1% = 0.01 = -40 db...in voltage ratio..you did mention a
power of 1 W...let s take it as an exemple..
1 W rms / 8 OHM load is 2.83 V rms...
distorsion at -40 db makes 28.3 mV...
(0.0283^2 )/8 = 0.0001 W = 0.01% of 1 W
Member
Joined 2009
Paid Member
And so, returning to the first post, the question was the value of a high Ft. If I understand at least a little of the past few pages then the benefit of a larger Ft is actually for use with NFB. The higher bandwidth, if supported by the preceding amplifier stages, allows the amplifier to retain good gain into much higher frequencies because now the output devices have more 'headroom' in terms of their speed so that even when they are operating outside of their ideal range in terms of current/voltage (such as Class B perhaps) they don't become too slow. This allows the NFB to reduce distortion at higher frequencies.
But I am now wondering whether amplifiers are being designed with adequate NFB factors and sufficient error amplifier accuracy at high frequencies to take advantage of this.
By some coincidence, I have a design on the drawing board with excessive OLG so I may do some experiments in this direction. It's essentially a JLH amplifier topology, so it has the benefit of a Singleton (as Anatech calls them) error amplifier, although doing the experiment properly means making a variant with low OLG and low NFB. I'ts perhaps a rather old experiment that many of you have already done, but I guess each of us has to hear it for ourselves at some time. If all goes well, it will be built sometime over the winter, when Class A brings fringe benefits to homes in the Great White North.
Last edited:
Often the concept of "introducing new distortion products" makes no sense though. Mathematically, if we have a static nonlinearity and can express the input-output relationship as a polynomial, we can by inspection observe the highest-order term of the polynomial and conclude that's the maximum order of the harmonic distortion product. Consider the tanh() function that shows up in BJT diff amp analysis. It has a power series representation, the Taylor series, but that thing has an infinite number of terms. So it generates an infinite number of harmonics. Even though the ones of very high order may be extremely difficult to measure, they are there nonetheless. Therefore the concept of "introducing new distortion products" makes no sense in this case, because a simple diff amp already generates an infinite number of odd harmonics. I haven't tried this, but I doubt that feedback would introduce even harmonics to the mix.
This is by no means an isolated example either. Ever do a DC sweep of an optimally biased class AB output stage in simulation and try to fit the resulting data to a polynomial using least-squares techniques? I tried and gave up at order 9. The results weren't even close. I don't have proof of this, but I've concluded that it's going to take a huge number of polynomial terms, possible infinite, to fit the curve properly. This is due to having small, fine-grained wiggles near the zero crossing (but not discontinuities), in effect straightening out to a gradual bend as the output voltage and current get larger. You can simulate the effects of feedback on this circuit as Bob originally did, producing a Baxandall-like plot of relative harmonic levels vs. feedback. The results of that analysis, found in the Cordell feedback thread, are counterintuitive. Even small amounts of feedback reduce all harmonics, and larger amounts of feedback reduce them more, in a monotonic fashion.
The conclusion is that the effect of feedback on individual distortion components is rather subtle to grasp, highly nuanced and sometimes counterintuitive, depending a great deal on the nature of the open-loop nonlinearity. I'm reminded of a saying by H.L. Mencken. "For every complex problem, there is a solution that is simple, neat, and wrong." Just replace the word "solution" with "explanation" and you describe much of what's been said here about feedback and distortion.
Last edited:
This is in accordance with partly of my observations, described here about post # 292.
Some statements do change off-topic in the meantime - regarded the NFB actually this is the right thread:
http://www.diyaudio.com/forums/soli...-nfb-negative-feedback-me-wrong-question.html
on the other hand, the clear assignment is problematic because of the numerous overlaps
Last edited:
Time is not anymore a problem in audio electronic, either analogical or digital (the so called "jitter problem", in the form as often publicized in audio reviews, is a not problem at all...).
This is indeed rubbish. The time domain is the most understudied one in all audio analysis. Thanks to Fourier.
And jitter is one of the biggest issues of digital reproduction. Have you ever asked yourself why Reel tapes sound better? No A/D and D/A conversion = most of the damages to the signal are avoided.
This is indeed rubbish. The time domain is the most understudied one in all audio analysis. Thanks to Fourier.
And jitter is one of the biggest issues of digital reproduction. Have you ever asked yourself why Reel tapes sound better? No A/D and D/A conversion = most of the damages to the signal are avoided.
Dear Troll,
You definitely need to go back to school - any basic EE Circuits and Systems course will tell you that any time invariant non-linear system is unable to generate non-harmonic components.
Otherwise, keep feeding here with your junk. As other mentioned, you are an invaluable deposit of confusions, misunderstandings, urban legends, with a certainly entertaining value.
Member
Joined 2009
Paid Member
In terms of entertainment value, check this out. It's a novel use of output devices if you ask me.
An interesting possibility opens up if the output devices driving the speaker have a different Ft than those driving the feedback loop.
The ALTMANN SPLIF Amplifier Topology
An interesting possibility opens up if the output devices driving the speaker have a different Ft than those driving the feedback loop.
The ALTMANN SPLIF Amplifier Topology
Lumba we sorted out the "other" frequencies just go back and read the thread. Someone should post an update of Otala's graph with all of them labeled. Please note, Otala himself did not instigate the absurd interpretation of his data.
Telstar,
This is not quite clear to me. It is not about noise floor, it is very importantly about harmonic spectrum.
That's the reality we are facing. The the ear`s distortion is not a disadvantage in sound reproduction.
No doubt. Several studies have been done showing that. The high frequency range must be in good shape for good performance in the audible range. Wide bandwidth is one condition of low high frequency distortion.