Well, call some of us CONFUSED. We thought USB was a "solved" problem?
I know that USB was -- say, prior to 2007 -- a compromise compared to SPDIF.
Enter Gordon Rankin and his "Streamlength" technology, adopted by Wavelength Audio and later (licensed by??) Ayre, AudioStream and others. I think BenchMark and Bel Canto and others had their own "in-house" USB solutions (asynch??), but I may be wrong.
I suspected SPDIF because of objections from way back (late 80s) -- well before USB -- when I2S was being pushed by manufs. like Perpetual Technologies, Audio Alchemy, et .al.
So:
SPDIF bad
I2S good
And that -- I think -- has been the consensus, on forums for over 10 years.
Okay ...settled ....but maybe not! ....because a few years ago, Schiit's Mike Moffat was openly stating that SPDIF was the better way to connect DACs because USB is inferior. Not sure whether he was referring to Schiit products, lower-end Schiit products, or all DACs (in general).
Is this hype, marketing, etc.?
It may have worked ....
Indeed, on certain forums (notably SBAF), members were CLAIMING the same thing .. spdif sounds better. did they drink the Kool-Aid?
And, a few years ago .... poof ... we have the solution...
Schiit's Unison USB.
And there are others, of course. Allo's usbridge, etc.
So what's the story???? ... many of us don't have hours and hours to kill on A/B listening tests. And, of course, there are many variables ... so let's make the issue testable and clear: say you own a Schiit Modi Multi (which does not incorporate Unison) and connect it via USB or SPDIF, which will sound better -- all else held equal?
If you don't have Schiit Unison device, pick your own favorite device with an "ideal" SPDIF and/or "ideal" USB-to-I2S. The best designs you've heard as of late 2020. Are they close ... just different ... or a clear winner?
I know that USB was -- say, prior to 2007 -- a compromise compared to SPDIF.
Enter Gordon Rankin and his "Streamlength" technology, adopted by Wavelength Audio and later (licensed by??) Ayre, AudioStream and others. I think BenchMark and Bel Canto and others had their own "in-house" USB solutions (asynch??), but I may be wrong.
I suspected SPDIF because of objections from way back (late 80s) -- well before USB -- when I2S was being pushed by manufs. like Perpetual Technologies, Audio Alchemy, et .al.
So:
SPDIF bad
I2S good
And that -- I think -- has been the consensus, on forums for over 10 years.
Okay ...settled ....but maybe not! ....because a few years ago, Schiit's Mike Moffat was openly stating that SPDIF was the better way to connect DACs because USB is inferior. Not sure whether he was referring to Schiit products, lower-end Schiit products, or all DACs (in general).
Is this hype, marketing, etc.?
It may have worked ....
Indeed, on certain forums (notably SBAF), members were CLAIMING the same thing .. spdif sounds better. did they drink the Kool-Aid?
And, a few years ago .... poof ... we have the solution...
Schiit's Unison USB.
And there are others, of course. Allo's usbridge, etc.
So what's the story???? ... many of us don't have hours and hours to kill on A/B listening tests. And, of course, there are many variables ... so let's make the issue testable and clear: say you own a Schiit Modi Multi (which does not incorporate Unison) and connect it via USB or SPDIF, which will sound better -- all else held equal?
If you don't have Schiit Unison device, pick your own favorite device with an "ideal" SPDIF and/or "ideal" USB-to-I2S. The best designs you've heard as of late 2020. Are they close ... just different ... or a clear winner?
Many variables indeed. Sometimes SPDIF (coax) is better, more often USB is a tad better when modern chips are used. Depends on which receiver is chosen. Some devices have high res capability via SPDIF and 16 bit via USB .... Some use an excellent receiver while others use a mediocre receiver. So I guess the point is that there are not many ideal situations. It is a surprise and some testing is required every time to discover the best interface in that situation with that device.
Thanks for pointing to Unison, I did not know that one. For years I hoped the industry would adopt the PS Audio I2S over HDMI system but alas.
Thanks for pointing to Unison, I did not know that one. For years I hoped the industry would adopt the PS Audio I2S over HDMI system but alas.
Last edited:
I2S
I made a mistake in my OP when I wrote:
I meant to say:
So:
SPDIF bad
USB good
So .... not much difference between the two so implementation configs (major or minor) made very insignificant sonic differences in the end product. And that explains the old-is-new-again phenomenon. Young audiophiles who've grown up with nothing but USB and wireless have discovered this vintage format that sounds "great" ... like tubes and reel-to-reel and Klipschorns. "They don't make em like they used to."
