The Oracle has spoken and it seems there are some grounds for optimism about this possibility. Would it be stretching too far to infer that one might be able to perform a software conversion on a hard drive (flac/wav file to DSF/DFF file) allowing the digital source at any later date to read the embedded dsd signal from an attached hard drive and forward to the dac as native dsd? Or does software conversion not satisfy the pre-requisites for success…
Not that I even understand how the 128 samplerate aspect fits into such a proposition and, after all, that seems to be the key to the success of the enterprise. If this question doesnt make any sense at all then I apologise to all readers in advance.
The differences we are talking about can be quite small. I’d strongly recommend that people try these free possibilities each way and judge for themselves. Personally I don’t like any sample rate conversion before the DS, but one could easily claim I’m biased. Also, just in case it’s not obvious: the rules are different for every DAC and possibly even each different input to a given DAC.
It would be interesting to know what the comparison was to. Bridge still seems like the best bet if you aren’t using a I2S native device. USB>SU-1>I2S should be properly compared to bridge or SACD I2S input if possible and I suspect that bridge is still better (though I certainly could be wrong).
BII vs MR (on LPS-1) & SU-1 (on stock PSU) - The SU-1 edged it slightly (for me anyway). Just more gravel or breath in the vocals and the instruments had a little more texture.
Then added the Curious USB cable in between the MR & SU-1 and it improved slightly more. My gf commented immediately that it sounded smoother (and she’s generally bored witless by all this stuff ).
Added the Wireworld platinum 0.3m HDMI cable and it improved by a little bit more.
Modded the SU-1 using the Kitsune kit so that I can power it by another LPS-1 and that made a big difference. Like immediately I heard it and had a big smile on my face.
Result is that the SU-1 path is now comfortably beating the BII. Sorry Paul and PSA team but to my ears it is. But the SU-1 path involves 2 Uptone LPS-1s, a Microrendu and two fancy pants cables and cost north of $2000 in total. The BII by comparison costs what, $700 and is no doubt a fantastic bit of kit. Also full MQA decoding is due to come to the BII soon which also increases it’s attraction. I’ll certainly be updating mine and giving it a listen.
Hope this helps.
Regards,
Alan
PS all my listening was 44.1 redbook CD’s ripped to a NAS. Not sure I get the “SACD I2S” comment. I don’t have any of my SACD’s ripped and my Oppo 105 doesn’t support I2S over HDMI.
If you look over at CA the SU-1 has a fairly loyal fanbase as a USB to I2S DDC that’s improving the SQ into I2S capable DACs like the DirectStream and the Holo Spring
I also use the SU-1. I don’t have the microrendu in the chain like Alan. I am running the SU-1 stock in between my computer and the DS with a wireworld .3m starlight? HDMI cable to my DS snr. Noticeable improvement in SQ. It has been in my system for a while but the main thing I remember is what would seem to be a lower noise floor which enables more detail to come through. I plan on doing the Kitsune hifi mod and picking up an Uptone LPS 1. Pretty much universal praise for this mod. I have one of the original versions of the PSA Bridge and was at first considering upgrading to Bridge II but I want to dip my toes into DSD and the Bridge II only does up to DSD 64 whereas the USB/Singxer doesn’t have this limitation. I will say however that I am intrigued at the possibility of full MQA unfolding on the Bridge since I am a Tidal subscriber and wonder what a full unfold would sound like.
For up to 24/192 its a tossup w/ the B II and a microrendu > Singxer SU-1 >I2S.
Both with Roon and Minimserver minimstreamer converting flac to wav.
For DSD the Singxer Su-1 > I2S really shines. For Roon native DSD and Minimserver passing native DSD I’m transported - amazing Huron sound.
Add an LPS-1 to power the Singxer SU-1 for extra dimensiolity - Jane-Selkye_Chris-Kee_Blue-Coast-Special-Event-7_Complete_2.8MHz-DSF from Blue Coast will make you smile…
Personally I don’t like any sample rate conversion before the DS
Isn’t that what the bridge is doing by downsampling the 128, 256 and 512 times DSD to 2.8224MHz? Wouldn’t it be better to feed these formats directly through the I2S connection if you have a full DSD streamer? Can you make the BridgeII ready for higher sample rates?
Ted Smith said
Personally I don't like any sample rate conversion before the DS
Isn’t that what the bridge is doing by downsampling the 128, 256 and 512 times DSD to 2.8224MHz? Wouldn’t it be better to feed these formats directly through the I2S connection if you have a full DSD streamer? Can you make the BridgeII ready for higher sample rates?
As I’ve stated before IMO if you have to use sample rate conversion (or chose to) before an input to the DS, I’d convert DSD to PCM (say, 24 or 32 / 352.8k (or 24/176.4k for the bridge)) unless you know for a certain that the eternal DSD to DSD converter does all of it’s processing at the sample rate of the higher DSD rate. As always tho, things like this are personal preference and you should pick which ever method that gives you the sound you like the best.
