Wishes Upcoming PSA DirectStream MKII

I may end up getting the new DS when it comes out (I already have one of my kidneys ready to be listed on eBay), but I currently have the Junior. And I’m having to send it back since I now have a second failed Bridge II card inside–the Ethernet connection quit working, just like the first one which was DOA when I bought it.

Since the Bridge II is on a removable card, I wonder if it would make much difference to the sound if I removed it. I still need to get it fixed (for future sale when I upgrade) but would consider giving it a try without the card in the meantime.

I still find it fascinating, that when you change several things in a given design, it doesn’t just get improved in certain characteristics, but also still has the same homogeneous sound…or how this sound at the end is “voiced into it” …

It is safe to remove it. But be very careful when you put it back in, there is no pin keying and it can easily end up misaligned. Also be sure to store it carefully. If you don’t have an electronics shipping bag (coated in, say nickel) you can take aluminum foil and wrap the bridge and press the pins thru the foil so they are all shorted together.

My experience is that technically best is (for most people) also sound quality best. I’m heading towards a well defined goal.

1 Like

Thanks Ted. I replaced the first card myself (my DS Jr. had a dead Bridge II when I first bought it) and agree it was a bit tricky. (Not like the old DIP ICs I used to work with back in the 80s!)

You make a really good point here, but also miss one or two other critical ones.

FPGAs aren’t by themselves a replacement for a DAC chip. (OK technically you probably could generate analog audio out of an FPGA but it probably wouldn’t sound any better than the cheapest of chip DACs.) The FPGAs are there to provide complex, high speed logical operations and in particular to perform sample rate conversion, digital low pass filtering, digital attenuation, and particularly in the case of Ted’s designs as well as this newest Teac it will also perform the modulation and noise shaping necessary to generate a 1-bit DSD output stream. The FPGA is a digital-to-digital converter in both the PS Audio and the Teac products. Yes they are competing in purely algorithmic terms to determine the better-sounding bits to output given the same PCM or DSD input, but they are also competing in terms of implementation expertise to minimise noise and jitter generated by the FPGA in the process. That noise and jitter doesn’t change the bits in the output stream but it does influence the result of the actual D-to-A conversion that follows.

The deeper innovation in Ted’s design relates to clocking, and based on what I see from the Teac layout I suspect he’ll keep his edge on this one. First there was the decision to optimise the clock not for long-term stability but for peak-to-peak consistency. Our ears can handle smooth frequency drift over a period of several seconds but we are insanely sensitive to frequency instability down at the microsecond-to-picosecond levels. Second is to have a single clock that’s an integer multiple of the output sample rate, placed immediately adjacent to the actual output switches, and from which every other clock in the system is derived – this really lowers the floor for the level of output jitter that can be achieved. And lastly I’ll mention the way that the DS DAC observes the incoming data rate with a degree of tolerance and gently adjusts its own master clock to match the average pace rather than locking to and being driven by the incoming signal’s clock and whatever jitter it contains. These design features contribute massively to the quality of the DS DAC’s output and I can tell from the dual mono circuit layout that Teac is not following particularly closely in Ted’s footsteps here.

Then we have the actual D-to-A conversion from 1’s and 0’s in a MHz-scale bitstream to a line level analogue audio waveform with sub-50kHz signal bandwidth. You can get a sense for how much this matters from the discussion of the DS Sr vs DS Jr vs TSS designs vs DS Sr mods with upgraded transformers vs the changes we’ve been told are coming in the DS Mk II. You need ultra-stable power supplies. You need switching devices with sufficient speed and consistency. You might have two, four, eight of them per phase per channel to increase SNR. Then there’s the analog low-pass-filter circuit, which is where the transformers come in on the DS Sr and above. If your switches and filter are too low level you might have to then run through some kind of line driver (which is what the Jr and the Teac do).

Although they both may be using FPGA to generate a DSD bitstream and then doing a direct analog conversion with some kind of low-pass filtering, the Teac product and the Ted Smith products are focused on very different priorities and a direct price/feature comparison is almost meaningless. Just about every “feature” on the Teac spec sheet degrades the ultimate sound quality that comes out of the device. “Dual mono” has some benefit in terms of power supply but it moves the D-to-A conversion some distance from the clock source and requires all these point-to-point wires that will radiate noise throughout the chassis. Bringing all that Ethernet connectivity and CPU-based sophistication inside the box also means more noise.

