WAV better than FLAC due to increased processing load on the CPU of the latter?
- Thread starter vln
- Start date
My experience suggests that the presence of the metaphorical waterfall, in this instance, makes no difference to the output*. This in turn begs the question - if a waterfall makes no difference to the output, what chance does a little drip have?
A good set of answers to my points, in general, but I have great concerns that there are many out there, perhaps not you, who are intent upon shrouding a perfectly simple concept in foo for their financial betterment.
* But I would dearly love to hear from those who find (and can demonstrate) that it does!
sq225917
Bit of this, bit of that
Item, you miss my point.
You sell dacs and audio computers, it is in your interest to make out that people need audio computers, not just any computer. Hence the proliferation of bullshit about tweaked mac mini's and the like that we see on these very pages, reflecting the sales pitches of people involved in the same area of business as you.
Sure there are dacs that don't render all sources the same, and some dacs that render some selected sources the same but not all of them. That's the nature of the beast, some designers are good at some things and others are good at some other things, no one doubts that. Only an unlimited budget would give them the ability to design multiple perfect interfaces.
But to make a system that made wav and flac sound different you would have to design it in such a way that the two file types were rendered by different processes, such as one going to dac as pcm and the other as flac, as in the case of the squeezebox. Even the simplest low cost pc can be made to spit out bits good enough to any half decent dac to render them identical. And to claim otherwise just highlights the how dubious the poster is.
Both my Audiolab and the Weiss 202 make flac and wav sound identical from my macbook, I've applied no tweaks to the macbook pro. If the dacs you sell aren't capable of rendering bit perfect streams into identical music then perhaps you should find some better ones.
I've not seen the Naim claim regarding flac and wav and replay on their own dac, but suffice to say for me that is just a big flag that reads ' this dac is not competent'. At least Linn are capable designers who saw a difference in power supply noise and investigated it through to find it made no difference.
You sell dacs and audio computers, it is in your interest to make out that people need audio computers, not just any computer. Hence the proliferation of bullshit about tweaked mac mini's and the like that we see on these very pages, reflecting the sales pitches of people involved in the same area of business as you.
Sure there are dacs that don't render all sources the same, and some dacs that render some selected sources the same but not all of them. That's the nature of the beast, some designers are good at some things and others are good at some other things, no one doubts that. Only an unlimited budget would give them the ability to design multiple perfect interfaces.
But to make a system that made wav and flac sound different you would have to design it in such a way that the two file types were rendered by different processes, such as one going to dac as pcm and the other as flac, as in the case of the squeezebox. Even the simplest low cost pc can be made to spit out bits good enough to any half decent dac to render them identical. And to claim otherwise just highlights the how dubious the poster is.
Both my Audiolab and the Weiss 202 make flac and wav sound identical from my macbook, I've applied no tweaks to the macbook pro. If the dacs you sell aren't capable of rendering bit perfect streams into identical music then perhaps you should find some better ones.
I've not seen the Naim claim regarding flac and wav and replay on their own dac, but suffice to say for me that is just a big flag that reads ' this dac is not competent'. At least Linn are capable designers who saw a difference in power supply noise and investigated it through to find it made no difference.
There are no elephants, and there is nothing paranormal about digital audio. Bits are bits, obviously. But any real-time digital conversion is heavily influenced by its local environment. How the bits are processed, clocked and moved makes changes to the local playback environment and affects DAC and amplifier performance. On the fly decompression is a factor - how influential is controversial - but - as Linn and Naim have demonstrated - you don't blithely ignore stuff like that. They're not paid to chase fairies.
High quality external clocks are a staple of the pro audio world, in which audiophools are mocked for straining at the gnat of power cables, but gulping down the camel of jitter. Even with an async DAC, you can't entirely escape Garbage In, Garbage Out, although the damage limitation is better.
High quality external clocks are a staple of the pro audio world, in which audiophools are mocked for straining at the gnat of power cables, but gulping down the camel of jitter. Even with an async DAC, you can't entirely escape Garbage In, Garbage Out, although the damage limitation is better.
SQthingy: your point about the relative competence of DACs is central to this discussion. What makes things extra-cloudy here is that every DAC manufacturer makes great claims about the ability of their converter to render the transport irrelevant. In practice, I have found these claims fall far short of reality.
Every week I get customers calling me complaining because their Benchmark DAC1 sounds different with different cables. Apparently, that shouldn't happen.
I'm still looking for the DAC that sounds identical with every transport and via each digital input.
