|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
24.6.199.104
I've been hearing a lot of stuff around here and other forums about USB audio interfaces and reactions to those statements. A lot of it is untrue or at least half truths, I thought I'd clean some of this up and hope to give a little dose of reality to the discussions here.To start off I have quite a bit of experience with this, I have been building my own DACs for the last several years including S/PDIF interface, USB to S/PDIF and USB to I2S direct to DAC chips. I have built many combinations and both listened and done measurements.
First a quick overview of USB modes, synchronous, adaptive and asynchronous. All of these are "isochronous" whih just means the bus has reserved bandwidth for the data. Synchronous is the really cheap, very high jitter interface, its not seen all that often today, when you see reaaly bad numbers for a USB interface its probably running synchronous. Adaptive, much better, it still has a PLL but its referenced to a local crystal oscilator, the jitter can be quite low if done right. Asynchronous, the receiver is in complete control, output clock comes directly from low jitter clock, no PLL involved. Very low jitter if done right. If done not so right can be about the same as adaptive.
Almost all my experimantation has been with adaptive mode. There are only a couple chips that can implement asynchronous mode and they take reprogramming to run in that mode, that programming is turning out to be very difficult, some day I hope to get it worked out.
The up shot of the tests shows that if you use asaptive mode direct to I2S and feed that into the DAC chips you can get in the 65 to 70ps range, very good but not perfect by any means. Note I did NOT say its jitterless, just quite low jitter. Yes there ARE 1KHz sidebands but they are very low if the interface is done correctly.
Any attempt to convert USB to S/PDIF and then to a DAC results in much greater jitter than this, in the 150 to 250ps range.
Thus the big advantage to using USB is if you do NOT have S/PDIF in the stream in any way. Thus using a USB to S/PDIF converter running into your super duper DAC is NOT going to see most of the advantage that's possible to achieve with USB. Many companies that are coming out with "USB DACs" are just sticking a USB to S/PDIF converter in front of their existing S/PDIF input DAC, people then buy these because USB is supposed to be better, then find it doesn't sound that much better, so they conclude that USB interface is a bunch of hype! The truth is that the USB interface was implemented such as to negate the advantage it DOES have.
The other thing I hear is that someone buys am inexpensive USB DAC and somehow expects it to sound way better than their current much more expensive "high end DAC". The interface is an important part of the package but by no means the only part. Even using a USB interface directly to DAC chips does not mean its going to sound fantastic. You still need to use the right DAC chips, output stage, power supply, decent quality components where they do the most good etc.
If attention to detail is applied to all stages of the device, and direct USB to DAC chips is used, the results can be spectacular, unfortunately at this stage of things very few such products exist. Most of what you buy today is going to be a compromise somewhere along the way. There is no magic genie such that the inclusion of this one thing automatically makes a device the best in the world. A correctly implemented USB interface is ONE of the parts that can be used to achieve very good digital sound reproduction, the rest have to be there as well.
I hope this clears up a few things, if anyone still has confusion about the subject I'll be glad to try and shed some light.
Follow Ups:
If attention to detail is applied to all stages of the device, and direct USB to DAC chips is used, the results can be spectacular, unfortunately at this stage of things very few such products exist.
The Wavelength DACs, the Empiracle direct I2S stuff for sure meet my criteria. The new DDDAC kit meets it, but its a kit!One I'm not sure of is the twindac+, I have yet to determine if the USB goes direct to the DACs or if it goes through S/PDIF. If anyone has one of these I would love to see a high res picture of the board, I could tell from that. There are a lot of things a like about the twindac.
The Apogee minidac may go direct, but I'm not sure, anyway I don't like a lot of the deign choices for that product, having heard ones done what in my opinion is "right" I think the Apogee falls short sonically.
"The new DDDAC kit meets it, but its a kit!"But the modules can now be bought pre assembled at an extra cost. This is what I’m getting. All you need to add is a chassis and DC battery w/charger!
