• Asynchronicity: A USB Audio Primer

    Recently the validity of USB as an audio interface has been called into question by some audiophiles. Adding to this was an all-encompassing statement in The Absolute Sound professing that USB interfaces are inferior to S/PDIF interfaces across the board. This had much of the computer audio world understandably bent out of shape. Instead of a disservice to the audiophile community I will attempt to provide accurate information based on facts and discuss different USB implementations. I'll focus mainly on the two different types of USB implementations asynchronous and adaptive. In my opinion any USB, Firewire, S/PDIF, or AES/EBU interface is capable of outperforming the other interfaces on any given day. None of these interfaces is inherently better or worse than the others. It's the implementation of the interface in each product that separates the men from the boys.


     

     


    Introduction

    Note: I am by no means a leading authority on USB audio and I relied heavily on engineers in the industry while researching this article. Some, but not all, of my sources were Gordon Rankin from Wavelength Audio, Charlie Hansen from Ayre acoustics, and engineers at Data Conversion Systems (dCS). I filter out all marketing terms and bias when analyzing my correspondence with all experts. This article has been in process for several months, long before the TAS article was published in print. This is not a response to the TAS article rather it's an attempt to provide facts about USB audio and arm consumers with more information. Like everything I write this article is wide open to comments and criticism from anyone in the world. I encourage everyone to leave a comment below.

     

    Universal Serial Bus (USB) is gaining in popularity by the minute among audiophiles seeking to connect a music server to their high-end audio system. One reason for this increasing popularity is the ubiquity of the USB interface. USB is available on virtually every computer manufactured in the last ten years. Plus, it's pretty easy to grasp the music server concept at a high level when all that's needed is to plug a cable into a USB port. Complexity, confusion, and a unique set of compromises arise when audiophiles involve internal cards like the Lynx or RME card that requires installation inside the computer's case. USB on the other hand is nearly fool-proof. A USB cable can only connect to a DAC and computer one way and once its connected the listener will have sound coming from the computer. Granted the configuration may need some fine tuning to get the best sound possible but nonetheless getting sound out of a USB port is quite simple.

     

    Many audio component manufacturers are currently building Digital to Analog Converters (DAC) with USB inputs. Some manufacturers are also building USB to S/PDIF converters that allow listeners to output audio from their computer's USB port and input that digital signal into a DAC without a USB input. Listeners have also elected to use a USB to S/PDIF converter if the USB implementation on the converter offers better performance or more sample rate options than the USB input on their current DAC. Like every other consumer product in audio and elsewhere, not all USB enabledDACs and converters are created equal. By far the most popular USB implementation method uses what's called Adaptive USB mode. The newest USB implementation used by a select few manufacturers is called Asynchronous USB mode. The technical differences between adaptive and asynchronous modes are very large. In addition there are differences between implementations within each USB mode. For example there are a few different adaptive USB implementations that differ widely in features and sound.

     

    Before delving into the adaptive and asynchronous USB details, here are some basics to keep in mind. The term USB DAC is a consumer friendly description of a digital to analog converter (DAC) with a universal serial bus (USB) input. This article is about USB inputs and their implementation withinDACs . One must separate the interface from the DAC as a whole to really understand what's going on and to make an educated purchase. A DAC with a so-called poor USB implementation may have the best S/PDIF implementation on the market and vice versa. Thus the sound of a DAC may vary widely based on the input used. The main thing to keep in mind when reading about adaptive and asynchronous USB modes is clocking. Clocking is extremely important with digital audio. Many digital audio experts agree that keeping the clock as close to the DAC as possible, or using a master clock for all digital components is the way to achieve the most accurate sound. In consumer high-end audio as well as professional audio clocking is a major concern and very often external master clocks are used to achieve the best sound.

     

    Here is one way to think about USB implementations that may help readers more familiar with S/PDIF. If I were a college Professor this is where I would tell my students to never repeat this and never write this on an exam. It is forillustrative purposes only.

