DSD vs PCM

Clearly you aren't thinking of air-to-air missiles...or maybe those are not thought of as "guided". Most air-to-air missiles are "heat seeking", and the old Sidewinders could legendarily fly up a jet's tailpipe. Longer range air-to-air missiles today use some form of radar guidance (heat seeking is mostly for somewhat close targets).

Think air-to-air missiles are fired at some fixed locus in space? A little speed and a few deltas, as it tries to get away, and we're gonna bring that sucker down!

Of course, not all "guided missiles" are of the air-to-air variety. From the Wikipedia article on "Intercontinental ballistic missile" there's this:

"Strategic missile systems are thought to use custom integrated circuits designed to calculate navigational differential equations thousands to millions of times per second in order to reduce navigational errors caused by calculation alone. These circuits are usually a network of binary addition circuits that continually recalculate the missile's position. The inputs to the navigation circuit are set by a general purpose computer according to a navigational input schedule loaded into the missile before launch."

"Differential equations"? Sounds a lot like "a series of deltas" to me.

 

I've done a direct comparison (using JRiver and the iDSD) of

PCM 16/44.1 --> PCM 24/352.8 (plus digital filtering)--> SDM --> analog

versus

PCM 16/44.1 --> DSD128-->SDM(?)-->analog

Again, the DSD128 sounds "better". Perhaps "better" because there's white noise to mask digital aliasing artifacts? My point is that "better" has to come from somewhere. And this personal experience suggests there may be some factual basis for the assertion that DSD can add euphonic distortion.

So then to extrapolate, the new PS Audio Directstream uses what could be called "DSD640". I can see how that would make 16/44.1 sound "much better".

 

Without knowing the characteristics of the DAC and the DSD128 conversion routines, it is hard to know what you might be hearing. However, consider that the sigma-delta noise in DSD128 is not going to start ramping up until about 40 kHz, and is likely to be something like -100 dB to -80 dB by 100 kHz. Moreover, the conversion from 8 Fs PCM to some form of sigma-delta will also involve the addition of noise at a similar low level.

Having done numerous mic. feed tests (console vs. DSD vs. 24/192), I'm at a loss to these claims of euphonic distortion or silent scream. DSD just sounds like the mic. feed to me.

 

I've also done these tests and find it sounds like the mic feed as well.

 

And, just to be clear, I'm talking about tests using the same converter in DSD and hi-rez PCM modes, so dCS, Prism or Mytek, say: not one company's DSD converter compared to another company's PCM converter, which can mean very different sound.

 

I guess what I was getting at is should DSD "sound like the mic. feed" if one simply upsamples 16/44.1 to DSD128? Or should it sound indistinguishable? There are other people over at Computer Audiophile reporting similar results with upsampling 16/44.1 to DSD128. Generally speaking, upsampling to DSD64 is much more difficult to distinguish from the native 16/44.1.

And just to preemptively address any attribution of this phenomenon to "mid-fi" equipment, my test environment is as follows:

MacBook Pro 7,1 running JRiver Media Center 19.0.119
Clicktronic USB 2.0 cable
iFi iDSD RCA out (volume knob maxed)
Schiit Asgard 2 headphone amp
Audeze LCD-2 headphones

 

The issue is: what should the 16/44.1 sound like? Few have non-oversampling DACs these days, and even fewer have non-oversampling DACs with good linearity. So, when it comes to playing back 16/44.1 the signal must be transformed into something else. Most DACs, as indicated above, use a two-step conversion process: first to 8 Fs PCM for digital filtering, and then conversion to some form of sigma-delta. Even if you did an interim conversion to 2 Fs or 4 Fs, the DAC would still upconvert to 8 Fs prior to final conversion to sigma-delta (alternatively, on some chips you can do the upconversion to 8 Fs externally, use your own digital filters, and then feed the 8 Fs signal to the sigma-delta modulator directly). Now, how DSD is handled depends on the chip. If it is a 1-bit converter, it should be converted directly. If it is a few-bit converter, it too will get converted to this form of sigma-delta.

It should, therefore, be obvious that there is the potential for audible differences, given the different conversion methods, and the fact that there are different modulator designs, different digital filter designs, etc.

I find it interesting that you state that there is less noticeable difference if you convert to DSD64. This could be indicative of something going wrong in the conversion to DSD128. Have you tried other alternatives (Korg?) for converting to DSD128?

Seems like the iFi may use a custom part from Burr Brown, so I'm not sure we can find a datasheet. Does anyone know which DAC chip is inside the box?

I assume JRiver have not published anything on their conversion routines either?

 

Don't have a Korg lying around, sorry.

