So a cable is a better filter than a... Filter. And whats ground conducted noise and how does a power cord deal with that?
Last edited:
They are delusional and either inexperienced or willfully ignorant to what the rest of the world of electronics is doing.
My usual question: what is your understanding of "noise floor modulation" in the context of a DAC? I know a spectrum analyzer noise floor can be "modulated", if the instrument has not enough dynamic range for the measurement, I know the noise floor can be in general modulated by a carrier single frequency, but I have no idea how to modulate the noise floor with a random (music) signal, of course without running the system into non linearities (like for example clipping). There is an old paper claiming that for multi-tone testing with a large number of tones, the IMD distortions could be much larger than the sum of individual distortions, but I don't think this is really anything outstanding, think of crest factor and the effect of clipping on IMD distortions and the associated noise floor "grass" (if that's what you think of "noise floor modulation").
Please clarify, or somebody may assume you are (again) making up pseudo-technical arguments. Note I did not even went into your second claim yet, the audibility of this "noise floor modulation", before understanding what it is, in your view.
Cortez, who are you asking? fagos measures the group delay with a ruler according to the frequency response and phase response, but he just flatters me by saying that I measure the THD spectrum with a caliper.
Normally we measure the distortion at the output of our amp, and the (implicit) assumption is that the input signal is free of distortion (or at least much lower than the amp distortion).
With a suddenly starting signal, there are many harmonics in the input signal, so when you measure the output of your amp, the task is to know which harmonics are from the input signal, and which are generated by the amp.
I can think of three ways to attack this:
1 - in some way, compare/subtract the input from the output, but the difficulty here is to exactly line up the two signals, normalize the levels and subtract a (complex) signal. Probably not doable with our equipment today, although there might be some digital processes that could do it like convolution;
2 - wait a relatively long time to make sure that the input spectrum has died out, then measure distortion;
3 - make sure that the input signal doesn't start suddenly by passing it through a low pass filter with a cutoff freq. just above the test signal frequency. That has the benefit that it also cleans up a bit the actual test signal.
I know you actually asked about measuring FCD but I don't know for sure what that is, what its definition is, so I can't answer it at this time.
If you tell me how FCD is defined, I will try again. It can be very complex, if you for instance define FCD as the distortion that the amp generates from the many harmonics in a suddenly starting signal, you are in effect doing a multitone measurement, but without knowing the input spectrum in detail.
Jan
Last edited:
Jan, you translated article by Hiraga, the following questions for you:
can you name the brands of amplifiers that were in the test?
What is the FSD level of the first amplifier and what is the second one?
Why does the second amplifier with TCD = 0.008% sound ugly?
Why does a high distortion tube amplifier have such a low RIMD (Reverse Intermodulation) noise floor? Much lower than an amplifier with THD = 0.008%
Jan, there is no distortion in the suddenly-onset signal at the amplifier input, they are only at the amplifier output!
and I'm tired of explaining why
can you name the brands of amplifiers that were in the test?
What is the FSD level of the first amplifier and what is the second one?
Why does the second amplifier with TCD = 0.008% sound ugly?
Why does a high distortion tube amplifier have such a low RIMD (Reverse Intermodulation) noise floor? Much lower than an amplifier with THD = 0.008%
Jan, there is no distortion in the suddenly-onset signal at the amplifier input, they are only at the amplifier output!
and I'm tired of explaining why
Attachments
Last edited:
Petr said: 'Jan, there is no distortion in the suddenly-onset signal at the amplifier input, they are only at the amplifier output!'
Correct, it is not distortion. But it is a lot of harmonics, caused by the wide spectrum of the modulating unity pulse. You showed it yourself earlier in this thread.
So as I said the task is to look at the amp output and decide whether the harmonics are from the (suddenly starting) input signal, or from the amp.
Thinking about it a bit more in the context of a multitone test, in a multitone test you select the frequencies so that the amp harmonics fall in FFT bins different from the input frequencies. In that way you can see what frequencies are from the input signal and which are from the amp distortion. With the spectrum from a suddenly starting signal, you cannot do that. I think.
Question to the signal processing pros: are the harmonics of a suddenly starting sinewave harmonics of the sine wave or is it more complex?
Jan
Correct, it is not distortion. But it is a lot of harmonics, caused by the wide spectrum of the modulating unity pulse. You showed it yourself earlier in this thread.