    S/PDIF has three main specs:
    1. RCA/BNC
    2. Toslink
    3. XLR AES/EBU


    USB Isochronous audio has three main transfer modes.
    1. Synchronous used primarily for ADC work.
    2. Adaptive
    3. Asynchronous


     

     



    Adaptive Mode USB

    Most USB capable DACs today use adaptive mode USB. This is commonly done using a PCM270x chip from TI and to a lessor extent the PCM290x or CMedia parts. The big plus for DAC Manufacturers when using this chip is that no programming is required. The chip can be "popped" into place without extensive R&D, USB audio programming skills, a lengthy time to market, and a substantial amount of money. Big drawbacks to this method are very limited sample rate support (32, 44.1 & 48k), maximum of 16 bit audio, and sound quality.

     

    Another less common adaptive USB implementation is done using a TAS1020 chip. Manufacturers then have a choice of implementing the chip exactly like the PCM270x without additional programming or possibly using the example code provided by TI, or the manufacturer can purchase code from CEntrance, Inc. to use with the TAS1020. Popular devices using the CEntrance code are the Benchmark DAC1 variants, Bel Canto USB Link, and the PS Audio Perfect Wave DAC. Using the TAS1020 and CEntrance code greatly enhances the USB interface and allows native 24/96 playback without the need for additional device drivers or special software.

     

    Some creativity is also used with each of the previous adaptive USB implementations. Some manufacturers use jitter reduction techniques such as adding an asynchronous sample rate converter. This can improve jitter measurements quite well but has also been reported to cause some fatiguing over extended listening periods. Some listeners report this as a Hi-Fi type of sound that is initially impressive, but long term listening may confirm otherwise. Another jitter reduction technique is to use an adaptive USB chip that converts directly to S/PDIF inside the DAC. The S/PDIF signal is then passed though theDAC's standard S/PDIF chip that has likely been refined for many years in countless audio products. This conversion technique can be a fairly good compromise between a simplistic adaptiveimplementation like the PCM270x chip from TI and a well done asynchronous DAC design.

     

    Using either of the aforementioned implementations requires adaptive mode USB. When using adaptive mode USB the computer is the master clock. In layman's terms the DAC is a slave to the computer and has absolutely no control over the timing of the audio. According to digital experts the USB frames in adaptive mode introducesubstantially greater jitter into the signal than asynchronous mode. "In Adaptive mode the computer controls the audio transfer rate, and the USB device has to follow along updating the Master Clock (MCLK) every one millisecond. The USB bus runs at 12MHz, which is unrelated to the audio sample rate of any digital audio format (i.e. 44.1K requires a MCLK = 11.2896MHz). Therefore Adaptive Mode USB DACs must derive the critical master audio clock by use of a complex Frequency Synthesizer. Since the computer is handling many tasks at once, the timing of the USB audio transfers has variations. This leads to jitter in the derived clock." Says Wavelength Audio's Gordon Rankin.

     

    Adaptive DAC information collected via USB Prober
    ____________________

    Audio Class Specific Audio Data Format
    Audio Stream Format Type Desc.
    Format Type: 1 PCM
    Number Of Channels: 2 STEREO
    Sub Frame Size: 3
    Bit Resolution: 24
    Sample Frequency Type: 0x04 (Discrete)
    Sample Frequency: 44100 Hz
    Sample Frequency: 48000 Hz
    Sample Frequency: 88200 Hz
    Sample Frequency: 96000 Hz
    Endpoint 0x01 - Isochronous Output
    Address: 0x01 (OUT)
    Attributes: 0x09 (Isochronous adaptive data endpoint)
    Max Packet Size: 576
    Polling Interval: 1 ms

    ___________________

     

     


    Asynchronous Mode USB

    Asynchronous USB capable DACs are few and far between. Currently Ayre, Wavelength, and dCS are the major manufacturers with asynchronous products on the market. In my opinion the reason for this lack of async DACs is simply because it's very difficult implement this technology. There is a specific skill set required to implement asynchronous USB and it's not common place in high-end audio. Implementing async USB requires a manufacturer to write its own software for the TAS1020 chip and invest thousands of hours on this part of the DAC alone. The limited number of manufacturers who've decided to take on this task instead of going with a plug n' play chip are doing it because they think the performance gains far outweigh the development pain.

