Testing audiophile claims and myths
Apr 22, 2015 at 3:15 PM Post #4,546 of 18,259
How about something guaranteed to be 101% PCM free - entirely DSD vs DSD comparison. Two ADCs, original recordings on both in DSD64, DSD256 recording on the more capable machine, DSD128 "bounced down from DSD256", DSD64 "bounced down from DSD256" using Sygnalist software, the same using Pyramix - IIRC.
The first DSD64 file was originally 0.34 dB too low in level being later "lifted by 0.34 dB" using Sygnalist soft. In short - 6 files, of which 5 are level matched < 0.1 dB, therefore directly ABXable - IF you have NATIVE DSD playback capability to DSD256 - otherwise to the DSD sampling your DAC is capable of playing back NATIVELY , using ASIO . ( DACs that can do it all start at  approx $200 - and up )
 
CAUTION - all six files , 4:53 in duration each, amount to 1.92 GB downloads. 
 
http://www.computeraudiophile.com/content/635-midsummer-night%92s-dream-compare-simultaneous-dsd64-and-dsd256-session-recordings/
 
Apr 22, 2015 at 3:27 PM Post #4,547 of 18,259
From math point of view DSD and PCM are fully identical in audible range.
 
I said about audible range due while no reliable proofs that we hear ultrasound.
 
Here can be differences due practical realisations - real devices / software.
 
It may be big differences :)
 
Apr 22, 2015 at 4:04 PM Post #4,548 of 18,259
Actually, you missed my point. Your proposed test would indeed do a very good job of confirming confirming whether digitizing an analog signal using a PCM ADC, then converting it back into analog using a PCM DAC, will result in any audible difference. However, that isn't the same thing as COMPARING DSD to PCM. In fact, what you're doing is comparing PCM to - nothing - or to a piece of wire. (Since I know that different PCM DACs do indeed sound slightly different, I'm pretty sure they each also sound different than a piece of wire.)
 
However, if we want to compare DSD to PCM, then what we need to determine is whether using a DSD ADC and a DSD DAC makes LESS of a difference to the sound than using their PCM equivalents. (If what comes out of that DSD DAC is a perfect reproduction of the original, then clearly we would strongly prefer not to alter it. However, if it's already been altered by being recorded and played back in DSD, then our only reasonable goal is to either compare the alterations perpetrated by DSD to those wrought by PCM, and figure out which is less bad, or to determine whether the additional changes introduced by the PCM process make a significant contribution to the errors already introduced by the DSD process or not.)
 
In other words, I really don't care much whether a PCM recording of a DSD recording can accurately reproduce the analog output of the DSD recording; I'm more concerned with whether a DSD recording is closer to the original source content than a PCM recording. (After all, the goal here is to reproduce the original, and not to reproduce the DSD recording.)
 
Quote:
You totally missed the point of my post.

There is zero need for digital conversion to or from DSD to PCM to compare them via the method I proposed.

You would take the analog output of the DSD DAC and run it into a comparator that can switch between a pass-through and a level matched PCM ADC/DAC loop.

It doesn't matter how the signal gets to analog before the PCM ADC/DAC loop. You could run the same ADC/DAC loop to do null testing against any source, be it a master tape, vinyl, high res PCM, etc.

Obviously you would be testing a whole ADC/DAC loop rather than one or the other. But since the whole point is to establish the audible transparency or lack thereof of PCM audio relative to some other format, this isn't contrary to the test goals.

 
Apr 22, 2015 at 4:21 PM Post #4,549 of 18,259
   
And people still probably won't be able to tell the difference just like they can't with vinyl.  (I totally forgot that link was in the OP... 
frown.gif
  Oh well, it's been almost 5 years.)
 
Part of the point is basically to cede the high ground to the format under test (DSD) and see if the PCM conversion screws it up.  If it doesn't PCM passes.  If it does, then further tests are called for.
 
No one experiment is going to tell you everything you want to know but this one will tell you something pretty useful.  Having an analog master and making a separate DSD and PCM recording of it it a fine idea too, but it's much less practical. I'd say that's it not even necessary unless someone can first ABX the PCM conversion in the tape loop and then ABX it in revers with a PCM  source and DSD conversion in the middle.

