange controversy !
https://www.diyaudio.com/forums/att...low-distortion-cfa-amplifier-def-cfa-tian-gif
Last edited:
Thanks a lot, Mooly.
To have done the file moove (If it was you) and for your encouragement.
I was really desperate, Only you and Aska seems to know what those results (not so bad for a simple circuit ?) request as work and subtlety (and, may-be a little experience ? ;-).
They has seen the sims, so few positive comments and encouragements in a lot of mud.
Is-it my ridiculous English that gives of me an image of a naive designer ?
I was so near to give-it up here.
Something is really sad in the way people behave on Internet, nowadays.
Nb: I have read the Dadod thread, this afternoon. An impressive work, impressive results, same kind of researches, same kind of reactions. Despesperating.
Last edited:
A couple of things I see
1) No need for series resistors in front of the gates?
2) Possible to move the gate clamp zeners to the bases of Q13,14, add reverse bias as Bob Cordell has shown to do in DH-220C to reduce their capacitance.
Hang in there Tryphon, we are not all a$$holes here. You have good ideas, but some people here (internet) have opinions that are not in good spirit.
1) No need for series resistors in front of the gates?
2) Possible to move the gate clamp zeners to the bases of Q13,14, add reverse bias as Bob Cordell has shown to do in DH-220C to reduce their capacitance.
Hang in there Tryphon, we are not all a$$holes here. You have good ideas, but some people here (internet) have opinions that are not in good spirit.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.