     

    Asynchronous USB essentially turns the computer into a slave device as opposed to adaptive USB which does the opposite. Thus, an asynchronous USB DAC has total control over the timing of the audio. One very important feature of asynchronous USB mode is bidirectional communication between the computer and the DAC. The computer sends audio and the DAC sends commands or instructions for the computer to follow. For example the computer's clock becomes less accurate over a given period of time and can send too much data too quickly and fill up the buffer. Asynchronous DACs will instruct the computer to slow down, thus avoiding any negative effects of a full, or empty, buffer which can manifest itself into audible dropouts and pops or clicks. According to Wavelength Audio the tail is no longer wagging the dog when using asynchronous USB mode. Plus all of this is done without additional device drivers or software installation.

     

    Asynchronous DAC information collected via USB Prober
    __________________________

    Audio Stream Format Type Desc.
    Format Type: 1 PCM
    Number Of Channels: 2 STEREO
    Sub Frame Size: 3
    Bit Resolution: 24
    Sample Frequency Type: 0x04 (Discrete)
    Sample Frequency: 44100 Hz
    Sample Frequency: 48000 Hz
    Sample Frequency: 88200 Hz
    Sample Frequency: 96000 Hz

    Endpoint 0x01 - Isochronous Output
    Address: 0x01 (OUT)
    Attributes: 0x05 (Isochronous asynchronous data endpoint)
    Max Packet Size: 588
    Polling Interval: 1 ms

    _______________

     

     


    Conclusion

    There you have it, my attempt to clarify a little bit about USB audio and explain why all USB implementations are not equal. To render an opinion on the state of USB audio one must research the different technologies and listen to different implementations of each technology. Currently in my listening room I have the Ayre QB-9 asynchronous USB DAC, WavelengthCosecant asynchronous USB DAC, dCS Paganini with Puccini U-Clock asynchronous USB converter, and a number of adaptive USB implementations including the Benchmark and Bel Canto implementations using CEntrance USB code. I am comfortable saying that USB is certainly an audiophile interface and it's ready for prime time. I am not comfortable making proclamations that USB is better or worse than the all other interfaces. There are alsodifferences within USB and I do think asynchronous can be better than adaptive USB implementations provided the implementation is impeccable. Readers considering the purchase of a USB DAC or converter must listen to as many products as possible before making a decision. Reading the TAS article and this article are only the tip of the iceberg. Take everything you've read with a bit of skepticism, but don't second guess what you hear while listening to a USB DAC demo. If it sounds go to you then it's good.



     

     



    Some Photos of my current Asynchronous USB selection

     

     

    Async Stack
    Async Stack
    click to enlarge


     

     

    dCS Puccini U-Clock
    dCS Puccini U-Clock
    click to enlarge


     

     

    dCS Paganini DAC
    dCS Paganini DAC
    click to enlarge


     

     

    Ayre Acoustics QB-9 DAC
    Ayre Acoustics QB-9 DAC
    click to enlarge


     

     

    Wavelength Audio Cosecant DAC
    Wavelength Audio Cosecant DAC
    click to enlarge


     

     

    dCS Volume Control Close-up
    dCS Volume Control Close-up
    click to enlarge



     