 
I would agree with you except for one thing - I think it's unfair to "cede the high ground to DSD". If we do the test this way, and we do hear a difference, people are going to interpret the results as "OK, we proved that PCM messes things up". Unfortunately, at this point, the DSD proponents will consider their point proven, and nobody will want to go on and test the rest of the possibilities - for example, that maybe DSD messes things up even worse than PCM.
 
I say that, if what we really want to know is "which format, if either, is more accurate" then we should set the test up that way. I simply see little point in proving "whether you can make a perfect PCM copy of a DSD source" until we've determined that this is a worthwhile goal.
 
My starting point for all this is that I happen to already know that I can hear differences between different PCM DACs playing the same PCM content - at least sometimes and with some DACs, so I have little doubt that a PCM copy will NOT be "perfect".... as they say in court, I'll stipulate to that. So my only real interest would be in knowing if DSD is better, worse, or equal. (If it turns out that DSD is no more accurate than PCM, then we're wasting a lot of effort even discussing whether we can make a perfect copy of an already flawed copy.)
 
Apr 22, 2015 at 4:47 PM Post #4,550 of 18,259
When you say that not all DACs "sound the same", what do you mean? It doesn't seem like people are able to ABX differences between *competent* consumer-level DACs, so I'd hope that pro-grade stuff is even more transparent.


-I'll come out and say it - given that my Nuforce uDAC ($100) is run at sensible levels into a not-too-hard-to-drive pair of headphones, I am unable to distinguish its output from that of my ($2,000) Benchmark DAC2 HGC.

For the record - that is as expected... :)
 
Apr 22, 2015 at 4:48 PM Post #4,551 of 18,259
  When you say that not all DACs "sound the same", what do you mean? It doesn't seem like people are able to ABX differences between *competent* consumer-level DACs, so I'd hope that pro-grade stuff is even more transparent.

 
I disagree entirely.
 
Excluding some niche audiophile products and options, any "competent" DAC is going to have a very flat frequency response and very low distortion and noise. The differences you hear are going to be due mostly to the differences in how they handle transients. Virtually all modern DACs use digital oversampling - and the benefits of doing so far outweigh the drawbacks. However, all digital oversampling filters create errors in transient response. (They reproduce a sine wave perfectly but, if you feed in an asymmetrical and irregular waveform, they change its shape. Mathematically, all of the energy is present at the right frequencies, but some of it arrives at slightly wrong times.) If you look at the brochures for a lot of DACs, they will show you oscilloscope pictures of transients... which look sort of like a short sharp spike with some ringing before and after the main peak. That ringing is energy that has sort of landed in the wrong place - or, more correctly, at the wrong time. Since it's pretty close to when it should be, we don't hear it as "extra sound at the wrong times", but it does affect the overall sound of what we're listening to. In fact, by designing the filters a bit differently, you can trade off various aspects of these errors, but you can't entirely eliminate it. (For example, you can mathematically "trade" "more post-ringing for less pre-ringing".)
 
Many DACs even offer you the option of choosing between different digital filters, each of which reproduces transients slightly differently. (Typically, if you were to feed a sharp spike into the DAC, one filter will reproduce it with some ringing before and some after, while another option will have no ringing before the spike, but at the cost of more ringing after it, and a third filter may have less ringing altogether - but at the cost of a less flat frequency response.) While certain filter choices may actually affect frequency response, most alter transient response while avoiding changing the frequency response significantly.
 
If you listen to a commercial DAC which offers multiple filter choices, and switch between them, you will indeed notice a subtle difference. It's quite subtle, and you may only notice it with some recordings, and only with some headphones and speakers, but it is quite repeatable when it's present. Listen to a good recording with wire-brush cymbals; with some filters they'll sound more like actual metal while, with others, they'll sound more like a steam valve going "tssss, tssss, tssss". In some cases, the sibilants that accompany voices will sound different - and more exaggerated with some filters than with others. Remember that, mathematically, all of these choices are "really close" but none of them are "perfect". Of course different DAC products also have different analog circuitry, which also accounts for slight differences in sound.
 
I would also remind you that these differences are very subtle - and far less than, say, the differences between speakers, or headphones, or phono cartridges. To be honest, if I switch back and forth between filters, while a song is playing, the differences seem obvious - eyes open or closed - but, if I were to walk out of the room and come back, I very much doubt I could tell you which one was playing; that's the degree of difference we're talking about here.
 
