Benchmark media systems DAC 1.
The DAC 1 uses a classic 1999 Analog Devices AD 1853. The best-performing DAC offered by Analog Devices, the 2003 TI, gains a bit on this device at twice the price, at least for devices in small quantities as indicated on the TI and ADI web sites. The AD 1853 DAC has mono and stereo modes. In stereo, only one chip is needed; in mono mode, a 3-dB noise improvement occurs when the two channels are combined. Typical signal-to-noise ratio in mono, with no weighting filter, is equivalent to 19 bits. No worst-case spec is given. Worst-case dynamic range in stereo is slightly better than 17 bits. Typical distortion for a 1 kHz full scale tone with the chip used in mono mode is 17.5 bits full-scale and a worst-case number is not provided. The worst-case stereo 1 kHz full-scale distortion falls slightly below 16 bits.
The typical distortion at 20 kHz with the signal -3 dB down from full scale is shown on a graph at 16.5 bits equivalent level. The complete DAC as measured on an Audio Precision (not just the filter characteristics themselves) is +/- 0.0002 and the stop band is -120 dB.
For the DAC 1 to deliver this performance, the PC-board implementation must be at least as good as the evaluation board used by Analog Devices to make the data sheet measurements. Most IC manufacturers supply the design of the board down to the level of the PC-board layout to assist users of the chip in achieving their performance objectives. Some companies will supply the actual board.
The clock signal is not supplied. However, in the lab, that is not a problem, since we can substitute a high-precision source. We need the clock to be very stable (the infamous jitter spec) to achieve the distortion levels of which the chip is capable. In designing a complete product for use by a consumer, the jitter of the SPDIF (Sony Philips Digital Interface Format) signal coming out of the back of a CD or DVD player must be dealt with. Benchmark solved the problem not by trying to recover a stable clock from the CD player but rather by not using the clock at all to drive the DAC. Instead, they use a sample-rate converter. This digital chip (to simplify dramatically) re-samples the incoming signal with a stable reference clock. Similar to a challenge commonly faced by CD-player manufacturers, which already have a fixed reference clock built in, the reference clock of the Benchmark One must be stable enough to hit the distortion figures. Benchmark's engineering team R&D efforts were rewarded with a stable crystal oscillator that produced a jitter level below what was required for the DAC performance to dominate clock deficiencies. According to Benchmark, this is an important design criterion of a crystal oscillator circuit that rejects power-supply noise.
John Siau, the lead engineer, has an impressive depth of understanding with respect to the performance characteristics of the sample-rate converter. Converters performance vary depending on the design used by the IC manufacturer, and the digital data that comes out of the chip can have more noise then the incoming digital data signal. This is not a clock issue; instead, it speaks to how the bits are manipulated through the digital signal-processing loop in the sample rate converter. Sample-rate converters are notoriously difficult to understand, and the wealth of operating options can be overwhelming.
A key "take away" here is that the specifications of the data converter may not be realized if the balance of the circuitry surrounding the chip is not properly implemented, and that includes chips that may only be crunching digital data. Now, the designer's conundrum is that in choosing a better DAC the other circuits in the complete system have to be good enough to justify the peak DAC performance, as specified in the data sheet. A converter with a worst-case performance of 15 bits is ten times less sensitive to performance degradation of the surrounding circuit than a DAC with 17.5 bit performance. Analog stages noise and distortion might have been low enough for a 15 effective number of bits DAC (ENOB) to dominate, but with a 17.5 bit DAC the analog section may have to be upgraded.
Although we would like to measure equipment of this kind in house here at The $ensible Sound, the pricey Audio Precision measuring device is the only instrument in the world that can test at the level of performance of today's best equipment. Audio Precision has had a market leading position for many years and no fully competitive product exists. Audio Precision can maintain its high prices, with no discount in sight. Other magazines have made this investment, but it is not on TSS' development agenda at this time.
In any case, we can still evaluate the Benchmark DAC 1 because they have an Audio Precision unit, and have documented their lab results. Assuming the legitimacy of the measurements, the signal-to-noise plus distortion ratio, at 1 kHz, shown in the Benchmark literature is 18 bits up to -4dB from full-scale. At full-scale the system operates at a bit better than the 17-bit equivalent level. These are typical figures. Benchmark guarantees nothing and, indeed, they are showing data that is below the limit of the worst-case performance of the Analog Devices DAC. They could be selective in their choice of chips, tossing ones that fail to operate to standard, but that could mean they are throwing out a lot of chips.
