|
|
发表于 2010-5-25 18:41:35
|
显示全部楼层
Only recently this went up as a reply to my query about what changes have been made to the DAC1. It is buried below (and so could easily be missed) but thought it to be of interest to many so I've taken the liberty of copying and pasting it here.
An inside look at the Benchmark DAC1 - from the designer:
All versions of the DAC1 use the following IC's:
AES Receiver:
AK4114 (AKM) - selected for its ability to accurately recover data in the presence of high-jitter, high-noise, and when signal levels are very low. The AK4114 is unsurpassed for data recovery. However, the clock recovery on this IC is very poor due to the wide bandwidth of it's PLL. The DAC1 does not use the AK4114 for clock recovery. The DAC1 only uses the AK4114 for input multiplexing, data recovery, and digital de-emphasis (for those rare CD's that actually use pre-emphasis).
Up sampler (ASRC):
AD1896 (Analog Devices) - selected for its low spurious tones, low distortion, and exceptionally low PLL corner frequency. Pin-compatible versions of this part exist, but should NEVER be substituted into a DAC1. The substitute parts have much higher PLL corner frequencies, and will destroy the DAC1's exceptional jitter attenuation.
D/A Converter:
AD1853 (Analog Devices) - selected for its exceptionally low distortion. In the DAC1, the AD1853 operates from a fixed-frequency ultra-low-jitter crystal oscillator that is well isolated from external clocks. The digital filters in the AD1853 are frequency shifted in the DAC1 to minimize passband ripple, and eliminate imaging and aliasing of near-Nyquist high-frequency audio content. The high-frequency transparency of the DAC1 is unsurpassed by anything we have measured. Most D/A converters suffer from aliasing, imaging, harmonic distortion (THD), intermodulation distortion (IMD), and jitter-induced distortion. We have gone through great pains to minimize these common defects.
While no audio device is perfect, it is possible to reduce the amplitude of error signal to inaudible levels: In the DAC1, jitter-induced sidebands never exceed -141 dBFS, spurious tones never exceed -126 dBFS, line-related hum never exceeds -126 dBFS, noise never exceeds -114 dBFS, and THD+N never exceeds -105 dBFS. This means that at a playback level of 105 dB SPL, the THD+N artifacts are at the threshold of hearing (if the room were silent), and the jitter induced sidebands are 36 dB BELOW the threshold of hearing (in a silent room). By any stretch of the imagination, jitter-induced sidebands are inaudible with the DAC1. So why do different transports sound different with the DAC1?
Transport issues: Some transports are not bit-transparent. Some DVD players have a digital volume control on the digital output. Some apply sample rate conversion to the digital outputs, and this conversion may be of a poor-quality. Also, most CD and DVD players are designed to mask or skip data when the disk cannot be read, some will attempt to re-read the data, but ultimately will skip or mask if subsequent reads fail. A dirty or scratched disk will play differently in different transports. A transport with dirty or misaligned optics will produce similar problems. Bottom line, the bits at the SPDIF or TOSLINK connector may not be identical to the bits on the recorded media. We have done some testing of transports and expect to post the results on our web site when we have more data. Some transports work very well, others do not. Unfortunately, we are finding that price is not a good indicator of transport quality.
A further word of caution when comparing digital interfaces with the Benchmark DAC1: If you connect a single transport with two different types of digital outputs to the DAC1 (SPDIF and TOSLINK for example), you WILL hear a difference between the two inputs when you flip the input selector switch on the front of the DAC1. The difference is real but the difference only persists for 500 mSec. Here is what you are hearing: When the input to the DAC1 switches, two soft-mute circuits activate to prevent pops and clicks. The mute/un-mute sequence ends in less than 100mSec (typically 60 mSec). During the mute sequence and for another 400mSec, the DAC1 UltraLock? circuitry is disabled and the audio is pitch shifted to adjust the internal buffer lengths. The pitch shifting is completed within 500 mSec after switching inputs. The red "ERROR" LED on the front of the DAC1 is illuminated while the mute/un-mute, and pitch-shifting operations are taking place. The pitch shifting sequence is either normal-up-normal or normal-down-normal. The first half of the sequence occurs during mute, but the second half is not muted. Consequently you will hear normal audio prior to the switch, followed by a 60-100 mSec mute, followed by new input pitch shifted, followed by a 400 mS adjustment back to normal pitch. The small but abrupt change in pitch after the mute is just enough to give the impression that one input is brighter than the other. The direction of the pitch shifting sequence is determined by the relative delay between the two digital inputs. The input with the largest delay will be perceived as brighter. If the optical interface has more delay than the coaxial, it will be perceived as being brighter. After 500msec, there is no difference. Kind of interesting!
Why so many OP-Amps in the DAC1? The answer is that the DAC1 uses parallel signal paths to keep the total number of OP-Amps to a minimum in any given path. The RCA and XLR outputs have identical but separate signal paths. The front panel gain and rear panel gain are separate but identical signal paths. This topology keeps the outputs well isolated an keeps each signal path to a minimum. Cost? extra Op-Amps.
Regarding regulator layout: The regulators are positioned at the noisy side of the box, so that only clean DC is distributed to the rest of the board using a star topology.
Regarding trace lengths: Obviously signals must traverse the box from the front-panel control to the rear-panel output jacks. Such a traversal must be made with care. All long analog traces are very low impedance, and are fully shielded in 3 dimensions by surrounding the trace on 4 sides with a shield. All of the shield components are "stitched" together at regular intervals using vias. This construction is very unusual, but is very effective. Check the crosstalk specifications if you doubt the effectiveness.
Regarding grounding: The transformer secondaries, rectifiers and bulk capacitors share a single ground point and do not introduce ripple currents into the ground plane. Line-related hum is -126 dBFS or better.
John Siau
Director of Engineering
Benchmark Media Systems, Inc.
Do not criticise the idiots in this world - we need them as they make the rest of us look so much better :-) |
|
|联系我们|有害信息举报：010-60152166 邮箱：zx@jd-bbs.com|手机版|Archiver|黑名单|中国耳机爱好者俱乐部
( 京ICP备09075138号 )
