Jump to content

Power cables - how to measure


Guest rmpfyf

Recommended Posts



11 hours ago, rmpfyf said:

The problem is your methods not being sufficiently sensitive to these ends.

I think comments like this make a lot of confusion for those who are not well informed..... and are a big part of the nonsense about "we can't meassure things we can hear" brigade that is very pervasive in high-end audio.

 

"Sensitive" is a super misleading word.    The measurement equipment is many orders of magnitude more "sensitive" than what is relevant.

 

What the test setup misses, potentially.... is transient events which could be caused by one of two main things.

  • Fluctuations in outside variables.....eg.  the mains power, or external interference, which comes and goes
  • Change in performance related to the input signal.    Large signals, complex/high-slew signals.

 

You don't need a "more sensitve" method to investigate these things.   You just need to repeat the measurements, and use various input signals..... and monitor the variation in the results.

 

It's a bit like "myth busters" style.    Ok, so a 1khz (and others) don't show anything significant...... What does it take to show something significant (wrt. outside forces, or different inputs signals).

 

 

It might soudn like I am being a pedant, by picking on the wording (especially sensitive) .... but I think it is very important.   People, otherwise, are left with the idea that we must be able to hear "better" than the posted measurements .... and therefore the "sensitivity of the posted measurements is not high enough"  .... ie. we can hear things below 120dB down  (or, even 100, really).    It's nonsense.

 

 

I'm not saying you won't find anything.... I just think it's bemusing that this seems like breaking new ground.

  • Like 2
Link to comment
Share on other sites

On 24/10/2020 at 9:29 PM, rmpfyf said:

If I can measure an output from a change in power cord at my DAC - given everything between the IEC socket and the audio output - then the difference is ****** severe.

 

I'm very confused about what you've said here.

 

You could measure differences which are 140dB below the peak signal..... given a good measurement device.

 

This isn't ****** severe.   This is the opposite of that.

 

 

Link to comment
Share on other sites

  • Volunteer
12 minutes ago, davewantsmoore said:

 

I'm very confused about what you've said here.

 

You could measure differences which are 140dB below the peak signal..... given a good measurement device.

 

This isn't ****** severe.   This is the opposite of that.

 

 

 

For some context… -140 db is the equivalent of someone shouting at over 10 thousand kilometres away

 

image.png.362eb7d9370223df7a55a13655034372.png

 

https://www.audiosciencereview.com/forum/index.php?threads/the-shoutometer.2555/

Link to comment
Share on other sites

46 minutes ago, davewantsmoore said:

I wasn't aware of that.

 

Although I guess "easily" is not well defined here.

 

Define audibility criteria for jitter. Go.

Or audibility criteria for waveform shape.

 

33 minutes ago, davewantsmoore said:

I think comments like this make a lot of confusion for those who are not well informed..... and are a big part of the nonsense about "we can't meassure things we can hear" brigade that is very pervasive in high-end audio.

 

"Sensitive" is a super misleading word.    The measurement equipment is many orders of magnitude more "sensitive" than what is relevant.

 

What the test setup misses, potentially.... is transient events which could be caused by one of two main things.

  • Fluctuations in outside variables.....eg.  the mains power, or external interference, which comes and goes
  • Change in performance related to the input signal.    Large signals, complex/high-slew signals.

 

You don't need a "more sensitve" method to investigate these things.   You just need to repeat the measurements, and use various input signals..... and monitor the variation in the results.

 

It's a bit like "myth busters" style.    Ok, so a 1khz (and others) don't show anything significant...... What does it take to show something significant (wrt. outside forces, or different inputs signals).

 

 

It might soudn like I am being a pedant, by picking on the wording (especially sensitive) .... but I think it is very important.   People, otherwise, are left with the idea that we must be able to hear "better" than the posted measurements .... and therefore the "sensitivity of the posted measurements is not high enough"  .... ie. we can hear things below 120dB down  (or, even 100, really).    It's nonsense.

