Power cables - how to measure

[davewantsmoore]

16 minutes ago, Stereophilus said:

Absolute SPL differences create an artificial sense of insignificance

Only to those who don't actually relate them to previous studies on what is thought to be significant.

 

I'm not saying that it won't be audible because "woah that's a really small/big number".   

 

That being said, the precice nature of the distortion is very important..... but as said, I think the types of distortion you will get are likely to be "particularly benign" in that regard.

 

 

That being said, low frequency distrubances could be particularly underrated.

[Stereophilus]

3 minutes ago, davewantsmoore said:

Only to those who don't actually relate them to previous studies on what is thought to be significant.

 

I'm not saying that it won't be audible because "woah that's a really small/big number".   

 

That being said, the precice nature of the distortion is very important..... but as said, I think the types of distortion you will get are likely to be "particularly benign" in that regard.

 

 

That being said, low frequency distrubances could be particularly underrated.

Do we then agree that these possibly significant differences are worth our time trying to measure?  Because, ya know... that’s why we’re here....

[MLXXX]

1 hour ago, rmpfyf said:

rmpfyf said:
If the time domain waveform is audible, all content above the floor is audible.

 

MLXXX replied:
Could you clarify what you mean here by the description "content above the floor"?  

 
(I note it is well known that low level content is often masked if much louder content is present at the same time.) 

 

rmpfyp then replied:-

 

Sample silence, process spectra - see what you get. What you have is 'floor'. This is a simple 'first cut'.

 

A second cut might involve creating noise and raising amplitude until audible, though this test has many facets and angles, and I'd argue it's not a relevant test for waveform shape audibility regardless.

 

 

I don't understand the relevance of the above to this thread.  It is clearly not relevant to an audiophile how the silence between CD tracks  "sounds",  because with a properly operating system they should hear nothing from their speakers at ordinary gain settings, when they are in their listening chair (not with an ear pressed against a speaker enclosure!). To then proceed to conjecture that very low level signals commensurate with system noise may become audible if piggy-backed onto an otherwise audible signal (e.g. onto the sound of a vocalist singing, or a piano being played) is a thought. It could possibly happen in special circumstances. 

 

However to conjecture that low level signals commensurate with system noise will always be audible if piggy-backed onto an otherwise audible signal (e.g. onto the signal of a vocalist singing, or a piano being played) would be a very long bow to draw indeed.  On the contrary, I would suggest that there is no guarantee at all that very low level signals piggy-backed onto clearly audible waveforms would make an audible difference.  It is a well known strategy of psychoacoustic codec design to discard low level signals masked by higher level signals. At high enough bitrates the processed audio becomes indistinguishable from the original, for human ears.  The codec is said to be "transparent". [The waveforms look decidedly different I might add, even at very high bitrates. It is perhaps surprising that the music can sound the same to our ears when the waveforms look so different to our eyes, when glancing at the timeline of an audio editor, such as the free software Audacity.]

 

I fail to understand the repeated references in this thread to very low level measurable variations possibly being audible, when the context is investigating power cords in an audiophile forum setting of (reportedly) clearly audible differences.  We should surely be looking for clearly measurable differences.  Measurement and analysis of variations buried in system noise would not be particularly helpful, in my opinion. They would cause delay, without a corresponding  audiophile benefit. 

Edited October 28, 2020 by MLXXX

[Guest rmpfyf]

21 minutes ago, davewantsmoore said:

When the magnitude of the added expectral components are small.... they're not large enough distortions to be audible.

 

Unless they draw attention to themselves somehow..... which is the disotriton audibility rabit hole.

 

Si.

 

21 minutes ago, davewantsmoore said:

For the types of disortion I suspect you to find from a power cord, it won't apply  (aaaand it will be v. small too).

 

We shall see. Much possibility here, not least as the weight of conventional, accessible tools and approaches doesn't show up anything. That's not to suggest stuff does or doesn't exist... just that we'd need to get creative. 

 

26 minutes ago, davewantsmoore said:

Ric'll stop talking to me, if he wants to.

 

... and then I'll stop replying.

 

Oh I'd need a far more interesting day job for that. You too.

[davewantsmoore]

39 minutes ago, Stereophilus said:

Do we then agree that these possibly significant differences are worth our time trying to measure?  Because, ya know... that’s why we’re here....

