Interview with Halide Design

August 15, 2011

The Halide Design USB to S/PDIF bridge is the most compact unit I am aware of. How were you able to incorporate all of the bridge circuitry in a slightly oversize silver RCA Eichmann RCA plug or BNC connector while your competitors need resort to a much larger enclosure?

First, keep in mind that we built the circuit as we thought it needed to be built, and then realized that it could be built into the plug after playing with the initial layout. That is to say, we didn’t skip on any parts in order to build it small; it just so happened that because of the nature of the circuit, we were able to push the design a bit, discovered that we could fit everything we wanted to right into the plug.

It required a bit of a push on the manufacturing technology, though. The enclosure is custom CNC milled from a single block of aluminum (similar to what Apple does for their laptops), with a small separate piece for the endcap on the back. This allowed us to make it exactly the right size, and thread directly onto the Eichmann RCA tip, which, by the way, is itself soldered directly to the circuit board.

For the BNC version, we actually use a 3D printer to make the adapter ring to fit the connector to the enclosure, and then jumper a very small distance with a silver wire. That seemingly innocent red ring you see at the end of the BNC version is actually made with some serious technology. The BNC connector itself is a standard part, but is machined on the back end in order to fit correctly onto the board. Again, without things like CNC and the 3D printing, which are not yet common in the audiophile world, this design wouldn’t be possible.

One quick point about building into the plug, and the reason we wanted to do this: it’s a known bugaboo of the S/PDIF format that the cable and termination is specified as 75 ohm impedance, but an RCA connector cannot pass 75 ohms, due to the geometry. This pretty much guarantees reflections from the ends of a cable, which can distort the waveform. The solution we have is just to skip the cable all together, and essentially inject the waveform directly at the DAC. That is, the best S/PDIF cable is no cable at all.

It’s not really fair to comment on the other manufacturers, and it’s always hard to second guess someone else. I think a number of these circuits could indeed in principle be fit into an enclosure similar to ours, it just takes some engineering (and of course adds to the final cost). Another well know S/PDIF device, for instance, does build the circuit into the plug, but at the USB end … in our view, that’s exactly backwards, since the cable you want to eliminate in the S/PDIF cable, not the USB one.

You have implemented an asynchronous data transfer interface in the Halide Design bridge. What are the advantages of asynchronous over other methods?

“Asynchronous” is a somewhat undescriptive term that’s used to describe USB audio systems where the master clock is on the device itself, rather than derived from the computer. The aysnch comes from the fact the audio clocks on the device run at exactly the speed they need to run at (some multiple of 44.1 kHz or 48 kHz), and the clocks on the computer (and the USB transmission) run at whatever separate speed they need to.

So, despite the terminology, it’s probably better to think of this as a synchronous technology, because everything on the audio board can run locked to a high accuracy master clock. This essentially puts the USB technology right back into solid audiophile territory, where the jitter on the system is just limited by the designer’s ability to make a clean, accurate clock, independent of what the computer is doing.

The other way, which at one point was the only way, is to take the computer clock, and try to smooth it out as much as possible, and use that. The chips that do this, the classic Burr Brown chips and the newer ones by Tenor, can actually do this much better than you may think. In addition, some designers add an additional layer of reclocking on the system, which can make the final jitter number come out quite low, but can add some unfortunate reclocking artifacts. (And of course, this adds to the size, weight, power consumption, and price.)

Like several other asynchronous USB products, you license the Streamlength asynchronous firmware from Wavelength Audio. Are they the only manufacturer of asynchronous technology? If not, what advantages do they bring to the table over other vendors?

Wavelength was the first to figure out how to do this, at least in a plug-and-play system. There are at this point some other manufactures doing this, such as dCS and those who license from them. In some sense, it’s astounding how quickly technology progresses.

One thing to keep in mind is that although an asynchronous system allows for incredibly low jitter, it doesn’t by any means guarantee it. You still are limited by the ability to generate a low-jitter master clock, the same limitation that has existed in digital audio ever since it was first introduced in the 1980’s. There are actually non-asynch devices out there with lower jitter than some asynch devices. These things depend tremendously on implementation.

In addition to utilizing asynchronous data transfer, what other steps have you taken to reduce jitter?

The most critical step is getting a clean master clock. Without that, all hope is lost – the rest of the circuit can only add to the error. The power supply must be as clean as possible, so the USB power is filtered fairly aggressively with some passive components before being down-regulated by an extremely low-noise regulator. The entire circuit is designed using RF (radio frequency) techniques, such as curved circuit traces and ferrite beads. A final trick that we learned involves using a tiny chip called a flip-flop to re-lock the S/PDIF signal to the clock before going out.

Jitter levels and their audibility can sometimes be a contentious subject. Can you explain what jitter sounds like and what audiophiles can expect to hear as jitter is reduced?

It’s funny, because jitter has taken on an almost mythical status in the audio world, so it’s worth stopping to note why. For one, it’s almost bizarre how sensitive the ear is to even small amounts of jitter. A nanosecond of jitter is considered fairly large, and is certainly audible. But this is actually a tiny number – a nanosecond is a billionth of a second, not enough time for light to even travel a foot. Still, listening tests (or some back-of-the-napkin calculations) reveal that at this point, you’ve essentially lost the ability to reproduce even 16-bits of resolution.

