AK4499 eval board data sheet has plots, compare to any other DAC data sheet 0dBFS vs -60dBFS.
You can work it out.
T
Last edited:
Hi Scott, have had a really good day and nothing is going to spoil it.
... as you want it to be.
But I hope you noted and understood the hypothesis, it is quite a real one.
Joe, much thought and study and measurement has gone into current drive for loudspeaker drivers.
Current-Drive - The Natural Way of Loudspeaker Operation
AES E-Library
I think some measurements have been posted on DIYaudio, also by me, comparing distortion between voltage and current drive. Recently, Pavel showed that the effect of current drive on intermodulation distortion might be more beneficial than on HD proper. And there exists a theoretical foundation behind how current drive can improve distortion caused by Bl fluctuation.
It is not a hypothesis, it is a fairly well researched topic. If you would study more, you might also improve your appropriate use of technical terms btw.
Current-Drive - The Natural Way of Loudspeaker Operation
AES E-Library
I think some measurements have been posted on DIYaudio, also by me, comparing distortion between voltage and current drive. Recently, Pavel showed that the effect of current drive on intermodulation distortion might be more beneficial than on HD proper. And there exists a theoretical foundation behind how current drive can improve distortion caused by Bl fluctuation.
It is not a hypothesis, it is a fairly well researched topic. If you would study more, you might also improve your appropriate use of technical terms btw.
Yes, it may be best to make it pseudorandom. However, even then you need to have some idea of what you are looking for in the output.mmerrill99 said:
I thought it had been known for a long time that signal-correlated noise might be more audible than random noise, although masking may play a role too.
Relevant, maybe; very significant, I doubt it.
You have to remember that real music made by real instruments played by real musicians already has various noises associated with the music. For example, most musicians move around while they are playing so this will modulate any 'noise' output from the instrument. Musicians themselves make noise e.g. as they move on their seats. Any playback noise is likely to be smaller than these performance noises so can probably be ignored.
I will disagree in the sense that I don't think this is what is being asked. Take the 32 tone multi-tone, by tuning the phase you can adjust the crest factor over a large range in fact for a given rms value (which is constant) you can have a well behaved system or one that is hard clipped. If you sat down and actually looked at the time domain signals possible from the same 32 tones you could convince yourself that some plausibly mimic these "music transients" people talk about, but in this case they fully obey Nyquist.
What MM proposes is not far fetched and IMO worth exploring at least out of intellectual curiosity. Create multi-tone time records with a wide range of crest factors out of the same tones and look at the result. You might see nothing of note or you might not be able to interpret what you see.
Hi Scott, have had a really good day and nothing is going to spoil it.
Not my intent, I can only assume that real measurements would only dilute your story. I remain disappointed that you still refuse to use basic complex algebra like 99% of the rest of the engineering community and continue to use your own personal made up terminology that is often wrong and meaningless to the rest of us.
Last edited:
You have completely misunderstood, maybe read it again because I suspect you just skimmed it.
This is not about current drive versus voltage drive, period!
In fact, read it again and you will see that current drive is not mentioned at all, in fact the whole topic was about voltage drive.
I am not interested in current drive. I have now said this numerous times, but getting used to not being heard.
Indeed, it needs some thought
And that is why I revert back to Audio Scene Analysis - it is the study of how we analyze the multitude of nerve impulses into audio streams - in other words we collate together the sounds from each instrument into a specific stream which we can focus on in isolation from other instruments/sounds. Any noise from shifting musicians is analyzed as NOT part of /is outside any audio stream - same as room noise/reverb/noise floor on recording, etc.
It's all got to do with HOW auditory perception works!!
What I think might be happening with noise floor modulation (NFM) is that this signal correlated noise can get analyzed into /confused as part of a specific audio stream. Perhaps, when NFM is present, it precedes the attack portion of sounds, particularly transients, leading to a perception of a less precise start to the transient elements in the music. This is not such a big deal when taken in isolation but when this perceived 'timing haze' in the sound is removed, it can lead to a much more natural, precise & interesting perceived presentation.
This can be perceptually significant for some/many people but we are so used to it that we don't consciously perceive it unless/until it's removed?
I know the next question is how do we know it is removed. That is what I'm trying to find - a measurement for analyzing this - at the moment my best guess is to use a multitone test signal with pseudo-random amplitude modulation of each of the tones
Last edited:
What I meant is that sine signal permanently changes amplitude from -Vp to +Vp with the defined frequency, so it has one value at defined time and the values are periodically repeated. So, the system has to respond and settle again and again. Only a screen time view is "frozen". As you already stated, it is the FFT setting that makes it averaged and apparently non-changing. But with the proper time and FFT length setting, windowing and averaging is not needed and with the signal like Demian's (or sine like 997Hz/44.1kHz) we can see that spectrum is not "steady state", it is changing as the system is responding to input changes. It is the averaging that makes it look "smooth like".
