I am confused by some of the things that I have heard about the Super DAC. One Big thing is Two chassis one analog and one digital. I find that very interesting as in all actuality there is no Analog stage in the Direcstream DAC. What it outputs is the result of filtering the digital wave produced by the video switch and presenting that to the preamp or amp directly. This is why I personally believe the Directstream DAC sounds so great. I just saw a video where Paul Mentions the use of a Fiber Optic connection between the two chassis. If that is the case the signal leaving the first chassis must be digital so that brings the conversion back to the second analog chassis. None this really matters to me as $6000 is the uppermost limit I would ever spend on a DAC talks of 15k, 20k or even 30k if way out beyond me. High end audio has gone insane!
I would suggest you read the two rather long topics which cover the TSS:
There are multiple things that need to be done in the FPGA:
- Input decoding
- PCM Upsampling (to 176.4/192, or 352.8/384 depending on the software release
- Misc: calculate sample rate, sample width, do deemphasis if needed, detect DSD, etc.
- Upsampling to 28.224MHz or 56.448MHz (and then back to 5.6448MHz or11.2896MHz) depending on the software release
- Sigma delta modulating to one bit
Outside the FPGA there are physical connectors, power supplies, the clocking and digital switches that convert bits to excursions from one power supply rail to the other at the right times and the low pass filtering to convert that to analog.
The FPGA is all properly digital, as are the input connectors, etc. and the power supplies for them.
I consider the clock, reclocking, digital switches, output filtering (and a few other things like the mute relay) and their power supplies to be “analog.”
Indeed inside the DS that’s exactly how the two boards are separated and in the TSS those two boards will be in separate boxes. The interface between the digital board and the analog board in the DS is a physical ribbon cable that connects the boards. In the TSS this will be an optical cable. But neither is more digital or analog than the other.
In the DS and TSS the analog board generates the clock and the clock is right next to the reclocker. There’s only one clock and everything on the analog board is synchronous with that clock, so filtering the clock signal out is very easy and there’s no beating between that clock and other digital things to make audio interference.
In the digital board (box) there’s a lot of hash, all of the various inputs wiggling at their own rates, multiple clocks used inside and outside the FPGA to support the digital functions, etc. Those various data streams and clocks can beat with the master clock and generate audio bandwidth noise/tones/chirps/birdies, etc. Keeping that stuff and (secondarily actually) the digital supplies separate from the analog and the analog supplies in the DS is much harder than it will be in the TSS by having two separate boxes.
Thank you Ted for clarifying that. The output chassis will be the part actually responsible for conversion from Digital to Analog. I can understand your wanting to keep the interference form the earlier stages away from the final output. I knew that you were not really following the video switch with an active analog stage. I would be concerned about the clock being located at a distance from the first chassis. I am sure you know what you want to achieve and how to get there.
The clock being a long way from the digital board/box is good, not bad: the only place the quality of the clock matters for audio is exactly at the point where there is no clock down stream: i.e… where bits become audio. So the final reclocking just before the output filter is where the master clock needs to be.
A system is better off with a single master clock driving everything (even tho other clocks may be derived from it.) Like the DS and DS Jr In the TSS the clock will be sent from the analog board/box to the digital board/box, in the digital board/box it “picks up the data” and returns to the analog board/box so that you never get a sample too early or too late. To do this in the TSS in the analog box there’s a phase alignment that compares edges of the incoming clock (from the digital box) to the edges of the master clock and shifts the phase of the clock from the analog box to the digital box to make sure things get back just in time. The optical connection can be any length whatsoever without causing clock timing issues.
Great stuff, Ted . Interesting and informative but concise. You always make it understandable.
I have a question about the use of fiber optic connections between the two chassis. I worked in telecommunications for 35 years. I am no audio engineer but i do remember that the conversion of electrical signals to optical and than back to electrical is a possible source for the introduction of jitter. Would that not be contraindicated in a DAC where you are doing everything possible to prevent jitter?
Ted, if the separation of the two boards is of importance, is there any aftermarket tweak DS owners could apply by their own, that better shields the one from the other? Mu metal etc.?
