What is the latest for support of DSD256 and higher sample rates?
The DAC market is growing and most are now supporting DSD256/512 as a minimum.
We support single and double rate DSD and haven’t added support for higher rates. There’s very little in the way of music out there and the vast majority of libraries haven’t much, though I know some people have it.
I’ll let Ted answer from a technical standpoint.
Short answer is that the DS is unlikely to support higher than DSD128.
Tho we could in theory downsample higher rate material that’s not what people expect and it would hurt the sound quality for everyone else. The DS was designed for double rate DSD thru and thru. From a theoretical point of view there’s a sweet spot in the sample rates. Slightly simplified: As you go higher in sample rate you get a better S/N ratio from the added resolution, but you also get more noise from the added jitter. There’s a crossover some where between DSD128 and DSD256 where the extra speed no longer gets more S/N than the added jitter takes away.
Most A/Ds and D/As out there that aren’t discrete are using single chips that support higher sample rates, but tho the sound may be different between DSD512, DSD256 and DSD128 they aren’t delivering lower S/N ratios at these higher rates. Pro converters from Merging tech, etc. get better sound from higher sample rates, but that isn’t necessarily because of those sample rates being inherently better.
I’m loath to add a feature just for nicer marketing that detracts from the sound quality compared to not having that feature available.
Ted, you are not alone in your assessment of the potential benefits of quad DSD’s S/N vs. the tradeoffs it presents in additional jitter. I have heard exactly the same explanation given from the designers of some other very high end DACs, although I would prefer not to mention any names since I was not given permission to share their thoughts. Like you they have decided against supporting anything higher than double rate DSD in their DACs.
Still, it would be nice if we had the option. I was able to listen to some QuadDSD recently, and it was pretty impressive. So much so, that I have now bought several downloads from 2L and Blue Coast, as I wait for delivery of my new Merging Technologies NADAC. Any day of the year, I would have preferred to add another DS for my headphone listening station, but I decidedly wanted the 256 option.
Isn’t it a question of what the recording format was, how important it is, that the higher sampling rates are provided?
I’m personally not sure when formats from single to quad rate are provided, which one is the original recording format and therefore best, and if the other sampling rates are produced just by forth and back converting between DSD and PCM.
Probably 2L, Blue Coast or Acousic sounds really record in DSD256 or begin to do now, but how are/were their lower rates produced then? To my knowledge it’s not possible to downconvert DSD without going to PCM and DSD again, right?
The discussion is interesting to me and I would have some opinions if I already understood the principle of DSD recording situation and how different DSD rates are produced. Maybe someone can clarify.
Things get pretty murky with sigma delta converters - tho we often speak about most DAC chips being PCM, they aren’t really. They use multi bit (5,6,7) sigma delta conversion and small DACs, sort of a hybrid between “pure PCM” and “pure DSD”. And on the other hand the older Merging Tech Pyramix editing stations do their work in DXD which is 352.8k PCM and but the results are released as DSD and many fine recordings were done this way.
Don’t assume that math or PCM is always involved when converting different DSD sample rates - some masterers use A/D -> analog console -> D/A to convert formats as well as for their mixing/mastering.
With higher rate DSD (say double and quad rate) there are some advantages to the A/D running that fast, but the current chips that support quad rate A/D are still using multibit sigma delta converters and then math to squeeze it to single bits at a high rate format (DSD) or to squeeze the results in to a lower rate but wider format (PCM) I don’t know the details of, say, Merging Tech’s quad rate A/D’s but the details I’ve heard imply that the conversion to single bit isn’t as direct as some high end single or double rate DSD A/Ds - that isn’t to say that it’s worse, but just that there are a lot of details that need to be taken care of at these frequencies and that a dogmatic view may restrict the possibilities too much for practical devices.
Tho I greatly respect David Robinson and his work for DSD I’m with Andreas on this one and I don’t think that we have a proper A/B comparison for comparing the best that double rate can offer vs the best that quad rate can offer. Some people still prefer music that was converted with an excellent single rate A/D over that that was recorded with a good quad rate A/D and IMO rates above double rate are too new to be mature enough to be excellent yet.
As a recording technology quad rate may have technical advantages over single or double rate, but as a delivery method I don’t know why we should look down on excellent conversion to double rate. As I’ve tried to explain elsewhere it’s not a question of the purity of one bit, but the high sample rate and gentle filters allowed by DSD style recording/processing/delivery that seem to really matter.
I want to be clear I am not “looking down” on anything here. I am a big believer in execution and results, not merely a singular focus on the technology used. As I said, I have heard some 256 playback that simply sounded remarkable. That could be a result of any one or many of the factors that Ted well articulated. But it was good enough to make me want to have 256 playback in my system. And it is entirely possible that I will put the NADAC in my system, discover that it doesn’t sound as good on PCM as my DS, and then have a difficult decision to make.
