ONLINE - ISSUE 41
Ten Questions about Computer Audio with Steve Nugent of Empirical Audio
Computer Audio has become the new rage in audio and for good reason: one has an easy and instant access to all their music as well the ability to search out countless other titles via the internet. The issue is simply where to start, though the answer is quite obvious: get a computer, rip and store the files, and then play them back to some DAC. Of course being audiophiles …err the nuts we are… the questions start to pile up rather quickly.
Mac or PC, and then once you got that settled, there is all the minutiae related to just setting-up that computer’s OS and configuration. Then comes how best to rip and how best to save the files, and then to where? Okay, so now that I have my files, how best to play them back and how best to get the files out of the computer and to what DAC? Yadda, yadda, yadda… each question leads to further questions to clarify the previous that then lead to other questions that suggests another question and …a downward spiral down into the rabbit hole we go.
So I went to the 2009 CES and found not only a wealth of information, but a wealth of confusion or at the very least, a wealth of disagreement among those that are either in the recording/software side, the hardware-side, and/or the "expert" sides of computer audio. Now CES is not the best place to get all the answers… time is an issue as is finding all the people to ask, so I came up with 10 ‘key" questions (these are my 10, you may have others or perhaps might not find these of any benefit to you, but I chose them because they are of interest to me and besides they reflect the most common or important areas that seem to pop-up whenever one talks about computer-based audio, so go pound silicon if they don’t work for you.) and emailed them to 12 people in the industry to answer. Their responses are here...
1. Let's start with interfaces; the obvious choices are USB, Firewire, Optical, and S/PDIF. What is your opinion on any of these interfaces? What if any, are the advantages or disadvantages of one over the others in terms of resolution, jitter, etc.?
The only interface of those listed that can have Async protocol, or handshake with flow-control is USB. Firewire does not support it and neither does AES/EBU or S/PDIF, coax or Toslink. Async protocol allows the master clock to be located at the designation device, either a converter, reclocker or the DAC itself, which is the optimum scenario. This establishes the master clock near the D/A conversion and then controls the source rate using a flow-control protocol.
However many USB and all Firewire interfaces do not use flow-control. The jitter characteristics with these is only as good as the circuits/chips that track the incoming stream from the computer. The master clock is established at the source, which is the computer, and then this must be maintained by all circuits up to the D/A conversion. The jitter from the computer must be dealt with by using PLL-type devices (Phase-Locked-Loop). These track the incoming datastream and synchronize to it, as the same time using filtering and other clever tricks to reduce the jitter of the original stream. Fortunately, there are a few very good chips available that deal with jitter very effectively, namely the DiceII chip for Firewire and the TAS1020 for USB. Even though these do not establish the master clock for the system and instead track the computer datastream clock, their jitter rejection is excellent. Current support for samples rates is: USB - 24/96, Firewire - 24/192 and S/PDIF - 24/192. Look for this to change in 2009.
Of the interfaces listed, an important one was omitted, and that is networked, otherwise known as LAN, Ethernet or Wi-Fi wireless network. The protocol of this network has inherent in it the flow-control and retry mechanisms that enable the optimum audio streaming scenario, as well as having the advantage of avoiding altogether the sometimes troublesome audio software stack of the computer OS. Using networked devices, either wired or wireless can be no different than sending data to a printer. The only concern is getting the data to the device intact. There is no timing information sent or implied. The data is not contiguously streamed at real-time speed as with USB, Firewire or S/PDIF interfaces. It is packetized and sent periodically in high-speed bursts over the network, whenever the network has an "opening". These packets are then collected in a buffer memory at the destination device where they can be clocked out to the D/A using a local low-jitter master clock. The fact that networked data flow incorporates flow-control and retry, and bypasses the computer audio stack makes it the superior method. The only disadvantage is that the player that interfaces to the network is currently a custom player, such as Squeezecenter or Sonos. Hopefully, in the future Microsoft and/or Apple will create more generic player software to drive a networked interface so that more player options will be available. As for bandwidth, networked interfaces can not only support the highest audio sample-rates, such as 24/192, but it can support multiple channels of this, allowing for multiple channel playback for software generated speaker crossover and even movie surround-sound.
2. With regards to software there are also strong opinions as to some being vastly superior (or for that matter, inferior) to others; people clearly hear differences in how files are being played back and therefore prefer one over the others. There is also a growing opinion that Pro software is the only way to go and that using iTunes, WMP, MAX, or other free software playback programs (FooBar, JRiver, MAX, etc.) is not the way to go. That is, these are sonically and musically inferior to the Pro software because the Pro software (say for example Amarra, Izotpe, etc.) is simply "better" at playing back music files. What is your opinion on what is going on here? That is, why would any of these programs be superior—or for that matter, inferior - to another with respect to say a .wav file in any resolution: 16/44.1, 24/96, or 24/192? Is it a matter of timing and jitter? Issues with the operating software and processing? The fact that some software runs "cleaner" than others—that there is nothing running in the background to muck things up? Or as some suggest that the "math" is simply better in some software than in others?
Software programs for playing back audio streams using the software stacks on both Mac and PC are still in their infancy IMO. User interfaces vary widely and capability for bit-perfect playback varies as well. Jitter from real-time playback of different CODECs, particularly compression algorithms is annoying at best. Latency varies a lot from one to the next as well, making some suitable for hi-res and recording studio work and not others.