Yes, that's what I've been saying all along.   Go ahead and figure out exactly what measurable differences in the audio signal there are from a power cord.

 

Someone already did it.... and we've been commenting on the results.

 

I've been saying that I don't think the presented results would be audible.   I've been accused of misunderstanding them (eg. frequency/amplitude, vs time/amplitude).  That isn't the case....   Then we kinda got to the point where "I should just admit that this is an evolving space and there could be something in there that's audible".   I'm yet to be convinced.

 

Oh ye of little faith (that's me).

Edited October 28, 2020 by davewantsmoore

[Guest rmpfyf]

5 minutes ago, MLXXX said:

I don't understand the relevance of the above to this thread. 

 

Spectra computed from a DFT will always have a floor. Always. 

 

5 minutes ago, MLXXX said:

To then proceed to conjecture that very low level signals commensurate with system noise may become audible if piggy-backed onto an otherwise audible signal (e.g. onto the sound of a vocalist singing, or a piano being played) is a thought. It could possibly happen in special circumstances. 

 

 

They're not signals that become audible, they're a mathematical expression, a decomposition of a signal as acquired. By definition if that signal is audible, so is whatever representation of it's content. Whether a listener has a relative sensitivity to that pressure waveform or something like but not quite it, different matter. What you are listening to is the whole thing.

 

1 hour ago, MLXXX said:

However to conjecture that low level signals commensurate with system noise will always be audible if piggy-backed onto an otherwise audible signal (e.g. onto the signal of a vocalist singing, or a piano being played) would be a very long bow to draw indeed.  On the contrary, I would suggest that there is no guarantee at all that very low level signals piggy-backed onto clearly audible waveforms would make an audible difference.  It is a well known strategy of psychoacoustic codec design to discard low level signals masked by higher level signals. At high enough bitrates the processed audio becomes indistinguishable from the original, for human ears.  The codec is said to be "transparent". [The waveforms look decidedly different I might add, even at very high bitrates. It is perhaps surprising that the music can sound the same to our ears when the waveforms look so different to our eyes when glancing at the timeline of an audio editor, such as the free software Audacity.]

 

You're into psychoacoustics here, and it is indeed complex. There's been no claim that small differences will always be heard, simply that it's incorrect to suggest that they always will not be heard. 

 

A good portion of psychoacoustic research goes beyond the notion that we can be tricked into believing two sounds are the same - with training, most can be taught or trained to pick differences.

 

1 hour ago, MLXXX said:

I fail to understand the repeated references in this thread to very low level measurable variations possibly being audible, when the context is investigating power cords in an audiophile forum setting of (reportedly) clearly audible differences.  We should surely be looking for clearly measurable differences.  Measurement and analysis of variations buried in system noise would not be particularly helpful, in my opinion. They would cause delay, without a corresponding  audiophile benefit. 

 

 

A clearly audible difference need not be one observable by conventional measures. What's already come up a few times is short-run, wide-ish-band discrete Fourier spectra. For various reasons different tools are needed.

[aussievintage]

7 minutes ago, rmpfyf said:

 

A clearly audible difference need not be one observable by conventional measures.

 

But I feel it is VERY likely to be.

[Guest rmpfyf]

2 minutes ago, aussievintage said:

But I feel it is VERY likely to be.

 

Depends which means.

 

Only as good as the method.

[MLXXX]

32 minutes ago, rmpfyf said:

A good portion of psychoacoustic research goes beyond the notion that we can be tricked into believing two sounds are the same - with training, most can be taught or trained to pick differences.

 

Studies involving the relative efficiency of audio codecs often do not go much beyond 128kbps because the more efficent codecs at that bitrate are approaching transparency for a large percentage of the population for most of the time.  Here's an example: of a compilation:

 

  

(There are many other similar compilations in graphical form or tabular form, on the net.)

 

 

If you read the literature in this subject area of audio codec transparency you will find it reported that at 256kbps even trained listeners will typically strain to hear any differences where a modern codec such as AAC is used for stereo music.  At 320kbps even trained listeners will typically need to throw in the towel, as only so-called "killer" samples, may remain distinguishable for them.  It is perhaps astonishing that if you compare waveforms as between a 24 bit 48kHz sample rate stereo LPCM file, and a 320kbps AAC version derived from it, those waveforms will look rather different [almost all the time], and yet sound the same even to trained listeners [almost all the time].   It can actually be hard to time align a lossy codec file with the original file by inspection of the waveforms.  And it can be hard to check for a level match.