But yes ... too bad about I2S not being the digital connection of choice. Tho' you have to be careful on terminating those properly. On that note, I think ARCAM, back in the early 90s, had some good ideas. The SAA7220 -->TDA1541A i2s connection was quite innovative. Still use that in some of my projects.
I made a mistake in my OP when I wrote:
I meant to say:
So:
SPDIF bad
USB good
Well, if true, then all the hoopla that Gordon Rankin raised was just that: hoopla! Not sure about that whacko .... but we don't hear much from him anymore. He made quite a bit off StreamLength to retire very comfortably. Especially after that Pono deal. Geeeeeeeeeeeeeeuzusssss!
So .... not much difference between the two so implementation configs (major or minor) made very insignificant sonic differences in the end product. And that explains the old-is-new-again phenomenon. Young audiophiles who've grown up with nothing but USB and wireless have discovered this vintage format that sounds "great" ... like tubes and reel-to-reel and Klipschorns. "They don't make em like they used to."
But yes ... too bad about I2S not being the digital connection of choice. Tho' you have to be careful on terminating those properly. On that note, I think ARCAM, back in the early 90s, had some good ideas. The SAA7220 -->TDA1541A i2s connection was quite innovative. Still use that in some of my projects.
Last edited:
To simplify stuff:
- early USB receivers were using an "adaptive" mode actually quite similar to spdif. Jitter of the best implementation was comparable to decent spdif receivers (maybe not the best).
- Things could be improved by slapping an ASRC after the USB receiver to get rid of most jitter and rely only on the DAC's internal clock (Benchmark DAC1, using a TAS1020B in adaptive mode, followed by a ad1896 iirc). Another option followed by some manufacturer who didn't like the potential for groundloops was to convert from USB to spdif (so that an isolation transformer could be used) and then to I2S (with eventually an ASRC, such as in the Bel Canto DAC3, where the chain was pcm2903-cs8416-cs8421). Btw, in these DACs, both the spdif inputs and the USB input were reclocked by the ASRC so good luck hearing a difference.
- Gordon Rankin created a firmware for the TI TAS1020B USB chipset, so that it worked in asynchronous mode. In asynchronous mode, the jitter again only depends on the DAC clocks' implementation, without the need for an ASRC (some claim they can hear it). The TAS1020B also allowed for higher resolutions signal than the pcm27** or pcm29** commonly used. Arguably, it is a more elegant solution but not necessarily better in practice than Benchmark's solution.
- Then arrived the XMOS chipsets that do asynchronous USB and high-res and are used in almost every commercial USB DAC or interface today.
In theory, there is no contest. Asynchronous USB is conceptually "perfect". But you cannot hear USB vs SPDIF in isolation: you depend not only on the quality of the implementation of the receivers themselves but also on how your DAC manages the digital flows and clocking (as well as power supplies, ground isolation and half a dozen other variables).
- early USB receivers were using an "adaptive" mode actually quite similar to spdif. Jitter of the best implementation was comparable to decent spdif receivers (maybe not the best).
- Things could be improved by slapping an ASRC after the USB receiver to get rid of most jitter and rely only on the DAC's internal clock (Benchmark DAC1, using a TAS1020B in adaptive mode, followed by a ad1896 iirc). Another option followed by some manufacturer who didn't like the potential for groundloops was to convert from USB to spdif (so that an isolation transformer could be used) and then to I2S (with eventually an ASRC, such as in the Bel Canto DAC3, where the chain was pcm2903-cs8416-cs8421). Btw, in these DACs, both the spdif inputs and the USB input were reclocked by the ASRC so good luck hearing a difference.
- Gordon Rankin created a firmware for the TI TAS1020B USB chipset, so that it worked in asynchronous mode. In asynchronous mode, the jitter again only depends on the DAC clocks' implementation, without the need for an ASRC (some claim they can hear it). The TAS1020B also allowed for higher resolutions signal than the pcm27** or pcm29** commonly used. Arguably, it is a more elegant solution but not necessarily better in practice than Benchmark's solution.
- Then arrived the XMOS chipsets that do asynchronous USB and high-res and are used in almost every commercial USB DAC or interface today.
In theory, there is no contest. Asynchronous USB is conceptually "perfect". But you cannot hear USB vs SPDIF in isolation: you depend not only on the quality of the implementation of the receivers themselves but also on how your DAC manages the digital flows and clocking (as well as power supplies, ground isolation and half a dozen other variables).
You have summed everything in your post perfectly.
Of course spdif can be brought to the conceptual perfection of async usb with a deep fifo. From the point of view of clocking both interfaces then become equal.
