SACD vs. Blu-Ray for audio quality/sonics

Some history from Philips research: https://www.philips.com/a-w/research/technologies/cd/technology.html

@rAndy-11 

Yes avoiding brickwall filters helps with passband ripple - this has been the long standing argument for upsampling. However randomization of DAC non-linearity is not well understood. Basically the higher frequency noise in high sample rates act like another form of dither that remove the quantization level errors of the DAC itself. 

Of course, the latest chips use other techniques too. The ESS chip in the Benchmark DAC 3 randomly selects which 1 bit sigma delta is used to convert (128 to choose from). This randomizes non-linearity. This is why I believe the latest devices sound like the best analog and why many like DSD or upsampling. or high sample rate files over redbook. The jitter monster was slayed about 15 years ago but only recently have designs addressed the inherent non-linearities of the levels in the DAC itself.

do you know how  ESS chip in the Oppo 205 is implemented?

@randy-11 

I think the Oppo has two of the 9038 chip - one used for multichannel and the other for stereo.  The noise reduction through random selection of sigma delta 1 bit from a multitude of 1 bit dacs is the standard logic used on this chip.

@randy-11

There is only so much to configure with the ESS chip. The 8 channels can be used individually or all summed or split into two groups of summed 4 channels. The summation reduces the channel count and increase the implementation cost but improves SNR. So 4 ESS 9028Pro chips may be the equivalent of 1 ESS 9038)

There are 7 filters built into the chip - most designers will select one. In all there only about a dozen options. (some of the subtle differences between DAC will be down to the filter choice)

Apart from the above, the analog and power supply circuitry surrounding the chip will the main difference in audio quality between different implimentations.

In the case of Benchmark they have built propriety digital and jitter rejection circuitry in advance of the chip. They upsample to 250 GHz digitally and then their Ultravlock adjusts timing at the 4 picosend level simply by a register shift. They then downsample to 211KHz. The choice of 211KHz is strategic in that it means the sample rate is well above the 192 KHz needed. This allows them to select the flatest linear phase option on the chip and the clever trick is that it shifts the filter corner point to be well away from the audio band (211 KHz has a Nyquist of 107.5 and this means the filter corner is 11.5 Khz above the NyQuist for 192 and 59.5 away from 96 and 85.5 away from 44 Khz nyquist and therefore will have no effect on the in audio band). So Benchmark have gone to effort to remove any audibke filter effects. Other designers may choose to tailor the sound using the filters.

Two main kinds of filter: Linear phase and Minimum Phase (a third kind could be everything else in between). Linear maintains the phase relationship between ALL frequencies - this best preserves the timbre of instruments. Minimum Phase screws up the phase relationship and changes the timbre but it eliminates pre-ringing. Since music is all about the relationship between various frequencies then Linear Phase filtering will sound the most natural and if well designed the pre-ringing will not be audible. Minimum phase makes no sense unless you look at waveforms and dislike aesthetically the pre-ringing.

That Bob Stuart is pushing minimum phase basically discredits him and MQA as a gimmick. He is pushing transient response (an engineering concept) over musical timbre (what we actually hear or how our ears work)