Bel Canto's take on DSD

[Superdad]

First order filter is 6 dB/octave and you need to reach at least 144 dB attenuation by new fs minus original fs/2.

 

44.1 = -6

88.2 = -12

176.4 = -18

352.8 = -24

705.6 = -30

1411.2 = -36

2822.4 = -42

...

some pretty big number (about 7.2 decades)

 

That's what I thought. So 2 to the 24th power is 16,777,216 times 44,100 = 739,875,225,600. Thus a first order filter will work nice on a 739.8752256 Gigahertz stream! I guess even HQPlayer on supercomputer is not yet up to that task!

[Miska]

That's what I thought. So 2 to the 24th power is 16,777,216 times 44,100 = 739,875,225,600. Thus a first order filter will work nice on a 739.8752256 Gigahertz stream! I guess even HQPlayer on supercomputer is not yet up to that task!

 

Good thing is that it gets easier pretty fast when increasing analog filter order and digital oversampling ratio together.

 

With third order analog filter already meets the spec with 128x oversampling. Of course running 24-bit R2R converter accurately at such speed is already a pretty tricky challenge, so practical way is to use less bits with noise shaping at that speed and add some noise suppression/EQ. Then you don't need to settle to 1/2^24 accuracy per each sample...

[Andrew Allen Ballew]

Um, you do realize that the Teac UD-501 uses Delta Sigma modulation for PCM? So the noise you are looking at has nothing to do with PCM.

 

The simple fact is PCM has nowhere near the ultrasonic noise content of DSD. Furthermore, you are looking at the level of noise AFTER low pass filtering!!! Of course it is that low in this case!

 

And everything that Bel Canto says about the technical aspect of DSD is correct. Now subjectively whether it sounds better played back as PCM or native DSD, that is debatable. It is true that depending on the hardware, DSD can and does often sound better converted to PCM first. Many of these hybrid chipsets that support both are optimized for PCM, and don't have optimal conditions for the best DSD playback. Heck, some of them (ESS, ahem) convert internally to PCM first anyway. Until ESS comes out and tells us otherwise, which they won't, since they reveal almost nothing about the actual goings on in their chipset, I will call foul on their 'native' DSD support. Volume Control, Oversampling, Filtering, and ASRC on DSD and it is still native? Yeah right.

 

Alright. Enough reality injection for today. Y'all carry on with your little arguments.

 

 

 

Optimal DSD playback, my ass. This isn't DSD playback at all.

 

As for the noise level in the inaudible spectrum, here's an interesting measurement Archimago made for his Teac UD-501 DSD DAC.

 

[ATTACH=CONFIG]11885[/ATTACH]

 

As you can see it's not all that much different from PCM, and is still lower than -100dB at 100kHz.

[InfernoSTi]

There is good and bad in every format: PCM, DSD, and even vinyl. It sounds like the "reasons" vinyl isn't as good as PCM are being recycled to DSD. Too bad…and it goes to show we don't listen to test tones but music, actually.

 

It seems that each has a "technical" reason why it does/does not sound as good as the other. Then there is always a group that says "it sounds better" despite the technical arguments. I am starting to move away from the "one formate is better" camp and into the "this is the golden age of music" camp.

 

I have never seen so much music available to so many for as little/as much as one wishes to pay. And the sound quality of an average system today is much, much better than an average system of the past. The old transistor radio is a far cry from today's iPhone/earbud arrangement.

 

So if you want to only go in 5/10ths of state of the art, you get great sound. If you want to be 8/10ths, you get amazing sound. In the past, it seemed that 7/10ths or better was needed to get such good sound. And lets not forget that we can have our collection with us where we want today.

 

With the right equipment, it will sound amazing with either format...

 

Best,

John

[Hiro]

Um, you do realize that the Teac UD-501 uses Delta Sigma modulation for PCM?

 

You do realize that all BD players, portable players, and 99% of DACs out there are using Delta Sigma Modulation for PCM?

 

Enough said.

[matthias]

The simple fact is PCM has nowhere near the ultrasonic noise content of DSD.

