OK. I apologize for harsh words. My position on this is that human experience is of prime importance, not computer programs. John Linsley Hood proved that sims could be wrong on at least one occasion.
Interesting stuff Richard - thanks for sharing.
This topology, although seemingly 'new' here on DIYAudio has been around for quite some time and is now very refined. What is important in my view for an amplifier designer, is that you work at your chosen topology and learn how to deal with the weak points and exploit the strong ones.
For VFA, I recognize the low noise benefits and the very low distortion potential, along with useful high input impedance properties. However, they can be complex and problematic to compensate to get best performance. CFA on the other hand may not beat VFA on absolute distortion performance (but anything below a few tens of PPM is more than adequate for audio - take a look at NP!), but they are easy to compensate, wide bandwidth and high slew rates. They also tend to be simpler (well, not if Edmond gets hold of them
As to exactly what makes a CFA and CFA and a VFA a VFA, my topological explanation is: is: In a VFA, the source and feedback signals steer a signal dependent portion of an LTP current source into a TIS input, generating a large output voltage swing, which after buffering drives a load. The key point is that the current source determines the maximum rate of change of the TIS output (I am considering a classic Cdom comp'd VAF here).
In a CFA, the maximum TIS input current value is is determined by the absolute values of the feedback network - there is no interposing current source to limit it. For this reason, the slew rates can be very high, ergo wide bandwiths. And, this is also why you can use the value of the feedback network to compensate a CFA (although with discrete designs, your options open up quite a bit). Of course, the divider action of the feedback network will cause a portion of this feedback current to be shunted to ground, but this is somewhat mitigated by keeping the output buffer degen resistors small in relation to the lower feedback resistor leg impedance - a trick easy to do in IC's and a bit more problematic in discrete designs.
When two instruments play the same tone at the same time in unison, like Dexter Gordon and Freddie Hubbard (one is reed and the other is brass) play the theme in Hancock’s Watermelon Man, you should clearly hear two different instruments. The instruments are played by humans, not by machines, therefore there is a slight delay between starting and stopping the tone. Besides, two tones have completely different color. The amp should reproduce these two tones so that we can hear these two different instruments and slight non-simultaneousness of players, the experience of reproduced music is much more vivid if the amp can do it properly. This is like IM distortion test, only with much more relevant and complicated tones (sine waves are just crude and simple approximation of instrument tones, in this case instruments play the same tone with different harmonic content). No matter how much complicated enhancements you add to VFB amp (aiming at low THD) you won’t force it to deal with this test as successful as CFB amp, because THD is more or less irrelevant. This is what Esperado calls separation of instruments. Superiority of CFB is direct consequence of better phase behaviour, and the better phase behavior is the consequence of superior bandwidth. By using CFB several important aspects relevant for audio are improved simultaneously. It’s the magic trick that works for audio just as much as for some other kinds of amplification.
There's probably 1000's of more proofs that humans generally get it wrong, so that doesn't help your case.
jan
Hi Waly,
What was it that you did not like about the class-D material? Is it class D itself, or the way I covered it?
Cheers,
Bob
I might ask as a friend of mine knows him ( Hiraga ) . Interesting you say that about Ken . To be honest I never got very excited about their amplifiers . Possibly things have changed ?
The Hiraga Le Monstre intrigues me . That one looks to let the music through .
I am very doubtlful people can hear amplifier with less than 0.1 % distortion sounding better ? That would be everyone alive now and everyone who every lived . So what is the point of doing it ? Bandwidth , Stability , Noise , Current delivery . Yes . 0.001 % distortion . What extreme madness is that ? If someone says because they can I will shake their hand for being honest . I strongly doubt most have speakers of better than 1 % THD . I do . I am happy with 1 % THD if nicely structured from an amplifier . I say to myself with my speakers it will still be lower than most people ever know . Our ears approach 30 % THD just before the point where we find it too loud ( Oxford University Radcliffe Hospital ) . The ear produces a spectrum not unlike SE valve amp . That type of distortion passes as 0 % if below 1% and exponential decay of harmonics ( harder to do than some think ) . I would frankly say any discussion leading away form that is the audio version of Homeopathy . However some suspect that lets say 0.1 % 5 th/ 7 th harmonic ( no other ) may be perceived as grossly more distorted than a good SE tube amp . That I might believe . Not thinking about hearing is like describing vision in the infra red or ultra violet . Interesting but not what humans can do .