When you say you've applied no tweaks to your MB, are you saying that you've tried them all and found no difference, or you've left that untried? I'd love to lend you the Aurender and see if you thought it sounded the same as the stock MacBook!
Every week I get customers calling me complaining because their Benchmark DAC1 sounds different with different cables. Apparently, that shouldn't happen.
I'm still looking for the DAC that sounds identical with every transport and via each digital input.
When you say you've applied no tweaks to your MB, are you saying that you've tried them all and found no difference, or you've left that untried? I'd love to lend you the Aurender and see if you thought it sounded the same as the stock MacBook!
...while my baby laptop was running a virus scan. Typically, this took the CPU usage up from 1% or 2% to 60% or 70% and the disk activity up from more or less zero to ...obviously quite intense.
Should I not have felt that the sound was different...flatter, narrower, less air...all that stuff...compared to listening when the computer is doing next to nothing apart from driving the USB port to which the DAC is connected?
If a difference is to be heard due to the nature and depth of the load on the machine, surely that little lot should have changed something?
Should I not have felt that the sound was different...flatter, narrower, less air...all that stuff...compared to listening when the computer is doing next to nothing apart from driving the USB port to which the DAC is connected?
If a difference is to be heard due to the nature and depth of the load on the machine, surely that little lot should have changed something?
sonddek
Trade: SUPATRAC
False.
False.
They stated publicly that they chased that one and found it to be a fairy. It's on this thread. Perhaps you should re-read the Linn quote.
I have seen the use of external clocks in studios explained in simple terms to you in several threads in this forum. I have no experience in studios but the explanations made sense to me. The explanation given to you was that multiple digital sources must be synchronised in the studio to ensure that their combination is correctly simultaneous. You persist in ignoring whoever kindly explained that. He also stated that the simple task of converting one bitstream back to an audio signal does not require multiple or external clocks, and that one good clock in the DAC is enough. Again, you don't read what patient and experienced contributors on this forum kindly explain, over and over again. If you don't understand what people write, just ask, because everyone here seems to be very helpful. There's no need to resort to mysticism with so many knowledgeable people around.
Ah, another challenge. OK, you're on. Send one to me, and I'll compare it to my 2008 mac and 2009 macbook. Oh, one small thing, it has to be able to run all my shop software at the same time as streaming to my hifi.
OK? Or is this another renege coming up?
sq225917
Bit of this, bit of that
Lend me what you like i'll give it a fair run and it'll get the blind ABX treatment with my trusty assistant YNWAN, who only does vinyl, pressing the buttons. I'll even take pics, post them all and capture the outs from the dac and host the files for all to compare.
I have tried almost every piece of playback software for the Mac, I still have Decibel, and Audivana+ on the system, primarily because they both offer 192khz playback which was needed for the Young, Decibel was my choice and Audivarna+ was a software volume control test, which Decibel doesn't offer. I use Itunes day in day out and only switch for hires files.
I disagree with your findings re local environment in the pc and dac replay. I do a fair bit of HD video editing, itunes never skips a beat.
I have tried almost every piece of playback software for the Mac, I still have Decibel, and Audivana+ on the system, primarily because they both offer 192khz playback which was needed for the Young, Decibel was my choice and Audivarna+ was a software volume control test, which Decibel doesn't offer. I use Itunes day in day out and only switch for hires files.
I disagree with your findings re local environment in the pc and dac replay. I do a fair bit of HD video editing, itunes never skips a beat.
Paul R
pfm Member
I've captured a couple of tracks played from wav and then flac. The ADC was running at 96k/24bits, so that the resolution of the capture significantly exceeded the resolution of the replay, and 96k means the the sampling rates are not closely related.
I can't hear a difference, but that's what I expect. So I tried something a bit more technical. AudioDiffmaker is supposed to be able to take two wavs and produce the minimised difference between them, so it accounts for small level changes, sample rate changes, time alignments, etc. Unfortunately it throws exceptions when given 24 bit files. So I truncated them to 16 in Audacity. The result is a null to about -85dB, the resulting difference file, when amplified, is just noise. I suspect much of the noise is a result of the dither used to truncate from 24 to 16 bits.
I'll upload the files tomorrow if anybody is interested. More investigation is required, but to a first approximation there is no difference at the output of the DAC in my system between a flac and wav.
I used my Lenovo X201i Thinkpad, freshly rebooted, the only 'foreground' app was Foobar2000, the output device is an M2Tech Hiface. The audio files were on the internal disk. DAC is a Twisted Pear Buffalo 2, ADC is from Uwe Beis with a low noise bipolar power supply. Capture was via an MAudio AP192 S/PDIF input and some home made software.