Why isn't correlated noise "ignored" in the same way as correlated reflections?
Sorry, but nonsense.
You've stated that "steady state" means the same as "time invariant" (your post #24521) and therefore I "had the nerve" to ask you for relevant references.
And I've quoted from the relevant literature (post #24647) :
Which shows that wrt LTI systems the terms "time invariant" and "steady state" are in no way synonyms.
Sorry, but nonsense again. It looks very much like projection as you've started the "stupid" game by claiming that:
can't be described as "steady state measurement" using sine waves. Even calling it "FUDing" when describing as such.
That wasn't the point, as I was referring explicitely to the usual measurements using distortion analysers or voltmeters.
Further it often helps to visualize certain properties in a different way, otherwise it would be sufficient to provide an impulse response/unity step response as it contains all the informations needed. (provided that our assumptions about sufficiently "weakly nonlinearity" and "mainly time invariant" hold true)
Last edited:
You have done this - you have such a dynamically changing multitone signal? Can you post a download link please - I would be most interested in using it?
Or are you guessing what the result will be?
So, if you say this is what you expect, noise floor modulating with the signal - as this is what all systems do, it would be interesting to see how some very low IMD amplifiers (Benchmark ABH2 & Purifi) perform in this test Vs 'normal' IMD amps & does it correlate to perceived sound quality?
I'm thinking that V.low IMD devices will show low NFM but I'm interested in how such IMD/NFM might be audible as the IMD tests (& even Demian's Triband test) show frequency products that are normally considered way below any possible audibility - so what gives - is the answer to be found in the NFM test I'm suggesting?
I don't yet know what the characteristics of such NFM might be but it is all to do with how auditory perception does it's job of splitting the nerve impulses (which have arisen from the sound at the eardrums) into a meaningful representation of the outside world of sound producing objects.
Auditory perception uses technique such as location from which the sound emanated, correlation of amplitude/frequency fluctuations & many other correlation techniques to categorise individual sounds into the sound stream from instrument A or instrument B. NFM may be so close to the signal stream's characteristics that it is analyzed & considered part of the same stream & is perceptually integrated into the audio stream
What is your definition of a "dynamically changing multitone signal"? According to you, is this a steady-state signal?
Attachments
@ syn08,
you objected to my assertion:
by stating:
Of course, one can blame me having not written "describing a condition of LTI systems" instead of "describing LTI systems" , but nevertheless it should have been very clear that "by _definition_ _excludes_ _transient_ _processes." refers to the term "steady state" but not to the definition of LTIs
Correct and I did not wrote something different.
As you're mainly responding to your bias driven interpretations it is just wasting a lot of time again and again.
I haven't claimed that steady state is used in defining LTI systems quite to the contrary.
you objected to my assertion:
by stating:
Of course, one can blame me having not written "describing a condition of LTI systems" instead of "describing LTI systems" , but nevertheless it should have been very clear that "by _definition_ _excludes_ _transient_ _processes." refers to the term "steady state" but not to the definition of LTIs
Correct and I did not wrote something different.
As you're mainly responding to your bias driven interpretations it is just wasting a lot of time again and again.
I haven't claimed that steady state is used in defining LTI systems quite to the contrary.
If I understand you correctly, you were stating that (one of the reasons) ABX tests often return negative results could be/is given by the time variance of the reproduction systems.
I just tried to point out, that we always are restricted to listen consecutively - when comparing something - and that we therefore are facing the same problems (i.e. resulting from any time variance) no matter if listening in an ABX or any other test situation (including "sighted" listening).
IMO he speaks about intersample overs, the issue known for long time and solution is headroom of digital reconstruction filters.
I was asking about your steady state and dynamic signals definition. To my best knowledge such definitions are not used and signals are divided as per several attached pages, original literature (book) is
Randall, R.B.: Frequency analysis, ISBN 87 87355 07 8
I do not think that this forum discussion will change my view of signal types.
Last edited:
Not at all, no reconstruction is involved, I need pictures. Give me an evening of fiddling to make some examples, I admit I was more motivated when it was my job but I think I can muster up a few examples.
Thinking that you can take the 32 tones of a multi-tone and give them random phase and get essentially the same result for crest factor every time is simply wrong.
Besides being total nonsense, why do you think people are confused?
- Status
- Not open for further replies.