The Hiraga Le Monstre intrigues me . That one looks to let the music through .
I am very doubtlful people can hear amplifier with less than 0.1 % distortion sounding better ? That would be everyone alive now and everyone who every lived . So what is the point of doing it ? Bandwidth , Stability , Noise , Current delivery . Yes . 0.001 % distortion . What extreme madness is that ? If someone says because they can I will shake their hand for being honest . I strongly doubt most have speakers of better than 1 % THD . I do . I am happy with 1 % THD if nicely structured from an amplifier . I say to myself with my speakers it will still be lower than most people ever know . Our ears approach 30 % THD just before the point where we find it too loud ( Oxford University Radcliffe Hospital ) . The ear produces a spectrum not unlike SE valve amp . That type of distortion passes as 0 % if below 1% and exponential decay of harmonics ( harder to do than some think ) . I would frankly say any discussion leading away form that is the audio version of Homeopathy . However some suspect that lets say 0.1 % 5 th/ 7 th harmonic ( no other ) may be perceived as grossly more distorted than a good SE tube amp . That I might believe . Not thinking about hearing is like describing vision in the infra red or ultra violet . Interesting but not what humans can do .
Bob, there is no way, on my opinion, a book writer about audio can forget class D analyses and explanations.
It is, obviously, the future. Will be better and better while power mosfets will increase their speed, and we will have more and more integrated circuits to simplify the design, as they are used exclusively in all cellular phones by milliards.
I don't know what you printed about, but, as you are so experienced, accurate and clever, i'm sure it is perfect, easy to understand and objective.
Of course, i suppose you had read the Bruno's papers and listened to a Ncore ?
It can make change the opinion of all those who still believe that Class D is only good for poor PA.
In fact i found them very similar to CFA in musical presentation, and my future active speakers will be class D for the bass/medium and CFA (of course) for the medium/treble (spherical horn 1500Hz to 16000).
One thing about class D .This week I knocked together a 5 transistor amplifier of 100 watts ( 2 x 2SA970 1 x 2SC2910 Exicon 10 N/P 16 ) . It is to be used between 15 Hz and 1 kHz for a 15 inch OB speaker . For fun I looked at it at 50 kHz . It was surprisingly good ( - 52 db second harmonic ) . I must thank D Self for the little tricks to make it so . I do use a little bit of VAS emitter resistance as I feel it helps ( 16 R 8 mA ) . The design will function at 0R .
The point is I couldn't do that with class D . It didn't have a PCB . It never once suffered instability . The distortion worryingly low .
Note this . Saying the 5 transistors you can probably draw the circuit . Cdom 15 pF . LTP 2mA . Gate stoppers 220R . Gain 34 . Bootstrap CCS .
The point is I couldn't do that with class D . It didn't have a PCB . It never once suffered instability . The distortion worryingly low .
Note this . Saying the 5 transistors you can probably draw the circuit . Cdom 15 pF . LTP 2mA . Gate stoppers 220R . Gain 34 . Bootstrap CCS .
I agree. A little harmonic distortion, when it increase with levels in a linear way is not so much a problem, and change only (slightly) the musical texture of un instrument. We sometimes use distortion to bring more presence or loud feeling on single instruments in records production.
Where most of the the problems lies, on my point of view is IM when several instruments play together (non harmonic products) that blur the scene, and instant dynamic. Transients make all the life, both for speakers and amplifiers.
I don't know how our brain+ears work together exactly, but i believe they are most sensible to the surface (energy) of a transient signal and phases than anything else.
I wonder why i'm still able to notice a difference when i add some correction to an instrument at 40Kz, while i'm no more able to listen a sin at 15Khz !
And youg musicians with brand new ears prefer my actual way to record cymbals at those i did when i was a lot younger. So strange.