 

Not correct:

Playback formats on a TEAC UD-501 (TI PCM1795) - Blogs - Computer Audiophile

 

KR

Matt

[matthias]

And everything that Bel Canto says about the technical aspect of DSD is correct.

 

The beauty of the conversion to 24/176.4 PCM is that all of the characteristics of the original DSD file are retained and you can get the sound characteristics of DSD without compromise to playback of any PCM files, either CD or at higher bit rates of 24/88.2 and above.

 

This is not correct. Conversion from DSD to PCM means you get PCM and the sound characteristics of PCM.

 

KR

Matt

[Superdad]

Not correct:

Playback formats on a TEAC UD-501 (TI PCM1795) - Blogs - Computer Audiophile

 

KR

Matt

 

Matt:

Those graphs of Miska's were from zero-to-2.5MHz. That scale is interesting on a macro level, but when most of us think about out-of-band noise, we are thinking about what goes on between say 22Khz 125KHz. It is more fair to look at graphs on that scale. My power amp may only be down about 2.5dB at 200KHz, but ears and tweeters are dead silent LONG before then.

 

I'm not anti-DSD in the least. It is just that making really good (measuring and sounding) filters is a challenge no matter what format we are talking about. It only becomes much easier once you get the data rate up in the 12-25MHz range (single bit or multi-level). That's where computing horsepower and smart s/w design come in. Just ask Jussi.

 

As much as I like convenience, I also like the idea of having two DACs: one optimized as a pure DSD unit (Miska's DIY or Sonore's PureDSD); and one optimized as high-speed, PCM performance machine (I'm not mentioning R2R for PCM because even that topology can be bettered with more discrete parts, leaving behind all standard DAC chips.)

 

Some of the debate reminds me of gas versus hybrid versus pure electric cars. No matter how you slice it, the hybrids are a compromise of several factors. Were I a rich man (I'm FAR from it), I might have a nice Porsche AND a Tesla.

[matthias]

As much as I like convenience, I also like the idea of having two DACs: one optimized as a pure DSD unit (Miska's DIY or Sonore's PureDSD); and one optimized as high-speed, PCM performance machine (I'm not mentioning R2R for PCM because even that topology can be bettered with more discrete parts, leaving behind all standard DAC chips.)

Some of the debate reminds me of gas versus hybrid versus pure electric cars. No matter how you slice it, the hybrids are a compromise of several factors. Were I a rich man (I'm FAR from it), I might have a nice Porsche AND a Tesla.

 

Alex,

 

yes, it is nice to have two DACs, but having only one DAC I would choose a DAC that favors DSD just a little bit more. With a good software player and a pure DSD-DAC up to DSD512 the drawbacks for PCM playback may be minimal or even non existent.

 

A Porscha and a Tesla is very nice match...........

 

ATB

Matt

[Paul R]

Hi Alex-

 

I am really coming to the conclusion that a DSD DAC plus software that can convert PCM to DSD provides about the best affordable sound available.

 

Perhaps I am wrong, because I don't have a Weiss DAC or dCS DAC available to compare to, but... Converting everything to DSD seems to really make a huge difference. Cleaner, more natural bass - better PRAT - cleaner treble - much better attacks/transient processing, and more. Native DSD or DSD128 is the closest to a really good vinyl rig I think Inhave ever heard. Better than most since there is no wow and flutter or surface noise.

 

Huh? This seems to be true despite knowing the "native" DSD I have was mastered in a PCM workflow. It even seems to be true when converting 192k tracks to single or double DSD. That really doesn't make sense, at least on the surface of things. Which always means there is something there I do not know.

 

But everyone here loves the "sound" of DSD. Stop dead in your tracks and listen kind of thing. That is true also with particularly well recorded PCM, which shows the least improvement when converted to DSD. But it also often true for plain old 16/44.1k redbook music converted to DSD.

 

All of which is making me suspect the much simpler filter needed for DSD is at the bottom of this. No proof yet, but it does seem like a reasonable path of inquiry. Filters really could be the biggest difference in the technologies,I think.