So as I said the task is to look at the amp output and decide whether the harmonics are from the (suddenly starting) input signal, or from the amp.
Thinking about it a bit more in the context of a multitone test, in a multitone test you select the frequencies so that the amp harmonics fall in FFT bins different from the input frequencies. In that way you can see what frequencies are from the input signal and which are from the amp distortion. With the spectrum from a suddenly starting signal, you cannot do that. I think.
Question to the signal processing pros: are the harmonics of a suddenly starting sinewave harmonics of the sine wave or is it more complex?
Jan
Last edited:
Sorry, I didn't quote, the question was asked from fagos
as he said you are measring the thing with the wrong tool.
Thanks Jan!
For me (at this point on the topic) FCD could be mean 2 things (a + b) but with 2 common criteria (1 + 2):
1) let the system settle to a calm state without any drive at the input
a) let's stimulate the amp with a step kind of "sine" signal starting from 0 to check the transient behaviour
(I know but yes still even if it's not a typical music kind of signal)
or b) apply a slower (full standard audio band speed) starting but a high level signal
(like the guitar string plucking as I linked a few posts earlier)
and check the response but from the very beginning 0 time point
2) Then we record/measure/check this exact one "period" (time section) output
response immediately after this stimulation and directly without any:
- waiting
- time windowing (taking multiple "periods")
- averaging
My assumption is that there can be some "nasty" transition artifacts after
a bigger (and faster) change at the input signal but only at the very beginning
which we can "miss" if we wait, average, etc.
Even if no one "belives" just for the sake of the game/topic let's
assume it can be true: so then how could we check/measure this?
I'm not sure that what was Graham's original idea from options a) or b)
maybe rather a) and b) is just my subjective interpretation which is a completely other topic.
Option a) focuses rather on the signal speed while b) focuses on the "state change" of the
whole system, the speed and cleanness of this setllement process but from the very beginning.
For example when we do some simulation often we have to wait untill the whole circuit settles
to its steady DC state operating point and then we can start to measure it with a test signal.
I guess the same is happening (but of course a much smaller level) when the average
"energy level" is changing during a musical signal.
I don't know whether I described this in an understandable way...
Last edited:
Cortez said: 'My assumption is that there can be some "nasty" transition artifacts after a bigger (and faster) change at the input signal but only at the very beginning which we can "miss" if we wait, average, etc.'.
That will certainly be the case, after all you are hitting the amp with a very broadband signal, much wider than audio, and an audio amp is not designed to handle such signals, so it will be a mess at the output.
If you can somehow measure it, maybe with a 1GHz bandwidth spectrum analyzer, how to interpret it, what to call it? It seems obvious that you should not call it 'amplifier distortion', that would give the wrong idea that the amp is distorting the signal itself. (Which it does of course, but that is drowned in the wide spectrum).
And I still cannot understand why anybody would want to test an audio amp with a non-audio signal and then draw any meaningful (in audio terms) conclusion from it. It is a completely artificial situation and after all this years we still haven't understood that it doesn't bring anything useful.
You can go back to the beginning, to Graham, and now to Petr, and the basic missing of the point that a suddenly starting signal is extremely broadband and not an audio signal. But, once you have dug yourself in, the only way forward is down deeper into the mess.
And then, when you run out of technical arguments in this technical discussion, you can confuse it all by 'I have amp x with p.pp% THD, and amp Y with q.qq% THD, and amp X sounds better to me!", which of course in the context of this discussion is totally irrelevant.
And so it goes.
Jan
That will certainly be the case, after all you are hitting the amp with a very broadband signal, much wider than audio, and an audio amp is not designed to handle such signals, so it will be a mess at the output.
If you can somehow measure it, maybe with a 1GHz bandwidth spectrum analyzer, how to interpret it, what to call it? It seems obvious that you should not call it 'amplifier distortion', that would give the wrong idea that the amp is distorting the signal itself. (Which it does of course, but that is drowned in the wide spectrum).
And I still cannot understand why anybody would want to test an audio amp with a non-audio signal and then draw any meaningful (in audio terms) conclusion from it. It is a completely artificial situation and after all this years we still haven't understood that it doesn't bring anything useful.