     
    Comments 93 Comments
    1. Lars's Avatar
      Lars -
      Thanks for the excellent article Chris. You are doing a great job.
    1. audiozorro's Avatar
      audiozorro -
      Chris, thanks for the audio primer.<br />
      <br />
      I have read that USB is capable of 24/192 if the manufacturer develops a special driver. Since the Four Musketeers of superior USB capable DACs (Ayre, Wavelength, dCS, and Empirical Audio) are capable designers and programmers, where are the 24/192 DACs from each?<br />
      <br />
      I also remember reading than ethernet is also a very excellent interface for computer audio but where are the ethernet DACs and I don't mean wireless?<br />
      <br />
      Clearly every personal computer built generally has USB and Ethernet connectors compared to the relatively few that have toslink, coaxial digital, or firewire. So is the problem that the DAC manufacturers have not fully embraced computer audio and are keeping one foot in the legacy door of CD transports and other audio products?
    1. DanRubin's Avatar
      DanRubin -
      Very nice job, Chris. Well done!<br />
      <br />
      I want to single out this statement you make: <br />
      <br />
      <em>A DAC with a so-called poor USB implementation may have the best S/PDIF implementation on the market and vice versa. Thus the sound of a DAC may vary widely based on the input used.</em><br />
      <br />
      This goes to one of the main bones of contention I have with the TAS article. In addition to its failure to discriminate (or even discuss) the different types of USB implementation, the article seems to lump all DACs with a USB input together in the category USB DAC. The way the ARC and Bryston review is positioned in the magazine makes it seem like these products are the ARC and Bryston entries into the USB DAC sweepstakes. I think that's misleading. These are their DACs, which happen to offer USB inputs. It's too late to put the genie back in the bottle, but we'd be better off if "USB DAC" was used only for products like the Ayre, Wavelength, and UltraFi that were designed specifically for USB. <br />
      <br />
      I know, I'm splitting hairs.
    1. borderdog's Avatar
      borderdog -
      Chris,<br />
      Great article.<br />
      My dog doesn't like to be wagged by his tail, so I bought the Ayre for him.<br />
      <br />
      Aaron H
    1. jivers's Avatar
      jivers -
      Great article, Chris. But I don't remember seeing or hearing any USB DAC's in your "state of the art" demos at the Symposium!<br />
      <br />
      Jeff
    1. Andrew S.'s Avatar
      Andrew S. -
      Thanks<br />
      very clear.<br />
      Wow - you got a dCS front end. Well done. I hope you have got a transparent system. You are going to need it ha ha ha...<br />
      Best<br />
      Andrew
    1. The Computer Audiophile's Avatar
      The Computer Audiophile -
      HI Guys - Thanks for the kind remarks thus far. <br />
      <br />
      @ audiozorro - DAC manufacturers have definitely embraced computer audio but many of them are still figuring out how they'll get into the game. Meaning, should they do Firewire, USB, 24/96 or higher etc... Right now the economy has some manufacturers in a holding pattern as well. I'm positive you'll see 24/192 USB eventually. The manufacturers I've talked to are going to do it right instead of rush it out the door just to have a component that says 24/192 on the box. Don't hold your breath for 24/192 USB, it may be a while. <br />
      <br />
      @ Soundproof - Thanks for the USB information. I wrote an article in April 2008 about problems with USB audio via certain USB ports on computers. It's definitely a concern that everyone should take serious, but steps can be taken to reduce and even eliminate the problem in many instances. Here is a link http://www.computeraudiophile.com/content/USB-Port-Not-USB-Port<br />
      <br />
      @ DanRubin - That's a hair worth splitting.<br />
      <br />
      @ jivers - You had to squeeze that in here didn't you. I would have liked to include some state of the art USB implementations in Studio A, but a line has to be drawn somewhere. I don't think the MBL / Magico v3 / Ayre QB-9 system in the lounge was a disappointment :~)<br />
      <br />
    1. astrotoy's Avatar
      astrotoy -
      I dipped my toe into the waters of hirez with a little E-MU 0404 USB DAC, which does up to 24/192 at 4% of the cost of the top CASH picks. It doesn't need a Lynx card, but is able to take hirez files directly. The only issue is that it uses Win XP and not Vista. It also won't do 192 or 176 with a Mac. It can play the RefRec HRx files at 176/24 with no problems. I am using Media Monkey as the software. Larry
    1. The Computer Audiophile's Avatar
      The Computer Audiophile -
      Hi Larry - There's no free lunch as I'm sure you know. The EMU DAC does support 24/192 and is only $200, but just because it goes to "11" doesn't mean much. I think there is good reason nobody else is supporting 24/192 via USB. While it's technically possible there are some major compromises. This particular DAC has very high noise related to the oscillators and the USB circuitry. Thus, the jitter is about 15x higher than some of the CASH list products that only support 24/96 via USB.<br />