Too, my speakers use a JBL driver + horn for the top half of the bandwitch.
High cutoff at 16Khz. And all listeners make remarks about the very natural sound in trebles, specially cymbals, where you hear the metal, the size, and the weight and speed of the stick impact.
I tried to add a super tweeter, it always killed the magic.
Now wally will definitively not understand why i'm looking for improvements in closed loop for my power amplifiers at 10 Mhz ;-). But i don't care.
It goes in the same sens, more the amp is 'fast', less you ear "trebles". Just instruments. We never hear those pshh-pschh treeble in real life that most VFA and soft dome tweeter produce.
The way you cover it. That's a way to interesting and complex topic to be covered in 50 pages. You did not even scratch the surface and the benefit for the reader is virtually zero. This deserves at least a separate book, in fact only the "feedback in switching systems" topic can, and already does, fill a full shelf.
In general, I take your book as a collection of facts, best practices and classic design ideas. When I need my memory refreshed on certain topics I go straight for it, and the quick/short answer is always there (example: Q: where is the lowest yield of a class B amplifier, A: at about 33% of the clipping power). However, for a more in depth answer (example: why is it 33%) I must go elsewhere. Hence my frustration for the lack of proper references.
Richard, please permit a naive question. What does this mean ? "..the input of a diff stage is compressive while the CFA is expansive"
Hi Waly,
Thanks for your feedback on the class D section of the book. I do plan to expand and further improve it in the Second Edition, as it is not only an important subject area, but also a fast-moving subject area. At the time I put it together, much of the class D information I could find on the various aspects of class D was spread out over many different places, and I worked hard to pull it together in one place.
I completely agree that an in-depth coverage of class D needs a dedicated book, and I don't claim to be the expert in that area to put such a book together. However, I disagree about there being little value in covering the subject as best as possible in a power amplifier book in 5 chapters and 50 pages that will increase to more pages in the second edition. For a book titled Designing Audio Power Amplifiers to not cover class D, even if that coverage is not as in-depth as one would like, would, I think, be a sad omission in these times when class D is becoming so much better and more dominant.
The key is for me to make it better - as good as possible - given the page limitations that we as authors always face. Perhaps I can take it to 100 pages. I welcome your input on specific ways I can improve those chapters, whether it is specific criticism of errors or specific suggestions for added/expanded material and references.
If you are highly knowledgable in class D, I would love to pick your brain on the subject. Send me an email.
Cheers,
Bob
I understand the Class D basic plot and the first level of complexity, but unfortunately I am not even close to an expert.
For quite some time I am yearning for a reference book on Class D principles and design. Perhaps Mr. Putzey will decide to put his knowledge in print...
You can use those 50 (or 100) pages in the new edition by expanding (e.g.) on loop gain analysis, stability criteria and analysis, Fourier analysis, current feedback amplifiers, etc... Things that the current edition is sorely missing.
Bob Cordell, your sentiments are perfectly made but there is so much education yet to be completed on mature technology. The AES, with its audience of professional engineers felt it necessary and helpful to invite d self to trot out charts from 1993 in order to teach folks the importance of biasing a LTP correctly in a 2008 "masterclass".
I realize that most of the debates you see here are old hat to you - but that is certainly not true for many. If a society of people who do this for a living need exposition of old tech - you can bet that the majority of amatuers do too.
There's so so much to cover - fet and jfet inputs, cfa, diamond buffers etc. do we really want to move on to the next Big thing when the old things are not even understood well yet by the vast majority of your audience ?
A (maybe even slim) volume on class D will be plenty attractive for those who want it. I for one, am not yet interested.
I realize that most of the debates you see here are old hat to you - but that is certainly not true for many. If a society of people who do this for a living need exposition of old tech - you can bet that the majority of amatuers do too.
There's so so much to cover - fet and jfet inputs, cfa, diamond buffers etc. do we really want to move on to the next Big thing when the old things are not even understood well yet by the vast majority of your audience ?
A (maybe even slim) volume on class D will be plenty attractive for those who want it. I for one, am not yet interested.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.