 

 

I'm not saying you won't find anything.... I just think it's bemusing that this seems like breaking new ground.

 

I think your arguments are the darling of people armed with a mic and scope. Let's put a front door on your argument here, Dave. 

 

'Sensitive' is misleading insofar as talking about sound pressure. We can certainly measure things we can't hear insofar as sensitivity to pressure. This gets misconstrued by many armed with a plot with dB in any axis as 'well we must be talking about sensitivity', 'jitter less than 120dB you can't year it' etc. Which, scientifically, is Bullshi-do, the way of bulls**t.

 

Here's an example:

 

15 minutes ago, sir sanders zingmore said:

 

For some context… -140 db is the equivalent of someone shouting at over 10 thousand kilometres away

 

https://www.audiosciencereview.com/forum/index.php?threads/the-shoutometer.2555/

 

A Bel is a unit of relative measurement. 140dB is a difference of 1*10^14 power ratio. That's it. Has f all to do with sound, kilometers or the like. 

 

As a relative measure of sound pressure then sure, that might hold true under whatever conditions it is calculated though that is not what a measure of relative power in spectra actually implies.

Link to comment
Share on other sites



38 minutes ago, davewantsmoore said:

 

I'm very confused about what you've said here.

 

You could measure differences which are 140dB below the peak signal..... given a good measurement device.

 

This isn't ****** severe.   This is the opposite of that.

 

Untwist your confusion, it's not hard.

 

If a power cord is creating differences (as in 'I can see it in a picture', 'its beyond the margin of measuring error') in spectra as posted by @kukynas (short run, limited resolution/averaging, wide windowing) as observed at the audio outputs of a DAC then the effect it is having is severe. Significant. Large. Etc. 

 

If the differences were down in the flys**t of -140dB somewhere well wide of the signal under evaluation it's likely an external effect. We're talking about data most relevant to the signal being evaluated. 

 

Please stop dragging arguments back to the corner space of 'but we can measure -140dB SPL and you can't hear it'. (a) I/we know and (b) it's irrelevant as an argument here. Don't feed the wolves. Appreciably they're very small yappy wolves but they already have all of the ASR forums as a playground, and I don't need to keep asking the mods to undertake an eradication over here.

Link to comment
Share on other sites

6 minutes ago, rmpfyf said:

Define audibility criteria for jitter. Go.

Perfect example.   No components added to the signal which are above (say) 100dB below.    That figure depends on gain structure through, and if you have less than perfect, then you will be amplifying them.... but this isn't the fault of jitter.

 

This leads to:

  • Using different test signals
  • Repeating tests under a lot of conditions

 

To try and find when things do pop up into the area where they can be audible.    One often overlooked area IMO is at low frequencies....    Under very complex sgnals is also very intersting.... not only multi (lots) tone, but under various data-dependant jitter possibilities (although specific to a certain DAC).

 

6 minutes ago, rmpfyf said:

'Sensitive' is misleading insofar as talking about sound pressure. We can certainly measure things we can't hear insofar as sensitivity to pressure. This gets misconstrued by many armed with a plot with dB in any axis as 'well we must be talking about sensitivity', 'jitter less than 120dB you can't year it' etc. Which, scientifically, is Bullshi-do, the way of bulls**t.

Nonsense.

 

If my max SPL in room is ~120dB at any frequency ..... and there are components in the signal which are 100dB quieter than this.... then they are below the threshold of audibility, even before we begin to factor in any potential for them to be masked.

 

9 minutes ago, rmpfyf said:

Please stop dragging arguments back to the corner space of 'but we can measure -140dB SPL and you can't hear it'. (a) I/we know and (b) it's irrelevant as an argument here. Don't feed the wolves. Appreciably they're very small yappy wolves but they already have all of the ASR forums as a playground, and I don't need to keep asking the mods to undertake an eradication over here.

 

I firmly believe that it's true.... and I don't understand how it's not relevant.

