Yes, my ideas were influenced by your comments on Cherry (AKA Output Inclusive Compensation for those not familiar with Richard's enthusiasm)
This is, in a sense a continuation of the idea to minimize the local feedback and expand the loop.
Cherry includes the VAS and OPS, why not eliminate the local feedback implicit in the EF too.
I also expected, based on what I had read, that the VAS was the main source of NMPhase, and that it would be minor.
But study of Damir's circuit lead to a surprise - that there could be far more obvious NMP from the compensation if it was more complex than simple Miller.
I finally understand this pretty well.
Not criticized, just observed😉, I don't claim to know what the practical limit is, yet.
I concur that many of the other ultra low distortion attempts also have similar ULGF, maybe not obvious.
ULDistortion requires lots of LG, so we either push up the ULGF or ramp the LG faster with frequency.
I am inclined to do some of both.
Best wishes
David
CE allows driving the output Q base with >100 mA Iout monolithic op amps which today can have order of magnitude higher BW than the Audio Power Q
I'd guess that together with the CFA? driving the CE output Q base, a 2nd monolithic input op amp selected for low noise, high linearity would give enough loop gain for shaping to whatever is needed
but that gain shaping will be nested loops of some or several flavors
my take on the relatively limited usefulness of Black's FeedForward for Audio Frequency improvements:
I'd guess that together with the CFA? driving the CE output Q base, a 2nd monolithic input op amp selected for low noise, high linearity would give enough loop gain for shaping to whatever is needed
but that gain shaping will be nested loops of some or several flavors
my take on the relatively limited usefulness of Black's FeedForward for Audio Frequency improvements:
Last edited:
I was not aware of that amplifier, at first impression it does look to have included several of the ideas I had in mind.
Thanks for the reference, I will study it a bit more.
Yes, that was the other idea I had😉
My first inclination is simply to buffer the CE with an EF and use an AD797 as the IPS to drive it.
Seems clean and uncomplicated, any reason not to do it this way?
Best wishes
David
As usual, I am substantially in accord with your ideas, as expressed in the quoted text.
In particular the concerns about the complexity and sensitivity of the coupler network, for inaudible benefits.
Last edited:
OLG of an AD797 is 80dB@10kHz and 74dB@20kHz.
If the VAS + OS arrangement has 0.5% THD and that the amp has a gain of say 30dB then it wont be possible to get lower than 30 ppm THD for frequencies above 10kHz, and that s assuming that you can apply 50 dB loop gain at 10kHz and as much as 44dB at 20kHz, wich is unlikely without serious stability issues.
Remember that I do have the CE OPS gain to use as well.
I have experimented quite a bit with Two Pole compensation, Bode steps and similar techniques in Spice.
I expect I can apply quite a lot of feedback without stability problems.
Best wishes
David
two pole compensation tricks let you expose the high gains available from the AD797 "decompensated" datasheet table at audio frequencies, and roll them off near typical PA unity gain itercept frequencies
http://www.diyaudio.com/forums/solid-state/123613-class-biasing-ad797-6.html#post1525757
shows plenty of excess loop gain while providing Av = 30 external gain
http://www.diyaudio.com/forums/solid-state/123613-class-biasing-ad797-6.html#post1525757
shows plenty of excess loop gain while providing Av = 30 external gain
Last edited:
Hi Wahab,
Your conclusion is based on a number of assumptions that are rather questionable.
Moreover, if we add feed forward error correction to the setup, some assumptions become irrelevant.
The picture is taken from "Design and Construction of a Feedforward
Error-Correction Amplifier", SUSUMU TAKAHASHI AND SUSUMU TANAKA
I know - feed forward is a little bit off-topic at the moment, however, I would keep it in mind, thinking about low-distortion amplification, as it really allows to break the limits, you can reach with the feedback.
Cheers,
Valery
Attachments
Last edited:
Dave is right to be interested in current drive of the output stage. When I developed my simple darlington amplifier, I was surprised how good the transient response was. This I later realised was partly due to the current drive of the output darlingtons . Baxandalls Cbc becomes a free Cdom for the Vas in my design. However I was not 100% happy with the regulators as a band-aid fix for the quiescent current variation with power supply voltage problem. For those who can tolerate the rambling narrative here is my design:
https://consort3.wordpress.com/2010/10/30/simple-cost-effective-darlington-amplifier/
https://consort3.wordpress.com/2010/10/30/simple-cost-effective-darlington-amplifier/
Fair point, there s a margin left at this stage, litteraly..
Certainly that optimisations can be worked out, after all there s some Pioneer car amp that use an AOP + VAS + OS topology and wich work very reliably with quite surprising perfs on the simulator, something like 1ppm@1kHz/25V, i ll post the schematic later albeit the OS is a classical EF and far fetched from your project, but this can be indicative given the compensations implemented.
@Vzaichenko
Of course i was talking of usual compensation schemes, i m aware of the methodologies you are talking about and their doability, so far Cherry s NDFL is the most efficient when it comes to linearity in the audio band, the practical implementation is not systematically satisfactory, though.
FTR i get lower stability than with usual designs, not bad but not good enough for my taste where unconditional stability should be met even without output inductance, and with about 1uF capacitance added to the load.
David, with the CE OPS gain goes higher distortion, and final result is very similar to EF or SF. The CE is not easy to bias and make thermally stable.
It's quite simple to have an amp with MOSFET SF OPS with high Loop Gain at 20 kHz, and very stable too, with appropriate compensation.
I see the audio power amp future in combination of analog Class A or AB and Class D. I don't think pure Class D will beat sonically Class A in near future. Combination of Class A and Class D could be powerwise efficient almost as pure Class D.
BR Damir
Yes, the extra freedom made possible by the AD797 "decompensation" option is precisely why I named it as my choice in the earlier post.
I will certainly look at this more closely if I ever want to do a sub-PPM-distortion amp to make a point.
But no audible benefit for the extra complexity, I suspect, so I want to concentrate on how simple I can make a low distortion amplifier.
Thank you, I will look at it ASAP.
wahab said:
I have some doubts about NDFL, despite my utmost respect for Dr Cherry.
Yes, this is really the key.
The theory says the OPS distortion should come out about equal between the CE and EF.
The distortion from the other sections should be lower with the CE but that distortion is small so there is only a minor benefit from that.
The main benefit I see from the CE version is that the earlier sections can run on lower rails so we can use an op-amp for simplicity, or faster transistors for performance.
The main "cost" is the complexity of a thermally compensated bias system, exactly as you say, and even the IC manufacturers admit this.
So this thread is to see if I can find a reasonable solution to that problem so that the cost does not exceed the benefits.
Otherwise I will just retain my EF OPS.
Yes, exactly my opinion also.
But first I have to work out the Class B+ part😉
Best wishes
David
Last edited:
CFP distortion at low level?
I gather Guru Zan is also considering CE output stages instead of the usual EF as part of this exercise.
I have 3 serious objections to CFP output stages ... which can be considered the simplest type of CE output stage.
Anyone have the latest version of his book? I asked him to do more work on this and it would be good to see if he has.
I apologise to Guru Zan if this is too far off topic 😱
I gather Guru Zan is also considering CE output stages instead of the usual EF as part of this exercise.
I have 3 serious objections to CFP output stages ... which can be considered the simplest type of CE output stage.
- They are difficult to bias properly. I've posted in several threads including Bob Cordell's what you find in MAY PAs with CFPs. When the amp is stinking hot, you often find HUGE xover distortion for a minute or two until it 'cools' down. Bob has seen this too.
- Just as important is that it is difficult to turn OFF CE O/P transistors .. especially after HF overload. Cherry's early 'CE' has each O/P 'pair' nearly as complex as my complete PA designs to deal with this.
- Lastly, something I haven't tested in real life. Doug Self in his 4th edition compares 'distortion with level' of amps with EF2 & CFP output stages in Figs 5.39 & 5.40 The CFP stage has a good deal more distortion than the EF2 at levels below 5W
Anyone have the latest version of his book? I asked him to do more work on this and it would be good to see if he has.
I apologise to Guru Zan if this is too far off topic 😱
Last edited:
I am interested in rail-to-rail CE OPS.
NOT interested in CFP, I share many of your concerns.
Best wishes
David
Interesting work on class B (AB) OPS, assisted by switch mode stage:
https://www.researchgate.net/publication/3280444_A_Class_B_Switch-Mode_Assisted_Linear_Amplifier
Also note the list of references in the end of the article.
Cheets,
Valery
Not really "current driven OPS", probably that topic deserves at least a new thread, maybe a whole sub-forum between "Class D" and "Power supplies"
Best wishes
David
Interesting but not so good result and to complicate.
There is simpler approach without jump in distortion when Class D kick in.
BR Damir
The answer must be the topology of one of the rail-to-rail OPAs.
Do you have one (or two) you favour at present?
This was one of my obsessions in Jurassic times as it was for Mike Albinson & Peter Walker.
It now moves down the list but I still cringe at those who drop the voltage (slightly?) to the VAS 😱 (Sorry Damir 🙂 )
I've spent a long time looking at fancy topologies .. but always came back to the IPS - VAS - EF (sortof) output topology in its various forms.
The only novel and useful to me things which have emerged over some 5 decades of agonising over topologies has been
- CFA inputs
- Vanderkooy & Krauel which may give you exactly what you want but needs a floating power transformer
Apologies to Cordell, Self & Prof Cherry but these gurus have merely led to a better understanding of 'standard' circuits. 😎
__________________
And does anyone have a copy of Self's latest PA book? 6th ed I think.
I want to see if he's investigated the increase of THD of CFP amps at low level compared to the equivalent EF2 I mentioned in #74.
Last edited:
- Status
- Not open for further replies.