The equipment needed to measure audiophile jitter levels, which may be a few hundred picosecond (peak-to-peak, across the audio spectrum), is extremely expensive. In many cases, the jitter on the test equipment will be higher than that of the audio device, which of course was designed to push the limits of jitter to as low as possible. In addition, there are a number of different techniques (such as the popular j-test, or measurements taken directly from the clock), as ways of analyzing the data (by measuring peak-to-peak vs. RMS, or various ways of bandwidth limiting), making comparison of two different numbers almost meaningless.

Because of this, the really high-end guys – those that understand the nuances – don’t bother to cite a number, because they know the numbers can be gamed. Some of the mid-fi guys simply claim “zero-jitter” or “jitter-free,” which of course is impossible, and completely misleading and unhelpful. It’s worth noting that Stereophile does a fairly good job in a difficult situation by using a consistent measurement setup, so that different devices at least go through the same test, and clearly explaining the test setup (and its baseline limitations).

In terms of sound, it depends a lot on the character of jitter. High jitter makes things sound, for lack of a better word, “digital.” There is a distinct, sort of harsh or edgy sound that poor CD players are known for, that comes largely from jitter.

Depending on the nature and frequency of the jitter, it can do anything from wash out the bass, blur out the instruments, or flatten or shrink the soundstage. Conversely, lowering the jitter has a similar effect of increasing the resolution of a system through other means – truer instrument texture, more resolving details (without increased harshness or brightness), and clearer instrument separation and soundstage.

What playback software do you recommend for Windows and the Mac, and why?

Windows is sort of notorious for screwing with the audio data coming out of the computer. I have no idea why they do this, but by default, Windows is not bit-perfect. We typically recommend JRiver, though there are other software packages out there that allow for bit-perfect playback.

Similarly, Amarra is popular on the Mac, though Macs tend to be much better about playing out the true audio. One advantage of these players is that they will automatically set the system output rate to match the audio rate (44.1 kHz, 88.2 kHz, etc.), so there is no resampling.

In general – and this is true for anything in audio – if you don’t hear a difference, don’t bother with it. In the case of something like JRiver on a PC, it’s an immediate, very obvious difference.

The Halide Design Bridge is an all-in-one product preventing audiophiles from trying different USB. What is your position on the efficacy of after-market USB cables?

Keep in mind that the entire design of the Bridge is to make the quality of the computer’s clocks irrelevant (because of the asynch), as well as the quality of the computer’s power supply (because of the filtering and regulation). So, similarly with the USB cable, which should have no effect on the circuit performance, by design. Regardless, the cable we use passes the USB 2.0 spec, which is actually more than can be said for many of the “audiophile” cables out there.

When we started making USB audio devices, there was no such thing as an “audiophile” USB cable. About a year ago, we took a look at what was out there, and got about half a dozen of the newer cables. Even after examining these things outside and in, it’s not clear what exactly many of these guys are selling. Some manufacturers make claims that are not accurate, though it takes a scalpel and wire clippers to find this out. Most of the cables range somewhere between a poorly made to well made, but very standard, USB cable. Some geometries are simply bizarre.

Regardless, we decided to try an “upgraded” USB cable for the upcoming DAC HD, which uses Wireworld’s Starlight cable. The difference is probably minimal, but customers had requested such a cable, and Wireworld was one of the few designs that we thought was compelling, since they separate out the data and power lines, and wasn’t absurdly priced. In this case, it allows for cable lengths up to 7 meters, due to the thicker wires (the DAC HD draws much more power than the Bridge).

At one point, it was on the to-do list to get the cables actually tested, and measure things like transmission line impedance, or noise-crosstalk into the power line, but this maybe fell a bit by the wayside due to disenchantment with the industry. We may do this at some point. We may go back and do this at some point, and post the results …

Although the built-in six foot cable should prove adequate in most situations, are there any caveats or limitations with respect to using USB extension cables?

Again, it should it should be essentially transparent, for the above mentioned reasons. Because the clocks run at the tip of the Bridge, extending out the cable doesn’t change the final signal. It’s easy to get this confused with running a longer S/PDIF cable, which can indeed degrade the signal.

Your sister company, devilsound labs, offers a small non-oversampled (NOS) DAC built into a USB to dual-RCA plug cable whose playback it limited to 32kHz, 44.1kHz and 48kHz sample rates, with no resampling, upsampling, or oversampling. What are the advantages of a NOS DAC?

Well, the short answer is just that they sound good. People debate why, but it’s sort of been an audiophile “secret” for more than a decade. NOS DACs output at full bit-depth on each sample, which makes them much less sensitive to jitter. They don’t have the high-frequency switch-noise that is somewhat difficult to deal with in a standard DAC design … or maybe it’s something about the extra ultrasonic energy that’s generated, which makes things sound natural. In the end, though, it’s about how it sounds, and the devilsound DAC sounds pretty damn good.

I understand you have a new 96/24 capable USB DAC in the works. What can you tell us about it?

This has been a long time in development, and should be out before the end of the year. The new DAC has the Streamlength code for extremely low jitter, as well as a number of other circuit improvement … details will be posted on the website as we get closer to the release.

Thank you for taking time to speak with me today.  Readers can see my review of the Halide Design USB to S/PDIF Bridge here.