The types of differences you hear between DACs are also rather different in type from thos you hear with amplifiers or speakers, so you may not notice them unless you actually listen for them. (And, yeah, they are about as miniscule as the differences I typically hear between DSD and PCM versions of the same recording.)
 
Apr 22, 2015 at 4:59 PM Post #4,552 of 18,259
-I'll come out and say it - given that my Nuforce uDAC ($100) is run at sensible levels into a not-too-hard-to-drive pair of headphones, I am unable to distinguish its output from that of my ($2,000) Benchmark DAC2 HGC.

For the record - that is as expected... :)

 
It really depends on what you're listening to and what you're listening to it on...
 
Since I happen to work for Emotiva, I have an Emotiva DC-1 DAC, and a pair of our Airmotiv powered monitors (which use the modern version of the air motion transformer tweeter). On those speakers, using the HDTracks 24/192 versions of the Grateful Dead albums, or the Eagles Hotel California, the DC-1 sounds very different from my Wyred4Sound DAC2, and both of those sounded very different than the DragonFly (v1) I had; likewise, all of those sounded different than my Benchmark DAC1Pre (the old series). However, on a pair of AKG K240 headphones, I couldn't hear any difference at all between any of them (on a pair of HiFiMan 500's you could hear the difference - but barely).
 
 
 
 
When I compare that DAC to my Wyred4SOund DAC2, on those speakers, the difference is pretty obvious, and both of those DACs sound pretty different than my old DragonFly (v1). Likewise, I used to have a benchmark Dac1Pre (the old series) - which I just sold - and it also sounded quite different.
 
Apr 22, 2015 at 5:25 PM Post #4,553 of 18,259
   
I disagree entirely.
 
Excluding some niche audiophile products and options, any "competent" DAC is going to have a very flat frequency response and very low distortion and noise. The differences you hear are going to be due mostly to the differences in how they handle transients. Virtually all modern DACs use digital oversampling - and the benefits of doing so far outweigh the drawbacks. However, all digital oversampling filters create errors in transient response. (They reproduce a sine wave perfectly but, if you feed in an asymmetrical and irregular waveform, they change its shape. Mathematically, all of the energy is present at the right frequencies, but some of it arrives at slightly wrong times.) If you look at the brochures for a lot of DACs, they will show you oscilloscope pictures of transients... which look sort of like a short sharp spike with some ringing before and after the main peak. That ringing is energy that has sort of landed in the wrong place - or, more correctly, at the wrong time. Since it's pretty close to when it should be, we don't hear it as "extra sound at the wrong times", but it does affect the overall sound of what we're listening to. In fact, by designing the filters a bit differently, you can trade off various aspects of these errors, but you can't entirely eliminate it. (For example, you can mathematically "trade" "more post-ringing for less pre-ringing".)
 
Many DACs even offer you the option of choosing between different digital filters, each of which reproduces transients slightly differently. (Typically, if you were to feed a sharp spike into the DAC, one filter will reproduce it with some ringing before and some after, while another option will have no ringing before the spike, but at the cost of more ringing after it, and a third filter may have less ringing altogether - but at the cost of a less flat frequency response.) While certain filter choices may actually affect frequency response, most alter transient response while avoiding changing the frequency response significantly.
 
If you listen to a commercial DAC which offers multiple filter choices, and switch between them, you will indeed notice a subtle difference. It's quite subtle, and you may only notice it with some recordings, and only with some headphones and speakers, but it is quite repeatable when it's present. Listen to a good recording with wire-brush cymbals; with some filters they'll sound more like actual metal while, with others, they'll sound more like a steam valve going "tssss, tssss, tssss". In some cases, the sibilants that accompany voices will sound different - and more exaggerated with some filters than with others. Remember that, mathematically, all of these choices are "really close" but none of them are "perfect". Of course different DAC products also have different analog circuitry, which also accounts for slight differences in sound.
 
I would also remind you that these differences are very subtle - and far less than, say, the differences between speakers, or headphones, or phono cartridges. To be honest, if I switch back and forth between filters, while a song is playing, the differences seem obvious - eyes open or closed - but, if I were to walk out of the room and come back, I very much doubt I could tell you which one was playing; that's the degree of difference we're talking about here.
 