I'm happy to adjust things to maximise the alleged difference and repeat the exercise. Or perhaps Item will let us measure one of his 'transports'.
Paul
I can't hear a difference, but that's what I expect. So I tried something a bit more technical. AudioDiffmaker is supposed to be able to take two wavs and produce the minimised difference between them, so it accounts for small level changes, sample rate changes, time alignments, etc. Unfortunately it throws exceptions when given 24 bit files. So I truncated them to 16 in Audacity. The result is a null to about -85dB, the resulting difference file, when amplified, is just noise. I suspect much of the noise is a result of the dither used to truncate from 24 to 16 bits.
I'll upload the files tomorrow if anybody is interested. More investigation is required, but to a first approximation there is no difference at the output of the DAC in my system between a flac and wav.
I used my Lenovo X201i Thinkpad, freshly rebooted, the only 'foreground' app was Foobar2000, the output device is an M2Tech Hiface. The audio files were on the internal disk. DAC is a Twisted Pear Buffalo 2, ADC is from Uwe Beis with a low noise bipolar power supply. Capture was via an MAudio AP192 S/PDIF input and some home made software.
I'm happy to adjust things to maximise the alleged difference and repeat the exercise. Or perhaps Item will let us measure one of his 'transports'.
Paul
dave
Plywood King
Did you leave more than one call or email?
Odd, I've never had a problem getting a reply after three decades or corresponding. I've even received a few email replies over the weekend.
Cereal Killer
432
QUESTION: can we hear everything we can measure, and can we measure everything we can hear?
I know you believe that now, but there will be an audition in your future where you will realise you were mistaken.
Linn did not map the differing rail noise and processor usage to an audible outcome. Naim, Wavelength and Empirical did. But the question is taken quite seriously by serious people.
I can tell you that what motivates equipment designers and back-room types to investigate these issues is not some implausibly concerted program of misinformation: they simply want to make their products sound as good as possible: above all, they don't want customers failing to hear their designs at their best because they forgot to recommend the right file type.
Equally patiently I explained that although multiple clock syncing is one application for such devices, the sonic impact of such devices is well known and appreciated - hence the sales of Rubidium clocks like the 10M: google it, or search YouTube. Ditto the Pyramix and Merging products. In their world, domestic digital audio is stuck in the 1980s.
No-one is resorting to mysticism. But people get sloppy when they bandy around terms like 'not audible' and 'imperceptible'. How do you measure perception?
The day job of an equipment tester or professional reviewer is to find in the measurements what their ears tell them, and vice versa. Frustratingly, they rarely correspond 100%: invariably they conclude that they can't measure everything that matters with sufficient accuracy. That's the bedrock, uncertain reality that keeps arguments like this one revolving. The U and D are a given, but there's no need to F.
Robert
Tapehead
Definitely not.
Yes, so long as you are referring to differences. For example if you hear something as 'bright' it must be referenced against something else.
Cereal Killer
432
So why do we put so much faith into graphs when its clear we cannot all hear the same and measurements go well beyond our hearing thresholds?
And are you sure we can measure and capture everything even harmonics in room and reflections?
(Just playing devils advocate
)
And are you sure we can measure and capture everything even harmonics in room and reflections?
(Just playing devils advocate
)Slightly off-topic, but the accusations made in this thread are lame.
There are many institutionalised rip-offs daily fleecing us: speed cameras, 'administration fees', anything a solicitor does, anything sold to parents for their babies, women's cosmetics, etc. There cannot be any justification for the price of those things. They are inexcusable.
In audio, there's only one product that in any way compares to those: high-end cables are priced cynically to exploit music lovers. A speaker cable should not cost more than a car. That's simply out of order.
Among other things we modify and assemble specialised audio computers. If you examine our pricing, you'll note the price at which we sell them is little more than the value of the parts. They are not sold via dealers or distributors. We're completely transparent about what goes into them, and we encourage people to build their own equivalents.
So we're absolutely Teflon-coated when it comes to criticism about product pricing, or mystification, or proprietary snake-oil, or rip-offs. It's all open-source. We know this stuff works. It's easy to do it yourself and come to the same conclusion. So please, enough cobblers about 'FUD' conspiracies.
You might not currently think digital audio is any more complex than threading beads on a wire: you're entitled to your opinion, but know that it will be opposed by most authorities on the subject, even if it's supported by the majority of folks in the street. In your world, all digital cables must be identical; all USB > SPDIF converters must be identical; all power supplies for digital products must be identical; all CD transports must be identical; all computer hardware and software must sound identical, etc. To me, that's a weird, weird world to live in, but good luck to you. In the nicest possible way, the ignorant sleep well!
There are many institutionalised rip-offs daily fleecing us: speed cameras, 'administration fees', anything a solicitor does, anything sold to parents for their babies, women's cosmetics, etc. There cannot be any justification for the price of those things. They are inexcusable.
In audio, there's only one product that in any way compares to those: high-end cables are priced cynically to exploit music lovers. A speaker cable should not cost more than a car. That's simply out of order.
Among other things we modify and assemble specialised audio computers. If you examine our pricing, you'll note the price at which we sell them is little more than the value of the parts. They are not sold via dealers or distributors. We're completely transparent about what goes into them, and we encourage people to build their own equivalents.
So we're absolutely Teflon-coated when it comes to criticism about product pricing, or mystification, or proprietary snake-oil, or rip-offs. It's all open-source. We know this stuff works. It's easy to do it yourself and come to the same conclusion. So please, enough cobblers about 'FUD' conspiracies.
You might not currently think digital audio is any more complex than threading beads on a wire: you're entitled to your opinion, but know that it will be opposed by most authorities on the subject, even if it's supported by the majority of folks in the street. In your world, all digital cables must be identical; all USB > SPDIF converters must be identical; all power supplies for digital products must be identical; all CD transports must be identical; all computer hardware and software must sound identical, etc. To me, that's a weird, weird world to live in, but good luck to you. In the nicest possible way, the ignorant sleep well!
Robert
Tapehead
So why do we put so much faith into graphs when its clear we cannot all hear the same and measurements go well beyond our hearing thresholds?
And are you sure we can measure and capture everything even harmonics in room and reflections?
(Just playing devils advocate
You don't need to put so much faith in graphs.
Measurements are most useful when establishing benchmarks, so the levels as which certain types of distortion become audible. I don't just mean THD but all forms of deviation from a known reference.
If a graph tells you that amp A is flat within 0.1dB and amp B gets that down 0.05dB, the latter is of no added value or worth since you cannot hear the difference. Same for illustrations showing amplifier THD at 0.01% and 0.001% - the latter holds no value to the listener, though of course it pleases the designer.
The problem with the effects being described by Item is that they fall into the 'pleases the designer' category at best. As an example, you could conceivably monitor the PSU rails feeding a PC CPU and track noise/fluctuations on those rails. You might well be able to produce two graphs illustrations showing a difference when playing different file types. This might interest some, but what matters is what appears on the analogue output much further down the line.
Measure at that point, and invite listeners to test the audibility with some degree of statistical relevance.
Nobody bothers, because actually the proponents of these arguments know full well that none of their claims stand up to any form of scrutiny. Simple as that I'm afraid.
Finally, the issue of stress has been suggested yet again as a reason for not performing blind tests. The implication is that listener stress impedes the ability to critically identify differences.
You can disprove this easily enough by introducing small, measurable differences to a system and observing the ability of the listener to pick with accuracy. If the stress factor is real, they'll fail repeatedly.
Robert
Tapehead
Yes, but they should cost about the same as a pack of cigarettes!
Audio is littered with rip-offs.
Special feet, spikes, damping stones and certain support systems spring to mind, but the biggest rip-off is intellectual. Taking the barely plausible and spinning it to influence buyers with no ability to analyse the science behind the claim and apply any relevance to audio.
We know they are different but the vexed question - with controversy inbuilt - is 'Do WAVs and FLACs sound different?'
The moment we move the discussion to 'do they measure differently?', we're into a whole other domain of trouble. Who's measurements? Which measurements? Where? How? With what equipment?
Especially when seeking something as subtle as this gremlin, the devil is really in the detail. A half-assed attempt at it is worse than none at all because it generates spurious, but authentic-looking, data. If issues like this could be settled with commonly available techniques, all such debates wouldn't everywhere rumble on. I would be quite surprised if all these different boxes turned out to be as identical as the measurements suggest.
With lenses, there's a similar issue: on one level, a lens is simple: all it does is bend light: it comes in one end and exits the other. What's the fuss? But there is inevitable transmission loss, expressed as Modulated Transfer Function.