The TSS DAC is an all-out effort on sound quality where the most important and expensive changes are what happens downstream from the FPGA. Put the master clock, the output switches and the analog low-pass filter in a completely separate box with uncompromising power supply design, and keep all the noisy digital pieces several feet or even tens of meters away at the other end of a non-conductive optical fibre. Yes the FPGA will be more capable and we’ll get some benefit from that, but it’s far from the most significant change.

The DirectStream MkII is halfway between DS Sr and TSS. More optimised power supplies, better output transformers (yes @tedsmith I’m pretending this is a done deal), more isolation from noise borne on digital audio connectors, the eviction of noisy ethernet things, a quieter and better isolated display and front panel controller, the ability to actually power down unused inputs like USB, switchable ground lifts all over the place. Yes it also gets the newer FPGA and will benefit from Ted’s experience in lowering the noise and jitter generated within the FPGA itself, but compared to everything else… that’s the least interesting thing about it.

The Teac product is interesting, and will probably make many customers very happy, but those are very different customers from the ones that PS Audio is targeting here. Just because it has an FPGA doesn’t mean it’s even close to being in the same category.

3 Likes

Yeah I don’t claim to understand the actual algorithm (haven’t followed Ted’s other threads and he wouldn’t disclose the secret sauce details anyway), and if the Teac really does sell for $3500 (less than DSJ did) it wouldn’t be fair to expect it to compete.

FWIW, the Teacs can accept an external 10 MHz clk which further improves performance - I use one with the nt505. Teacs goes for about 1300. So in total it would be a 4800 (3500+1300) solution.

None of these proprietary conversion scheme companies publish their conversion algorithms for obvious reasons, so the proof will be in the pudding (listening)

I just realized that a bunch of discussion on other fpga dacs recently was actually in another thread, not this one, so probably I wasn’t providing full context with comments above and maybe seemed a bit out of left field. Oh well

What do you think that external clock is buying you??

1 Like

A 10MHz input to the DS would lower performance. Using a 10MHz clock to other units (or not) in your system and letting the DS do what it does is definitely lower jitter.

[Edit: speed-racer beat me.]

2 Likes

Again, I don’t know the algorithms. DCS and Chord use external clocks to improve performance. Is the suggestion that they and others who do are out to lunch?

No, external clocks are needed in pro systems no matter what they do to sonics. For personal playback systems that aren’t ever involved in mastering you can do better without them: tracking multiple clocks, e.g. the 10Mhz as well as any inputs is always worse than tracking one. To really use a 10MHz clock well every digital thing in your system needs to be using it so only one clock is needed. I’ve simply chosen different things to optimize than MSB, dCS, etc.

2 Likes

All good. I didn’t suggest anything was lacking with any dac that didn’t use an external clock, or dispute any of the contentions made in the post I responded to

You just suggested that a 10MHz clock improves performance with no qualifiers, that’s false. It can help, but for many DACs it makes things worse if the rest of the system isn’t driven off of that same 10MHz clock and further, converting clocks rates to the one really needed instead of 10MHz is jitter inducing. It certainly can help a DAC but the DAC could be even better without it.

2 Likes

Ted: Do you have detailed access to the proprietary algorithms that DCS, Chord, MSB, Teac delta sigma will use?

Do you have their fpga code? I don’t

No, but physics puts a limit on possibilities. Nothing I said about clocks depends on proprietary algorithms, just physics.

If I’m being too defensive and you aren’t trying to poke me: MSB and dCS earlier marketing materials implied worse algorithms related to sound quality than I hope they are using now. I do know about the limits of upsampling and sigma delta modulation and once again I’m optimizing different things than they are. I wouldn’t be surprised if I could learn some things from them, but the other way is probably true as well. This is all just engineering and isn’t rocket science. It’s just that different people are solving slightly different problems constrained by their history, marketing, price and cost choices, etc.

6 Likes

Fair enough. That’s all I’m saying also. Lots of different ways to do conversions as evidenced by more and more dipping their toes in the non-chip pool. Time will tell and competition is a good thing in most areas of life. Good time for digital audio. As always, the market will decide.

Kinda reminds me of the many dual suspension designs in mountain biking 15-20 years ago

1 Like

That’s innovative! Could be a very convenient feature. :wink:

It’s not my idea. I forgot where I heard about it, but it seems like a convenient idea for those who don’t want to hook up a network to the DAC just for upgrades.

1 Like

I remember that type of update on Audiobyte products. Had black dragon in the past and currently hydra converter I updated same way.

Not sure if this has been discussed before or not. What about adding a filter to the DAC that can be applied to help ease the pain on poorly recorded material? A feature like that would would go a long way in persuading me to upgrade from DSD mkI.