The types of differences you hear between DACs are also rather different in type from thos you hear with amplifiers or speakers, so you may not notice them unless you actually listen for them. (And, yeah, they are about as miniscule as the differences I typically hear between DSD and PCM versions of the same recording.)

Type of filtering is the core of the Sygnalist HQ player http://www.signalyst.com/consumer.html . You can choose pre-ringing, post ringing - etc - according to the known defects of CDs that were recorded on known equipment producing these undesirable effets - and software will add exactly the opposite pre or post- ringing, effectively restoring the desired response. It takes one hell of a computer to do it right ( IIRC - 12GB RAM is minimum requirement ) - actually, they offer a music-only computer, optimized to cover for any setting that may actually tax that same computer to the max.
 
There is one fly in that ointment - you have to know on which machines your CDs have been made - and which "correction" filters are to be applied. About as useful and widespread as various non-exactly-RIAA-equalization curves for vinyl - some are known, most aren't.
 
Enter the simplicity of the DSD512 - NO filtering required anymore, no guesstimating which equipment has been used to produce your file. At least in digital domain - analog sections will still be the limiting factor and will ultimately set the SQ. Digital can be dirt cheap - analog most definitely not - if it wants to be really good. For fact - not in mfr's brochures or internet only.
 
Apr 22, 2015 at 5:29 PM Post #4,554 of 18,259
If you hear these differences under DBT conditions will all possible confounding conditions accounted for, then ok, otherwise it's hocus pocus. Even then, any differences could be unnecessarily deliberate (like if someone wants their fancy filter to sound better). In either case of upsampling or downsampling, the ringing artifacts should have energy concentrated towards 1/2 the lower Nyquist rate. To me that means people should be able to hear them in a downsample from hi-res formats to Redbook, and they don't (at least in DBT conditions), even in music with lots of bashing on cymbals.
 
Apr 22, 2015 at 6:22 PM Post #4,555 of 18,259
  Since I happen to work for Emotiva, I have an Emotiva DC-1 DAC, and a pair of our Airmotiv powered monitors (which use the modern version of the air motion transformer tweeter). On those speakers, using the HDTracks 24/192 versions of the Grateful Dead albums, or the Eagles Hotel California, the DC-1 sounds very different from my Wyred4Sound DAC2, and both of those sounded very different than the DragonFly (v1) I had; likewise, all of those sounded different than my Benchmark DAC1Pre (the old series).

 
Which one was accurately doing the d/a conversion? They can't all be performing to spec if they all sound different.
 
It's much more likely that high end DACs are jury rigged to alter the sound than midrange ones. Try comparing to the line out from an iPod or a decent blu-ray player and see which one matches those. Those are most likely to be accurate. I have an Oppo HA-1 and iPods and iPhones and Macs and an Oppo BDP-103D and a low end Sony blu-ray player... all of those sound exactly the same. If I bought something that wasn't audibly transparent with all of them, I would start worrying and would want to know which component was performing out of spec.
 
Apr 23, 2015 at 4:11 AM Post #4,556 of 18,259
Just to get some idea what is pre-requisite to listen to really small differences in sound ( "anything" - from ADCs, DACs, analog sections within, amps, etc ) , here interview with the designer of one of the best loudspeakers ever made : Andrew Munro 
 

 
I have auditioned this system about two years ago - and it is SUPERB - near field monitor. Trouble is, simply too small to be used in home, at normally larger listening distance from that just behind the mixing desk in a studio as seen in the video.
 
However, after having lounched an even smaller Egg, my hopes skyrocketed at the confirmation that Bigger Egg is scheduled to appear in forseeable future:
 

 
Now, I can "see" the Technics coaxial driver capable of 100 kHz  http://www.technics.com/us/products/c700/sb-c700.html response "mated" to the superb enclosure of Egg - the bigger, upcoming one - as the ultimate loudspeaker for serious work - AND home listening. Wishful thinking? - hope NOT !
 
Here the presentation for the original Egg 150 : 
 

 
The Audio Amateur and later The Speaker Builder magazines have been publishing articles on difraction effects in speaker (boxes) in 70s and 80s - and is really nice at least one manufacturer has been able to put a practical egg shape enclosure in regular production. Prior to that, "eggs" have been an exclusive domain for the "lunatic carpenter fringe" - and only few are known to be ever actually made - none I am aware of went into production due to the prohibitive cost.
 