 

Link to comment
Share on other sites

1 minute ago, davewantsmoore said:

Perfect example.   No components added to the signal which are above (say) 100dB below.    That figure depends on gain structure through, and if you have less than perfect, then you will be amplifying them.... but this isn't the fault of jitter.

 

This leads to:

  • Using different test signals
  • Repeating tests under a lot of conditions

 

To try and find when things do pop up into the area where they can be audible.    One often overlooked area IMO is at low frequencies....    Under very complex sgnals is also very intersting.... not only multi (lots) tone, but under various data-dependant jitter possibilities (although specific to a certain DAC).

 

Dave - please contribute any agreed, scientifically-derived standard for as much. Not a scope for an experiment. Many have and continue to experiment here.

 

You're welcome at any stage to admit that no such standard exists and it's an evolving space. This is not SPL sensitivity, which is well defined.

 

2 minutes ago, davewantsmoore said:

Nonsense.

 

If my max SPL in room is ~120dB at any frequency ..... and there are components in the signal which are 100dB quieter than this.... then they are below the threshold of audibility, even before we begin to factor in any potential for them to be masked.

 

Nonsense yourself. 

 

'Components in the signal' = a 1kHz tone at 120dB with a 900Hz component at 100dB under is not the same, nor is it percieved identically, to two unique tones representing both compoenents. That's an oversimplification we afford ourselves for many purposes, though it doesn't fit here.

 

5 minutes ago, davewantsmoore said:

I firmly believe that it's true.... and I don't understand how it's not relevant.

 

We all firmly believe it's true Dave. You get no points for opining on the blueness of the sky nor the highness of the moon, dude. 

Link to comment
Share on other sites

  • Volunteer
40 minutes ago, rmpfyf said:

A Bel is a unit of relative measurement. 140dB is a difference of 1*10^14 power ratio. That's it. Has f all to do with sound, kilometers or the like. 

 

It's translating it into distance (much like the 'how much power do I need" calculators take into account how far you sit from your speakers).

 

In any event I've stayed out of this thread largely because of your swearing…. looks like that's not changing anytime soon… see ya

  • Haha 1
Link to comment
Share on other sites



1 minute ago, sir sanders zingmore said:

It's translating it into distance (much like the 'how much power do I need" calculators take into account how far you sit from your speakers).

 

I get it, though this is dB in context of time-domain SPL. Not necessarily what spectra implies.

 

1 minute ago, sir sanders zingmore said:

In any event I've stayed out of this thread largely because of your swearing…. looks like that's not changing anytime soon… see ya

 

 

If I diminish it to -140dB would that change your.... but your podcast name offends me too...

Ta ta.

Link to comment
Share on other sites

44 minutes ago, rmpfyf said:

 

Dave - please contribute any agreed, scientifically-derived standard for as much. Not a scope for an experiment. Many have and continue to experiment here.

 

You're welcome at any stage to admit that no such standard exists and it's an evolving space. This is not SPL sensitivity, which is well defined.

What are you talking about?.... a "standard for as much" ie. how far distortion components need to be under a signal to be audible?.... or something else.

 

I can explain it.... I just need to be sure I'm answering the right question.

Quote

'Components in the signal' = a 1kHz tone at 120dB with a 900Hz component at 100dB under is not the same, nor is it percieved identically, to two unique tones representing both compoenents.

I don't understand what you're trying to say here.

Quote

We all firmly believe it's true Dave. You get no points for opining on the blueness of the sky nor the highness of the moon, dude. 

My comment still stands.   I don't undrstand how it's not relevant.

Edited by davewantsmoore
Link to comment
Share on other sites

18 minutes ago, rmpfyf said:

I get it, though this is dB in context of time-domain SPL. Not necessarily what spectra implies.

What do you mean by that?

 

18 minutes ago, rmpfyf said:

offends me

I find all this very difficult to believe....  perhaps I don't get out enough.  ?

Link to comment
Share on other sites

3 minutes ago, davewantsmoore said:

What are you talking about?.... a "standard for as much" ie. how far distortion components need to be under a signal to be audible?