I’m carefully using ECL interfaces for the optical transmitter and receiver to lower jitter. ECL is a non-saturation switching technology as opposed to CMOS, etc. Coming out of saturation is a process that adds jitter to most digital logic. In a handwaving sense you can think of ECL as “class A” digital (linear, but always conducting and always power hungry) as opposed to class D digital which is lower power (only uses power when things change) but noisier.
But jitter isn’t nearly as important for the incoming optical on the analog box because there are multiple levels of jitter reduction (reclocking) after that. Deserializing the data is a reclocking operation (the edges of the parallel output data are controlled by an accurate local clock. After that there are two more ECL based reclockers, each with their own power regulators and clock drivers. The final reclocker is the one right next to the master clock to get the lowest clock jitter.
Mu metal shields magnetically, which isn’t really relevant for the digital/analog separation, but it could help shield the output audio transformers from the AC transformer, etc. On the other hand our experience is that any shielding (mu metal or Faraday) near the audio transformers hurts sound quality more than it helps (it really mutes the highs.) The digital and analog boards already use the external copper and via stitching at the edges to lower electrical field emissions and reception, but there’s nothing that prevents more electrical shielding between the boards, but most shielding uses metal and that metal needs to be carefully insulated to avoid accidents later, say, as the box is being moved or jostled. The electrical connection (ribbon cable) between the digital and analog boards is probably the worst link between the boards, it ties their grounds together, but otherwise the signals are all ECL, balanced and capacitor isolated to help lower EM radiation and EM reception. Going to optical in the DS aftermarket isn’t a good idea: the timing of the signals over the interface assume a particular time delay. Probably the 2nd biggest interaction of the digital and analog boards is thru the top level power supply. Having a better regulated 12V for the analog supply (but sharing the same ground as the rest of the system) is probably the best upgrade for isolating the digital and analog cards.
Thanks Ted, so the only meaningful measure seems not recommendable to someone without an electronic engineering degree, so I’m out
I think you might find a few posts from people asking about shielding in the DS and/or DS Jr - I don’t know if they reported back their success or failure.
Are you planning to use the same Crystek clock as the DS in the TSS DAC or have you found something newer and better since the original DS design?
Also, the last time I asked if you could consider a headphone output, I don’t think you said no but you did explain the things you would have to consider.
Any updated plans there regarding a headphone output or is that now off the roadmap?
There are more really good clocks out there, but none that meet my needs better than the one I’m using. If I used asynchronous sample rate conversion I could use a fixed clock that was a little better. I don’t think one would hear the difference that that better clock made, but I’m sure most people wouldn’t like the difference using ASRC made.
Given the architecture of the DS/TSS a headphone output couldn’t be of the same quality as the regular outputs even with a lot more cost, I could get something with similar quality as the DS Jr output for a headphone output, but most customers with harder to drive headphones would be better off with a quality headphone amp and the current architecture.
Thanks Ted, understood.
Are you still using the Oppo HA-1 for headphones use?
Yep, I’m not likely to change it because the headphones are essentially a backup system to check whether a newly found problem is in the DS or my system (if I can’t figure that out with a scope.) I’m glad the local dealer recommended it, I probably would have overlooked it otherwise.
Maybe already discussed in the forum and I missed it. The TSS will be amazing.
As a DS dac owner I am super happy, cannot imagine what more the TSS can do. But
my question is will there be a new reference Transport down the road. A $6k transport
with a $20k+ Dac is a bit of a mismatch. It’s like me holding Kate Beckinsale’s hand
walking down red carpet.
I don’t fall for the High price means better performance thing. The DS DAC and DMP beat many much more expensive offerings. Yes I am sure Ted’s TSS will be better but not a step up equivalent to the cost difference.
Actually, a new transport is in design. Not so much to match the TSS, mostly because the OPPO drive in the DMP is no longer available. Paul has said it will sound better than the DMP, but will not play SACD like the DMP. As always, PS Audio is learning things as they develop each new product and will put all their best effort in the new disc transport to make it a top line performer. Of particular note, Paul has said they have a new design Digital Lens that sounds much better than the old one used in the DMP as well as other products. Just don’t expect the new transport any time soon, lots on the table at PS Audio, most notably the TSS and new speakers!
Don’t sell Kate Beckinsale short. She would do fine as your escort.
Oh I would not complain at all.