Spdif and especially optical may now have the upper hand in terms of isolation.
TOSLINK
Remember, years ago, when it was heavily criticized? I think the argument was that its LED driver was way too slow to create a clean, fast signal necessary for SPDIF transmission.
And then, some units featured laser-based TOSLINK that was supposed to solve all these "optical fidelity" problems. And, then, supposedly faster/better LEDs (and/or their send/receive units ... dunno which) also improved things. I've heard high-end manufs, like PS Audio, claim their TOSLINK is comparable to other inputs (???? wha ???).
Again, it's a fun "new" toy ready to sell to Gen Y ... ooooh .... ahhhhh .... flashy fiber-optic cable with pretty red xmas lights.
But is it a serious contender? Really?
Optical (TOSLINK) SPDIF is another point of confusion.
Remember, years ago, when it was heavily criticized? I think the argument was that its LED driver was way too slow to create a clean, fast signal necessary for SPDIF transmission.
And then, some units featured laser-based TOSLINK that was supposed to solve all these "optical fidelity" problems. And, then, supposedly faster/better LEDs (and/or their send/receive units ... dunno which) also improved things. I've heard high-end manufs, like PS Audio, claim their TOSLINK is comparable to other inputs (???? wha ???).
Again, it's a fun "new" toy ready to sell to Gen Y ... ooooh .... ahhhhh .... flashy fiber-optic cable with pretty red xmas lights.
But is it a serious contender? Really?
If one wants to transfer PCM data from place A to place B and are content with 192/24, Toslink is an excellent transport resource given that there are suitable actions taken at the receiver end to cater for the degradation in timing that a heavily BW limited link induces.
The upside is supreme galvanic isolation and very low BER.
Parse that boldwin!! :-D
//
The upside is supreme galvanic isolation and very low BER.
Parse that boldwin!! :-D
//
Using my Linux PC (Intel P4, Linuxmint 32-bit xfce v19.3), I spat out both (but not concurrent) USB 2.0 and spdif (coax) to an unmodded Schiit Modi Multibit.
The RCA-coax spdif jack on the Linux PC is on a Asus Xonar ST soundcard.
The Linux PC won't allow me spit out usb and spdif concurrently. Both have to be manually config'd (which may be a good thing because "switches" stay clean and isolated).
The USB (which is not Unison on the Modi MB) was subjectively superior. But some of that may be due to USB having higher gain (for whatever reason). In fact, the Schiit Modi MB output level is a bit higher than many dacs and cd players.
Have not measured anything.
The RCA-coax spdif jack on the Linux PC is on a Asus Xonar ST soundcard.
The Linux PC won't allow me spit out usb and spdif concurrently. Both have to be manually config'd (which may be a good thing because "switches" stay clean and isolated).
The USB (which is not Unison on the Modi MB) was subjectively superior. But some of that may be due to USB having higher gain (for whatever reason). In fact, the Schiit Modi MB output level is a bit higher than many dacs and cd players.
Have not measured anything.
Last edited:
My enthusiasm for coax and optical spdif is reserved only for devices with a fifo. Clearly your dac is not of those.
I literally can't believe we're discussing SPDIF in 2020. I boxed up may Newava pulse transformers and SPDIF re-clockers on vero-boards, way back in 2009.
SPDIF input
In case there were any doubts, I just found it in deep storage, attached to an old diyparadise Monica DAC ... whipped out my iPhone and took a snap.
Correct me if I'm wrong, but methinks this was one of the Jocko contributions from decades back....
I literally can't believe we're discussing SPDIF in 2020. I boxed up may Newava pulse transformers and SPDIF re-clockers on vero-boards, way back in 2009.
In case there were any doubts, I just found it in deep storage, attached to an old diyparadise Monica DAC ... whipped out my iPhone and took a snap.
Correct me if I'm wrong, but methinks this was one of the Jocko contributions from decades back....
USB and SPDIF are not comparable. USB is a packetized data communication interface. It has inherent latency issues. SPDIF is the digital equivalent of an analog audio cable designed to run long distances with embedded clock so that transmitter and receiver can operate synchronously with minimum latency.
Huh?
SPDIF at receiver end has some logic RAM buffering. And it may be a LOT of buffering if, say, a "Time Lens " strategy is used.
And even a "pure" digital transmission (USB data lines) is ... well, you won't believe me ...... so I'll let Dr Ledermann (Soundsmith Audio) 'splain:
RMAF 2018 - Why Analog is Digital and How to Fix It - Peter Ledermann
RMAF 2018 - Why Analog is Digital and How to Fix It - Peter Ledermann - YouTube
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.