Wishes Upcoming PSA DirectStream MKII

Only that I don’t know how much more my enjoyment will be when I get a chance to listen to music from the PST to my future DS MKII. Can hardly wait.


Amen, Brother!

1 Like

As are most things, when qualifying as “competitive best”. IME digital is enjoyable and convenient, vinyl just sounds better. I neither case did I have to spend ridiculous amounts of money to satisfy my listening taste. Is there better, certainly. I must add the opportunity to stream has been a game changer, not sonically but as an expanded opportunity to explore new music. :wink:


Well said, it comes down to pursuing good music followed by good sound. A matter of priorities in my case. Formats serve different needs, and some music is unavailable in various formats. I must say I still don’t miss cassette tapes, reel to reel maybe slightly.


Hi @Ted: is the bill of materials now steady for the DS MK II? Or do you expect small changes in components like regulators, opamps, cables or connectors in the coming months?

So far there isn’t any need for changes. We haven’t thoroughly tested every piece of hardware yet, but so far things are working as designed.


My money for the MKII is in my savings account; can’t wait to experience the effect of the – 20DB to -30 DB noise floor…


Ted do you have a prototype that you have been able actually listen to? If so any comments on what you hear so far?

1 Like

No listening yet. It’s been at the top of my priorities for quite a while, but other things keep happening. On the other hand the new tools I’ve written, along with the new FPGA, etc. give me much better visibility into the digital side of things and I’m quite pleased there.


For example this is the digital input and output of the sigma delta modulator in the FPFA with full scale 8820Hz Sine wave input but the volume control set to attenuate by -120dB (roughly the analog noise floor (-120dBFS would be -140 on the DS’s volume control).

Tho the sigma delta modulator noise floor is increasing as expected it’s still below -198dB at it’s peak in the near audio bandwidth. FFTs of noise show the noise at a lower level than they show the signal. The point of this test is that there aren’t spikes, birdies, idle tones, etc that creep out of the SDM noise and get close to the analog noise floor even with low level input signals. (Tho I’ll be working on the bump at about 1.8k @ -240dB it’s not anywhere near significant.)


The input signal, is that 20x64x44.1kHz @ ~56 bits?
And the SDM is 4x64x44.1kHz @ 1 bit?

This looks… astonishing. It seems to imply a post-SDM noise floor in the audio passband at somewhere around -160dB down to -330dB, in the abstract information domain at least. Have you got an idea of what the analog noise floor will look like at the summed output of the digital switches prior to the analog filter, if you fed the converter with this same SDM bit stream?

I’m not expecting anything like those figures in a physical system of course. I’m curious about what the limits are, and the reasons why. Clock stability, DC supply stability, and internal switch noise are the three factors that first come to mind.

FFTs are great for lowering the apparent noise so that you can see, say IM distortion, jitter or other nonlinear distortion (for which sigma delta modulators are famous.)

Any white noise power is spread evenly over all of the frequency bins of an FFT while periodic signals have their actual power in the expected bins that cover them. So you can lower the apparent level of white noise as much as you want by using more samples in the FFT. For this work I’m using ten times oversampling of one second of audio: 112.896 million points. Each doubling of the number of points lowers white noise by 3dB so this FFT is lowering noise by about 136dB so I can see any crud that might affect the audio near -120dB… The SACD ultrasonic noise limits are -40dBFS so I’m comfortably clear of that.

Right now I’m receiving 30 bits from the upsamplers, I add six more bits on the bottom and I’ve got the volume turned down 120dB so the noise is down there around 56 bits. I.e. the volume control doesn’t add noise.


Oops I forgot part of your questions.

In the DS Jr the noise in the voltage reference for the regulators was at about -144dB (of 2.5V). The regulators themselves add noise, but quiet rails for the analog help a lot with overall analog output noise. The noise of the jitter from the VCXO and reclocker is way below -120dB, but I don’t have the tools to measure it. The biggest sources of analog noise are the “digital switches” (video opamps.) Video opamps have a lot of advantages, but low noise isn’t one of them. Higher speed implies higher noise… The DS Mk II opamps are about 6dB quieter than the ones used in the DS and the better audio transformers will have less noise/distortion so the analog noise floor should be noticeably lower.