Manufacturers publish MTF charts that appear to be a complete description of how that lens resolves. But . . . each manufacturer has a slightly different methodology and way of communicating the results, so it's not possible to compare a Nikon lens with a Canon one exactly. Two or three different frequencies are sampled, leaving crucial gaps unmeasured. Then, there are question marks over sample variation: the test results are only valid for that lens: the one you buy will almost certainly not give exactly the same performance. And anyway, how exactly does that squiggly line on the chart translate to actual performance? The MTF data is absolutely truthful within a tiny band of usefulness.
Other losses are measured: geometric distortion and various types of chromatic aberration are mapped very precisely, though of course with slightly incompatible measurement systems, and rarely full spectrum. With all this data, do we finally know how the lens draws a picture? No.
Because the eye/brain is sensitive to all kinds of crucial stuff they don't, won't or can't measure: rendition of defocused areas, and the transition between focus and defocused zones is critical to the 'look' of the image. Not measured. Microcontrast and macrocontrast behaviour isn't usually evident from the published data. Different lenses have different depth of field effects at the same aperture. They are differently prone to flare and ghosting: again, no measurements available. Each lens renders colours differently but again you don't find measurements for that. There may be further 'intangibles' that comprise the 'signature' or 'character' of a lens even less susceptible to description or reductive analysis.
The bottom line is you just shoot with it and check the results.
Fortunately, the results can be digitally captured 'asynchronously' and examined at leisure. Here's the final rub, though: the inspection of the lens is limited by the resolution and characteristics of the imaging system. In audio equivalence, the ear/brain is an entirely different kind of 'imager' to an ADC, and the 'experience' can't be recorded with a PET scanner and better interpretative neurology than we currently have.
Fascinating though the attempt is, there may always be a gap between what is commonly measured and what we hear. Our mind may or may not prove to be an unreliable witness.
The problem with mechanical measurements is the noise floor - what's +N? - the scope and quality of measurements made, and interpretation of the data. The problem with listening is 'reporting' and our gullibility. Nothing's perfect.
The moment we move the discussion to 'do they measure differently?', we're into a whole other domain of trouble. Who's measurements? Which measurements? Where? How? With what equipment?
Especially when seeking something as subtle as this gremlin, the devil is really in the detail. A half-assed attempt at it is worse than none at all because it generates spurious, but authentic-looking, data. If issues like this could be settled with commonly available techniques, all such debates wouldn't everywhere rumble on. I would be quite surprised if all these different boxes turned out to be as identical as the measurements suggest.
With lenses, there's a similar issue: on one level, a lens is simple: all it does is bend light: it comes in one end and exits the other. What's the fuss? But there is inevitable transmission loss, expressed as Modulated Transfer Function.
Manufacturers publish MTF charts that appear to be a complete description of how that lens resolves. But . . . each manufacturer has a slightly different methodology and way of communicating the results, so it's not possible to compare a Nikon lens with a Canon one exactly. Two or three different frequencies are sampled, leaving crucial gaps unmeasured. Then, there are question marks over sample variation: the test results are only valid for that lens: the one you buy will almost certainly not give exactly the same performance. And anyway, how exactly does that squiggly line on the chart translate to actual performance? The MTF data is absolutely truthful within a tiny band of usefulness.
Other losses are measured: geometric distortion and various types of chromatic aberration are mapped very precisely, though of course with slightly incompatible measurement systems, and rarely full spectrum. With all this data, do we finally know how the lens draws a picture? No.
Because the eye/brain is sensitive to all kinds of crucial stuff they don't, won't or can't measure: rendition of defocused areas, and the transition between focus and defocused zones is critical to the 'look' of the image. Not measured. Microcontrast and macrocontrast behaviour isn't usually evident from the published data. Different lenses have different depth of field effects at the same aperture. They are differently prone to flare and ghosting: again, no measurements available. Each lens renders colours differently but again you don't find measurements for that. There may be further 'intangibles' that comprise the 'signature' or 'character' of a lens even less susceptible to description or reductive analysis.
The bottom line is you just shoot with it and check the results.
Fortunately, the results can be digitally captured 'asynchronously' and examined at leisure. Here's the final rub, though: the inspection of the lens is limited by the resolution and characteristics of the imaging system. In audio equivalence, the ear/brain is an entirely different kind of 'imager' to an ADC, and the 'experience' can't be recorded with a PET scanner and better interpretative neurology than we currently have.
Fascinating though the attempt is, there may always be a gap between what is commonly measured and what we hear. Our mind may or may not prove to be an unreliable witness.
The problem with mechanical measurements is the noise floor - what's +N? - the scope and quality of measurements made, and interpretation of the data. The problem with listening is 'reporting' and our gullibility. Nothing's perfect.