This review page is supported in part by the sponsors whose ad banners are displayed below

Today’s witches’ brew of faster sampling rates, computer networks and streaming point-to-point digital protocols such as AES and S/PDIF makes jitter control a major issue. Given the low noise floor in high-quality 24-bit recordings, new artifacts emerge from their former hiding places. The JET PLL used in the INT 202 is a new design and includes both software-based and analogue elements. The jitter attenuation and clock regeneration process requires a stable crystal-controlled reference clock; a numerically controlled oscillator and filter that produce an internal intermediate clock; a higher-bandwidth voltage-controlled oscillator (VCO); feedback loops and filters. The interaction of all these ultimately controls the timing of the AES and S/PDIF signals.

Adjusting the parameters of the software filters and hardware feedback control loops optimizes the jitter attenuation. The flexibility of this software-mediated process would be very difficult to achieve with purely analogue engineering. The numerical parts of JET technology run in the ARM processor, which maintains 96 binary bit accuracy. The result of all this is a PLL that allows elaborate tuning of its performance, which in turn minimizes jitter across the desired range of audio frequencies to a much greater degree than before.

Software-based level control:
The INT 202 employs a dithered digital attenuation technique whereby the level of the incoming digital audio signal is reduced by software signal processing algorithms. This approach has several advantages compared to switched resistor networks and especially ganged analogue potentiometers. First, it doesn’t age or wear out. Second, linearity and exact tracking between left and right channels are easier to maintain. Third, an all-digital implementation reduces the number of discrete parts required, thus reducing overall real estate.

By adding attenuation control to the playback chain, it’s possible to eliminate a preamplifier in many cases. DACs can often drive an amplifier directly, which reduces cost and complexity. This approach also has the potential to deliver superior performance. One of the fundamental catches in digital audio in general and software-based signal processing (here attenuation) in particular is that when representing or quantizing an analogue signal with a finite number of binary digits, the conversion is seldom perfectly accurate. Quantization errors during the original sampling by the A/D converter, subsequent processing and conversion back to analogue eventually show up as objectionable artifacts.

This happens because the input signal is rarely exactly equal to one digital number or another. It’s almost always somewhere in-between. 16-bit technology for example has a numerical palette of only 65,536 quantization levels. The representation of analogue waveforms is somewhat crude as a result. Going to 20 or 24 bits greatly increases the available range of numbers—good—but on average the digital equivalent is always going to be one half bit off, either slightly too low or high. The quantization process introduces inherent distortion and noise that’s highly correlated with the input and difficult to remove using analogue components.

To solve this, signal-processing scientists and engineers figured out early in the digital era that injecting small amounts of random digital noise or dither resulted in a statistically better representation of the original signal. This general approach discussed here in simplified terms ultimately allows the quantization noise to be uncorrelated with the input, shaped and essentially parked (but not entirely disposed of) in unfashionable remote portions of the audio spectrum. There’s no magic bullet here because these errors can’t be made to truly disappear; however they can be put some place where they won’t bother us as much.

Bit-perfect data transfer:
Weiss supplies twelve special .WAV files (one for each of the sampling rates and sample word lengths supported) that confirm bit-perfect transmission from source to INT 202.  Selecting the bit-transparency mode mutes the INT 202 and special logic inside the INT 202 inspects the incoming bit stream. If all is well, the front panel indicator lamp keeps flashing.

This is more of a problem than one might at first imagine. All too often even the most experienced digital audiophiles will find that something is a little bit off during critical listening or measurement sessions. In one recent exchange on the forums, an audio engineer (not a hapless consumer) discovered that a test signal was about 1dB lower than expected. It took several days and help from several knowledgeable readers before everything was sorted out. Somewhere along the line, an operating system based audio equalizer (OS X in this case) had unintentionally kicked in to the playback chain.

At the root of such difficulties is the fact that both Microsoft Windows Vista and Windows 7 as well as Apple OS X contain software stacks that incorporate signal processing, dithering, equalization and more which in turn result in non-bit-perfect digital output. Depending upon your application, these audio processors are actually very useful especially for home-theater and portable players. For audiophile scenarios however, they’re anathematic.