Thanks as always.

Is the random/white noise additive, or does the highest floor win? ie if the SDM produced white noise at a level of say -120dB and the analog components were at -110dB, would the output noise be -110dB or something a little higher?

Approx -109.586dB.

convert dBs to ratios
-120dB to 1e-6
-110dB to 3.162e-6

Square them

Add them

Square root that

Convert to dB


I just want to come back to one of my inputs about the naming of future PS Audio DACs inside the “DirectStream MKII” discussion.

I would love to see a “Ted Smith Signature” Series of DAC’s, just like the BHK Signature Amplifiers and Preamplifiers. When I see how much engineering art Ted Smith has put into the development of the upcoming DirectStream MKII DAC, I wish that his name would be honored on this system as well and not only on a future “no limits” absolute HighEnd DAC. So here my suggestion:

Originally planned name: DirectStream DAC MKII
Suggested name: Ted Smith Signature TWO | TSS TWO

Originally planned name: Ted Smith Signature DAC (TSS DAC)
Suggested name: Ted Smith Signature ONE | TSS ONE

We all know, that Ted Smith stands for outstanding DAC’s. So just skip the word “DAC” in the name of the product. :sunglasses:

Would this not be a great idea? What do you think about my suggestion?
I might never be able to afford a TSS ONE, but I would be very proud to own a TSS TWO…


And when will that be? Anytime soon? :drooling_face:

My wish would be for an upgraded version available that supports 5 channel

1 Like

If you haven’t seen them, check out the prototype photos in this earlier post:

While hypothetically the digital board could get revised to have fewer input connectors and focus on multi-channel formats, the current two-channel analogue board is really big, and it’s that way for multiple reasons that contribute to sound quality – eg numerous discrete power supplies, layout decisions to minimise cross-talk and noise, and the eight digital switches plus large output transformer per channel. It’s hard to imagine that board being made significantly smaller without compromising performance, and it’s hard to imagine six analogue channels plus a digital board physically fitting into a standard DS chassis.

The other critical thing to understand is that a key part of the DS performance is having the master clock crystal physically very close to the digital switches that amplify the output bitstream from chip-logic voltage levels to analogue audio signal levels with sufficient current to drive external cables and amplifiers. It would be wickedly hard to cram four additional channels around the crystal, and having them at any distance would degrade their sound quality.

So I suspect your best option for multichannel solutions with the DS is to follow the path already taken: combine three stereo DS units with a digital processor that splits your 5.1 signal into three stereo PCM pairs appropriately assigned to individual DAC and amplifier channels.


Some preliminary measurements:

All with 1,048,576 point FFTs on my PicoScope 4262 (16 bit high resolution oscilloscope)
0dBFS is 4VRMS.

You’ll have to read the fine print to see the levels and frequencies at the cursors.

Analog noise floor wide spectrum, tho it looks a little uneven most is below -150dBFS:

Analog noise floor close in spectrum, the 60Hz noise is below about -110dBFS:

Sine 1kHz -90dB, no noticeable spuria, white noise floor even with lots of active signals:

Sine 50Hz -3.77dB, the low frequency harmonic distortion:

(Compare to DirectStream Mk I (from New Firmware: Measurements | Stereophile.com)

19kHz and 20kHz intermodulation, no fuzz:

(Compare to DirectStream Mk I (from New Firmware: Measurements | Stereophile.com)

Unit on the bench with the digital board mounted upside down on risers so I can probe the analog card:

It sounds like I predicted it would sound: more ease, blacker background, longer decays, cleaner more accurate bass, more accurate timbre, less overall distortion which encourages you to listen louder but even so is less tiring with long term listening. I’ll leave it to those at PS Audio to describe what they hear when their units are not in heavy use in engineering. I’m enjoying it (and more importantly my wife is enjoying it.)

There’s still work to do but things are looking and sounding good.

[Edit: this is post 1000 on this thread :slight_smile: ]