Let the Games Begin


I connected an older jukebox style Sony CDP-CX200 to my higher end stereo. I ran an optical cable through a two way optical switch to a Schiit Bifrost Uber (about 7 years old now) into my Musical Fidelity integrated amp. Surprisingly, it sounded very good?????

I compared it to my Jay’s Audio transport going into a Denafrips Pontus 12 anniversary DAC.

A very interesting video, it's educational for sure.....

Hearing is so weird. When I was working in the studios, I had the chance to work with one of the first Sony digital multitrack machines. I think it was 1” tape. We were all amazed that the playback off the tape was identical to the recording sound and we all pronounced it to be “better “. Later down the road many of us decided we liked the sound of analog multitrack with its added noise and compression. Go figure .

There are a few key things left out of this analysis. But I should preface this "critique" by stating I'm a fan of digital music and no longer own a turntable or other purely analog source.

This analysis assumes that the analog source signal has no content above 20Khz. In reality, the signal will have many harmonics well above 20Khz, which have to be filtered out to prevent aliasing artifacts. This is not easy to do. Higher sample rates make it easier, but still not trivial. 

A practical high-order low-pass filter will still introduce phase shifts and frequency response aberrations, and require a fair amount of circuitry (usually done with op-amps) which introduce their own distortions. 

Once the signal is digitized at a high sampling rate, it's fairly straightforward to digitally filter the data and down-sample back to Redbook, but this will no longer be an "exact" copy of the original input signal. 

16 bit samples are probably adequate to handle the dynamic range of the source content (particularly with all the compression that is often used), but this assumes that the full 16 bit dynamic range is utilized. In many cases the input signal is captured at well under it's maximum signal level since it is virtually impossible for the recording engineer to anticipate the maximum signal level that the musician will create. 

I've talked to a number of recording engineers who say they try to set the levels based on a sound check, and then the musicians play much louder during the recording session resulting in clipping. And, of course, musicians do not like it when the recording engineer tells them they have to do it again because of a mistake they blame on the engineer. As a result, most content is captured significantly below the maximum digital signal level. 

If the content is captured with 24 bit samples to start with (best case is probably closer to 20 bits of real signal-to-noise), then the recording and mastering engineers have more room to adjust levels and achieve closer to the Redbook dynamic range potential. 

If everything in this process is done optimally, I think it's possible to convert to Redbook and achieve stellar results. With well-recorded and mastered content, Redbook audio can sound fabulous. 

Since everything is rarely done optimally, having a little headroom in both sample rate and bit depth can be helpful. 

Converting the digital back to analog can also create artifacts. If the output is not bandwidth limited, the conversion process will include high-frequency "mirrors" of the original signal. Many listeners find these artifacts fatiguing. 

Most DAC manufacturers have resorted to oversampling the digital data and implementing these filters mostly in the digital domain. 

So we end up sampling the input waveform at several multiples of the Redbook frequency, then downsampling to 44.1Khz to store the data. Then on playback, we end up upsampling the data back to a multiple of redbook so we can implement the reconstruction filter largely in the digital domain. 

It seems logical that avoiding the down-sampling and up-sampling steps would result in a purer reconstruction of the original content. 

Some of these replies are technically over my head, so I will give a simpleton's explanation of what I am hearing in my latest system which I have been running for about 6 months now.


Jay's Audio CDT3 Mk3 CD Transport

Holo Audio May Level 3 Kitsune tuned DAC

Primaluna Evo 400 Preamp

Primaluna Evo 400 Power Amp

Klipsch Cornwall 4 Speakers

Rel S812 Woofers (2)

Most reviewers agree that the Holo May sounds best in NOS mode (though it will upsample quite high if desired,) so I leave the dac on the NOS only setting. However the Jay's Audio transport has a toggle switch that allows for 4x upsampling (176.4) processed by the Transport, and I am finding that to be an intriguing setting with certain music and maybe with all music. The Holo May immediately locks on to the 176.4 sample being performed by the CD transport but is itself not doing any upsampling.

My system became more accurate and discriminating after I attached some home made signal "grounding boxes" to the CD Transport, DAC and REL's and built a heavier and more rigid component rack. All of these tweaks have brought the system to a higher level at which point I could switch back and forth between Redbook and 176.4 and hear more of a difference between the two. However, the 176.4 definitely sounded smoother with slightly better instrument separation though accuracy of timber and tone may be somewhat distorted. The mids - especially vocals - are more holographic and affecting at 176.4.

I.E., I am apparently wishing for a blend or marriage of these two modes, but wondering if it is possible and how I might go about it.