Well said, Demian. Everyone, even though this is not my specialty, I KNOW that we have researched and debated this for the last 30 years. Some here should get up to date, by reading up, if possible, before we have to answer very early questions and assumptions.
While I realize that not everyone here has been in audio design, or on this thread for an extended time, please don't get miffed if we don't go overboard to explain everything once again from the beginning, just because you don't want to read the prerequisites on any subject that we discuss.
While I realize that not everyone here has been in audio design, or on this thread for an extended time, please don't get miffed if we don't go overboard to explain everything once again from the beginning, just because you don't want to read the prerequisites on any subject that we discuss.
Johnloudb said:
All i´m saying is that jitter in the digital domain doesn´t have any impact as long as the data remains unchanged.
Everythings changes if digital analog conversion is done.
It depends strongly on the clock reconstruction strategy of an d/a-converter unit, if an digital cable might produce any audible difference.
And of course via the cable two quite strong hf generators are coupled, that may lead to audible differences too.
Jakob2 said:
Let's cut it short. Deterministic vs. nondeterministic jitter maps precisely (in the frequency domain) to pink noise (bandwidth limited) vs. white noise. This can be mathematically proven, see Shannon et. al. Take a look at this: http://cp.literature.agilent.com/litweb/pdf/5989-9364EN.pdf
Now, the question precisely translates to: what is the impact of pink noise on the signal spectra? Of course the answer depends on both the signal and the pink noise bandwidths, and on the amplitudes, but this doesn't make the effect non harmonic. Only the white noise, having infinite bandwidth, doesn't impact the individual signal harmonics.
Jakob2 said:
Nothing to worry about as he writes next:
"It is one very good reason for passive bandlimiting at the input of every active amplifier stage. Otherwise, the effect is cumulative - and demonstrably ugly sounding. "
The same can also be said of power supply artifacts. PSRR generally degrades very roughly at about the same rate as the open loop gain in a fed-back amplifier. Power supply artifacts often extend up to beyond a MHz - easily measured with a spectrum analyzer. Won't at least some of this junk mix down to the audio band?
syn08 said:
Please refer to page 4 of the linked agilent pdf; on the right side of the diagram two mechanism of deterministic jitter are mentioned, first data dependent and second periodic jitter (for example sinusoidal jitter).
If there is deterministic (in this example sinusoidal jitter) jitter present above the audio band, it depends on the dac structure and the signal content in the audio band which effects it will have.
The effect of sinusoidal jitter above the audio band on a audio signal of 19 kHz will be an intermodulation product inside the audio band if the sinusoidal jitter is "appropriate" .
I feel that we have a misunderstanding at this point, but do not know which way to handle it.
Jakob2 said:
I don't see any misunderstanding; sinusoidal jitter maps in the frequency space to limited bandwidth noise around the fundamental. If this bandwidth overlaps with the audio spectra (which, in turn, depends on the signal content), you got harmonic effects. Otherwise, you don't get any effect. Yes, the DAC structure may interfere here as well. At the very limit case, if "music" is a single tone outside the sinusoidal jitter limited bandwidth, you got 0 (zero) harmonic effect.
Jakob2 said:
On a second thought, I think we have a misunderstanding here. Following your definition of "non-harmonic distortions", any intermodulation product between two uncorellated signals (one could be pink noise) leads to non-harmonic distortions, if it falls within the signal bandwidth. I disagree with this definition of "non-harmonic distortion", in fact, this concept could be as useless as the concept of "linear distortions". BTW, following your definition, is the 19+20KHz IMD test a measure of an amplifier's "non-harmonic distortions"?
Ad797
Hi Scott,
Are sure that this is the real schematic (AES preprint 3231, fig. 3a)?
There are a few things that puzzles me and/or makes no sense:
1. Q61, unless the bases of Q3 and Q4 are tied to the collector of Q61 instead of to the collector of Q7.
2. Under normal conditions Q50 is conducting, which spoils the whole thing. If I break the connection between the base of Q15 and the emitter of Q50 the circuit (almost) behaves as expected.
BTW, is Q50 meant to ease the recovery from overloading?
3. Bootstrapping the collector of Q19 instead of Q18. The latter makes more sense as no non-linear Cob loading (by Q18) of the input of the output stage.
Please, see below for my suggested mods and, of course, correct me if I'm wrong.
Regards,
Edmond.
scott wurcer said:
Hi Scott,
Are sure that this is the real schematic (AES preprint 3231, fig. 3a)?
There are a few things that puzzles me and/or makes no sense:
1. Q61, unless the bases of Q3 and Q4 are tied to the collector of Q61 instead of to the collector of Q7.
2. Under normal conditions Q50 is conducting, which spoils the whole thing. If I break the connection between the base of Q15 and the emitter of Q50 the circuit (almost) behaves as expected.
BTW, is Q50 meant to ease the recovery from overloading?
3. Bootstrapping the collector of Q19 instead of Q18. The latter makes more sense as no non-linear Cob loading (by Q18) of the input of the output stage.
Please, see below for my suggested mods and, of course, correct me if I'm wrong.
Regards,
Edmond.
Attachments
Wavebourn said:
Yes they are. This is just another non-linearity type (not even new, Self is mentioning about and its pretty famous in IC design, requiring placing the input stage devices on izotherms, etc...).
For the same reasons, an amplifier with excellent THD20 and very poor IMD 19+20KHz (or the other way around) doesn't exist.
Re: Re: Ad797
Before even looking at, I may fully agree with the conclusions. Its not even big news, it was previously reported that humas subjectively enjoy distortions (2nd harmonic?).
The problem is to objectively evaluate this kind of preferences and quantify in a general acceptable form. Then engineers could easily pick up and design according to these guidelines.
But I am afraid its going to be pretty hard to set such generally accepted guidelines, for controlled distortions content. As such, it is naïve to imagine that an amp with personalizable distortion content will ever be economically designed and manufactured.
dimitri said:
Before even looking at, I may fully agree with the conclusions. Its not even big news, it was previously reported that humas subjectively enjoy distortions (2nd harmonic?).
The problem is to objectively evaluate this kind of preferences and quantify in a general acceptable form. Then engineers could easily pick up and design according to these guidelines.
But I am afraid its going to be pretty hard to set such generally accepted guidelines, for controlled distortions content. As such, it is naïve to imagine that an amp with personalizable distortion content will ever be economically designed and manufactured.
1audio wrote
I think this is a very poor example. I believe that you are referring to spread spectrum clocks being used to to spread clock harmonics over a wider than usual spectrum in order to pass the FCC emissions test. Any decent spectrum analyser available in the past 20 years (apart from the cheap garbage based on tv tuners) can see this effect.
The spectrum anlyser is NOT at fault. The FCC test is based on the average signal level in a narrow sweeping band. This is not an example of inadequate equipment but of an outdated mandated test regime. Or rather a perfect example of commercial greed leading to some clever design methods that satisfy the letter of the law while comprehensively breaching the spririt.
I think this is a very poor example. I believe that you are referring to spread spectrum clocks being used to to spread clock harmonics over a wider than usual spectrum in order to pass the FCC emissions test. Any decent spectrum analyser available in the past 20 years (apart from the cheap garbage based on tv tuners) can see this effect.
The spectrum anlyser is NOT at fault. The FCC test is based on the average signal level in a narrow sweeping band. This is not an example of inadequate equipment but of an outdated mandated test regime. Or rather a perfect example of commercial greed leading to some clever design methods that satisfy the letter of the law while comprehensively breaching the spririt.
- Status
- Not open for further replies.