I don't understand what you're trying to say here.

My comment still stands.   I don't undrstand how it's not relevant.

 

Point is there's is no accepted limit for audibility of jitter, waveform shape, etc. I'm not suggesting 'anything goes'. Just pointing out that there's an evolving space here and criteria for audibility are not so simple. 

 

Pointing to audibility criteria for time histories of SPL as a way of interpreting components in spectra has some significant limitations. Take the previous exaxmple with the 1kHz tone and a 900Hz component - you couldn't hear the 900Hz tone in isolation. Though as a Fourier component of a tone you can hear - does it make that tone audibly different? The answer isn't 'no, because 120dB difference in spectral energy, and you can't hear -120dB SPL' - for it is not SPL. It's a ratio of relative energy in spectral content as computed by whatever implementation of Fourier methods employed in the analysis. Of course there certainly is a point below which it is inaudible but that isn't defined by the ratio of spectral energy in any components alone. In a broadband sense it can make for a very useful tool. If we were talking about the differences in power cords making for anything able to be captured by common applications of these tools (as intended for, let's call it 'relatively broadband' analysis), well... if it showed up that easily then I guess we'd all have fancy power cords. 

 

Let alone getting into the vagaries of how sensitive we might be on an amplitude and frequency basis orwhat, in a stereo context, channel-to-channel differences might make etc...

 

...point is audibility is evolving. I'd suggest this to be an application that plays closer to the limits than it does in the 'meat of it'. 

 

1 minute ago, davewantsmoore said:

I find all this very difficult to believe....  perhaps I don't get out enough.  ?

 

Being in Victoria I certainly don't get out enough ?

Link to comment
Share on other sites

15 minutes ago, rmpfyf said:

Point is there's is no accepted limit for audibility of jitter

Of course there isn't.    "Jitter" is a very abstract topic..... and every converter behaves differently when faced with specific jitter.   "Behave" in this instance means "results in a certain analogue output signal".     The output signal being what is actually audible, or not.

 

15 minutes ago, rmpfyf said:

, waveform shape, etc.

Sure there is.

 

It's just extremely complicated.....  as there are a lot of permutations.    Absolute levels and "slope" (ie. non-linear distortion)..... and both of these under a whole gamut of input signals (complexity, size, etc.).

 

.... but how much of these is audible, is reasonably well understood, with a few caveats (and I think, open secrets.... eg. errors correlated to a specific channel).

 

15 minutes ago, rmpfyf said:

criteria for audibility are not so simple. 

Indeed they are not simple..... but that is simply due to the potential possibilities in the inputs signals, and test conditions, which need to be evaluated before "proof" can be approached.

 

It's been ongoing forever.

 

15 minutes ago, rmpfyf said:

Of course there certainly is a point below which it is inaudible but that isn't defined by the ratio of spectral energy in any components alone.

What else is it definied by?

 

Other components (of course) ... and their levels.... and?!

15 minutes ago, rmpfyf said:

if it showed up that easily then I guess we'd all have fancy power cords. 

So, overay the unwindowed signals.

15 minutes ago, rmpfyf said:

Let alone getting into the vagaries of how sensitive we might be on an amplitude and frequency basis orwhat, in a stereo context, channel-to-channel differences might make etc...

Vagaries?

 

Of course, it is quite complex ......  but it is the place to look.   It is the signal that we hear.

Link to comment
Share on other sites



25 minutes ago, rmpfyf said:

Point is there's is no accepted limit for audibility of jitter, waveform shape, etc. I'm not suggesting 'anything goes'. Just pointing out that there's an evolving space here and criteria for audibility are not so simple. 

There's no accepted limit for the point at which newspaper headlines become illegible because of insufficient contrast, insufficient illumination, or lack of clarity in the font design or the spacing of the characters. However we readily accept that at a distance of 10,000km the headlines will not be readable by a naked human eye.

 

It is only when dealing with stimuli around the threshhold of human perception that it becomes important to tease out the precise limits of perception, and investigate further.

