TAS 07/08 2011 "The Berkley Audio Design Alpha USB Interface" just hilarious

michkhol -- Sat, 07/16/2011 - 09:39

.. USB interface audibly degrades the signal passing though it...
... it is a "packetized data" format in which data are split up into discrete chunks, wrapped up with information about those chunks, transmitted, and then put back together at the receiving end.
Dear Mr. Harley,
if it were really so, no one would be able to use a single external USB drive. So if there is any degradation, it is outside the USB interface. Unless of course it's a sales pitch.

Sam -- Sat, 07/16/2011 - 23:49

I don't understand what you are saying Mike?

Jon2020 -- Sun, 07/17/2011 - 23:32

Although digital ones and zeroes get through to their destination from point A to B through a USB interface, what Mr harley is referring to is the degradation of the sound, not the 1's and 0's. The degradation of digital audio through any interface, as it finally sounds from your speakers is well established. The ones and zeroes may remain intact. 

Chris Martens -- Mon, 07/18/2011 - 12:02


I fully understand your point regarding data integrity, per se, and agree that USB does, as you rightly point out, a fine job of keeping the 1's and 0's straight. But that said, there are also at least two other ways that interfaces can cause signal degradation--at least for purposes of playing digital audio files. One problem is added noise, and the other is added jitter (that is, timing errors).

I think the benefit of doing a well-executed USB-to-S/PDIF (or USB-to-AES) conversion is that coaxial S/PDIF and blanaced AES interfaces both have (or at least are thought to have) a better track record than USB in terms of minimizing noise and jitter problems. Also, an awful lot of higher-end DACs are designed around the working assumption that your main digital interface will be coaxial S/PDIF or balanced AES. In this way, the Berkely serves as a gateway between your PC (or Mac) and higher-end DACs.
Does that help explain the value proposition of Berkeley's Alph USB interface?
Best, Chris Martens


Chris Martens
Editor, Avguide.com/Playback/The Perfect Vision 

michkhol -- Tue, 07/26/2011 - 14:28

Frankly I thought it would be obvious. But the case seems to require more explanation:

First about the USB bus clock. It operates at 1KHz and data sent each 1ms is called a frame. Each frame contains a special Start Of Frame packet at the beginning so any device could synchronize with the bus clock. While the clock rate can vary, it is NOT the jitter (something causing sample read errors) because the timing of the frames changes with the clock but the data in the frames remain intact.

There are two ways the audio data is usually transferred via USB, interrupt and isochronous.The former was used when the isochronous mode was not yet supported. In the interrupt mode the host periodically polls the device if it needs any more data and each data packet requires a confirmation. Thus the data integrity is guaranteed but the timing is not. Given the large enough buffer on the receiving side the resulting delays could be avoided at the cost of increased delay between sending data and hearing the actual sound, but no one seemed to bother. Instead people started to use

The isochronous mode. In this mode no confirmation packets are necessary, neither any device polls. Into each frame the host inserts the exact amout of data needed to play audio for 1ms. Then the frame is sent off. It is guaranteed that each frame will contain the data for the device, it is NOT guaranteed that the packet in the frame will be delivered. So we may lose audio data in 1ms chunks. This may be the cause of the "noise" you are talking about.

I hope now you see, in the isochronous mode no USB DAC, neither for $10 nor for $100000 can prevent losing data packets because it happens even before they reach the device input. And hence no isochronous USB DAC can guarantee 100% audio quality.

Now about the S/PDIF (AES) conversion. What actually happens is that the signal from the digital domain is converted to the analog domain. Bits and bytes become different voltages at the output with the generated clock to be able to read them back. Add here the self-noise of the signal output stage. As a result you get the guaranteed jitter and noise when the S/PDIF (AES) signal reaches the destination.

So the Berkeley's Alpha USB interface may be a decent converter for those who does not have a USB input in their DACs but it is not able to remedy any data loss which happens before the data reaches the interface itself.