 

As for the general population in Australia, we have been watching movies on television for years, blissfully unaware [most of us anyway] that 24fps movies have been shown at 25fps, raising the tempo and the pitch of the sound.  I can tell when this happens myself, despite lacking a sense of absolute pitch. For my ears, the sound is thinner, the speech is a little hurried, and the vowels have a slight chipmunk flavour to them.  (This is the so-called PAL speed-up effect.)

 

Sorry to sidetrack onto lossy codecs and PAL speed-up, but I really think we should be looking for clear measurable differences if the topic is comparing power cords, because the audiophile context is that a significant number of people have reported hearing clear differences, not faint differences.   This in my humble view is not the thread for chasing esoteric differences at the borderline of acoustic science.  If anyone wants to do that, e.g. investigate the audibility of very low level jitter, let them start up a thread for that purpose.

[davewantsmoore]

4 minutes ago, MLXXX said:

I really think we should be looking for clear measurable differences if the topic is comparing power cords

Just "look for" whatever data happens to be recorded.

 

Then analyse, and close the loop (ie. do more testing, cos the first one wasn't the end).

Well, that's what everyone else does, anyways.

[Red MacKay]

9 minutes ago, MLXXX said:

As for the general population in Australia, we have been watching movies on television for years, blissfully unaware [most of us anyway] that 24fps movies have been shown at 25fps, raising the tempo and the pitch of the sound.  I can tell when this happens myself, despite lacking a sense of absolute pitch. For my ears, the sound is thinner, the speech is a little hurried, and the vowels have a slight chipmunk flavour to them.  (This is the so-called PAL speed-up effect.)

 

 

I agree with you 100% here.  To be truthful, I have never noticed it with movies, BUT, I do notice it when playing vinyl.  My ears are not golden any longer as I get older, more brass these days - but my sense of pitch has not diminished one bit.  Everyone cracks on about how important it is to get your TT running at 33.33 rpm perfect.

 

Well - what a load of cods wallop!  When I play my records at 33.33rpm, to my ears the music is too slow!  My aural memory clearly remembers songs played on the radio back when I was a teenager.  If I speed my record playback speed up a touch - then the aural memory fits perfect.  Maybe the pitch of the voices might be incorrect running faster - but once again, that is how I remember it and how I WANT to hear it played back.

 

Now get back on topic you guys!

[Guest rmpfyf]

2 hours ago, MLXXX said:

Sorry to sidetrack onto lossy codecs and PAL speed-up, but I really think we should be looking for clear measurable differences if the topic is comparing power cords, because the audiophile context is that a significant number of people have reported hearing clear differences, not faint differences.   This in my humble view is not the thread for chasing esoteric differences at the borderline of acoustic science.  If anyone wants to do that, e.g. investigate the audibility of very low level jitter, let them start up a thread for that purpose.

 

Don't apologise for being O/T, it's been a wide ranging discussion and for the most part a civil one, which is good. 

 

I would offer that yes, much of the time a high-bitrate compressed file is virtually indistinguishable and there's even stuff like MQA which deliberately trades bandwidth in regions of likely known inaudibility. I wouldn't though put it past people with adequate training - which may simply be familiarity in their own listening environment - to be able to discern further. Trying to tune a piano first time is daunting; not impossible but unfamiliar. With training, the ability to discern accurate pitch increases dramatically. Frequency sensitivity varies with waveform too - the just-noticeable threshold decreases 3x with complex waveforms against pure tones. It's a fascinating field of study.

 

I think you're conflating 'significant' with 'significance in measurement' where measurement is a particular take on spectra. That won't show up much IMHO. It is possible to have significant differences that are not visible in typical spectral measurements. There's nothing wrong in hypothesising how something could manifest itself, and then tailoring an experiment to suit - it's pretty standard stuff. 

[andyr]

8 minutes ago, rmpfyf said:

 

That won't show up much IMHO. It is possible to have significant differences that are not visible in typical spectral measurements.