 

Just observations and opinion above at this point. Especially given the radical cost differences between PCM and DSD DACs at this time. The cost difference is really starting to make sense from the AMR viewpoint to me now.

 

We do not think much of downloading, much less storing on disk, a few gigabytes these days. Such would have been almost unthinkable for a home system only 20 years ago. Prohibitively expensive at least. Perhaps DSD has always been "the best" technogy, but was not implementable because of technological costs of storage space?

[mav52]

Hi Alex-

 

I am really coming to the conclusion that a DSD DAC plus software that can convert PCM to DSD provides about the best affordable sound available.

 

Perhaps I am wrong, because I don't have a Weiss DAC or dCS DAC available to compare to, but... Converting everything to DSD seems to really make a huge difference. Cleaner, more natural bass - better PRAT - cleaner treble - much better attacks/transient processing, and more. Native DSD or DSD128 is the closest to a really good vinyl rig I think Inhave ever heard. Better than most since there is no wow and flutter or surface noise.

 

Huh? This seems to be true despite knowing the "native" DSD I have was mastered in a PCM workflow. It even seems to be true when converting 192k tracks to single or double DSD. That really doesn't make sense, at least on the surface of things. Which always means there is something there I do not know.

 

But everyone here loves the "sound" of DSD. Stop dead in your tracks and listen kind of thing. That is true also with particularly well recorded PCM, which shows the least improvement when converted to DSD. But it also often true for plain old 16/44.1k redbook music converted to DSD.

 

All of which is making me suspect the much simpler filter needed for DSD is at the bottom of this. No proof yet, but it does seem like a reasonable path of inquiry. Filters really could be the biggest difference in the technologies,I think.

 

Just observations and opinion above at this point. Especially given the radical cost differences between PCM and DSD DACs at this time. The cost difference is really starting to make sense from the AMR viewpoint to me now.

 

We do not think much of downloading, much less storing on disk, a few gigabytes these days. Such would have been almost unthinkable for a home system only 20 years ago. Prohibitively expensive at least. Perhaps DSD has always been "the best" technogy, but was not implementable because of technological costs of storage space?

 

Paul what are you using for paying DSD files these days ? the iDSD

[Paul R]

Paul what are you using for paying DSD files these days ? the iDSD

 

Most of what we have here right now is borrowed, but:

 

Schiit Loki

iDSD (plus iUSB Power & cable)

Teac UD-501

Benchmark HGC-2

 

We also have a few DSD capable AVRs here, which is mildly astonishing. They need to be fed from a network server, not a computer via USB though.

 

We evaluated a MyTek as well, and are waiting on a Korg and borrowed Ayre QB9 w/DSD upgrade.

 

The reason we are going through so many is pretty simple Chris. I had a lot of trouble believing that DSD could sound so darn good, "native" format or up sampled via JRMC. When I say "so darn good" think jaw dropping astonishment. The other reason is that price does not seem to be as good an indicator for DSD as for PCM sound at this time. The best sound here seems to be coming from the iDSD unit, which is almost the least expensive of the bunch. That doesn't make a lot of sense, unless there is an operator in effect I am not accounting for.

 

When I listen to these DACs and here them all sound better than in PCM mode, every single one of them, I start thinking my assumptions are incorrect. The iDSD in PCM sounds almost as good as the Proton, though perhaps it does not get the PRAT quite as well. In DSD mode, up-sampled to DSD64, it sounds much better. Haven't made up my mind about up sampling to DSD128 yet- it definitely sounds different, but it also sounds like it has lost something.

 

DSD64 "native" music I have around sounds fantastic. It utterly draws me in and makes me forget I am supposed to be listening to the equipment. I tend to play album after album. Definitely has the undefined "something" that really good vinyl has. (grin)

 

More, one of the touchstones for how well a DAC performs (at least to me) is how different each track on a album sounds. With PCM, I can usually tell they sound different. With DSD, there is not question about each track sounding uniquely individual. I believe it makes it possible to hear straight through to the recording, even with very modest equipment.

 

Now, in theory, 24/176K and DSD64 should sound the same, but whoo boy, they do not.

 

Too many questions, and it takes time to dig though all this and find the answers, or at least, some of the answers.