But I do believe that folks like Blue Coast and 2L will continue to push the technology deployment envelope on the recording side, and that along the way, there will be lessons to learn about delivering that stretch into the playback side. The more flexible our home tech can be, the longer we can reap the benefits of the wonderful sound PSA is delivering. I for one have strong faith in the ears at PSA, and am very hopeful for a long-time-frame use of my DS. Of course, there is nothing to keep me from using it for years regardless of what other additional technologies come into play. i would just love to think of the DS remaining at the heart of it.
Ted Smith said Things get pretty murky with sigma delta converters - tho we often speak about most DAC chips being PCM, they aren't really. They use multi bit (5,6,7) sigma delta conversion and small DACs, sort of a hybrid between "pure PCM" and "pure DSD". And on the other hand the older Merging Tech Pyramix editing stations do their work in DXD which is 352.8k PCM and but the results are released as DSD and many fine recordings were done this way.Don’t assume that math or PCM is always involved when converting different DSD sample rates - some masterers use A/D → analog console → D/A to convert formats as well as for their mixing/mastering.
With higher rate DSD (say double and quad rate) there are some advantages to the A/D running that fast, but the current chips that support quad rate A/D are still using multibit sigma delta converters and then math to squeeze it to single bits at a high rate format (DSD) or to squeeze the results in to a lower rate but wider format (PCM) I don’t know the details of, say, Merging Tech’s quad rate A/D’s but the details I’ve heard imply that the conversion to single bit isn’t as direct as some high end single or double rate DSD A/Ds - that isn’t to say that it’s worse, but just that there are a lot of details that need to be taken care of at these frequencies and that a dogmatic view may restrict the possibilities too much for practical devices.
Tho I greatly respect David Robinson and his work for DSD I’m with Andreas on this one and I don’t think that we have a proper A/B comparison for comparing the best that double rate can offer vs the best that quad rate can offer. Some people still prefer music that was converted with an excellent single rate A/D over that that was recorded with a good quad rate A/D and IMO rates above double rate are too new to be mature enough to be excellent yet.
As a recording technology quad rate may have technical advantages over single or double rate, but as a delivery method I don’t know why we should look down on excellent conversion to double rate. As I’ve tried to explain elsewhere it’s not a question of the purity of one bit, but the high sample rate and gentle filters allowed by DSD style recording/processing/delivery that seem to really matter.
Thanks much for this insight Ted!
The only thing I’m not sure I got right is:
Do you say, up/down conversion of DSD data is not necessarily a “bad thing” and that it doesn’t matter in which sample rate a recording was made? So what I read out of another of your Posts, that there are more advantages for single/doublerate than for higher sampling, combined with the previous assumption means, that a single or double rate download would theoretically sound better or very similar to a higher rate file, no matter in which rate it was recorded or how it was converted (at least if it’s a file provided by the Producer)?
How about downconverting higher rate DSD files to single or double rate with Jriver to make it playable with the DS or the Bridge: with Jriver this means -6dB by the PCM steps involved as I understood. Good or bad? Any better conversion option?
Thanks so much again.
jazznut saidThere's a big difference between theory and practice - in theory sample rates above quad rate are noisier with equivalent technology than single or double rate. In reality it all depends on the particular converters, the skill and care of the person doing the conversions and of course the idiosyncrasies of the DAC...Do you say, up/down conversion of DSD data is not necessarily a “bad thing” and that it doesn’t matter in which sample rate a recording was made? So what I read out of another of your Posts, that there are more advantages for single/doublerate than for higher sampling, combined with the previous assumption means, that a single or double rate download would theoretically sound better or very similar to a higher rate file, no matter in which rate it was recorded or how it was converted (at least if it’s a file provided by the Producer)?
How about downconverting higher rate DSD files to single or double rate with Jriver to make it playable with the DS or the Bridge: with Jriver this means -6dB by the PCM steps involved as I understood. Good or bad? Any better conversion option?
If the sample rate remains high and all filtering is gentle up/down conversion isn’t necessarily evil. As I hinted above the biggest contributor to a great recording is the performance and the care of the recording engineers, etc. There are no simple rules about which conversions are good or bad. The devil is in the details. Personally I’d not use JRiver’s DSD conversion (unless you like the sound or course) - the DS is better at upsampling PCM or single rate DSD to double rate DSD than JRiver or most other options available.
The -6dB comes converting PCM to DSD: SACDs define 0dBFS as 1/2 of all 1’s or all 0’s. Ironically tho you should be able to get this back when converting back to DSD I don’t think that’s an option in JRiver.