Better double check that. I know for a fact that Doede doesn't want to get into the assembly/support bsuiness and sells only kits. A "pre-configured" kit is not a "pre-assembled kit"Only the SMD components are assembled in his kits as far as I know.
On the other hand - these things aren't that difficult to build if you have a basic understanding of how these things work.
I am currently using Doede's USB board with a modified version of his previous DDDAC. After sorting out some ground loop issues in my system, I couldn't be happier, and the DAC isn't even finished yet (heat sink is clamped to the chassis with a vise, etc ;-))
Peter
Well I did manage to put a Radio Shack crystal radio kit together once! And I used to be pretty darn good at doing those connect-the-dots puzzles as a kid! Looks like everything has code numbers, which is comforting, but I'm a little worried about the little "things" that may need to be unidirectional. Are there many of these?
not a single diode in these DACs unless you build your own AC-DC power supply. There are a few polarized electrolytic capacitors in there, but the PCB is labeled for +/- and the caps should be clearly labeled too. Make a mistake and it will "vent" anyway :)The bigger issue will be that there's no "do this, then do this, and then this" instruction manual. You have to work from the schematics, parts list and description he posted. The soldering of the DAC chips may be a little tricky for somebody who hasn't done much with the soldering iron in a while, but it really isn't a tough project.
As long as you put a fuse in the wire to the battery (its cheaper than other parts that will go up in smoke if you make a mistake), you should be able to put the kits together pretty easily. They are much easier to build than the original DDDAC which hat no text screened on the PCB.
The kit doesn't include a battery checker anymore (but its easy to build from scratch), plus you need to get switches and line out and power connectors, coupling caps and other parts on your own. Don't underestimate the cost of these items.
Peter
And how about the AQVOX USB 2 D/A?
Thank you for your very informative article, that would help many people to make their decisions.I have heard that the new "Microsoft OS Vista" will support "isochronous data transfer" through USB/Firewire , do you think that would make the manufacturers design jitterless DACs'?
All the operating systems already support isochronous modes, all audio modes use isochronous transfer, which just means the bus is reserving bandwith on the bus.Are you getting confused with asynchronous mode? This is one of the isochronous modes which lets the reciver be in complete control of the timing and sends commands back to the computer to speed up and slow down to match whats happening at the receiver.
john, thank you for the very informative post. do you feel that any of the jitter reducing boxes such as a monarchy dip would help one using a usb to coax device in bringing down it's output jitter much?
When I first heard about a USB (or Firewire, for that matter) DAC (or s/pdif converter) I thought the idea was a good one. My assumption was that the data would be streamed into a large buffer on the receiving end, then output with very precise timing (controlled by a fancy clock that was powered in a very clean, un-computer-like way) to a DAC (or directly to the DAC chip that was in the device itself). The assumption was that this buffering would make all talk of jitter (related to the sending and receiving ends of the USB connection) irrelevant. I imagined it was something like how a hard drive must work...you send it data, but it has to buffer it so that it can write it, with very precise timing involved, to that platter that is rotating at very high speeds.Why don't any of the USB solutions I've heard discussed implement an approach like this?
FIFO buffering is used in high end audio and does reduce jitter significantly. Problem seems to be that COST is much more of an issue to computer people. Perfection can't be achievedfor nothing!
The problem with these types of block transfers and large buffers is that it creates latency. This interferes with the usualy CD-player type controls, such as FF FR, skip to next songe etc.. Not that it cannot be done, but implementing these functions in software and hardware with a fully buffered system is non-trivial and requires the buffers to be flushed etc., without causing pops and ticks in the playback, or unusually long delays.