The Benchmark users manual shows the distortion of a 10 kHz signal with a very long cable to degrade the SPDIF signal (no, fellow enthusiasts, this is only a function of the cable's RLC characteristics, cable from Beldon works as well as something your audio dealer would love to sell you). No significant distortion products are seen other than the second harmonic distortion, which is not related to jitter. Jitter would cause sidebands to be present around 10 kHz. The sidebands around the signal result from the signal being modulated by the clock timing variations driving the DAC. Benchmark provides an AP plot of total harmonic distortion as a function of clock jitter levels demonstrate that THD is invariant to jitter because the sample rate converter is rejecting the jitter in the incoming signal.
The Benchmark website also shows other measures of jitter performance that require measurements be made inside the box. In my opinion, these are meaningless to the end user. Only what comes out of the box counts. Benchmark does not report full-scale distortion above 1 kHz, though they do provide an Audio Precision graph at -3dB down, and that shows it is holding at 17.5 equivalent bits at 20 kHz. The overall curve is a bit better than what is shown in the ADI data sheet, indicating that ADI was being conservative in the data sheet or the parts are selected by Benchmark. Perhaps Benchmark designed a better board to support the chip. (Do not hold your breath on this one, given the resources that Analog Devices has in its application engineering department, and the fact that the designers of the chip interface with their application people to ensure optimal chip performance.)
From AP plots provided by Benchmark, frequency response measurements, including the frequency response of the analog reconstruction filter, show a very small roll-off starting at 5 kHz and ending with a -0.25 dB amplitude at 20 kHz. Benchmark even supplies data for five production line units to show variation in the analogs filters performance with component variations in the filter. Line hum, as shown in an AP graph, is suppressed to more than a million times relative to full-scale.
Benchmark's measurements demonstrate that the performance of the complete DAC 1 is dominated by the Analog Devices AD 1853. Unfortunately, you cannot take full advantage of the performance of the DAC 1. SACD and DVD-A digital data transports are limited to 16-bit, 48 kHz performance. We are patiently waiting for a digital link standard for these media, but a cost-effective solution is not in sight. It is comforting to know that the Benchmark DAC 1 exceeds the performance of what comes out of your CD player. However, to hear the SACD and DVD-A media at the performance levels set by the best DACs you will need to find a DVD player with these DACs inside, which is why we are spending so much time describing what we are finding in our DVD player reviews.
If you have a 5.1 setup, the DAC 1 is a no-go, because it is a two-channel device. Even if you were willing to purchase three DAC 1s, no digital link exists to connect them to an AV receiver or processor. Please note that the digital link required here is different from the link need for getting SACD or DVD data across TO an AV receiver or processor. Rather, we want to get processed data FROM the AV receiver or processor. For all 5.1 formats, the AV receiver must decode the incoming data (lossy encoding as in Dolby Digital for a standard DVD or lossless encoding as used for a DVD-A), add the appropriate time delay to each speaker, perform bass management, equalize the channels, level adjust each channel and, if desired, add in sound-field synthesis. We would need a link after all this processing. That will come as custom installers want to transport 5.1 signal in digital form to multiple rooms, but it is yet to be a reality.
In summary, the DAC 1 is a cost-effective device when listening in two channels to CDs and there is an interest in exceeding the CD's inherent performance capabilities. This is so, even if the CD has been noise-shaped to present resolution at about 17 bits to about 4 kHz at the expense of lower resolution performance at higher frequencies.
The question of audibility need not be addressed here. If it measures better, then it is better. Is it worth the money over a lower-priced (say $200) DVD player for what may be a slight incremental improvement in sound or none at all? I cannot answer that, but I do aver that it is not sensible to spend any more money for an outboard converter than the price of the DAC 1.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||digital to analog converters|
|Date:||Jul 1, 2005|
|Previous Article:||Surveying the soundscape.|
|Next Article:||A new kind of woofer.|