But getting lost in your informative explanation of "Fs" and how they figure into sigma-delta modulation is my initial subjective description that DSD128 sounds "better". If PS Audio's Directstream DAC (for example which upsamples everything to what could be called "DSD640") sounds subjectively "better" to others, will you be as seemingly convinced that something is amiss? Forgive me if that sounds like a loaded question. I originally said it sounded "better" but after many hours of listening in that mode, I've reverted back to 16/44.1 native mode and I will revise the word to be "different". The whole core premise of my observation is the notion of "conjured resolution", which is a claim that PS Audio is making about the Directstream:

Burr Brown DSD1793

Sorry, not aware of any published technical information from JRiver.

Thank you for taking the time to provide some technical information. I did the "just listen" thing that seems to be valued by many here. I would like to understand the underlying technical reasons that 16/44.1 upsampled to DSD128 sounds "better" at first listen, at least to me.

 

I agree with all your points, but I'll just comment that I don't believe that Pyramix or Sonoma are still selling any of their SACD workstations anymore, or if they are, they're custom-made, one-of-a-kind units. I think the whole DSD platform imploded five or six years ago (at least).

I agree, but I think the results sound good because of the mastering engineer making the decisions... not because of DSD per se. I suspect it would've sounded good in any format, including CD.

 

I don't think that is true.

 

Sorry, I meant the Korg Audiogate software:

http://www.korguser.net/audiogate/en/index.html

You can use it for free provided you have a Twitter account. Not sure what it automatically tweets, though!

Another alternative is foobar:

http://diyaudioheaven.wordpress.com/digital/pc-software/foobar-2000-for-dummies-part-2/

scroll down to 'real-time conversion from PCM to DSD'. DSD128 is supported.

These might offer you an alternative 'flavor' of DSD128.

It is not that I find the notion that one conversion path can sound better than another a problem, but the assumption that the DSD128 path introduces 'euphonic distortion'. This suggests that the DSD128 path is not accurate, but sounds nice.

With regard to PS Audio, they seem to have gone over the top with the hyperbole for their new DAC which is a shame, since Ted Smith, the designer, is a top guy, and has been heavily invested in DSD from the beginning (owning all the state-of-the-art DACs at various times). As I wrote in a thread on this DAC, my good friend, Gus Skinas, who has heard all the top DSD DACs heard a prototype in his fully tricked-out room in Boulder, and thought it was the best DAC he has ever heard. So, knowing Gus and Ted like I do, I'm prepared to believe that it is something special (I am waiting to hear it for myself), and if that is the case it will not need any marketing BS to sell it.

I also need to check with Ted why it upconverts to DSD640 before downconverting to DSD128, rather than simply upconverting to DSD128. I'm sure there must be a good reason.

I'm not sure that PS Audio is making a claim of 'conjured resolution' in the quote above. I read it as this converter removes an extra veil, allowing you to hear detail previously masked. That makes sense to me. I don't see anything about trying to restore the HF, for example, removed by 44.1 anti-aliasing filtering. Now, that would be a gimmick!

Hmmm! The online datasheet for that part does not seem to jive with iFi's specs. (32-bit, DXD, etc.).

It could simply be a case of preferring one set of operations on the data compared to another. Do we know the precision with which things like filter coefficients, etc. are calculated in JRiver compared to in the Burr Brown chip, etc.? It is possible to subtly change the audio within the same DAC just by changing the 8 Fs digital filters, for example. So, I can imagine more audible changes existing between different conversion paths using different filter/modulator designs.

Which is more correct? You would need to be familiar with the master to answer that! (And even then a new reference converter like the PS Audio may have you re-thinking things!)

 

I checked the Pyramix and Sonoma websites, and they look very sad and old, like they haven't been updated in years. I think everybody doing high-res mastering is just doing it all at 192kHz/24-bit PCM in a traditional DAW, and then SRC-ing it to DSD as a final step. I don't think anybody cares, as long as it's not an uprezzed 44.1kHz master.

 

Well for sure not "everyone" is using 24/192. Many are, but some small labels are different ; Reference Recordings - 24/176.4 and 2L - 24/352.8 for a couple, and as already mentioned Audio Fidelity analog tape direct to DSD. I'm sure there are more exceptions although you may be correct in a general sense.

 

All this remind me of the early and not so early years of CD where there was an incredible enchillada of possibilities with the implementation of the SPARS Code. The only code that seemed to make sense was AAD, what meant that the album was tracked and mixed down to analoge, and mastered to digital (using analog consoles and A/D converters at the very end) for CD release. Everything else could mean anything, beginning with DDD recordings that were not really such a thing as they had to be tracked to digital, converted back to analog, mixed-down with analog consoles and converted back to digital. ADD was even worse as one didn't know if the album was tracked to analog and mixed-down to digital, or the individual tracks were all-analog and just used digital equipment to make an edited master with all the tracks on it.

 

oh...

so that's it.

sigh. I am SUCH a complete naif.

thinking that PSA had (perhaps again) rewritten decades of science and information theory...

this time with remarkable!, groundbreaking!!, Heretofore Unheard-of!!! success in mining new resolution out of existing audio files...

all resulting from clever digital signal processing architecture, embodied in what seems (from a crumpled document alleged to have been found blowing across a parking lot in Boulder) to be referred to as the novel

"HyperBole with DSMS*" audio processor.


*(DirectSpin MarketingSpiel)


silly me...

 

If this has any relevance here, I was considering purchasing a Schiit Loki because my Arcam IrDAC does not handle DSD. Among my other questions, Jason Stoddard gave me an unbelievably honest answer to this question:

2) Right now I primarily buy PCM downloads from HDTracks, Acoustic sounds etc but there are quite a few DSD downloads being released these days. I've bought a couple and use Jriver media center to convert the DSD to 192/24 which allows me to play it from computer, through ARCAM IRDAC and into system. Wondering if I'd be better off going from computer through a Loki and therefore not have to convert. Would I achieve significantly better sound on DSD downloads by NOT converting to PCM?

His Answer

2)Not in our opinion.


I then asked him again because I figured he just read the question wrong but he reiterated that, in his opinion, there was no reason to get a DSD only DAC. I was so surprised and grateful by his willingness to lose a sale that I'm now considering the possibility of auditioning the Gungnir.

 

Implying that everyone with the requisite experience has the same views, opinions and/or preferences doesn't make it true, either.

Problem: You have a variable equaling a finite value in your equation, which disqualifies it.

If you were to say DSD64 is at least as good as 24/88.2 or 24/96, I would say, "...can be...". It still depends on one's preference.
If you were to say DSD128 is at least as good as 24/176.4 or 24/192, I would say, "...can be...". It still depends on one's preference.
But,
To say DSD??? is at least as good as 24/192, I would have no response as the statement is unspecific.

Given the same source, all things being equal, generally speaking, my preference is for 24/192 over DSD64, DSD128 over 24/96. So on and so forth...

 

Indeed. From the horse's mouth:

The quote is from this thread:

http://forums.stevehoffman.tv/threads/sacd-fundamentally-flawed.26075/

It's worth reading the first 100 posts or so at least, many of the technical points discussed in this thread were also covered there.

(Lee, you should remember, you were there... )

 

It's fairly well accepted in the recording community that DSD is on par or, among several, a bit better with 24/192. That matches my own personal experience trying different things and listening to mic feeds. Both are very good when done well so it's not worth having a fierce debate over.

 

That was from 2003, sort of an eternity in digital terms. So it's pretty telling that Bruno is now making DSD A to D converters for Grimm Audio.

http://www.grimmaudio.com/pro-products/converters/ad1/

Bruce Brown thinks this is the best converter on the market.

 

A couple of good points, the first up is from our host Steve on the CCR improvement from the SACD layer.

http://forums.stevehoffman.tv/threads/sacd-fundamentally-flawed.26075/page-7#post-482300

Second, Bruno made a key point earlier in the thread where he mentions that "good electronics" often plays a big role in our evaluations.

Bruno, again this is 2003 so his thinking may have evolved, suggests that DVD-Audio has not had good implementations (he implies that several times) so it may be possible some of us prefer DSD due to electronics implementations, not the superiority of the format. I think Bruno even says that had Meitner done a DVD-Audio process then that may have been the preferred format sonically. Quite a compliment for Ed Meitner.

Thanks Don for digging up the old thread. My thinking has evolved a little as well and it's neat to see the old discussions.

 

I am not comfortable with the Grimm guys calling DSD 1-bit PCM (see the DSD FAQ section on the Grimm website - use Lee's posted link). DSD is Pulse Density Modulation (PDM) and is NOT PCM.

 

True.

 

Not necessarily. I've seen long on-line discussions by "experts" at WBF and Audio Asylum arguing this point with no consensus reached. More than that, saying the "recording community" has accepted that DSD sounds better than 24/192 PCM would require a very selective sampling of recording engineers, since most act as if DSD doesn't even exist, and some very well regarded engineers (e.g. Barry Diament and Keith O. Johnson) apparently prefer PCM.

 

Nevermind. My concern was with the generalization(s) used: "DSD" with no specified resolution, "recording community" with no specified application. As stated, it's impossible to engage in debate, fiercely or otherwise. I was hunting for clarity.