What you describe is called bulk transfer.- Bulk transfer : Data are transmitted by bursts and arrive on a big buffer. They’re then checked by an error correction system and the sink will ask for retransmission if needed. The complete data are finally reconstructed (and clocked if needed) by the sink. At first sight, this solution is perfect for audio transfers; the data are guaranteed to arrive and the clocking is done only by the DAC, with no jitter. However, this solution has been widely rejected. Bulk transfer as it is implemented cannot offer any guarantee of timing and would require big buffers. It would raise the costs but also mean that the user would only be able to control his DAC with a few seconds of delay. Practically speaking, this means that bulk transfer has been reserved to printers and the like, that need large files to transmitted without errors but with little considerations of timing.
Yes, there would be a delay. But, given that the amount of timing error generated by USB jitter is probably extremely small, wouldn't something like a half-second or so pretty much cover it? The buffering that is happening on the software/driver side is at least that long, in my experience.Again, I'm not an expert on this type of thing (I'm a software engineer), but it certainly seems like there is a workaround for jitter in a USB connection that will allow an external device to completely re-clock the received data.
It indeed exists and it's called asynchronous transfer. ;)Bulk mode wasn't designed with audio in mind. There is actually no mention of it at all in the paper on usb audio on usb.org.
Interesting measurements. Thanks a lot.The 75ps figure for direct I2S is strikingly similar to the 75ps of the DIR1703 spdif receiver, which claims similar result and is based on TI's spact system. A system originally designed for adaptive USB.
For the rest, nice summary. I did a similar exercise a few months ago here : http://www6.head-fi.org/forums/showthread.php?t=141238 The replies there point however to the possibility of achieving asynchronous transfert.
There are also the (dubious ?) claims of Wavelength (Gordon Rankin), saying that it is possible to get asynchronous transfer out of a PCM2707. See here http://db.audioasylum.com/cgi/m.mpl?forum=pcaudio&n=228
I am using the 2706 which does use SPACT to generate the clock, so the 65 to 70ps I'm getting from that is quite in line with the SPACT in 1703.The only chips I'm aware of that support the asynchronous mode in hardware are the 1020 and the 3200, both of whih require firmware changes to do so. I do not know of anyone that has successfully done this programming. I've been working on it for almost a year now and have yet to succeed. I'm going to be working on it this year and might get it working.
Creative used asynchronous mode in the 2NX, but they did not use an audio specific USB chip. They wanted to use USB2.0 so they picked a generic USB 2.0 interface chip and wrote their own firmware for that, implementing aynchronous transfer while they were at it. Even though they used asynchronous they still did not achieve really low jitter performance, it was not a criteria for the board. Thats what I mean by saying that if you use asynchronous you STILL have to do everything else right to get really low jitter.
The results from direct I2S adaptive mode are very good, much better than I expected, but of course there is still room for improvement by going to asynchronous.
Good work at clearing up the "USB DAC" myth. Your follow-up post also clarified the 2NX design.A few questions:
- what USB mode does Transit use?
- for a reclocking DAC like Benchmark DAC1, does a couple hundred ns jitter lead to any perceptible degradation to sound?
- Benchmark gives jitter tolerance in terms of "UI Sine". Can this be converted to "ns jitters"?
- what's your take on USB DAC chips like PCM2702E? Do you see a high-end USB DAC chip in the future?
The transit uses the 1020 in adaptive mode. The way the 1020 implements adaptive mode is different than the 270X chips, I'm not sure which is better though, I'm going to be doing some head to head comparisons later this year.An ASRC does NOT completely eliminate jitter from the input, I've gone over the details of that before. It can decrease jitter to some degree, but it cannot eliminate it. So even with an ASRC you CAN hear effects from upstream interfaces.
The 2702 is the same USB engine as the 2706 that I use, but it does not have I2S output, you have to either use its own built in DAC or the S/PDIF output.
If I can get the firmware properly written to run asynchronous mode with the 1020B then it will be an extremely good interface.
Has anyone tried or worked with the DIR1703 spdif receiver?If it has a similar SPACT circuit, isn't it possible to be as good as the TI USB SPACT implementation? Does anyone know a commercial DAC that uses it? I see Digi-key carries these chips for around 5.
