Tino,
yes, I have compared the Sony with the Magic DAC II using the Sony as a source for both- the Magic DAC in minimum phase mode definitively adds transparency and a sense that "the timing is right" at high frequency. As soon as the DAC Magic is switched back to "half band filter" (Filter #1), the effect at high frequency is back, leading me to believe that what I am hearing on the Sony is the 44.1K brickwall. However, you are describing a "lack" of high frequency that I have not noticed; this could be attributed to many effects, one of them would be a noticeable decrease of distortion compared to your previous player? For most of us, distortion = loudness (at least frank, top-bottom clipping). You *could* borrow a DAC Magic from one of your favorite stores for comparison purposes. I also noticed that you used an Acoustic Zen loudspeaker which sports (if my memory serves me well) a... ribbon tweeter. As much as I like the delicacy of those devices, I also usually think they are lacking slam in the lower part of their spectra (3~6kHz) and will sometimes overemphasize the 10K region, especially since the midrange of the Zen is quite a large driver.
Now, could Sony update their digital filters? Well, that's a tricky one- I do not know what processing abilities are left in the DSD ASIC, but this is how I would do it:
- Migrate the first two stages of digital filtering to the Sony ASIC; this stage would perform the critical 44.1K -> 176.4K upsampling. I would use minimum phase filtering there (many good papers to chose from in order to design the filter).
- Set the DAC for a 176.4K input, and use the "slow slope" filter as the upper region of the spectra will already be empty (20~80kHz).
There would still be pre-echos in the output stream as some filters are still half-band FIRs, but the longest filters (x2, then x4 have the largest number of taps) would not have pre-echos, thus limiting the duration of this effect in time. Now, does Sony read this column? I am ready for a flash update CD:)
Regarding modding the player- most dual opamps have the same pinout, so you should not have any issue trying some new friends. Keep in mind that opamps used in various locations need a different subset of optimizations, in particular:
- I/V converter needs fast settling time and slew rate, low current noise, and stability when a capacitor is connected from inverting input to output. This means that current feedback opamps are not recommended in this location. Sony selected a FET input device (OPA2132) for this application, also bringing very good DC characteristics and possibly removing the need for AC coupling capacitors between this stage and the LPF/Buffer stage. I would look at high performance, high slew rate opamps from ADI and National here. Please keep in mind that the noise of this circuit will be largely defined by the value of the feedback resistor (the lower the better), so output drive of the opamp also matters.
- Filter / buffer: low voltage noise, high slew rate and high output drive are a must to keep the performance up. I do not know the exact topology of the Sony's output stage, but I would also consider keeping good DC characteristics in order to minimize offset. A new, fully differential audio opamp was just released by National- it may just be the right device for this player, as nothing indicates that they are using an "auto-balancing" ciruit on the XLRs... This however would require the addition of a small PCB as it has a non-conventional pinout.
A side note regarding the NE5532 / 5534: they definitively bring a signature to the signal- signature that we have been hearing for over 25 years. Aren't we ready for something a bit more transparent? :)
Another side note regarding Tube output stages: properly designed SRPP stages will likely be the only topology that will keep the inherent noise floor and low distortion of this player, while adding the "glow" so many people seem to love. I would shy away from SET circuits as there is almost no voltage gain needed (the I/V brings the signal to the proper amplitude, not the output filter/buffer). The supply for a SRPP stage will likely be high (250 ~ 300V) and will make the integration in the player difficult. Another valid approach would be to use a discrete transistor based output stage, which would not require a very high supply...