On Windows Vista and Windows 7, users can install the $49 J.River MediaCenter (JRMC) or similar software, select Windows Audio Session Application Programming Interface (WASAPI) as the preferred audio subsystem to bypass the Windows Media Player and other internal code and they’re done. Note that while JRMC contains various signal processing applications, once WASAPI is set, they’re disabled by default. Getting things correct in iTunes is a bit more complicated but it will eventually work as well.

Absolute phase and muting controls: With higher-end gear, reversed polarity in recordings is often all too apparent. Naturally produced audio waveforms are generally not particularly symmetrical with respect to their amplitude. Positive pressure waves in the concert hall become negative pressure waves when reproduced in the home and such phase reversal can do strange things to the soundstage and transients. A fair number of recordings have inverted absolute polarity and in the worst cases, it may even vary from track to track. Some recording and mastering studios and even a few consumer components intentionally invert phase for technical reasons. In other cases it’s usually a matter of negligence. Having the ability to toggle absolute polarity from the comfort of the listening chair is an important creature feature and an essential tool when sorting out one’s system.

Single-wire/dual-wire output modes: Digital PCM audio data departs the INT 202 as balanced AES/EBU signals via a pair of XLR connectors and as unbalanced S/PDIF (the consumer-oriented analogue of AES) on RCAs. In normal single-wire mode, every one of the connectors carries the same digital stereo data stream, which allows four external devices to be driven simultaneously. A second configuration available for 176.4 or 192kHz sampling rates is the dual-wire mode, which supports single-channel monaural DACs (i.e. one channel per chassis) such as the Esoteric D-01. Some stereo DACs require dual-wire inputs at these sampling rates. In either case, channel 1 (nominally left) goes out through one XLR connector; channel 2 (right) goes out through the second. The same arrangement applies to the RCA connectors.

Setup and integration
: All that’s required to install the INT 202 to the media server is to connect the power and FireWire cables, then to install the FireWire drivers and device management control applets. These are supplied on a DVD that accompanies the unit. Driver updates and firmware revisions may be downloaded directly from the Weiss website. The software allows control over all settings and provides detailed status information displayed on various applet panels. Weiss’ recording industry heritage and the robust nature of FireWire are apparent in the range of available parameters and configurations. The current sample rate, bus mastership and other key settings should be inserted automatically as the media player software streams tracks to the unit.

After the driver installation, the J.River MediaCenter had no difficulty establishing the connection with the INT 202. Compared to death battles with previous FireWire drivers I’ve experienced using other products, this time everything went exactly as planned. After selecting WASAPI in JRMC, it was easy to demonstrate bit-perfect communication using the test files provided.

One applet panel monitors the performance and integration of the operating system drivers, the FireWire bus and the devices on it. Technically we’re checking the deferred procedure call latency (DPC), which is a key factor in determining real-world end-to-end throughput. It’s usually on the order of 500 microseconds or so, half a millisecond.

In normal operation, this latency will fluctuate somewhat. If peak latency crosses certain thresholds, the software will recommend one of three safe mode levels, which can then be set manually in the bus parameters panel. Safe Mode 1, which derives from standard Windows Device Model nomenclature, is preferred as a starting point. The higher Safe Mode levels apparently introduce longer latencies but for playback-only applications and current dual-core processors, this should not be an issue.

With the bit-perfect test files loaded onto the host media server or computer, one flips the TRSP CHECK toggle switch to its ‘on’ position, which immediately mutes the INT 202 once again without any annoying ticks. As soon as each of the 12 test files starts playing—there is one for 16-bit and 24-bit samples at each of the sampling rates supported—the power LED starts flashing and you’re ready to go. If not, one or more setting in JRMC must be corrected.

Chief culprits are likely to be that WASAPI was not selected correctly (DirectSound is the default); the wrong volume mode was set (select Application); or the JRMC DSP module is engaged, which may activate upsampling, apply equalization or otherwise corrupt the bit stream. The default DSP settings are set up to avoid any adulteration so if there’s a problem getting bit-perfect output, rest assured that it’s entirely your fault but will work once you correct it. With OS X, similar issues exist when using iTunes especially for higher-resolution file formats but once again, with a little care it is possible to get bit-perfect output. Just keep an eye on it. Once bit-perfect communication has been established, turn off the TRSP CHECK switch and start listening.