 

 

Impossible!    If the humble ear can hear it ... a device must be able to measure it!  

 

Otherwise EEs are out of business. 

 

Andy

 

[Guest rmpfyf]

2 hours ago, andyr said:

 

Impossible!    If the humble ear can hear it ... a device must be able to measure it!  

 

Otherwise EEs are out of business. 

 

Andy

 

I remember asking a power engineer colleague troubleshooting a total PITA of a problem at site 'Joe, what do you want for your birthday'?

 

Reply? 'A scope with infinite precision!'. 

[Guest rmpfyf]

Weeeeeeeell that went full d**khead moment for me. 

 

The plan was to hook up both d10's with their supplied cables and get a baseline, swap the cables and firm the baseline, swap one of the cables for a spanky one and see what's what if anything. Sample the whole thing 4x over simulaneously. Run each test and hour so 60 samples total, process deltas. 

 

Pulled out my Scarlett interface and realised I have no f cables for this thing to connect to my DACs. Proper genius moment. The cabling is -180dB to useful right now i.e. not distinguishable as useful in any way.

 

Sooooooooo if anyone out there is less impaired in the 'where is my stuff' department than I, I'm happy to send my test file, you send your results, I'll do some processing. Preferaby not @Ittaku becuase his DAC is proper amazing stuff that's very robustly engineered, we're looking for something more plebian that we might see a difference on  

 

Con all jokes asides could you run this thing quickly if I sent it to you? I'm keen to see if I can acquire the impulse I've put in there to sync acquisition in post processing.

[MLXXX]

8 hours ago, rmpfyf said:

I think you're conflating 'significant' with 'significance in measurement' where measurement is a particular take on spectra. That won't show up much IMHO. It is possible to have significant differences that are not visible in typical spectral measurements. 

 

I'm merely trying to avoid the exercise getting bogged down in trying to find correlations of the output waveform with one variable (changing the power cord) that are "swamped" or confounded by the presence of other variable factors.  The further down into the noise that you try to look for correlations, the more time consuming the task and the greater the risk of false positives. 

 

If there is a desire to  drill down to data variations at the limits of resolution of measurement devices, then it would become important to hold as many of the variables constant as possible that are not the "power cord in use" variable. One obvious factor to control would be the level and quality of the nominal 230V 50Hz supply.

 

A potentially powerful experimental approach would be test two units, of the same model device, simultaneously. (I think that the testing of duplicate devices may have been mentioned already in this thread.)  It would probably entail duplication of the capture devices.  Or, you could send out of phase signals to the duplicate DUTs, and test their outputs for a null, using a single measuring device.  As room temperature, mains voltage, and EMI levels varied, these variable factors might be expected to affect both DUTs in a similar way, making it easier to identify any variations in device output related to the "power cord in use" variable.

Edited October 28, 2020 by MLXXX

[Ittaku]

5 hours ago, rmpfyf said:

Sooooooooo if anyone out there is less impaired in the 'where is my stuff' department than I, I'm happy to send my test file, you send your results, I'll do some processing. Preferaby not @Ittaku becuase his DAC is proper amazing stuff that's very robustly engineered, we're looking for something more plebian that we might see a difference on  

 

Con all jokes asides could you run this thing quickly if I sent it to you? I'm keen to see if I can acquire the impulse I've put in there to sync acquisition in post processing.

Sorry I tuned out completely the last few days. What exactly do you want me to do? Have you written a scripted set of tests or something you want me to run? Oh and measurements I took were with a Matrix Audio DAC, not my MSB DAC.

[Guest rmpfyf]

2 hours ago, MLXXX said:

I'm merely trying to avoid the exercise getting bogged down in trying to find correlations of the output waveform with one variable (changing the power cord) that are "swamped" or confounded by the presence of other variable factors.  The further down into the noise that you try to look for correlations, the more time consuming the task and the greater the risk of false positives. 

 

I get that, and I'm not proposing a testthat's minutae, simply one that assesses potential differently to spectra.

 

2 hours ago, MLXXX said:

If there is a desire to  drill down to data variations at the limits of resolution of measurement devices, then it would become important to hold as many of the variables constant as possible that are not the "power cord in use" variable. One obvious factor to control would be the level and quality of the nominal 230V 50Hz supply.