 

Obviously, time spent testing 100 subjects for whether they can read newspaper headlines at 9,000 km, 8000km, 7000km, 6000km and 5000km, is time wasted.  

Link to comment
Share on other sites

@davewantsmoore (you are the most exciting work I have today which says something) we essentially agree. There are no defined limits in some aspects of audibility, testing has been been going since time immemorial and will likely continue henceforth. 

 

15 minutes ago, davewantsmoore said:

What else is it definied by?

 

Other components (of course) ... and their levels.... and?!

 

Frequency, etc. When comparing two signals with an identical core tone and a component of additional energy at a differing frequencies (noting identical energyis not identical amplitude but is a funtion of frequency), we're talking different (in time domain being what we hear) waveform shapes.

 

17 minutes ago, davewantsmoore said:

So, overay the unwindowed signals.

 

Which brings us back to the proposed experiment. Two signals, time domain, deltas between them. Agreement!

Link to comment
Share on other sites

4 hours ago, rmpfyf said:

There are no defined limits in some aspects of audibility

I think that very much sends the wrong message.

 

It implies (or may) that "the teeniest tiniest thing could be audible, nobody knows"..... where as the majorty of the debate is back at the threshold of audibility (ie. that quite gross errors are not audible).

 

... but that's all acedemic.   Step 1, is collect data about the audio signal.

 

4 hours ago, rmpfyf said:

Frequency, etc. When comparing two signals with an identical core tone and a component of additional energy at a differing frequencies (noting identical energyis not identical amplitude but is a funtion of frequency), we're talking different (in time domain being what we hear) waveform shapes.

I already said that, when I said "other components" ..... ie. other (additional) frequency components.

 

ie.   non-linear disortion.

 

The different waveform shape is non-linear distortion (and potentially a level error mixed in).  Which is what has already been posted in this thread, and pooh poohed. 

 

As I said... it could be investigated further using differing inputs signals (as the one posted earlier was very simple), or methods which seek to uncover other sort of variabilities.

 

4 hours ago, rmpfyf said:

Two signals, time domain, deltas between them.

Yes.... this is essentially dropping the time window to the smallest possible.

Link to comment
Share on other sites

50 minutes ago, davewantsmoore said:

I think that very much sends the wrong message.

 

It implies (or may) that "the teeniest tiniest thing could be audible, nobody knows"..... where as the majorty of the debate is back at the threshold of audibility (ie. that quite gross errors are not audible).

 

... but that's all acedemic.   Step 1, is collect data about the audio signal.

 

You say wrong, I say nuanced. It's important to understand dB in the context of SPL and non, and to understand that all audibility isn't defined by sound pressure levels that can't be heard. 

 

This, and that there's absolutely nothing wrong in (correctly) stating that the sum of audibility is not a SPL threshold, and that the limits of audibility are not a clearly defined thing - that it's an evolving space in relevant science. To get comfortable with as much, and exhale.

 

51 minutes ago, davewantsmoore said:

I already said that, when I said "other components" ..... ie. other (additional) frequency components.

 

ie.   non-linear disortion.

 

The different waveform shape is non-linear distortion (and potentially a level error mixed in).  Which is what has already been posted in this thread, and pooh poohed. 

 

Too much here is undone by poo poo. https://www.youtube.com/watch?v=QeF1JO7Ki8E#t=0m29s

 

 

NLD is simply a differing relationship between input and output. Frequency spectra is a means to capture it, within the accuracy and resolution of the method used. It is, at the fine end, a very difficult way of doing so owing to the inherent margins in spectral methods. Time history deltas are arguably more accurate. 

 

57 minutes ago, davewantsmoore said:

Yes.... this is essentially dropping the time window to the smallest possible.

 

Where the FFT is perfect (i.e. infinite signals) then yes. For FFT's implemented as Discrete Fourier Transforms (DFTs) then no, this is not the same. See previous comment.

Link to comment
Share on other sites

15 minutes ago, rmpfyf said:

It's important to understand dB in the context of SPL and non, and to understand that all audibility isn't defined by sound pressure levels that can't be heard. 

I'm not so sure about that, for the most part.

 

Small changes in waveform shape aren't audible on their own.  That being said, if they're strongly correlated into one channel, then they can mess with "imaging" due to the timing mismatch.

 

Link to comment
Share on other sites



7 hours ago, davewantsmoore said:

It's just extremely complicated.....  as there are a lot of permutations.    Absolute levels and "slope" (ie. non-linear distortion)..... and both of these under a whole gamut of input signals (complexity, size, etc.).

 

.... but how much of these is audible, is reasonably well understood, with a few caveats (and I think, open secrets.... eg. errors correlated to a specific channel).

I'm not sure that is true.  Do you have anything to support this?  

 

We do have a lot of experimental data on audibility.  A lot of it exists at a very basic level.  Not a lot of data exists about audibility of complex sound waveforms with constant variability of amplitude, frequency and harmonics.  This crosses over with the debate about the subjectivity of hearing and double blind tests and test subject control.

 

"Understood" is, in my view, a very large assumption based on inadequate data.

Edited by Stereophilus
Spelling
Link to comment
Share on other sites

18 minutes ago, Stereophilus said:

I'm not sure that is true.  Do you have anything to support this?  

A lot of the level errors parts is contined with research on speech intelligibility and audibility.

 

Shape/slope.... is spread wider.    There's studies on timing distortion of single channel signals....  studies on actual magnitude (ie. how much non-linear distortion is audible)

 

.... and then studies which look at "imaging" (ie. distorting one channel), which applies to level and shape.   Almost all imaging studies I've seen are to do with level ...... although I know of a lot of people who propose that non-linear distortion that is correlated in one channel is very important too.    I've seen no formal studies of it, except for annecdotal evidence, and logic around why you might expect it to work like that.

 

If you're looking for papers with "the complete answer", and/or concise numbers/guidelines, etc..... then you won't find this.  What is audible depends on a lot of complexities like the content (masking), and levels.

 

I don't keep a list around, as I only spend time on things I think I should be looking into.....  not things I've decided to leave alone.

Edited by davewantsmoore
Link to comment
Share on other sites

22 minutes ago, davewantsmoore said:

I'm not so sure about that

 

Oh?

 

22 minutes ago, davewantsmoore said:

for the most part.

 

Thank you for proving my point :D 

 

22 minutes ago, davewantsmoore said:

Small changes in waveform shape aren't audible on their own.  That being said, if they're strongly correlated into one channel, then they can mess with "imaging" due to the timing mismatch.

 

And again :D :D :D 

Link to comment
Share on other sites

21 minutes ago, Stereophilus said:

Not a lot of data exists about audibility of complex sound waveforms with constant variability of amplitude, frequency and harmonics. 

It would be interesting to know how hearing aid manufacturers or the likes of Cochlea approach those problems?

Link to comment
Share on other sites

27 minutes ago, allthumbs said:

It would be interesting to know how hearing aid manufacturers or the likes of Cochlea approach those problems?

Hearing aids are, at their simplest, a microphone + amplifier + speaker... amplifying all noise outside the ear and concentrating onto the tympanum.  The better ones selectively amplify the speech range.  The super deluxe ones cancel out ambient noise.

 

Cochlear implants work on the same principle.  Instead of a speaker they use an implanted electrode in the cochlear  with small electrical impulses to excite nerve endings within the cochlear directly, bypassing the the tympanum, middle ear and basilar membrane.  As the positioning of the nerves in the cochlear determines their frequency sensitivity, the computer inside a cochlear implant is simply a FFT distributing electrical impulses to positions along the implanted electrode.

 

Edit: For the sake of being obvious, these things have no meaningful relationship to our discussion about audibility in the context of this thread.

Edited by Stereophilus
Link to comment
Share on other sites



  • Recently Browsing   0 members

    • No registered users viewing this page.




×
×
  • Create New...
To Top