 

Oh another thing, all the DACs sound much more alike playing DSD than they do playing PCM. I don't know what the means exactly, but it is another datum. I really hope companies lie Bel Canto get "on the bandwagon" soon. It think DSD in DACs may open up a whole new environment for Audiophiles. We need companies like Bel Canto, and PS, and Berkely to really pin down the best it can be.

 

-Paul

 

Oh yeah, this is definitely exciting too.

[mav52]

Thanks Paul....

[Superdad]

Hi Alex-

 

I am really coming to the conclusion that a DSD DAC plus software that can convert PCM to DSD provides about the best affordable sound available.

 

 

Paul:

I am inclined to agree with you, but having being schooled by John Swenson about the limitations inherent, I do not think that the currently available, conventional chip-based DSD DACs bring us to what is really possible. I think when people get to hear something like Miska's new DIY boards or Sonore's PureDSD board, then even more ears will be open to the format.

 

I am concerned that too much weight is being given to "1-bit DSD" as opposed to the fact that the benefits are coming primarily from getting to very high rates which can be more easily filtered. What I am saying is that I think the real future is with high-speed, multi-level "DSD" (ala Miska's board, though it uses opamp output stage, so perhaps is more demonstration than ultimate statement). Doing the delta-sigma modulator (of PCM material) to 22.5792MHz (aka DSD512) in s/w certainly saves on DAC processing h/w cost, but properly done multi-level handling (registers, flops, isolation, and especially well-designed filters!) is not going to show up in $500 products.

 

(And of course the new Ted Smith PS Audio piece is an interesting and ambitious case, though I hesitate to call it a "pure" DSD solution, given its multiple modulators that even DSD material passes through.)

 

Yet coming back to your statement, I do agree that PCM>DSD in s/w, when fed to a cleanly done, simple DSD-only DAC--with some care put into the output stage--should offer great sound at affordable prices. HQPlayer on today's cheap computers sure replaces a LOT of expensive silicon in the DACs of the past decades! (I don't think I'd want to be dCS.)

[matthias]

(And of course the new Ted Smith PS Audio piece is an interesting and ambitious case, though I hesitate to call it a "pure" DSD solution, given its multiple modulators that even DSD material passes through.)

 

According to the comment of Ted Smith on the PSAudio forum we have to assume that it is not a pure DSD DAC:

 

It's important to not get religious about any particular part of an audio system, there are many ways of putting a fine system together: it's the final result that matters.

 

As far as we can tell it's not that PCM is evil, it's that a sample rate that's too low rate looses some information that matters. For the final conversion to analog a single bit signal has big advantages in simplicity and purity of the analog output. We take advantage of the strengths of both fast DSD (>22MHz) and wide PCM (>24bits).

 

In more detail:

 

As J.P. says the DirectStream uses at least 24 bits for the PCM upsampling to 176.4k or 192k. There are a lot more bits used in the filters, but there's no reason to keep all 60 or 66 bits between stages... The final upsampling of the 24 bit PCM or the incoming single or double rate DSD involves an wide IIR filter. We keep a lot of bits when applying the IIR filter, then doing the volume control and then all of the way thru the sigma delta modulator that produces the final double rate single bit DSD.

 

In the case of the DirectStream we know that we couldn't have a clean output stage without double rate DSD being the final result. We also know that converting to single bit over and over again along the processing path would compromise the signal. By allowing the signal to get both faster and wider we loose no information. If we'd have gone to, say quad rate DSD we'd have more jitter so double rate DSD seems to be the sweet spot for this style of implementation.

 

-Ted -

[MikeJazz]

HQPlayer on today's cheap computers sure replaces a LOT of expensive silicon in the DACs of the past decades! (I don't think I'd want to be dCS.)

 

Superdad,

if you really think that the dCS level of quality can be obtained by a combination of HQPlayer and a "purist" DSD DAC, I will be very very interested in testing this in the future. Because some of the best digital I heard come exactly from dCS and desperately out of my reach!!!