Ted, I do get your comment about sample rate conversion, particularly down sampling DSD and making sure that the process is done at a rate at least equal to the higher sample rate. I do have one question about your comment. You suggest down sampling DSD 128 or higher to 176.4k. Is it of particular significance to keep the rate change ratio an integer value, or would down sampling to 192k be as good an option? I suppose that part of the answer depends on the resampling algorithms, but is there a more generic answer. Perhaps it is just easier and safer to keep the ratio to a power of two integer value.
Thanks in advance, and I look forward to yet more enlightenment and education. Brilliant engineers who can also communicate clearly to less educated tech and non-tech types are a rare breed and a treasure to share space with.
Short answer: going to 192k can be done as well as 176.4k, but the code is much harder to write from first principles and more complicated if written naïvely. I.e. there are a lot more broken (or poorly implemented) 192k downsamplers out there than 176.4k downsamplers.
More detail:
In the abstract there exist arbitrarily accurate resampling algorithms from any rate to any other rate. But simple upsampling and downsampling by an integer is much easier both in the programming and the amount of arithmetic. If you want to avoid weird interpolation (which will introduce some distortion (tho limitable) the sanest way of upsampling and downsampling consists of finding LCM(in rate, out rate) (the least common multiple of the input and output sample rates), zero stuffing the input to get to that sample rate, filtering to half of the lowest of the incoming and outgoing sample rates and throwing away all but 1 sample for each LCM(in rate, out rate) / samples. (Then there’s the significant optimization of not doing the multiplies for the incoming zeros and not calculating the filter outputs that you’ll be throwing away.) Here’s a medium level lecture about this: https://www.youtube.com/watch?v=qca4jIoxDtY
In the case of your question that means that going from DSD 128 down to 176.4k is a simple filter keeping up to 88.2k and throwing away all but 1 in 32 samples (obviously you don’t need to calculate the samples you’ll be throwing away). To get to 192k you need to upsample to 28.224MHz ( = 10 x the DSD64 rate, sound familiar?), filter keeping up to 96k and then downsample by 147 to 192k. The difference is that the 2nd path clearly requires five times the math for a naïve implementation or much more careful coding for an implementation that doesn’t do any multiplies for the incoming 0’s or the unused outputs. Both can be done with the same amount of math if you are careful, but the chance for bugs in the 2nd case is a lot higher.
Thank you Mr. Wizard for that informative response. You confirmed my thoughts on subject and elaborated nicely, adding to my understanding of it. I have not yet watched the video, but do intend to. I just need to block off a chunk of time to do it justice.
Are you able and willing to offer an opinion on the commonly discussed products (J River, Foobar2000, MinimServer/MinimStreamer, …) and the quality of their sample rate conversion software? I understand and accept if you choose not to offer specific opinion on products, mostly just curious.
I know that JRiver added SOX based filters in release JRiver MC 22 ( (or something like that)) which is a good implementation.
Similarly you can find foobar2000 plugins for resampling based on SOX and dBpoweramp’s SSRC (both probably better than the older PPHS implementation)
I don’t know anything about Minim…
Things are a little more complicated when DSD is added to the mix - there I definitely suggest at least trying the DSD to 24/176.4 options available rather than relying on less well documented DSD to DSD converters (or even PCM to DSD converters.)
On the MAC I would trust BitPerfect’s conversion to DSD to do a good job, you can test for yourselves whether you’d like to use it compared to using the DS’s versions.
Many people speak highly of HQPlayer, since it’s redundant with things I use I haven’t tried it or looked at its documentation about it’s algos.
In all cases the differences can be small enough that it’s pure personal preference which paths you’d like to use. I can say that the DS’s filters for PCM don’t alias and have as wide of bandwidth as practical to be at -144dB at the Nyquist frequency. Some prefer slower rolloffs for “simpler” music (by simpler I mean a small number performers/sources, not a statement about the music itself.)
Wow. Thanks again for the wealth of information. I hope that we are not boring anyone. For myself, I am thoroughly enjoying this lesson / discussion.
Curiosity assuaged, I will say that even before this day of lessons I would have been strongly inclined to go with integer ratio sample rate changes where possible or practical, for exactly the reasons that you indicated. Still, this is not for naught as I have come out of this with a bit more knowledge and understanding of the subject. I hope that others here have also benefitted from this discussion.
I am glad to hear that dBpoweramp is likely good quality, as that is my primary tool for sourcing digital files for my library. Of course, 99.9% of the time it is just used for ripping CDs, but knowing that it should be good for quality sample rate conversion is comforting to know should said function become necessary. I’ll have to look into plugins for Foobar2000 and check up on MinimStreamer’s conversion tools to make sure that I have all bases covered.
).