      <br />
      That said, this DAC does have a big following and many people are really enjoying the sound.<br />
      <br />
      <br />
      <br />
      <br />
    1. astrotoy's Avatar
      astrotoy -
      Chris, thanks for the explanation. As this was a test for hirez, I've decided to see what the fancier setups are like. I very much liked what I heard at the Computer Audiophile Symposium and today I had Tim Muratani and Mike Romanoski over to my home to scope out setting up a demo of the new Amarra Model 4 DAC and the software package that was demo'ed at the Symposium. It runs on a Mac through firewire, not USB. I am very interested in copying a fair amount of my vinyl onto 24/192 and use the software to remove clicks and pops. It looks like I would be able to generate the RIAA and other EQ curves in the digital domain as part of the A to D conversion process. I have a lot of old records (mostly EMI and Decca classical from the late '50s through the '60's that use non RIAA EQ. I'll let you know how it turns out. So far I have been very impressed by the expertise of Tim and Romo and how to optimize this all in my system. Larry
    1. Soundproof's Avatar
      Soundproof -
      A problem with audio on some USB-ports. Well done with the write up, Chris.<br />
      You have to ascertain which of your USB-ports is suited for audio, as not all are.<br />
      <br />
      <br />
      The use of USB Audio Devices on the Left-Hand USB Port Is Not Recommended. This applies to ALL MacBook Pro Models (Core Duo and Core 2 Duo).<br />
      <br />
      • The 15" MacBook Pro models have 1 USB port on the left side, and one USB port on the right side.<br />
      • The 17" MacBook Pro models have 2 USB ports on the left side, and one USB port on the right side.<br />
      <br />
      Due to the current USB configuration of the 15" MacBook Pro under OS X, use of USB audio devices is supported on the right-hand USB port only. Use of such devices on the left-hand USB port(s) is not advised because it may cause audio interrupts and/or dropped samples. However, the use of an iLok on the left-hand port has been qualified and is fully supported.<br />
      <br />
      Due to the current USB configuration of the 17" MacBook Pro under OS X, use of USB audio devices is only supported on the right-hand USB port, and the left-hand USB port farthest from the screen. Use of such devices on the left-hand USB port closest to the screen is not advised because it may cause audio interrupts and/or dropped samples. However, the use of an iLok on the left-hand port has been qualified and is fully supported.<br />
      <br />
      These USB port recommendations are specific to USB Audio devices only (such as the FastTrack USB or Audiophile USB). USB Keyboard and Control Surface products do not have a recommended USB port at this time.<br />
    1. The Computer Audiophile's Avatar
      The Computer Audiophile -
      Hi Larry - Wow, what a small world. I just got off the phone with Tim about thirty minutes ago! I think Romo is a very knowledgeable and great person to work with as well. Please let me know what you think of the demo once you have it running in your system and once you have a chance to convert some vinyl. <br />
      <br />
      Sounds like you have some very fun listening ahead of you.
    1. riderforever's Avatar
      riderforever -
      Just some questions: <br />
      <br />
      * what does it mean Polling Interval 1ms, in both adaptive and asynch mode?<br />
      <br />
      * can the oscillator of the motherboard influence the SQ of adaptive USB dacs?<br />
      <br />
      * can a greatly optimized PC audio reduce the difference between adaptive and asynch USB dacs? I'm thinking to a Linux based machine, with no tasks to do but MPD to play music. I think in this scenario the timing should be improved considerably<br />
      <br />
      * what about the buffer of the TAS1020 chip? Shouldn't it help when dealing with jitter? If data coming from PC is used to feed this buffer, and the DAC clock is used to acquire audio samples from this buffer rather than from USB directly, I see no issues in the adaptive implementation.<br />
      <br />
    1. Wavelength's Avatar
      Wavelength -
      Rider,<br />
      <br />
      <cite>* what does it mean Polling Interval 1ms, in both adaptive and asynch mode?<br />
      <br />
      You can see the Polling interval is also set on the adaptive. This is when the computer sends the Start of Frame (SOF) and is programable from 1-32ms<br />
      <br />
      <cite>* can the oscillator of the motherboard influence the SQ of adaptive USB dacs?<br />
      <br />
      Not really though if it does change the variation of the SOF frame then yes it would effect adaptive. But I have not seen that happen in my experience.<br />
      <br />
      <cite>* can a greatly optimized PC audio reduce the difference between adaptive and asynch USB dacs? I'm thinking to a Linux based machine, with no tasks to do but MPD to play music. I think in this scenario the timing should be improved considerably.<br />
