You seem to have a real talent to drop atomic bombs without war declaration
Look here how closed loop in audio bandwidth is affected by what happens between 1 Mhz (vfa) and 10 Mhz (cfa).
http://www.diyaudio.com/forums/soli...-self-wants-your-opinions-11.html#post3585406
35dB Gain x 10MHz BW = 560MHz GBW for a power amplifier. Try building that and come back when you are done. I'm not holding my breath.
Otherwise, 10MHz closed loop bandwidth for audio is anyway nonsense. Not only because it wouldn't make a difference even for bats, but also because you need an input filter to get rid of the RFI. That would limit the bandwidth to a few 100's of KHz.
You provide brutal unjustified assertions (witch goes against all i had found during 40 years of experience, both as an audio designer and as a sound engeneer, hating voodoo).
I provided elements of discussion and *measured differences* in the audio bandwitch . It applies at feedback closed loops where the differences are situated between those two frequencies.
Of course, it is a lot of work to open the post in reference, look at the images, and try to understand what they means.
It looks like impossible to have a serious technical discussion in this forum dedicated at "true" false" kind of arguments. As if the truth was depending on the numbers of true or false believers.
As many of you know, I published the design of the CFA forrunner just before Elantec et al did it. in fact I was called and discussed it with mfr/designers working on same approach. Now that was a long time ago. I was asked why I used that topology instead of the VFA type I had used before. It was for the high freq linearity at higher signal levels that i was after. It was not designed for wider bandwidth - though that has gained the upper hand in CFA useage. That original CFA circuit did well beyond 10MHz. But it was not designed with that in mind. it was a side effect of making a very linear circuit with low impedances.
Thx-Richard Marsh
Thx-Richard Marsh
Last edited:
I'm really, tired, tired, tired.
I filter my input signal, indeed, at around 200Kz -3db. And, despite this, noticed a real audible difference between the two amps, correlated by everybody witch was present when i made demonstrations between those too cfa / vfa same devices amps, no bats.
The day i will see a flat bandwidth in the input of my darlingtons in closed loop, i will be satisfied. Even with my 10Mhz amp, i'm far away. As you can see on my images.
There is a clear audible difference, too, between OPAswith 13V/µs (supposed overkill for 20Khz bandwitch) and 500 Vµs, ones. I just tried to propose a track to explain why. You don't even try to discuss about.
i'm not an audiophile, just an electonician and sound engineer, i'm not using Bybee stuff, or exotic cables in pure cryogenic golden unoptainium. But i believe in what i hear, even when i cannot explain. The difference is, on my side, i try to understand.
This is, on my point of view all the differences between CFA and VFa, that people here with ears and a CFA at home seems to have noticed , on a sonic point of view, while others just do not try because they do not 'believe' in teh way our ears hear transients. and high harmonic distortions out of their audibility of pure sinusoidal signals frequency .
It is, here, a question of speed between error and correction by the feedback, or phase, as you prefer.
So am I.
Claiming to hear the positive effect of a 10MHz closed loop bandwidth is, to me, a total show stopper.
Good luck in building that 500MHz GBW power amplifier and make sure you don't hear the sound of instability instead of the sound of 10MHz bandwidth.
P.S. Hints:
- Fix your models. According to the datasheets, the 2SK135 has a Ciss of 600pF, while the 2SJ50 has a Ciss of 900pF. You are using 220ohm gate resistors, this will make a pole in the forward path at under 1MHz, which you obviously did not model correctly (otherwise you won't get this amp decently stable, with or without a CFA, with 20dB feedback). Although you did not post any relevant information about the amp stability, I assume you know how to analyze the loop gain.
- You have no other frequency compensation that I can see, beyond the pole above and the small phase lead caps. This is a safe recipe for bad sound - the Ciss defining the dominat pole is, unfortunately, voltage (Vgs) dependent. Which leads to a textbook example of TIM (transient intermodulation distortion) - the pole frequency is modulated by the input signal.
Yes...my models....blabla...
They are from Bob Cordel site... Lucky, he is reading this thread.
I'm listening to this amp, right now. Bandwidth was 5Mhz in real world. No instability, no overshoot on square waves, and, believe-me NO audible TIM (that's why i made this mod) and very very litle IM when i measured-it. Better than the original VFB version.
You are right, is true that the darlington can run out of current in hf at full levels. Happily, there is no risk of high levels components after 40Khz, and the slew rate is ~1000V/µs. Go figure.
You still don't comment the phase effects of reduced bandwidth of the feedback loop in the audio range when closed loop witch was the subject (not my amp).
Nonsense. The amp will be completely stable. Make the amp, test it, than speak.