You can go back to the beginning, to Graham, and now to Petr, and the basic missing of the point that a suddenly starting signal is extremely broadband and not an audio signal. But, once you have dug yourself in, the only way forward is down deeper into the mess.
And then, when you run out of technical arguments in this technical discussion, you can confuse it all by 'I have amp x with p.pp% THD, and amp Y with q.qq% THD, and amp X sounds better to me!", which of course in the context of this discussion is totally irrelevant.
And so it goes.
Jan
Last edited:
Then I ask Petr: is it necessary (for you) to use this step kind of sine wave or is it
enough to use jsut a very fastly increasing signal started from a small sine wave?
Like I draw earlier? (attached again)
as it's not the normal drive/load yet it can give some extra (and usefull) info about the behaviour of our amplifier
that wouldn't come up when testing it with a max 20kHz sine wave and a perfectly resistive load, that's all.
But the specialty about this "first cylce" idea (for me at least) is that it focuses on the very
early reply and not on the averaged response after some delay as with real music it's also the
requirement that the system should perform perfectly from the very first stimulations and not just
after a settling time and on an avareage level.
enough to use jsut a very fastly increasing signal started from a small sine wave?
Like I draw earlier? (attached again)
I guess it's similar case when doing a test using a continouos square wave (and/or even some capacitive load)
as it's not the normal drive/load yet it can give some extra (and usefull) info about the behaviour of our amplifier
that wouldn't come up when testing it with a max 20kHz sine wave and a perfectly resistive load, that's all.
But the specialty about this "first cylce" idea (for me at least) is that it focuses on the very
early reply and not on the averaged response after some delay as with real music it's also the
requirement that the system should perform perfectly from the very first stimulations and not just
after a settling time and on an avareage level.
Attachments
Let's imagine this might matter, what is the temporal response of your hearing?
What I meant was that even a normal music signal contains a lot of "random" and high transient
changes (signal / energy levels) and the amplifier should follow it is close and clean as possible.
And a continuous sine wave test (even if it's higher then 20kHz) doesn't represent this well.
Specially when we measure the distortion after a time and on a lot of periods and doing an averaging.
Of course I don't know but that's true for the opposite possibility as well.
Maybe it counts a lot... Or you can guarantee 100%-ly that it doesn't matter?
Then why we do all our developments after the age we got our 0.000x THDs?
There is always something to improve I guess, then why not this aspect?
This is a misunderstanding. Random and high transients are just a bunch of coming up and decaying sinewaves within the audio bandwidth. If an amp does well on the signals in the audio bandwidth, it does well on music. The amp doesn't know what it gets, it just sees a continuously changing level. As long as that changing level stays below clipping and below 20kHz 'speed', there is no difference in this respect.
The difference that is there is that if two or more frequencies are present in the signal, they may intermodulate and create sum and difference frequencies. That is just another manifestation of non-linearity, just as HD is a manifestation of the same non-linearity, but measured in a different way.
But as far as the frequency domain is concerned, a single sine wave or music doesn't make a difference for the amp.
Jan
Last edited:
It's easier to try to solve it then start a new "profession".
Could you please be more specific? In what context you are asking the question?
Of yourse I know that a step signal contains an (infite in theory) fast
component which couldn't be a real requirement for an audio amplifier.
This is want you meant?
I think it could, but this also is just an assumption of course.
My guess is that when there is a bigger and faster change there could be more
transient artifacts / side effects produced as the system "slips" a little bit.
Like the overshoot on a square wave but not directly on the signal but on the
"state" of the system which can modulate the amplification, the feedback, etc.
I mean something like the memory distortion or the heat of the components for example.
These are relating mostly to the average energy level of the system.
As we need some time after swtiching on the amplifier to achieve the normal
state (DC currents, temperatures, etc) the same is happening during a signal.
When driving it with a higher level of signal the whole system should be "warmed up" while the signal is going on/further.
That's why I like the FCD as it (maybe) can also cover this effect while a continuous sine based test can't.
That's why I'm not focusing "just" on the frequency reponse when stimulating with a sudden test signal but on the "initial state" change.
Once again the base question (for me) regarding the FCD:
What if there is some transient/temporary distortion with non-periodic highly
changing signals but only at the begining because the system settles in the first few periods..?!
(Let's assume we can hear it...
)
In the context of music and our hearing. The bandwidth is far lower that infinite, do you have an idea of how low it is?