Since I primarily use a DirectStream DAC I’m not buying quad rate recordings right now. Most of the people who produce quad rate recordings will also release them in a variety of lower rates and they are already the kind of people that take care when recording and processing.
If your DAC supports quad rate, you can, of course make up your own mind about the relative sound quality for you with various rates.
Thanks Ted, just to clarify two possible misunderstandings:
I use the DS, no other DAC and I don’t up or downconvert what could be done by the DS
I only downconverted a DSD128 file I had to DSD64 to prepare for the Bridge use. That’s where i recognized the -6DB and due to the Jriver downconversion from DSD128 to DSD64 by internally going over PCM (so I understood), I guessed there might be other shortcomings in that conversion, that make the resulting DSD64 file sounding worse than an official DSD64 download.
Do you know a better Option to downconvert existing DSD>64 to DSD64 for Bridge use than with Jriver?
I don’t know if AudioGate has that DSD to DSD conversion but if it does it may well be a better converter. Foobar2000 has an on-the-fly DSD->DSD convertion plugin (ASIOProxyInstall from the foo_input_sacd sourceforge project) but I don’t know how good it is. As always with these things you just need to try them yourself to find the one that you like best - there’s little cost involved unlike trying a bunch of DACs…
But if you are using JRiver already, as far as convenience is concerned, doing the conversion in JRiver wins.
I have several dacs that play it in native and honestly I hear no difference period
I have tried with headphones and speakers
about all I can say is perhaps it’s ambience that’s improved but I cannot nail it down.
The files I have tried were downloaded in all three formats and I can tell 64 to 128 but not 256
the ds dac shows any change more expensive dacs do in format changes so for me it’s not worthwhile to have
Hi Ted,
If you would just suppose that dsd256 could pass the bridge, fpga etc.and you could just feed it directly into the passive filtering, what would happen to the sound? Would it be worse than the normal dsd128 in that stage of the signal path? Or I could extend this, what is the benefit to downsample anyway from dsd640 to dsd128?
berlin
With no upsampling the most obvious problem is no volume control.
The less obvious problems with the current analog card are:
The software in the FPGA, the hardware from the digital card to the analog card and the whole design of the analog card were optimized for 128FS, the clock selection, the most effective bypassing frequency of all capacitors, the effective range of the power supplies, etc. I.e. higher frequency inputs would be noisier unless things were changed in the current hardware.
The hardware is built assuming a particular maximum density of ones and or zeros: if the input signal were from DAC chips that allow a higher density the output would be distorted (or worst) when it got loud.
As I mentioned if I were doing everything from scratch we might make different decisions.
Ted Smith said I don't know if AudioGate has that DSD to DSD conversion but if it does it may well be a better converter. Foobar2000 has an on-the-fly DSD->DSD convertion plugin (ASIOProxyInstall from the foo_input_sacd sourceforge project) but I don't know how good it is. As always with these things you just need to try them yourself to find the one that you like best - there's little cost involved unlike trying a bunch of DACs...But if you are using JRiver already, as far as convenience is concerned, doing the conversion in JRiver wins.
I’m somehow still stuck to Jriver because Foobar seems to not support manual file conversion from DSD128 to DSD64 and Audiogate seems not to offer any kind of conversion (especially not to higher sampling rates) in the free version and can only be bought together with a whole professional suite for a lot of money.
Does anyone have a hint to get the manual file conversion from any of those?
How do companies who offer downloads in several DSD sampling rates convert to lower DSD rates from their original recording sampling rate? By loosing 6dB and going over PCM, too?
You said it yourself: they don’t use free tools.
1 Like
Ted Smith said You said it yourself: they don't use free tools.right
I just was not sure if converting from DSD to DSD by -6dB and PCM is a question of commercial or free SW or not
No, in general you don’t need to go to standard PCM in a DSD to DSD conversion.
As mentioned above the final mixing might go to multiple resolutions at the same time so no conversions are required at all (either from DSD to DSD or DSD to PCM, etc.).
If you are converting DSD to DSD and do any math, you can have another bit or two on the top end so that there are no extraneous -6dB (or other) scalings.
It’s only when you are using tools that don’t have direct DSD to DSD conversions or tools that aren’t designed to work together that compromises creep in.
FWIW foobar2000’s (and I assume JRiver’s) on the fly DSD to DSD conversions don’t have -6dB scaling problems either.
Got it, Audiogate really does a direct DSD conversion without -6dB, great!
By the way, you all might know the following page, I just found it now (offering a comparison of DSD256, 128, 64). Havent tried it yet.