You would be surprised if ever someone do a simulation of this circuit.
It works but let say that it s not exactly hifi...
Eh, simulators! Instead of serving people they are enslaving people. Dangerous toy for adults. Powerful narcotic... mind narrowing...
Of course that some people can not hear the differences, if they do not listen at all and instead spend too much time in front of PC monitor...
Actualy simulators are not even needed in this case ,
basics calculations can be made in seconds , wich lead
me to conclude that few people can actualy rapidly
estimate if a given circuit components values are
within tolerances that yield a functionnal amplifier.
I was wondering about this, Pioneer was using diamond buffer input configuration in cfa audio amps long before comlinear and elantec applied for various patents with the circuit. They also applied for the same patents in the USA before Comlinear and Elentec. The outcome of this was the Comlinear and Elentec patents were granted 2 years ahead of the Pioneer ones but Pioneer had applied 2 years before they did. The worst is the fact the pioneer patents were granted as well. What exactly is going on in the patent office ?? The wording and view of the operation is different, as also can be seen here in this forum where some use the Term VAS and others TIS but the building block is the same. Where exactly did this circuit block originate ??
That is not a fair comment. There are several posts that claim that sims prove that CFAs are 'better'. You can't have it both ways.
jan
This post had our whole office in laughter, we were burning the midnite oil and the discussion was interesting enough so it was displayed on big screen for all to see. Later though I was trying to figure where you were coming from.
You sure talk the talk but you havent walked the walk.
Even suggesting the use of miller compensation on CFAs had our toes curling. You can maybe get away with it with designs like the VSSA but diamond front end forget it.
Miller compensation with the topology causes peaking, you dont seem to know this because youve never built such a amp or even simmed one, I deduce. You can try till you blue in the face but some peaking will persist no matter what you try.
Due to the class AB drive nature of the diamond stage theoretically there is no limit on the current to charge the compensation and this too has implications regarding the choice of optimal compensation scheme. If you had walked youd have known that what you call brutal is equivalent to using around 18 pf miller compensation in a CFA design but with disasterous consequences as you would have noticed with the first point. Besides peaking that maybe you could somehow get away with youd end up with much reduced phase and gain margins. Distortion is increased contrary to your assumption that it will decrease and made worse by the next point. In case youd want to make use of a triple EF youd require the dreaded CB driver compensation as well as base resisters to be able to maintain stability.
TPC has the same disasterous consequences although youd be able to lower THD, but lower THD compaired only to using just plain miller. Horror story for you I suppose, shunt can actually outperform TPC. See some comments and suggestions by keantoken how shunt compensation acts in a CFA, one member Ive seen thats figured it out as has homemodder. Further the pole created by the capacitance of large geometry devices used in a outputstage can actually be a blessing in compensating a CFA but youll only know this once you walked.
Before proposing unneeded lectures you should rather actively persue some yourself using not only books but including actual components.
Since we talk different languages there is no point in further discussion so lets drop this too.
Never mind that CPUs and GPUs comprising as much as billions
transistors are first simulated before being manufactured and
are working perfectly audio fans will tell you that basic circuitry
using a dozen active devices are out of reach for said simulators..
More seriously , of all the circuits i simulated the best complexity/perfs
ratio is obviously hold by the usual ubiquitous discrete 741 , that is a single
differential + current mirror + enhanced VAS , better perfs can be extracted
with a symetrical differential input amplifier but at the expense of simplicity.
You seem to be unaware of the fact that there are plenty of monolithic op amps of the so-called "CFA" type that use Miller compensation.
This is surprising for someone who claims to have a great deal of experience with this lamentably inferior topology.
Last edited:
I will do too.
Just one remark: it seems that people (always the sames) who publish very intelligent things like: "lamentably inferior topology"(mouaaaarf) ,"don't need such a slew rate", "1Mhz has no meaning in audio" had not the slightest experience and comprehension of CFA and had never listened to any of them. As you point out, it is obvious they just don't know what they are talking of.
Funny to see that they are all this kind of people listening to music across simulators ... or books !!!!
Time for them to order a *cheap* VSSA, if they know how to use a soldering, and listen. This said, i'm sure they will never do it, for obvious reasons.
EOT for me, it is always the same story, each time CFA vs VFA comes in the discussion, a pure waste of time and a disagreeable moment.
Really, point them out.
Your now familiar attempts at changing details and focus points in posts and turning them into a farce to suit your views is overbearingly irritating and fruitless. Could you also back off a wee little bit from repeating and beating around a dead horse for ages, the threads are clogged up and annoying to read.
Even so Ill indulge in the hope of seeing some new topology or building block that would allow implementing Miller in a more successfull way with CFAs.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.