myhrrhleine said:
A great deal of work in this regard has been done in the digital audio arena, where concern about the LSB noise has been considered. Of course, the LSB noise is not simply white noise that is uncorrelated with the signal. For this reason, it has been found that quantization noise in digital systems can be audible even if it is below the system's white noise level.
Since distortion components are correlated with signal, it is reasonable on the same basis to believe that they will be audible even if they are below the noise floor of a system.
Cheers,
Bob
LSB noise, also called quantisation distortion, is indeed correlated to the signal and sampling frequency. It gives a bunch of low level spurs. The remedy for this is to ad dither, a white noise signal with a gaussian or triangular distribution and an amplitude of 1 LSB.. Dither noise must be applied before the quantisiser (ADC) to take effect. The dither randomises the spurs and spread their energy evenly over the whole audio band. The result is a slight increase of wide band noise but the spurs are vanished. Another effect is that the resolution limit of 1 LSB is removed and signals at lower levels than 1 LSB can be resolved.. Nowadays nearly all CD’s are dithered.
janneman said:
Yes Jan,
Most interesting things happen when subjecting hearing to various effects! I think though that in the case you mention the signals are rather louder than threshold of hearing. You might comment.
One other interesting phenomenon of many that I recall, is the case where a piece of music is played with a sharp rejection filter somewhere in the audio band. If the filter is sharp enough, the effect is not really heard. But if the filter is cancelled after having been active for several minutes, a very audible peak is suddenly evident at that frequency!
Jcx,
Thanks for that reference. These folks certainly have something to say, which could also be relevant to analogue only response. It would appear that at least some of the resistance to CD music arises from background effects due to the digital process, that have not been sufficiently investigated.
Question for John Curl
Hi John,
I have a question for you about the JC-1.
Assuming that the NFB is about 34 dB at LF, and about 20 dB at HF, with a 4 kHz open-loop bandwidth, why did you not choose a 20 kHz open-loop bandwidth that would have yielded 34 dB NFB all the way out to 20 kHz?
Would the amplifier have sounded better under these conditions? I realize that this would imply more NFB at 20 kHz, and that you shy away from more NFB, but you already have the more NFB at low frequencies. Besides, I would guess that you would feel that the increased open-loop bandwidth would benefit the amplifier.
I realize that pushing the open-loop bandwidth out to 20 kHz and increasing the 20 kHz NFB to 34 dB would increase the gain crossover frequency to about 1 MHz, but I would think that your use of the fast Sanken RET output transistors would enable you to safely go out to 1 MHz.
I'm not trying to suggest you change your amplifier. I'm just hoping you will share with us your rationale and experience here.
Thanks,
Bob
Hi John,
I have a question for you about the JC-1.
Assuming that the NFB is about 34 dB at LF, and about 20 dB at HF, with a 4 kHz open-loop bandwidth, why did you not choose a 20 kHz open-loop bandwidth that would have yielded 34 dB NFB all the way out to 20 kHz?
Would the amplifier have sounded better under these conditions? I realize that this would imply more NFB at 20 kHz, and that you shy away from more NFB, but you already have the more NFB at low frequencies. Besides, I would guess that you would feel that the increased open-loop bandwidth would benefit the amplifier.
I realize that pushing the open-loop bandwidth out to 20 kHz and increasing the 20 kHz NFB to 34 dB would increase the gain crossover frequency to about 1 MHz, but I would think that your use of the fast Sanken RET output transistors would enable you to safely go out to 1 MHz.
I'm not trying to suggest you change your amplifier. I'm just hoping you will share with us your rationale and experience here.
Thanks,
Bob
Hi Bob,
regarding new professional products, I am quite surprised when I see NFB designed like this:
http://stereophile.com/solidpoweramps/207classe/index3.html
(especially Fig.5)
Would you have any explanation why a designer made it this way?
Thanks,
Pavel
P.S.: probably not only NFB problem.
regarding new professional products, I am quite surprised when I see NFB designed like this:
http://stereophile.com/solidpoweramps/207classe/index3.html
(especially Fig.5)
Would you have any explanation why a designer made it this way?
Thanks,
Pavel
P.S.: probably not only NFB problem.
PMA said:
Hi Pavel,
In fairness to this amplifier, it is not a lot worse than many other amplifiers, and is better than many, be they with or without NFB (at least as far as the Figures 5, 8 and 9 go).
It is doing THD-20 of 0.03% into 8 ohms and 0.05% into 4 ohms at 17W into 8 ohms. Not great, and we could do much better, but not terribly bad compared to most of the competition. I'd like to see the THD at higher power levels, of course. The fact that the distortion rises as the frequency goes up is not uncommon, and is partly a reflection of the fact that their NFB keeps it very low at lower frequencies. If this were a so-called low-feedback amplifier, the distortion across the whole band would be almost as high as the THD-20.
Yes, crossover distortion is plainly visible, but keep in mind that the plot is not to scale and that the measured value is about 0.006%. Yes, it is only at 1 kHz and would be likely worse at 20 kHz. That XO distortion would not "look" as bad if the amplifier were putting out 0.1% of second harmonic at 1 kHz, would it? I'm not apologizing for the amplifier's performance, just reminding everyone that it all has to be kept in perspective.
