DirectStream Max PCM Rate


#1

@tedsmith

What is the maximum PCM input for the DirectStream, running Red Cloud, via USB? As best as I can tell, it is 352.8 as I can not get 384 to work. If it is 352.8, why is it not 384?


#2

384k is not supported (yet). Until Huron it couldn’t be (at least without downsampling first to 192k.) For USB we’d need to change the PS Audio USB driver and/or the XMOS code to allow it (tho the newer Windows 10 USB Audio 2.0 drivers might support it.) I’ll talk to PS Audio engineering and see if they can allow it as I make sure that the FPGA code doesn’t have some bugs in the not fully tested 384k path. It will be easier to do in the new next release of the FPGA code I’m working on.


#3

When I was trying to find files in 352.8, all that I could find were 384.
Ted suggested converting them with dbpower amp to 352.8, everything sounded excellent.
So in the mean time, you may want to try that route to play your 384 files.


#4

The most common 352.8k files are DXD files. You can try some free ones from http://www.2l.no/hires/index.html


#5

I don’t have any 384k tracks. I am playing around with HQPlayer upsampling and was trying to figure out what the max input sample rate of the DSD is. I was surprised that it would do 352.8k but not 384k.


#6

I am mystified why anyone would want to upsample lower-res music before sending it to a DSD, since the DSD essentially does a far better job of upsampling than you can accomplish externally.

Could someone explain to me the reasoning for doing this at all?


#7

Choosing your upsampling algo is like picking cables - it’s almost purely personal preference and you can tailor the resultant sound by making various changes… Unlike most DSP, since there’s no technically perfect filter, any given filter is a set of compromises. Different people might hear these differences and have a personal favorite. This is fine. Personally I wish that people would at least try the DS’s upsampling, but I’m not offended if they don’t.


#8

Thanks, Ted. So, as I expected, there is no real technical reason for spending any time on this, and I’m unlikely to hear “improvements” from the “differences”.
I quit spending time on this sort of thing after a brief flirtation with tube-rolling a while ago. I found it far more enjoyable to spend my time listening to music rather than listening to search for imaginary flaws in reproduction.


#9

If you didn’t hear differences with tubes you’re not likely to hear different upsampling filters before the DS.

I do hear both and enjoy changing filters as it’s so much easier than changing tubes. I use HQPlayer embedded and while there are numerous filters and dithers to chose, I prefer a couple over the others, but none are better or worse just slightly different. I also listen straight allowing the DS to upsample native.

Different music benefits from different filters (different compromises). Chamber music is very different than modern post punk electronica. Some filters favor faster transients, etc. It’s all in the ear of the beholder.


#10

You go right ahead and be mystified. I don’t know why you think that the DSD must do a better job of upsampling than can be accomplished by HQPlayer. There is no logical or technical reason to believe this. Nothing Ted wrote in his reply would support your position. Nor would it support you assumption that there is “no technical reason to spend time on this”. Just because you found no value in tube rolling does not mean there is no value in tube rolling. The same goes for upsampling filters.

I have a tube preamp and amp and spent a little time trying different tubes until I found a combination I liked. I will do the same with HQPlayer and it’s filters. One man’s imaginary flaws are another man’s real flaws. If you don’t want to spend a little time and money pursuing better sound, that’s fine. But don’t mock people that do.

Finally, the major benefit I have found with using HQPlayer Embedded with Roon is that it bypasses Roon entirely when communicating from the Roon Core to my ultraRendu streamer. The major knock many people have with Roon is how it sounds compared to other player applications. Well, HQPlayer doesn’t use Roon or RAAT and talks to the ultraRendu acting as an NAA directly.

Anyway, if you haven’t tried it, I don’t see how you can get all indignant with your mystification with any credibility.


#11

Indignant is what I do when someone accuses me of things I didn’t say.

I did not say I didn’t hear differences, I said I didn’t hear “improvements”, just differences. You may prefer one flavor of distortion more than another, and even a different flavor for different music, but it doesn’t make either any closer to the original. I’m presently using an Auralic Aries G1 with my DSD, and I see no reason to tinker with the products of real audio engineers.

I gave up on Roon for its minimal jazz and high-res catalogs, and MQA is just another lossy codec. Try QOBUZ if you want good streaming sound.


#12

My understanding is that the DS(J) upsamples everything to 20xDSD and then brings it all down to 2xDSD regardless of whether the user applies separate upsampling ahead of the DS(J). So I think the question is whether music sounds better upsampled twice versus upsampled once. I have no opinion on this question, and I believe the people who say their systems sound better with or without preliminary upsampling ahead of the DS(J). But I also believe that all of us are experiencing the DS(J)'s upsampling algorithm whether or not we apply pre-DS(J) sample-rate conversion.

(I do not use pre-DS SRC, but I do have MinimServer set to convert 16-bit FLAC to 24-bit WAV in my Melco, which feeds a DS. I find that this makes an audible improvement, perhaps because the processing requirements of the FLAC-to-WAV conversion are being handled ahead of the DS, thus sparing the DS from having to work on that task.)


#13

Well phrased. That was my thought when I suggested that the DSD internal conversion to 20xDSD is going to be more accurate than the result of externally computing upsampling and then using the DSD.


#14

What does that even mean in regards to upsampling? Accurate compared to what?


#15

For what I understand, and I am sure @tedsmith can clarify, upsampling is an iterative process so nothing in the DSD is upsampled just once regardless of whether or not some upsampling is done prior to the DSD receiving the track.


#16

By once versus twice, I simply meant applying one SRC algorithm versus applying two separate SRC algorithms. All of us experience the DS(J) SRC algorithm. A subset of us apply another SRC algorithm ahead of the DS(J) SRC algorithm.

I used to own a TEAC UD-501 DAC with built-in upsampling that could be defeated by upsampling to 24/352.8 or higher ahead of the DAC. As I recall, the consensus of UD-501 owners (myself included) was that the DAC sounded best by avoiding TEAC’s built-in upsampling algorithm, which I achieved by upconverting everything to 24/352.8 in JRiver and Audirvana.

The DS(J) is very different than the TEAC UD-501 in that pre-DAC SRC does not defeat the DAC’s internal SRC algorithm. As such, I am content to let the DS handle the SRC. To paraphrase the guys at Schiit (who are anti-SRC), ‘just because you have to cook your turkey in a microwave does not mean you have to run it through two food processors first.’


#17

I am quite happy that bootzilla and eldrick are content with having their DS DACs do all the upsampling of their content. More power to them and I would never say that they are wrong for doing that. Heck, that’s exactly what I did for the last year.

But suggesting that having a quality product, like HQPlayer, do upsampling is inferior just because it’s not built in to the DAC is technically indefensible and simply wrong. The arguments presented by those two are just plain invalid. If they want to leave the filters up to PS Audio, great. But don’t make up “technical” arguments to justify their positions.


#18

I never wrote that using HQPlayer is inferior to leaving the SRC entirely to the DS(J). On the contrary, I stated that “I have no opinion on this question, and I believe the people who say their systems sound better with or without preliminary upsampling ahead of the DS(J).” If you like using HQP ahead of the DS, then by all means continue to do so.


#19

Yet you continued with your upsampling twice can’t be good position…


#20

I didn’t say double SRC can’t be good … I simply said I don’t do it. If is sounds better to you, then it is worth doing. Again and for the final time, I do not have an opinion on the topic.