Sorry Mr. Marsh, that's not it. Learning about CFAs is definitely a good exercise. They have to be in the toolbox of every EE, circuit or system designer.
What drives me nuts is when (say) Mr. Marsh claims that CFAs would have any advantage in audio reproduction (or even worse, general statements about CFAs having lower distortions, higher loop gains, easier to compensate, etc...). In fact, CFAs are not the audio magic sauce, or the best thing since sliced bread, they are simply a design option with it's own trades, compromises and limitations (P.S. and with certain application domains where their properties shine, audio not being one of them).
In fact I can quote you some audio guru posts in the diyaudio archive claiming that CFAs (specifically the Alexander amplifier) sound "horribly metallic" and should not be considered for any SOTA audio reproduction. Which makes the CFAs in audio as much a fashion issues as full complementary designs, no global feedback, etc...
Oh, and I believe we have a vastly different view of "learning". I'm personally not shy of grabbing a pen and paper and spending hours of calculating, or firing up a simulator and trying to understand the properties of a circuit topology. I would say, you should try this sometime, it could be an enlightening and revealing experience.
Last edited:
I'll have to call you out on that, If there is not some kind of protocol on that listening test "horribly metallic" is about as much BS as anything here. Good luck getting Mr. Marsh to even read relevant links.
Last edited:
Exactly my point, BS transformed in a fashion. I'll bet in 10 years from now we will read again about how inappropriate CFAs are, for audio.
Could be you are correct in 10 years. But I'd bet you a dollar it isnt. Early on, I observed that many (most) high-end from manufacturers are the CFA or CFA-like topology and low end of their line is regular VFA. I suspect there is a good reason other than advertising --- most don't advertise the design issues to lay consumer. So, it is an engineering and listening decision. Or, coincidence? I dont think so.
I have built and SIM'ed a few circuits and built a few of this and that. I understand the 'slewing' condition and that is generally not happening with any modern design of good design. An input filter can prevent it in any case.
Me thinks you ascribe way too much importance to my rhetoric.
I do tend to exaggerate a bit at time when I get excited about good products and their sound. The CFA I heard from DADoD, is head and shoulders above what i used before (VFA) - which was no slouch - trust me. Yes, it does does not have any funnies when clipped.... which is also hard to do with high power (>>50W/8) and high eff/sens speakers. I have promoted higher power to prevent clipping and audible compression which in part lead to buying the JBL M2. However, I pay particular attention to PA performance and sound at low power (<10-20W).
Because of my observations, there were at least 2 forums created just to do that study; One for explaining and understanding the differences.. pro/con. The other to sim. build and listen to the various topologies. A good book on audio might glean some useful info from both.
Enjoy,
Richard Marsh
I have built and SIM'ed a few circuits and built a few of this and that. I understand the 'slewing' condition and that is generally not happening with any modern design of good design. An input filter can prevent it in any case.
Me thinks you ascribe way too much importance to my rhetoric.
I do tend to exaggerate a bit at time when I get excited about good products and their sound. The CFA I heard from DADoD, is head and shoulders above what i used before (VFA) - which was no slouch - trust me. Yes, it does does not have any funnies when clipped.... which is also hard to do with high power (>>50W/8) and high eff/sens speakers. I have promoted higher power to prevent clipping and audible compression which in part lead to buying the JBL M2. However, I pay particular attention to PA performance and sound at low power (<10-20W). Because of my observations, there were at least 2 forums created just to do that study; One for explaining and understanding the differences.. pro/con. The other to sim. build and listen to the various topologies. A good book on audio might glean some useful info from both.
Enjoy,
Richard Marsh
Last edited:
Hi Bob, I'm not angry with you, I thought after posting that maybe there were already many things competing for your attention. I probably need to think about how to explain things without causing too much confusion. What I wrote reflected my state of mind, after my computer dying again and some other stuff. But even then my irritation was at what you wrote, not you as a person.
250W Exicon Lfets.
That part you got right. I used the words "strongly suspected", which taken literally as intended does not convey absolutes. You can think about it this way. Just because a trace resonates with a MOSFET doesn't mean they will oscillate. Many people assume it is the MOSFET that causes the oscillation when it is actually the feedback loop, in an indirect way. The gate stopper is a possible solution for both problems, but a better solution is to understand the problem better.
This is not something I usually do, but my guess would be yes, although I would make it 10 times to be sure. For local oscillation you are really looking for the frequency to be constant regardless of feedback compensation. If you change the compensation and the oscillation frequency changes, then you know the feedback loop is strongly influencing the oscillation. This is confounded by the fact that oscillation is a form of overdrive however, and this in itself tends to shift the frequency.
It's not compensation if it doesn't improve stability. Some amps react positively to it, many negatively. But in this case the better solution is still to remove the offending lead capacitor, rather than to increase gate/base stoppers.
250W Exicon Lfets.
That part you got right. I used the words "strongly suspected", which taken literally as intended does not convey absolutes. You can think about it this way. Just because a trace resonates with a MOSFET doesn't mean they will oscillate. Many people assume it is the MOSFET that causes the oscillation when it is actually the feedback loop, in an indirect way. The gate stopper is a possible solution for both problems, but a better solution is to understand the problem better.
This is not something I usually do, but my guess would be yes, although I would make it 10 times to be sure. For local oscillation you are really looking for the frequency to be constant regardless of feedback compensation. If you change the compensation and the oscillation frequency changes, then you know the feedback loop is strongly influencing the oscillation. This is confounded by the fact that oscillation is a form of overdrive however, and this in itself tends to shift the frequency.
It's not compensation if it doesn't improve stability. Some amps react positively to it, many negatively. But in this case the better solution is still to remove the offending lead capacitor, rather than to increase gate/base stoppers.