Now look at my favorite chart, the twin-tone test in Figure 9. This is not bad compared to many other amplifiers. What I immediately look for is for the sidebands to be down at least 100 dB. With the exception of the third-order sidebands at 18 kHz (about -90), this amp just about achieves that. However, it does show sidebands out to 13th order (at 13 kHz) down only by 105 dB, and I would like to see these higher-order components disappear more quickly. This is most likely the signature of the crossover distortion.
Also note that the sidebands about the 19+20 kHz signals do not have "dirty skirts". Dirty skirts are when you see other distortion spectra climbing up the sides of the distortion spectral lines, making them appear fatter near the bottom. Dirty skirts are an indication that more nonlinearity is going on. We want to see clean, well defined, narrow distortion spectral sidebands on either side of the stimulous.
Notice also that second order nonlinearity at 1 kHz and 4th order at 2 kHz are also at -105 dB, which is quite good, and reflective of the action of the presumed higher amount of NFB at the lower frequencies.
This is an unremarkable amplifier, but certainly not terrible, in my opinion.
Cheers,
Bob
Bob said it.
Just to add then, that like PMA, I am also sometimes worried about the rising distortion, indicating a questionable phase angle involved. What is going to happen with a loudspeaker load? As said before, we compare performances and invariably favour the best thd, irrespective of what will be audible (as Bob said: Perspective). I would prefer an amplifier with rather 0.03% of constant distortion (no phase angle) and wider open loop response than is indicated here. It is also a question of what is being generated intermodulation wise above 20 KHz. I have too often seen this sort of NFB pattern (yes PMA, it occurs quite often) achieved in order to be able to quote impressive figures at 1 KHz, regardless of what happens higher up. I am not accusing this manufacturer - as Bob said, reassuring two-tone performance.
Perhaps also, distortion % can be misleading; absolute figures for cross-over products would be more informative. Percentage of what? Referred to full output, quite audible cross-over products could still look acceptable.
Just to add then, that like PMA, I am also sometimes worried about the rising distortion, indicating a questionable phase angle involved. What is going to happen with a loudspeaker load? As said before, we compare performances and invariably favour the best thd, irrespective of what will be audible (as Bob said: Perspective). I would prefer an amplifier with rather 0.03% of constant distortion (no phase angle) and wider open loop response than is indicated here. It is also a question of what is being generated intermodulation wise above 20 KHz. I have too often seen this sort of NFB pattern (yes PMA, it occurs quite often) achieved in order to be able to quote impressive figures at 1 KHz, regardless of what happens higher up. I am not accusing this manufacturer - as Bob said, reassuring two-tone performance.
Perhaps also, distortion % can be misleading; absolute figures for cross-over products would be more informative. Percentage of what? Referred to full output, quite audible cross-over products could still look acceptable.
When I trim a bias for a power amp from 0mV (drop at RE), the distortion at the distortion-meter is declining. At 14mV it reaches the lowest distortion. But when I continue to raise the bias (17, 20, 25 mV at RE), the distortion dial is rising, the lowest is at 14mV. What happened here, what makes the distortion rises after 14mV drop at RE?
lumanauw said:
This is normal.
The crossover distortion that you are I suppose measuring is caused by variation in output resistance with load current level ( at low current).
The optimal biasing for minimal distortion requires that R * gm =1
(see Oliver's paper or my post 2162 in BJ vs Mos thread)
where R is the total ohmic resistance seen from the emitter towards the base of one output emitter follower.
This resistance is : Re + re + Rs/(beta +1)
where Re is the emitter resistor on which you measure Vq
re is the parasitic internal emitter resistance
Rs is the total base + source resistance
If Re is dominant then the condition becomes gm * Re = 1 which means Io * Re = Vt = 26mV. This is the usual condition ( Self)
because gm = Io/Vt where Io is the bias and Vt is the thermal voltage 26 mV at room temp.
If Re is made smaller to avoid losses, then re may not be neglected in front of Re and gmR=1 becomes gm ( Re + re ) =1
This means Io Re + Io re = 26mV and you are measuring the first term. ( see my post 2162 in BJ vs Mos). It is then normal that at optimum IoRe should be lower than 26mV.
JPV
JPV said:
I can confirm a similar experience with my output stage with the Sanken SAP16's. For zero bias vs temp drift, I should set bias current to 40mA which across the 0.22 ohms nominal Re's is about 18mV. However, min distortion is at about double that... Another compromise
Jan Didden
Nelson Pass said:
That's a subjective call. It is the experience of many that bias
current higher than the measurement "sweet spot" sounds
better. I don't doubt there are at least as many in this forum
who will disagree.
Interesting
An explanation could be:
Let the bias current Io be a little bit higher (at room temperature ) than what the condition gm*R =1 or equivalently Io*R=Vt = 26mV would require for minimal distortion, it means gmR>1. Let say Io*R=31mV which is Io 20% higher.
Vt increases by 86 microvolts/°C. Therefore a 60°C increase of the die temperature gets Vt from 26 to 31 mV, a 20% increase.
If Io is too high at room temperature and if the Vbe multiplier tracks well enough to keep Io constant with temperature, an increase in operation of 60°C brings gmR back to 1 which is minimal distortion.
It sounds better when it's hot
JPV
