|
Computer Audio Asylum Music servers and other computer based digital audio technologies. |
|
|
|
Computer Audio Asylum Music servers and other computer based digital audio technologies. |
|
In Reply to: Tony's Player posted by John Swenson on July 13, 2012 at 00:15:47:
Comments re DSD
My reservations with DSD concerned the noise floor, specifically the rising high frequency noise and the general level of noise. Even more important than the general level of noise is the effect of noise modulation. With multi-bit signalling dithering can be performed to remove noise modulation. (Rectangular dither corrects the mean, triangular dither corrects the next higher moment. Or subtractive dither might be used to eliminate all possible signal related modulation.) With one bit signalling dither can not eliminate noise modulation, as the total energy in the signal is constant and louder than the music. Rather, all that can be done is to filter out as much of the noise as possible.
Obviously, if the noise is 190 dB down and modulates with the music plus or minus 6 dB the modulation will be irrelevant. But the problem is that the noise is not down that much. Indeed, the noise in the "audio band" below 20 KHz is not much better than -120 dB (as claimed by Sony and Philips). If you look at sample sigma delta modulators, you will see that these produce peak signals that are not even 16 bit accurate. I did this by running one of these and just measured a DC signal, one can clearly see how the noise level (peak or RMS) vary with the DC input, looking only at the output after filtering with a 64K sinc filter with 22.05 kHz transition. So I still don't care for DS modulators. However, when run at 5.6 MHz all of these effects are substantially reduced.
More important, I have run listening tests with my Mytek Stereo192-DSD DAC and compared upsampling in the computer with sending straight PCM to the DAC (which uses the SABRE chip). The sound, in increasing order of quality is as follows
1. 44/16 file, sent at 44/16 to DAC
2. 44/16 file, converted to 176/24 and sent to DAC
3. 44/16 file, converted to DSD128 and sent to DAC.
Or with high res,
1. 192/24 file, sent at 192/24 to DAC
2. 192/24 file, converted to DSD128 and sent to DAC.
These are not claimed to be scientific tests as I have not done level matching other than manual volume control adjustments which come in 1 dB steps. The gain through the DAC in DSD mode is not the same as with PCM, as it depends on the level of modulation in the SDM. I use HQPlayer to do all the sample rate conversions as well as the sigma delta modulation.
It is also possible to upsample DSD64 files to DSD128 before sending it to the DAC, but I didn't find any obvious benefit from doing so. I may do this later, should the software get the capability to do room convolution at DSD64 rate, which would add low order information that might benefit from the higher speed.
Comments re Memory player and Shift Register DAC
The main reason for suggesting a 1 bit DAC was actually with a shift register implementation, not a memory player. The hope was to build a 1 bit DAC that provided complete isolation through a well engineered shift register, operating as a single clock domain. There would be no LSI chips, just MSI or discretes. The interface would consist of three pairs, a clock pair going from the DAC to the transport, and a left and right pair sending one bit pulse streams. The interconnect cable would be required to have a specific length, and the timing arranged at the transport to ensure that the phase of the clock signal at the DAC matched the data stream. As a result, the entire DAC would consist of a single clock domain. Not many flip flops or gates are required in this design, so the designer would have no excuse if the resulting device didn't isolate from the input and/or didn't produce good sound. Note, in this design every single transistor in every single gate and every single wire, etc. must be modeled as analog components. Getting the shift register to attenuate all legal variations on the input stream (within the receiver's eye pattern spec) to where they are at least 160 dB below peak output in the output will be a bit of a challenge. One does not want any extra components to make this more difficult than possible. Presumably a simple computer interface could be built to interface the three twisted pairs.
IMO if this can be done it would be a much more practical device than the memory player we've been discussing, because of latency and other user interface issues. It will be a much harder sell to skeptical audiophiles than the memory player, of course. I would actually want to have both of these around, since the memory player will make it easy to test how well the player and amplifiers are affected by running but disconnected computer systems and this may be a good tool for testing power supply isolation, cable isolation, etc. It might be a good idea to put the memory player function in a separate box at the end of the twisted pair to allow both methods of operation.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
This post is made possible by the generous support of people like you and our sponsors:
Follow Ups
- RE: Tony's Player - Tony Lauck 08:10:40 07/13/12 (0)
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2025, Audio Asylum, all rights reserved ]