Apr 23, 2015 at 5:43 AM Post #4,557 of 18,259
DACs that "sound the same" is not exact term for technical testing.
 
Then wtite methodics of testing must be noted how differ (%%, dB, V, ...) devices that used for test.
 
These devices measured by certified measurement tools with precision from 3 times more.
 
In this case then we read results of test, we can say, as example "for DACs with maximal difference in frequency responce 1 dB we can't listen difference".
 
If don't do it we can use 2 DACs what have difference 5 dB (even from one manufacturer) and say about listening of difference.
 
Or we use DAC with 1 dB difference, but our friends use with 5 dB. We will say "no difference", our friends will say "there are difference".
 
 
 
What is scientific experiment? It is experiment with repetable results independent of experimetators and number of performance.
 
Possible test one or two devices. It interesting. However here impossibly make general resume.
 
Apr 23, 2015 at 6:54 AM Post #4,558 of 18,259
Actually, you missed my point. Your proposed test would indeed do a very good job of confirming confirming whether digitizing an analog signal using a PCM ADC, then converting it back into analog using a PCM DAC, will result in any audible difference. However, that isn't the same thing as COMPARING DSD to PCM. In fact, what you're doing is comparing PCM to - nothing - or to a piece of wire. (Since I know that different PCM DACs do indeed sound slightly different, I'm pretty sure they each also sound different than a piece of wire.)
 
However, if we want to compare DSD to PCM, then what we need to determine is whether using a DSD ADC and a DSD DAC makes LESS of a difference to the sound than using their PCM equivalents.

 
If a PCM A/D->D/A loop (or DSD->PCM->DSD digital conversion) cannot be distinguished from the analog (or DSD) original in extensive blind testing, then that already strongly suggests that PCM is indeed good enough for audibly transparent reproduction. What would be the point of DSD making less than already zero audible difference to the sound, then, considering the practical disadvantages (such as more difficult and computationally expensive DSP) compared to PCM, and the less efficient usage of a given bit rate ? DSD64 requires 4 times as much space as Red Book, which could spent on more channels - a real, audible improvement - and still better than CD quality sample rate and resolution.
 
Apr 23, 2015 at 8:26 AM Post #4,559 of 18,259
  DACs that "sound the same" is not exact term for technical testing.
 
Then wtite methodics of testing must be noted how differ (%%, dB, V, ...) devices that used for test.
 
These devices measured by certified measurement tools with precision from 3 times more.
 
In this case then we read results of test, we can say, as example "for DACs with maximal difference in frequency responce 1 dB we can't listen difference".
 
If don't do it we can use 2 DACs what have difference 5 dB (even from one manufacturer) and say about listening of difference.
 
Or we use DAC with 1 dB difference, but our friends use with 5 dB. We will say "no difference", our friends will say "there are difference".
 
 
 
What is scientific experiment? It is experiment with repetable results independent of experimetators and number of performance.
 
Possible test one or two devices. It interesting. However here impossibly make general resume.

 
I agree. I could probably have put my question better as:
If one is really hearing differences among (pro-grade) ADCs/DACs, what are these differences, and are these differences expected given the specs? Moreover, are these differences still detectable in blind testing circumstances?
 
The answer I got, that I'm still not quite sold on, was pre-echo due to digital filtering. My doubt is due to my own testing of down-sampled (even sub-44.1) material where I can't hear such artifacts, and due to the theory about the frequency characteristics of pre-echo in cases of resampling to/from standard PCM formats. I of course could be convinced instantly if a hi-res sample could be provided that caused audible pre-ringing when down-sampled to Redbook.
 
Apr 23, 2015 at 8:26 AM Post #4,560 of 18,259

 
If you want to know whether there is a point to DSD, or rather want to disprove supremacy of DSD, you don't really have to compare it to anything at all.
All you need to do is to show whether the current system of digital capture and playback is transparent, which maverickronin's set-up lets you do. Whether the source is a DSD output, tape, vinyl, or directly from a pair of mic pre-amps in a concert venue, if you can't differentiate it from the PCM AD/DA loop, then that's that. You can't improve on transparency. One kind of homeopathic remedy doesn't work any better than another kind of homeopathic remedy.
 

Users who are viewing this thread

  • Back
    Top