R.Cordell used that phrase in the "interview" series he did with us.Hi Bob;
thank you for your "The sound of the amp depends on how it misbehaves" phrase during your lecture on BAF.
It was great how you put this simple truth in words, as if I finally heard my thoughts aloud.
I did see a similar anecdote some years earlier, but can't remember whose typewriter it came from.
To me it makes a lot of sense.
The Baxandall test and maybe that of others to set up different amplifiers and tailor the frequency responses to be exactly equal makes the amplifiers indistinguishable if you ensure they never misbehave. This supports the Cordell contention.
Not strange at all, most of the headphone drivers in high end cell phones are CFA based recycled xDSL drivers. This has been going on >10yr. now in USB headphone amps.
EDIT - The spec are, should we say, nothing to write home about. I have no idea what would make the THD fall apart before 10kHz. IIRC our own jcx had a thread long ago on headfi with superior results. Jan also produced a nice DSL driver based PA.
Last edited:
Hi Bob, Kean
I have been interested in the problem of multiple, possibly nested loops for a while and your "parasitic oscillations" are simply a common instance.
The difficulty is that the simple Black feedback model does not handle this even close to accurately.
Bode prepared an elaborate mathematical framework that can handle arbitrary multiple loops, does the simple loop theory, and then, just as the preliminaries are finished, stops - says he was too tired.
So there is not much theory, probably explains why you two have had a communication problem.
Dr Ed Cherry has covered this a bit but he's not easy to read.
I have started to work on the problem in the style of Mason and Signal Flow Graphs.
The results are similar to Middlebrook's GFT.
Do either of you have any other relevant references or information on the problem?
Best wishes
David
2 reasons for the more frequent use of CFA -- Could it be?:
1. In the practical world of small package sizes or portables, current on demand I/O stage now dominates. High idle etal currents are not an option in many device packaging such as IC/opamps. So, the CFA with current on demand I/O is used more and more.
2. The typical VFA used in low cost power amps and receivers et al found in retail stores, does not have adequate SR or margin. Not that they couldnt be designed to.... the SR just has been typically low on the majority of new sold amps. Substitute a CFA and it sounds better and the HF THD is also better. If VFA were universally avail with higher currents (>SR) they might not be judged inferior to CFA.
THx-RNMarsh
1. In the practical world of small package sizes or portables, current on demand I/O stage now dominates. High idle etal currents are not an option in many device packaging such as IC/opamps. So, the CFA with current on demand I/O is used more and more.
2. The typical VFA used in low cost power amps and receivers et al found in retail stores, does not have adequate SR or margin. Not that they couldnt be designed to.... the SR just has been typically low on the majority of new sold amps. Substitute a CFA and it sounds better and the HF THD is also better. If VFA were universally avail with higher currents (>SR) they might not be judged inferior to CFA.
THx-RNMarsh
Last edited:
No, it is not. Although this was told a gazillion of times, somehow these simple facts don't compile for you:
1. Current on demand is not a CFA specific. There are VFAs (example, the Stochino amplifier) that are implementing current on demand.
2. "Typical VFA used in low cost power amps and receivers et al found in retail stores, does not have adequate SR or margin." Purely nonsense, 90% of the high end audio amplifiers on the market are VFAs and except for maybe a few grossly incompetent designed, they all have enough SR to make the indistinguishable from any CFA you may think of. "Substitute a CFA and it sounds better and the HF THD is also better." "Sound better" - a fashion statement, without any substantiation. The rest is exactly what I mentioned in my previous post as driving me nuts: CFAs can't provide a better linearization through negative feedback, at least because the Bode maximum feedback theorem doesn't make any distinction between VFAs, CFAs or WMYHFAs (Whatever Makes You Happy Feedback Amplifier).
Long forgotten PFA (Or just PA, POWER amplifier)? ;-)
"Voltage Amplifier Certified" engineers keep forgetting about non-linear and frequency dependent currents, while "Current Certified Engineers" keep forgetting about non-linear and frequency dependent voltages.
So, let's finally return to power amplification. Everywhere, from microphones to speakers, and be happy!
"Voltage Amplifier Certified" engineers keep forgetting about non-linear and frequency dependent currents, while "Current Certified Engineers" keep forgetting about non-linear and frequency dependent voltages.
So, let's finally return to power amplification. Everywhere, from microphones to speakers, and be happy!
Waly, please cut out the personal comments. You are free to challenge ideas, but not the person in a rude way. Try and keep it clean.
Really, I would appreciate if you could point me to any "personal comments" that I made lately. I'm willing to listen and learn, a private message is ok.
Hi Waly,
Basically, any comment you would be unhappy receiving shouldn't be posted. You can disagree over ideas, points of view or a given "fact" or statement, but do not comment on others in an unkind manner. Your posting history has many examples and I figured a word might be enough.
-Chris
This rates.
Basically, any comment you would be unhappy receiving shouldn't be posted. You can disagree over ideas, points of view or a given "fact" or statement, but do not comment on others in an unkind manner. Your posting history has many examples and I figured a word might be enough.
-Chris
Some practical reasons --
Small packages with limited package dissipation issues.... IC packaging... which needs higher peak levels or low Zo drive would find current-on-demand a no-brainer. Its somewhat of a moving target ... whether CFA or Modern VFA (MVFA).... are quit popular.
There is still the curiosity of the higher end makes and models moved to CFA topologies -- Sony, Marantz, Yamaha etc and Krell, Quest and many other and smaller companies.
I have also speculated why that might be based on practical grounds (low cost, heat, idle current etal) while keeping high performance. If there is a good reason for the higher power higher end models use CFA topologies, it would be good to know and include in the book as a pro or con point of view.
THx-RNMarsh
Last edited: