Speaking of quotations, here are a couple that were not made in the context of audio but I think are very apt in the audio world:
“If you can't explain it simply, you don't understand it well enough”
“Unthinking respect for authority is the greatest enemy of truth.”
Guess who?
These are both true statements, but let's remember the following. These are my responses to the above.
If you CAN explain it simply, you MIGHT understand it well enough. Simple is not always correct. I've met many who can explain things in a simple and concise manner, while still getting it WRONG.
Unthinking DISrespect for authority is also a great enemy of truth. Too many on this forum (I'm not implicating any specific person) are too willing to tell the pros and profs that they got it wrong. Questioning authority is neither good nor bad, because the critic could be right, or maybe wrong.
Those who encourage others to question authority should be equally willing to have their own beliefs questioned. Unfortunately, the most vocal critics of authority have little tolerance for their own critics. That's how it is, sadly.
Still, your comments are appreciated.
Agree, but telling ANYONE that they're wrong is not disrespect, *provided* the telling is backed up by factual arguments (which may or may not be correct).
But of course!
jd
Do you mean that if you wanted to "improve" the amp by changing the output transistors, should you seek ones with better high current gain or with higher ft?
Member
Joined 2009
Paid Member
GREAT!!!!
Hi
Piercarlo
Yes. I guess you could call this belaboring the obvious, but what's not always obvious is that many circuits will have an infinite number of harmonics, yet it's sometimes claimed that "circuit X has no harmonic components above the Nth", when what is meant is "circuit X has no measurable harmonic components above the Nth". That is, "no harmonics" gets treated as being equivalent to "no measurable harmonics", yet increasing signal level (but below clipping) will cause the high-order components to show up in the measurement. One very, very smart person who claimed that source degeneration in a JFET circuit didn't increase high-order harmonics was fooled because he didn't crank the signal level up high enough to see them above measurement noise.
Not what I'm saying. Bob found for the specific case a class AB BJT output stage with a typical bias that this was true. What I'm saying is that each case needs to be looked at individually. In fact, Baxandall's BJT example is a good counterexample to the "a single-tone will be monotonically reduced as NFB factor is increased" argument. BTW, my tanh() example was a bit silly. Since exp(x) is an infinite series, it could be argued that in the absence of some (unknown to me) technique for canceling the high-order harmonics, any BJT circuit will have an infinite number of harmonics.
Yes, but maybe "complicated" is not the best possible word or phrase. Certainly it's a mess mathematically. Baxandall was only able to get a closed-form solution for the amplitudes of all harmonics for the case in which the open-loop amplifier produced only second harmonic. So it means simulation or measurement becomes necessary to find the distortion components for non-trivial circuits. This tends to prevent the kind of clean prediction that can be obtained from a closed-form mathematical solution. It's also messy because there doesn't seem to be a clear rule that can be applied in advance to determine what's going to happen. People naturally gravitate to these clear rules, but the sometimes surprising effects of feedback on different open-loop configurations tend to violate such rules. For example, Bob's findings for the BJT class AB stage shows the "feedback increases high-order harmonics" assertion to not hold in all cases.
Last edited:
I think go for the higher FT, another caracteristic I look at is lowest Cob, if you need higher power capability use 2 pairs or more, it would give lower distortion too.
I have found that I do like no global feedback designs but this is only for mid and high frequencies, for bass I find a feedback circuit to have a better controlled bass.
Last edited:
I´m just asking if the necessity of very high fT devices at output is something that must be placed ahead of other questions like better gain characteristics etc. Sometimes it seems to be the case on discussions.
regards,
'belaboring' or 'belabouring', great word. Never seen it used before.
I agree with your point. The point I was making is that in cases where an infinite series of harmonics are not present in the OL system, application of NFB can create new harmonics.
Ok. I would add that I consider single-tone analyses pretty irrelevant.
"Mathematical mess" rings true. I don't think Baxandall had the miracle of LT Spice
. There are exceptions, but I think we agree that one cannot make a blanket statement like "NFB never adds new harmonics".The cautious, maybe even belaboring, comment I am making to folks is not to assume that stable NFB always makes a circuit perform better in all respects. Rather, NFB changes the nature of the performance and one can choose certain aspects/measurements that improve and certain ones that deteriorate. In particular, changing the spectral content, including adding new spectra, are common characteristics of the application of stable NFB.
Because a power serie is made of derivatives, this mthat aceans that the non linear model must be very accurate at the derivative level to give accurate distortion results. Do you believe that the models uscuraced in simulation have
Okay, but such systems are rarer than one might think. In fact, one could argue that they're nonexistent. BJT circuits are out. And the square-law formula for JFETs is just an approximation. Baxandall shows a phenomenon similar to MOSFET sub-threshold conduction in JFETs (non-zero current for Vgs beyond Vgs(off)). This involves exponential functions, thus an infinite series of harmonics.
Okay, but for static nonlinearities, if the single-tone behavior is known, we also know the IM behavior.
"Mathematical mess" rings true. I don't think Baxandall had the miracle of LT Spice
Yes. The point is that it's a fuzzy situation. One could say that "NFB increases high-order harmonics" and "NFB never adds new harmonics" are both wrong in general. The truth of the matter needs to be found on a case-by-case basis.
And speaking of LTspice, I investigated Baxandall's results for the case of an open-loop circuit having square-law behavior. Baxandall used a 10 percent open-loop distortion figure. I did his plots for 10 percent, 1 percent, and 0.1 percent open-loop distortion. The feedback-generated harmonics are extremely low when the open-loop distortion is low. These higher-order components are lower than one could expect for any practical circuit in the absence of feedback. This was mentioned earlier in this thread and can be found starting here.
Last edited:
For JFETs, the answer is "no", because the models are too crude. For BJTs the situation is much better, but there will of course be limitations to accuracy.
FFT analysis seems to be pretty tolerant with problems related to derivatives. The level 1 through 3 SPICE models for MOSFETs have discontinuities in their second derivatives at the threshold voltage, yet the FFT doesn't choke when the circuit is analyzed. If one tries to use these models in a harmonic balance simulator (which depends on continuous derivatives), the simulation will not converge. This problem came up at a simulator vendor where I used to work. We had to disallow the use of the level 1 through 3 SPICE models in harmonic balance for this reason.
Imho current gain linearity of output devices is to be privileged as choice criteria of power transistor. However current gain linearity or even Ft are, by itself, only details of a project. They count for their own in the performance of amplifier circuits of course but don't count alone.
Hi
Piercarlo
The quest for high fT does seem to be a bit of a religious crusade with some.
IMO the priority is linearity. If your amplifier and speakers and cables were all linear then fT would be irrelevant and NFB would be unnecessary.
So I argue that high fT is a means to an end. Either fT needs to be correlated to linearity in an open-loop system or fT needs to be high to enable stable NFB loop gain if you are trying to control the non-linearities away. So as Bigun said 'it depends' on what your approach to linearizing the system is. And this depends upon the specific system and parts you are using.
And some forget that we can only hear up to about 20kHz on a good day.
It would not surprise me at all to hear a better sounding amp with much lower fT output transistors than another. fT isn't the goal.
Last edited:
I believe you but I have a difficulty in understanding why the FFT would be more immune.
Let suppose that I have a system and a model of that system. If the system is let say weakly non linear and if the model has some inaccuracies so that derivaties are way off, then a calculation of the distortion components via power series or FFT of the output of the wrong model should give the same results and if not how can we say that the FFT result is better than the calculated one.
JPV
Single-ended triode designers might argue with you.
I'm not really sure what point you are trying to make. In the limit, when one scrapes the bottom of the noise floor one will find every possible frequency present so you will always be right in a sense.
I agree. What I should have been clearer about is that the FFT is tolerant of discontinuities of derivatives while harmonic balance is not. If the derivatives in the model are continuous but of the wrong value, then the results will of course be wrong for either technique.
Here's an example of discontinuous second derivative in the level 1 MOSFET model.
Id = 0 for Vgs <= Vto
Id = K*(Vgs-Vto)2 for Vgs > Vto
The second derivative of Id with respect to Vgs is zero for Vgs <= Vto and 2K for Vgs > Vto. This causes harmonic balance to bomb.