The TI DIR1703 is officially discontinued; it had problem locking when frequencies were changing. You needed a reset at that point.The Bel Canto DAC2 uses it iirc.
Question:What are some of these USB DACs that actually go direct from USB to the DAC chip? You know of any that are coming out?
Unfortunately there aren't very many. The Wavelength USB dacs are direct I2S, Emprical Audio showed a direct I2S box to a P3A and mentioned that he was going to go to a Benchmark DAC1 as well. I've heard both these and they are very good, Empiracal doesn't have the I2S stuff on the web page so you need to talk to Steve directly to find out what he charges for those.The one I'm not sure about is the twindac+, I have yet to find out if its direct I2S or if it goes through S/PDIF. The new Stello DAC is also up for grabs on this as well. Since someone here jut bought one we can probably find out real soon.
Every other one I know of uses USB to S/PDIF.
I actually heard this system at CES, it was probably the highest resolution system I'd ever heard. You could hear EVERYTHING in the recording. It was frightening.Well I found my next DAC. I'll get a perpetual tech 3a and then get a wireless/usb> i2s unit at some point.
"Every other one I know of uses USB to S/PDIF."Does the Apogee Mini-DAC use USB to S/PDIF as well? Maybe I have missed it, but I haven't heard much about the innards of this DAC.
Apogee uses tusb3200 + ad1896. In other words, they apparently use an ASRC to get rid of jitter, using I2S directly.
I'm not aware of any USB-I2S-DAC solutions from large companies. You pretty much have to make your own, as Mr. Swenson has, or buy one from Empirical Audio. Empirical now offers USB-I2S interface box, but the DAC choices get limited to Perpetual P3A and Northstar 192 DAC, both of which has I2S interface. I believe Steve still recommends both DAC's I2S input to be modded for best results. Benchmark DAC-1 should be getting an I2S input mod in the months to come from what I understand..
"To start off I have quite a bit of experience with this, I have been building my own DACs for the last several years including S/PDIF interface, USB to S/PDIF and USB to I2S direct to DAC chips. I have built many combinations and both listened and done measurements."Do you have a website detailing some of your work? If so, I'd be interested in checking it out.
I've got an I2S compatible DAC and have been searching for an affordable USB-to-I2S option. I've seen some of the kits posting here on AA and I don't mind doing some soldering, but I'm not looking to build it from scratch or solder tiny traces (plus I haven't seen any posts on how the kits sound). I'm interested in stock or modded options (I've got a USB Audiophile and Squeezebox2 as possibly candidates). Any suggestions?
If you have a decent DAC that supports I2S input, here's a really affordable USB-I2S solution. I recently added this to my DAC and except for volume control issues (the PCM2707 USB chip is not very happy with ASIO and I have yet to make anything come from it with kernelstreaming), as well as ground loop issues in my building, I have been able to achieve the absolutely highest quality playback from my system so far. I also own an M-Audio Audiophile USB and a modded Transit, but since both of these go to SPDIF, they are in the parts box now. No comparison.the actual DAC on that board is "ok" but I wouldn't want to call it audiophile. The key of that USB-I2S board is that it creates the direct interface between PC transport and a high end DAC.
Gonna have to build it more or less from scratch, though - except the SMD components are already on the board.
Peter
Thank you for this; it's the sort of info one hopes to seek in a forum. Unfortunately one rarely gets it, and rather less from this board.One aspect you have not covereed is the poor and dodgy software driving some usb interfaces - more hassle than it is worth. Another aspect is the less than perfect usb interfaces found on pcs. It can take a lot of messing about to get extraneous noise suppressed properly.
One last aspect has nothing to do with usbs; it's simply that noiseless computers are very difficult to implement. With 24/96 audio I expect better than the CD thd+n of 95dB or so. This is very hard to get on a computer as many sound cards are just not good enough.
This post is made possible by the generous support of people like you and our sponsors: