Hi-Res - Does it matter? Blind Test by Mark Waldrep

[mansr]

1 hour ago, marce said:

A good example of expectation bias, was people being given cheese and onion in a salt and vinegar packet, they all got the flavour wrong!

Which brand? If it was the boring Walkers, I'm not surprised. There's hardly any correlation between the flavour name and the actual ingredients.

[mansr]

7 hours ago, Jud said:

If an impulse lasts less than 1/20,000th of a second, is it inaudible?

The duration is unimportant. The frequency content is determined by the rise and fall times. It starts at zero and extends higher the steeper the slopes are. There will thus be some audible frequencies also in a very short pulse.

[mansr]

42 minutes ago, Rt66indierock said:

They don't add up to anything but a rounding error in the market.

DSD is made of rounding errors.

[mansr]

8 hours ago, Allan F said:

Would you care to share what evidence you have to support your contention that "chip manufacturers never could give a good reason why they included DSD"? I strongly suspect that you are presenting your opinion as fact.

Many DAC chips likely support DSD because there's sufficient demand for the feature that the relatively small effort needed is worthwhile. These days, they probably simply paste in their existing DSD handling block with minimal tweaking for each new chip.

[mansr]

10 minutes ago, Jud said:

I was under the impression chips with sigma-delta modulation, which may produce DSD or other formats, originally took over from chips without because the chips with SDM enabled the total package, including hardware, to be less expensive.

I was talking about DAC chips accepting DSD input (in addition to PCM), not internal implementation details. There is still a relationship between the two, though. A PCM DAC using a sigma-delta architecture can be adapted to accept DSD simply by feeding the bitstream directly into the D/A stage, bypassing the oversampling and sigma-delta stages. If a small effort can increase sales by even a modest amount, it may be worthwhile. Contrast this with something much more difficult to implement, say HDCD. Nobody is including this because the world market of perhaps 10 such chips would never pay for the investment.

[mansr]

7 minutes ago, gmgraves said:

I really don’t understand your use of the term “vaporware”. In the computer world, vaporware is a product that developers/manufacturers keep promising to bring to market, but never do. If I’m not wrong, you are using the term to mean a product that doesn’t perform to it’s makers’ promises. Is that right?

Until Tidal started streaming MQA, vapourware was an appropriate term. A handful of demo tracks does not a product make.

[mansr]

5 minutes ago, jabbr said:

Thats not what we in the technical world consider “vaporware”. Vaporware is a product that has been promised by marketing people but doesn’t exist as a working product. It only exists as the vapor coming out of marketing folks mouths.

A classic example is Duke Nukem Forever, although it did eventually materialise. Another one is Project Xanadu, started in 1960 and having yet to deliver anything.

[mansr]

5 hours ago, gmgraves said:

It’s been a long time, but I seem to recall that the. Kernel of OS X was the one used in the NeXT  computer. NeXT was BSD UNIX. One thing that the UNIX core of OS X does is make the system bulletproof. While the Mac interface is logical, fun to use, and robust, what makes it really powerful is the easy access to the CLI.

OS X uses the Darwin core system consisting of the XNU (X is Not Unix), a Mach-derived microkernel, with some BSD bits glued on along with some of Apple's own pieces. The userspace shell utilities (ls, cp, and friends) came from BSD, but I don't think much else did. The executable file format is the Mach-O monstrosity (from NeXT) rather than something sane like ELF (used by most Unixes except OSF/1 and AIX) or even ECOFF. Considering all the disparate parts mashed together, it's a wonder it works at all.

[mansr]

1 hour ago, gmgraves said:

Looks like your XXHighEnd software is Win-Blows only, no Mac version

Apple won't allow GUIs that ugly on Mac.

[mansr]

23 minutes ago, sandyk said:

Diament's recordings in 24/192, with genuine musical content to >55kHz (!)

How do you know it's genuine when you can't hear it?

[mansr]

11 hours ago, sandyk said:

 It is easily verified in an Audio Editing program that it isn't noise.

How do you know it's not distortion?

 

11 hours ago, sandyk said:

In this case Barry Diament also uses microphones that are only 1dB down at 40kHz.

What is the frequency response of your speakers?

[mansr]

If you have a hearing impairment at specific frequencies, then obviously sounds at those frequencies will not be able to mask (weaker) sounds that would be imperceptible with normal hearing. This could have interesting implications for perceptual codecs like mp3 that rely on masking both to discard parts of the input and to hide artefacts they introduced. When the masking frequency is removed, what remains may well be quite distorted by the encoder. None of that suggests in any way that ultrasonic frequencies are ever audible to anyone.

 

Regarding normal age related hearing loss, a person with an upper limit of, say, 12 kHz who is trained in listening for distortion may well be better than an untrained young person at detecting it as long as it falls in the frequency range they can still hear. There is nothing remarkable about that. It also does not suggest that the older person would somehow be able to hear ultrasonics at 50 kHz.