 

Mentioned many times... which is why a nice spanky power meter is about to be ordered. I've found a few at an absolute ripper price - if you want one too please let me know.

 

2 hours ago, MLXXX said:

A potentially powerful experimental approach would be test two units, of the same model device, simultaneously. (I think that the testing of duplicate devices may have been mentioned already in this thread.)  It would probably entail duplication of the capture devices.  

 

 

Read my suggested example in my last post! It's exactly this. 

 

The both out of phase suggestion is difficult to implement for small differences in timing to how they get data. USB is a serial bus and building a digital splitter for an e.g. optical stream would take time.

Edited October 28, 2020 by rmpfyf

[Guest rmpfyf]

27 minutes ago, Ittaku said:

Sorry I tuned out completely the last few days. What exactly do you want me to do? Have you written a scripted set of tests or something you want me to run?

 

For now if I send you a WAV can you just play it e.g. 30x over (looped in VLC, aplay'd in bash in a while loop or similar), record the output and send it here? Want to test something. It's only a minute long.

 

27 minutes ago, Ittaku said:

Oh and measurements I took were with a Matrix Audio DAC, not my MSB DAC.

 

 

That's still not sufficiently plebian, Con  

[Ittaku]

33 minutes ago, rmpfyf said:

For now if I send you a WAV can you just play it e.g. 30x over (looped in VLC, aplay'd in bash in a while loop or similar), record the output and send it here? Want to test something. It's only a minute long.

I could but am a bit busy at the moment so it could take me a while. You may have found a cable first.

[Guest rmpfyf]

1 hour ago, Ittaku said:

I could but am a bit busy at the moment so it could take me a while. You may have found a cable first.

 

Maybe. Want the file anyway? Let the eBay deliveries race your availability  

[Ittaku]

42 minutes ago, rmpfyf said:

Want the file anyway? Let the eBay deliveries race your availability  

Go on then.

[Guest rmpfyf]

13 minutes ago, Ittaku said:

Go on then.

PM'd

[MLXXX]

5 hours ago, rmpfyf said:

Mentioned many times... which is why a nice spanky power meter is about to be ordered.

What model is this? I presume it will  monitor the incoming AC for voltage and frequency.  Will it also capture harmonic content and/or general "hash"?

 

It is not the same as locally generating an AC supply, which is what I had in mind to really pin down that variable.

 

On the other hand using the actual mains allows the ingress of hash (at whatever level and in whatever form it may take from time to time at the power point in use, at the test site), and that hash one suspects is a potential trigger for aberrant device output and possibly the reason some audiophiles have reported improved performance when fitting a power cord that in some way suppresses hash.

 

5 hours ago, rmpfyf said:

Read my suggested example in my last post! It's exactly this. 

 

The both out of phase suggestion is difficult to implement for small differences in timing to how they get data. USB is a serial bus and building a digital splitter for an e.g. optical stream would take time.

I posted very late last night and at the time didn't fully digest the part of your post where you described a thwarted plan to test two "d10's".

 

There are other situations where a two devices fed out of phase under test method could work fine. Many DACs have balanced outputs. It would be easy to wire up a circuit to feed that analogue signal as an unbalanced input to each of two preamplifiers or main amplifiers.   Or even easier, keep the feed balanced if the devices under test have balanced inputs and just swap the phases around for the input to one of those devices.

 

Of note would be the fact that any inherent mismatch between two amplifiers in frequency response would show up after adjusting for a null at one frequency (say 1kHz). And then there would be a change in overall gain of amplifiers as they heated up. However if you used the same model device you'd stand a better chance that the underlying frequency responses would be a close match, and that their internal temperature related changes would be similar.

 

The null method is extremely sensitive as it shows up even the slightest differences. So if the power output of one power amplifier altered by .001% there'd be a chance of capturing that.  You could choose a test file consisting of a singe fixed frequency to facilitate finding the very smallest possible discrepancies in power output (or gain) or distortion products.  

 

Or if the slew rates of the amplifiers under test differed slightly, a fixed high frequency test tone might be the go, to lay their relative performances open for detailed examination.  You'd be heightening the chances of identifying a measurable difference after swapping power cords on one of the amplifiers, switching the amplifiers back on, waiting for them to settle, and measuring for a second time the extent of departure from a perfect null as between the outputs of the two amplifiers.

 

Another advantage of the null method is that it avoids any need for the ADC to have exceptional ("atomic clock"!) standard long-term clocking stability, or laboratory-grade short term stability.  Even a garden variety ADC would do to capture the differences in output of the devices under test.  I am assuming here that the devices under test do not themselves subject the incoming signal to digital processing, but merely amplify.

Edited October 29, 2020 by MLXXX

[Guest rmpfyf]

2 hours ago, MLXXX said:

What model is this? I presume it will  monitor the incoming AC for voltage and frequency.  Will it also capture harmonic content and/or general "hash"?

 

Got a lead on some used AcuVim IIDW's and some Schneider ION PM8000s. Both 10Hz, 1/2/3 phase, both Class 0.2S, both 63rd harmonic accurate (may estimate 2x w/software), both Ethernet, both waveform capture. AcuVim's at USD$150ea +shipping. You will need to get accurate CT's and a sparkie. Contact me if interested. Awaiting price on the Schneiders. 

 

2 hours ago, MLXXX said:

It is not the same as locally generating an AC supply, which is what I had in mind to really pin down that variable.

 

I want one of these https://www.regatron.com/product/overview/programmable-bidirectional-ac-power-sources/tc-acs-series/ but they're pricey at over $100k  

 

2 hours ago, MLXXX said:

On the other hand using the actual mains allows the ingress of  hash (at whatever level and in whatever form it may take from time to time at the power point in use, at the test site), and that hash one suspects is a potential trigger for aberrant device output and possibly the reason some audiophiles have reported improved performance when fitting a power cord that in some way suppresses hash.

 

I planned to kill the HVAC and leave it on overnight, and then (if repeated with metering) discard anything with a weird transient throughout the test.

 

2 hours ago, MLXXX said:

I posted very late last night and at the time didn't fully digest the part of your post where you described a thwarted plan to test two "d10's".

 

No probs.

 

2 hours ago, MLXXX said:

There are other situations where a two devices fed out of phase under test method could work fine. Many DACs have balanced outputs. It would be easy to wire up a circuit to feed that analogue signal as an unbalanced input to each of two preamplifiers or main amplifiers.   Or even easier, keep the feed balanced if the devices under test have balanced inputs and just swap the phases around for the input to one of the devices under test.

 

Of note would be the fact that any inherent mismatch between two amplifiers in frequency response would show up after adjusting for a a null at one frequency (say 1kHz). And then there would be a change in overall gain of amplifiers as they heated up. However if you used the same model device you'd stand a better chance that the underlying frequency responses would be a close match, and that their internal temperature related changes would be similar.

 

No bal output devices here. 

 

The d10's are aluminium-cased; it's easy enough to strap 'em together for heat transfer. Mind you these are among the last DACs I'd expect to have any useful response on this experiment.

 

2 hours ago, MLXXX said:

The null method is extremely sensitive as it shows up even the slightest differences. So if the power output of one power amplifier altered by .001% there'd be a chance of capturing that.  You could choose a test file consisting of a singe fixed frequency to facilitate finding the very smallest possible discrepancies in power output (or gain) or distortion products.  

 

Or if the slew rates of the amplifiers under test differed slightly, a fixed high frequency test tone might be the go, to lay their relative performances open for detailed examination.  You'd be heightening the chances of identifying a measurable difference after swapping power cords on one of the amplifiers, switching the amplifiers back on, waiting for them to settle, and measuring for a second time the extent of departure from a perfect null as between the outputs of the two amplifiers.

 

Another advantage of the null method is that it avoids any need for the ADC to have exceptional ("atomic clock"!) standard long-term clocking stability, or laboratory-grade short term stability.  Even a garden variety ADC would do to capture the differences in output of the devices under test.  I am assuming here that the devices under test do not themselves subject the incoming signal to digital processing, but merely amplify.

 

 

The amps I have handy aren't amenable to this, I'm in a DAC-type situation. Ideally I'd feed them optical so that there's no serial bus alignment issues, though then powering them could be a little prickly. Ideally I'd RTFM and realise it's an optical output.

Edited October 29, 2020 by rmpfyf