Jon2020 -- Tue, 07/26/2011 - 23:17

Good points raised by your good self. This is the very reason why isochronous USB is passe for audio data while any USB converter worth its audio salt today will have to be asynchronous. And so it is for the Berkeley Alpha USB DAC.
Cheers! Jon.

discman -- Wed, 07/27/2011 - 05:32

If Mike is trying to impress us with how much he knows, which is my understanding of his acerbic thread title and insulting reply opener, I first suggest some study of communications. Like Sam, I don't understand some of what he is saying, and I've read it twice. I get his meaning, but it is easy to be a jerk without impressing or imparting information.
Thankfully, he provides enough information that I can make an educated guess. Start with:
-- Asychronous (interrupt) mode USB can transmit data faithfully, but can alter the timing of the bitstream
-- Isochronous mode USB can lose data in  a way that the data cannot be recovered
-- S/PDIF (AES) mode introduces potential voltage and timing degradation to the bitstream
As Jon points out, in this case we can forget about isochronous mode. Berkeley doesn't use it, so it really isn't relevant here.
In the case of asynchronous mode, we are left with the impact of the timing variations introduced by the interface. Harley says the interface audibly degrades the signal. I think, from past readings and discussions of this, that Harley's point is based on the (possible) impact of these timing differences on the D/A conversion process (i.e. timing errors lead to analog signal errors). Mike, I believe, thinks that these timing variations can be dealt with flawlessly on the receiving end. If I understand all this correctly (and I may not!), then Mike's original beef is with Harley's statement that the USB interface causes audible degradation. I think Mike would be happier with a statement like "timing variations introduced by the USB interface mean that insufficiently well-engineered DACs may generate audibile signal inaccuracies".
At first I thought this was just a pedantic attack on the common choice of a consumer writer to smooth out his sentences with some loss of precision. But I think Mike may be raising two interesting points. I phrase them here as questions to indicate that I don't know if this is correct, it is what I surmise Mike and Harley are trying to say:
1. USB per se is a perfectly useful interface? The tendency to use wording that suggests an intrinsic flaw, when in fact only an engineering issue has been raised, is problematic for consumers trying to decide what to buy.
2. USB creates issues that are so commonly not dealt with, or are extremely hard/impossible to deal with, that consumers should not casually choose this interface for the highest quality music systems?
Often the conversation about point 2 gets lost in a shouting match about how digital systems don't lose data and thus are provably perfect. I would be curious to know more about 2 and thank Mike for starting a thread that leads there (I think).

michkhol -- Wed, 07/27/2011 - 10:37

All information I provide here is freely available on the Internet, I only searched for it. The fact that it may impress someone I take as a compliment :)
Before I return back to the technical discussion I'd like to reiterate the point I was trying to make in the first post. The statement "USB interface audibly degrades the signal passing though it" is wrong on at least two counts. First, the signal(s) passing through the USB interface are not audio, they just represent audio _data_ among many other types of data used in the USB communication. The audio degradation (if any) happens only and if the digital signal from the USB interface is passed through a poorly designed USB-DAC link. Second, taken in the context of the article the statement implies that conversion to S/PDIF produces a "better" audio signal, or in oher words a USB interface is apriori worse than an S/PDIF. Well the fact is (see my previous post) any conversion to digital-through-analog introduces at least some jitter and noise _in addition_ to any deficiencies already present in the signal.

Back to technicalities. First my apologies for not paying attention. I completely missed that the Berkeley USB interface is of asynchronous type. The asynchronous type is one of the isochronous types (and not the interrupt mode). The one I described is called adaptive and is used in cheap USB interfaces. The difference is in the clock source. In the isochronous mode of asychronous type the device (not the host) is the clock source. The implementation is more complex, because _both_ host and device should be able to support it and no USB chip supports it out-of-the-box. It has to be reprogrammed. That's why asynchronous USB interfaces are more expensive. Essentially the asynchronous type solves the problem of the unstable host clock. What I could not find out is, whether it can ask the host to retransmit a lost packet. If it can, than the answer to your first question is yes.

My personal answer to your second question is using a professional FireWire interface. What's good for recording music obviously can't be bad for playing it back. You have to be careful though choosing the FireWire controller for your PC. Macs do not have this problem, all of them work fine.

All content, design, and layout are Copyright © 1999 - 2011 NextScreen. All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.