      <br />
      Actually not, see it has to do with the variation in timing between the Adaptive device. Optimizing your computer will only effect the variables inside the computer which makes things sound different. It will not effect the adaptive device any.<br />
      <br />
      <cite>* what about the buffer of the TAS1020 chip? Shouldn't it help when dealing with jitter? If data coming from PC is used to feed this buffer, and the DAC clock is used to acquire audio samples from this buffer rather than from USB directly, I see no issues in the adaptive implementation.<br />
      <br />
      Actually the way the code is setup in the TAS1020, NO. See if you look at figure 2-1 on page 25 of the TAS1020B you can see that the the SOF goes into a 16 bit timer and this is output via ACGCAPL/ACGCAPH. These values are used to dertermine the difference between the SOF timing and the internal timer and the difference is applied as the Adaptive difference into the Master Clock (via Frequency Synthesizer).<br />
      <br />
      See there are two jitter problems with Adaptive:<br />
      <br />
      1) The inherent change in the Master Clock due to the way Adaptive protocol works.<br />
      <br />
      2) The Frequency Synthesizer it self has significant jitter even if you set it and left it there. More than 100x that of an external fixed oscillator.<br />
      <br />
      ~~~~~~~~~<br />
      <br />
      24/192: Gang this protocol requires Class 2 Audio support which is not available in Windows in Linux. When they get closer to implementing it we will have a plan to support it.<br />
      <br />
      Thanks<br />
      Gordon
    1. riderforever's Avatar
      riderforever -
      thank you for your prompt response and continuous knowledge sharing.
    1. Andrew S.'s Avatar
      Andrew S. -
      G'day Gorden<br />
      great to see your dacs have found their way down under at long last.<br />
      <br />
      re: "Actually not, see it has to do with the variation in timing between the Adaptive device. Optimizing your computer will only effect the variables inside the computer which makes things sound different. It will not effect the adaptive device any."<br />
      <br />
      A fortiori are you saying that the only thing effecting jitter is the usb implementation?<br />
      <br />
      Cheers<br />
      A<br />
    1. Wavelength's Avatar
      Wavelength -
      A and the rest of the Gang,<br />
      <br />
      Correct... a computer cannot add jitter to the system what so ever. It is only when the PCM data is converted to I2S (or L/R justified) that jitter will occur.<br />
      <br />
      The implementation of this will be critical to the overall jitter of the system.<br />
      <br />
      Thanks<br />
      Gordon
    1. SidneyStencil's Avatar
      SidneyStencil -
      I'm not sure I understand this jitter problem. <br />
      <br />
      Firstly, if the PC sends samples from its memory buffer at 1ms intervals (as per USB specification) to the DAC's own buffer, and the DAC, using its crystal oscillator as a clock, then processes those samples, where do the variations in timing arise?<br />
      <br />
      Secondly, and more important, does it even matter? According to the AES, jitter below 20ns is inaudible, and even something as dependable, common and cheap (and therefore inherently non-audiophile) as the M-Audio Transit measures a mere 2ns.<br />
      <br />
      All answers gratefully received.<br />
      <br />
      Sid
    1. The Computer Audiophile's Avatar
      The Computer Audiophile -
      Hi Sid - Welcome to Computer Audiophile. Please post your second question about jitter in a new thread on the forum. I don't want to derail the discussion covering a USB Audio Primer with very technical jitter talk. We all know jitter conversations carry on forever and nobody comes out a winner. <br />
      <br />
      In the new thread can you also post the AES document that states jitter below 20ns is inaudible? There are several types of jitter, some matter, some don't. Measuring jitter is very difficult as well. I am really interested in the AES testing methodology in how they determined the 20ns inaudibility.<br />
      <br />
      Thanks a ton Sid!
    1. SidneyStencil's Avatar
      SidneyStencil -
      Oh dear.<br />
      <br />
      I've stumbled into The Computer Audiophile Jitter Wars, with my very first comment. I'm sorry; I should have spent more time in the forums first to get the lay of the land. <br />
      <br />
      So, very briefly, here's the AES paper:<br />
      <br />
      <a href="http://www.aes.org/e-lib/browse.cfm?elib=8354">E. Benjamin, E. Gannon, Theoretical and audible effects of jitter on digital audio quality (1998)</a><br />
      <br />
      And here's a more recent study published in "Acoustic Science & Technology":<br />
      <br />
      <a href="http://www.jstage.jst.go.jp/article/ast/26/1/50/_pdf">K. Ashihara et al., Detection threshold for distortions due to jitter on digital audio (2005)</a><br />
      <br />
      Phew.<br />