[Paul R]

Superdad,

if you really think that the dCS level of quality can be obtained by a combination of HQPlayer and a "purist" DSD DAC, I will be very very interested in testing this in the future. Because some of the best digital I heard come exactly from dCS and desperately out of my reach!!!

 

I would dearly love to hear some of the inexpensive DSD DACs up against the top level PCM DACs like dCS. From memory, I think they would be close - way closer than I would have believed a year or two ago. I would also like to hear them up against really top flight vinyl.

[Hiro]

I would dearly love to hear some of the inexpensive DSD DACs up against the top level PCM DACs like dCS.

 

I would like to see a shoot-out between the 5bit/64fs dCS and 1bit/128fs Lampizator. According to Bruce Brown the DSD-only Lampi bested the $13000 Playback Designs DAC (http://www.stereomojo.com/Lampizator%20DSD%20DAC%20Review/LampizatorDSDDACReview.htm). Would it be able to successfully compete with even more expensive dCS?

[Paul R]

Well, I walked into a lucky buy yesterday, found a Bel Canto 2.5 DAC for sale at a neighbor's garage sale. When I mentioned they could get a lot more money for the thing than the price tag asked for, they asked if I had two twenties or not? So... a really *really* good deal. Looks to be in all working condition, but I am a little disappointed in the USB. Seems to be limited to 24/96K?

 

In any case, I plan to try listening to this DSD version of The Carpenters - Singles 1969-1981-DSD, purchased from AcousticSounds tonight through the Bel Canto (down sampled as noted on the Bel Canto website), the Wavelength Proton (down sampled the same as for the Bel Canto) and through the iDSD at DSD2X resolution. The metadata in the music files tell me it was created with Korg's Audiogate software, and the literature claims it is from the original DSD.

 

We shall see - or rather - hear.

 

-Paul

By the way, least anyone think I am insanely greedy, I doubt I will keep the Bel Canto DAC, but it will make a nice donation to our church's spring bazaar.

 

[Superdad]

Superdad,

if you really think that the dCS level of quality can be obtained by a combination of HQPlayer and a "purist" DSD DAC, I will be very very interested in testing this in the future. Because some of the best digital I heard come exactly from dCS and desperately out of my reach!!!

 

Well I did not exactly say that Mike. We all know therer are many other elements to a great DAC besides its digital format topology. Power, interstage isolation, clocking/reclocking, output stage, etc. It is not as if great performance has suddenly become cheap and easy! I just meant that now the computational workload can--at least for computer-willing audiophiles--be shifted to software.

 

--Alex C.

[wisnon]

I would like to see a shoot-out between the 5bit/64fs dCS and 1bit/128fs Lampizator. According to Bruce Brown the DSD-only Lampi bested the $13000 Playback Designs DAC (Lampizator DSD DAC Review). Would it be able to successfully compete with even more expensive dCS?

 

I own a similar version DSD-only Lampi to what Bruce reviewed, but with NOS Tungsram tubes and Duelund copper caps. It is incredible at DSD128...untouchable, really.

 

Simple LP Filtering, well done, is the key...along with a hugely overspecced PSU.

[Paul R]

Well, I walked into a lucky buy yesterday, found a Bel Canto 2.5 DAC for sale at a neighbor's garage sale. When I mentioned they could get a lot more money for the thing than the price tag asked for, they asked if I had two twenties or not? So... a really *really* good deal. Looks to be in all working condition, but I am a little disappointed in the USB. Seems to be limited to 24/96K?

 

(*sigh*) My Daddy was right, if it seems to good to be true, it is. There is no usable output from this DAC, the output section has been throughly fried. Oh well, I will contact Bel Canto and see how much a repair would cost, but I bet it will be close to the price of buying the DAC new. Sorry guys... Guess one really does get what one pays for.

 

Paul

[wisnon]

Sorry to hear that....

[Stiv]

I would dearly love to hear some of the inexpensive DSD DACs up against the top level PCM DACs like dCS. From memory, I think they would be close - way closer than I would have believed a year or two ago. I would also like to hear them up against really top flight vinyl.

 

yawn.

[Paul R]

Have you tried Melatonin for that?

 

yawn.