Both Mac and PC software has gotten a lot better sound quality wise, although not necessarily easier to use. Windows Kmixer, which was part of the PC audio stack for XP machines and required users to jump through hoops to bypass has been eliminated in Vista. Vista is still not bit-perfect however, unless your files are 24-bit. iTunes is bit-perfect on a Mac, but not necessarily on a PC.
Player software has to deal with all of these issues:
1) Bit-perfect playback
2) Achieving low-latency
3) Support for lots of formats
4) Legacy XP support
5) Interface support
Almost all of these problems have to do with the audio software stack of the OS on the computer. These are mostly avoided if one uses a networked solution, although drop-outs can occasionally occur if there is Wi-Fi interference or lots of network traffic.
Players that I like the SQ of are:
Foobar 0.8.3 and SRC upsampled to 24/96. If one uses Kernel Streaming or un-maps the device on the PC, the SQ can be outstanding. I prefer .wav files to anything compressed, even lossless. Other good players for PC are Jriver and XXhighend.
3. Let's move on to ripping. As with the above, there are proponents that claim only certain software, and optical drives for that matter, can "accurately" rip a CD. That they can clearly hear differences between rips via different means; even though the rips are bit for bit perfect. Any thoughts on what is going on here? Is there an advantage to using specific ripping software or drives over another? Say iTunes, WMP, Max or whatever when compared to say EAC?
EAC is the best ripper IMO because it calibrates to the CD-ROM drive "offsets" and generates a checksum on every track doing multiple reads.
There is a growing body of evidence that ripping with iTunes, particularly to Apple Lossless may compromise the sound quality
4. File formats. Any reason why a .wav, AIFF, or FLAC file is better than say Apple Lossless? Again people suggest a strong preference for one over the others, so something must be going on here?
I prefer AIFF and .wav files. If disk is cheap, why bother with 50% compression that you get with a lossless format?
Even though they generate a bit-perfect uncompressed copy, my customers have reported that the lossless CODECs sound different than the same .wav file on playback. My theory is that these CODECs causes significant jitter with real-time playback.
5. There is also a movement towards Pro DACs. Naturally there are DACs of varying quality and performance, but is there any reason why a PRO DAC would be better than a DAC made by a manufacturer from the audio community? Say ones of comparable quality and build?
Absolutely not. It is more a function of the designer's talents than whether it is pro or consumer gear IMO.
6. Along those same lines, what makes one DAC a better choice for computer-based audio than another? Jitter reduction, chip sets, power supply, etc?
Jitter is the number one issue with digital playback, not just from computers. When one has a low-jitter source, all DACs start to sound more similar, with the simpler ones taking the lead. I am convinced that this is why there are so many conflicting reports on a given DAC.
The second important issue with DACs in general is digital filtering. Without digital filtering I suspect that many DACs would sound significantly better. There is a growing following among audiophiles for NOS DACs, and it makes perfect sense. Some don't find them interesting, but I believe the problem is driving them with a low-jitter source. This evens the playing field dramatically.
Other critical ingredients that make an excellent DAC include (in no particular order):
1) D/A chip choice
2) PC board layout
3) I/V conversion
4) Analog filtering
5) Power delivery to both the D/A chip and the analog output stage
6) Simplicity of the analog output stage
7) Digital signal paths to maintain low-jitter
7. What do you see as being the most important factor in getting the best sound in computer-based audio? That is what should the consumer address with the greatest concern when setting up a computer-based audio system?
Jitter by far is the most important issue for digital playback sound quality, both CDs and computer audio.
8. Along with that, what do you see as being the most important factor in NOT getting the best sound in computer-based audio? That is, what can have the greatest potential to adversely affect the sound in computer-based audio?
The thing to be aware of and avoid is the pitfalls of Windows Kmixer and compressed audio formats, such as MP3.
9. Some suggest that they computer must be audio dedicated. That is it must be "built" or configured for the specific purpose of only playing music and that any and all non-audio related programs and such must be eliminated. Your feelings on this? Is it important or not, and why so?
Depends on the playback interface. If it is networked, then the computer can be used for lots of other work.
If you are playing back using USB or Firewire and using compressed formats and on-the-fly upsampling, then you need really low latency and a lot of megaflops out of the CPU. In this case, I recommend a dedicated computer.
10. Where do you see the greatest impact to come in computer-based audio for the future?
Hi-resolution formats transmitted over networks to end-point devices that support multiple streams. These technologies are just around the corner. I see the speaker of the future having no crossover. The customer loads the crossover info into his computer for his particular speakers. Then the S/W player streams three stereo streams either wired Ethernet or Wi-Fi at 24/192 to a receiver device, which clocks three stereo DACs at low-jitter. Then the DACs feed six digital amplifiers which in turn drive two multi-conductor cables to the stereo speakers. This is the future of 2-channel IMO.
Steve Nugent has 26 years experience developing digital interconnects and I/O circuitry in the computer industry at companies including: Intel Supercomputers, Intel Corp., Unisys, Siemens, StorageTek, Burroughs and Honeywell. Steve currently is the designer of Empirical Audio's line of USB converters, DACs, reclockers, and cables. For more on Steve, read an interview that ran in Issue 7 of PFO.
Use the links below to read other responses to these questions