 

When the CD format was created, the parameters (sample rate and resolution) were chosen such that it could capture any sound reasonably audible to humans. Had studies shown that humans could hear up to 30 kHz, they would have used a higher sample rate (and perhaps made the discs larger). Why is this so difficult for some to accept?

[mansr]

55 minutes ago, John Dyson said:

The only way that soundwaves combine nonlinearly (that is, produce other, audible frequencies) is if the air or the eardrum is driven into nonlinearity.  Nonlinearities are tantamount to nonlinear distortion, therefore are erroneous results, because there is NO WAY to predict the results in the recording studios.  That is, the possible nonlinearties are not part of authors intent unless it is a carefully controlled science experiment.

Air itself is ever so slightly non-linear, but not enough to matter at the frequencies and levels we're talking about. Besides, any audible non-linear effects in the air would have been already captured by the recording microphone, and we wouldn't want them to happen again on playback, would we?

[mansr]

5 hours ago, audiobomber said:

Ideally shouldn't the speakers reproduce exactly what the microphone captured?

Indeed, and any audible distortion created by ultrasonics between the instrument and the microphone will be recorded just like any regular sound. Additionally capturing and reproducing those ultrasonics would cause the same distortion products to be created once again, adding to the already recorded (and thus reproduced set). In this sense, filtering out any ultrasonics before playback is the more correct approach. However, the non-linear effects of air at normal music levels are negligible, so none of this actually matters.

 

That leaves non-linear effects in the ear. Let us assume, for sake of argument, that they are strong enough for ultrasonics to make an audible difference. That raises a few questions. Do microphones capture the ultrasonics accurately without themselves introducing distortion of a similar kind? (Probably not.) Do the speakers reproduce those frequencies cleanly? (Nope.)

 

Insofar high-res playback is distinguishable from CD quality, there is little reason to believe the differences represent greater accuracy. More likely, any differences perceived are the result of various unwanted distortions. CD quality may not be perfectly accurate, but it is still more accurate than a similar level plus additional random distortions. Above 25-30 kHz, playback equipment (especially speakers) is all over the place. They can't all be right. Better to not try.

[mansr]

5 minutes ago, ralphfcooke said:

Surely the bit depth only relates to the noise floor, it's the sample rate that determines the frequency range?

That is correct.

[mansr]

14 minutes ago, audiobomber said:

I agree, the two conversions ares a problem, especially since 96 to 44.1kHz is a non-integer conversion, requiring interpolation.

Non-integer ratios are not a problem in the slightest.

[mansr]

17 minutes ago, audiobomber said:

Yet many DAC manufacturers use separate clocks for 44.1 vs. 48kHz streams.

What has that got to do with anything?

[mansr]

14 minutes ago, John Dyson said:

Just requires more software than trivial conversions...

Specifically, a polyphase interpolation filter. It's marginally more complicated but no less accurate.

[mansr]

Just now, John Dyson said:

More math means less accuracy, especially tricky if using 32bit signed math (I hope not).

The number of arithmetic operations per output sample is the same.

[mansr]

2 minutes ago, lucretius said:

Is it not the case that with floating point math, you are going to get a small truncation error for each operation? (Maybe too small to matter?)

Floating-point or fixed-point (of which integer is a special case) doesn't matter. Whenever precision is limited, some operations will produce an unrepresentable result that has to be rounded. As long as the cumulative error remains below an acceptable level, there is no problem. For the task of resampling audio with a final output as 24-bit integer, doing the calculations in 64-bit floating-point provides more than enough headroom to ensure accurate results.

[mansr]

This was about accuracy of resampling with integer vs non-integer ratios. There is a widespread belief that non-integer-ratio resampling is somehow inferior to integer-ratio. This is false. If a particular integer-ratio implementation is good enough, it can be easily extended to non-integer ratios using a polyphase filter bank without impairing the accuracy. How you ensure that accuracy in the first place is a completely different discussion.

[mansr]

1 minute ago, pkane2001 said:

So getting back to this blind test. Do you see anything wrong with the methodology? Mark used Sonic Studio’s PROCESS tool to do the conversions. The maker claims this to be an "ultra-resolution" set of DSP tools, whatever that means. Any issues that you or anyone else is aware of with this tool that would invalidate the test?

It's easy enough to grab the files and check for any anomalies.

[mansr]

1 minute ago, pkane2001 said:

How would you detect the loss of computational precision that John is describing other than as part of the overall error? I have the files and have submitted my test results.

Put them through your Deltawave software and look for differences that might be audible according established thresholds.

[mansr]

8 minutes ago, John Dyson said:

Oh, that is your hot button -- about the integer vs. non-integer ratios.

I was responding to this:

2 hours ago, audiobomber said:

I agree, the two conversions ares a problem, especially since 96 to 44.1kHz is a non-integer conversion, requiring interpolation.

[mansr]

Just now, audiobomber said:

It means that some designers disagree with you. I'm well beyond believing that just because someone is a math-head they know all there is to know about great sound.

Clocking of DAC chips has nothing to do with the question at hand.