Topic: ADI-2 DAC question about "Steadyclock FS" femtosecond jitter spec

Hello,

I'd like to know why the new RME ADI-2 DAC "Steadyclock FS" device is, on the one hand, boasting about jitter so small it's measured in the femtosecond range (ie. anything smaller than one picosecond), hence the "FS" name, yet on the other hand, in the published tech specs we read...

* Jitter when synced to input signal: < 1 nanosecond
* Low Jitter Design: < 1 nanosecond in PLL mode, all inputs
* Internal clock: < 800 picosecond Jitter, Random Spread Spectrum

...which seems like a gross contradiction to me?

Also, those published specs for ADI-2 DAC look exactly identical to the ADI-2 PRO version specs as well.  Correct?
Are these published specs for  ADI-2 DAC model actually the correct values in the correct SI units?

If both DAC and PRO versions have identical 800 picosecond jitter spec from the crystal clock, how come only consumer version offers the "Steadyclock FS" facility?  Why didn't the PRO version offer it?
Or rather, how *CAN* the "Steadyclock FS" facility even possibly be achieved from such a jittery master clock?
ie.  How can the DAC ever achieve femtosecond clock stability from its internal reference crystal clock given it has up to 800 picosecond intrinsic jitter....?

The Mytek Brooklyn DAC in comparison, for instance, is also a "femtosecond clock" but uses a crystal clock with jitter spec of 0.82 picoseconds ( = 820 femtoseconds) which seems an appropriate figure, whereas RME's spec of 800 picoseconds (= 800,000 femtoseconds) appears to be almost 1000x times more jittery.  That's what I don't understand!?  Are RME's units simply wrong in a typo error?  ie.  They meant "femtoseconds" not "picoseconds"
If true, 800 picoseconds is very bad IMHO.  It's maybe okay for a consumer level computer soundcard, but not a reference hi-fi or recording studio D/A converter.  Even my twenty-year-old Apogee PSX100 converter which I bought in 1998 has got clock jitter spec of 22 picoseconds.  So RME's 800 picosecond spec has me very worried, otherwise the product looks very attractive, feature wise and price wise.
I'd appreciate if somebody could explain this disparity in the specsheet, or if that's a typo, please give us the correct figures.  Thank you.

2

Re: ADI-2 DAC question about "Steadyclock FS" femtosecond jitter spec

The manual's entry is the same since 10 years or so, and refers to clock jitter that can be measured on external outputs like the word clock one with a standard DSO. The relevant sampling jitter is not given, nor do we boast with quite meaningless numbers of crystal jitter, as the only relevant jitter is the one that you finally see in the AD and DA conversion. And that one looks extremely clean, as upcoming reviews will show.

But if it makes you feel better: the crystal used in the DAC is referred to as the lowest jitter crystal currently available, outperforming the famous Crystek ones.

And neither ours nor Crystek give a single jitter number (which has no meaning as it doesn't tell where and how it was measured), but a diagram of jitter over frequency in dBc, with values far below -100 dBc.

Still we will not enter this useless number throwing game. SteadyClock FS uses a 'femtosecond' clock (marketing hooray) and has less self-jitter than the former version (marketing not excited). More info is not available from RME at this time, sorry.

Regards
Matthias Carstens
RME

Re: ADI-2 DAC question about "Steadyclock FS" femtosecond jitter spec

The jitter value quoted by Mytek is only called "internal jitter", and does not refer to external sync and actual conversion. No other (comparable) figures are provided...

Regards
Daniel Fuchs
RME

4 (edited by propianist 2018-02-07 20:58:41)

Re: ADI-2 DAC question about "Steadyclock FS" femtosecond jitter spec

Hi Matthias, thanks for your reply.

MC wrote:

The manual's entry is the same since 10 years or so…

Why publish info that you know is 10 years out of date?

MC wrote:

The crystal used in the DAC is referred to as the lowest jitter crystal currently available, outperforming the famous Crystek ones.

If you’re using today’s latest crystal, why are the specs still the same as 10 years ago?

MC wrote:

...it refers to clock jitter that can be measured on external outputs like the wordclock one with a standard Digital Sampling Oscilloscope.

Neither the ADI-2 DAC or ADI-2 PRO actually have any external wordclock output.
PicoQuant, Keysight, Standford Research and others do make time interval counters accurate down into the small picosecond ranges.  Surely there is a way to measure the internal performance to better precision than <1 nanosecond.
If you’re only measuring to approx 1 nanosecond resolution at the outputs, how can you quantify subtle improvements to the design like the “SteadyClock FS” enhancements?
Surely you must also need to measure with much higher precision to assess any improvement the femtosecond clock makes?  How else can you know if your “SteadyClock FS” is operating within its target spec or not?  Why aren’t you publishing those figures?

MC wrote:

Still we will not enter this useless number throwing game.

If quoting the crystal clock jitter specs is a “useless number throwing game” why are the industry’s most respected converter companies like Prism Sound, dCS, Antelope, Mytek, Lavry, Apogee, Benchmark, Lynx, Black Lion, Forssell Technologies, Crane Song, etc., all treating it seriously?

MC wrote:

Neither ours nor Crystek give a single jitter number (which has no meaning as it doesn't tell where and how it was measured), but a diagram of jitter over frequency in dBc, with values far below -100 dBc

Okay, so print the graphs as well as publish the figures.
Page 75 of ADI-2 Pro manual has a jitter graph at half the Nyquist freq of 11.025kHz, sampled at 44.1kHz, but considering the converter goes up to 768kHz sampling rate, the proportional impact of that jitter relative to the 768kHz sample period and its waveform reconstruction would be made over 16 or 17 times more severe!

Futhermore you’re claiming an audio frequency response up to 180kHz @ -3dB for A/D and 115kHz for the D/A, which means a single fullscale cycle of a wave at that frequency must race across the 24 bit quantisation levels at a much faster pace because of it’s shorter duration at higher frequency than your graphs’s 11.025kHz wave, so the number of discrete quantisation levels covered during each sample’s timing jitter uncertainty period is a whole lot more, at least 10 times greater, as well, which will effectively lower the precision resolution in quantisation bits at those higher frequencies.  Certainly not able to define the signal amplitude to the full 24 bit depth at those frequencies.  The lower bits would be just erroneous random data.  Hence the picosecond jitter value is of importance to know potentially how far the converter can resolve or reconstruct the audio signal.

The fact it produces sidebands on a freq domain graph is just another way of viewing the same information.

If the crystal jitter figure is 0.7 picoseconds or something, that would be amazing, and would prove to me mathematically that the converter clock can resolve tighter timing steps than 24 bit quantisation level steps over a fullscale 20kHz audio speed bandwidth, which is more than good enough for me – I can’t hear to 180kHz anyway.  But given a specsheet jitter figure of 800 picoseconds only, and no further information, I’m left guessing.

MC wrote:

More info is not available from RME at this time, sorry.

That’s a pity because I was hoping to purchase 3 of these. (6 channels for a 3-way digitally FIR processed tri-amped monitor system.)
I'm quite willing to believe the RME ADI-2 DAC really is a fantastic product, and most reviews I’ve read online seem to agree.  I don't own it and have never heard it myself.  I’m not criticising or doubting the actual hardware is top notch, I’m just asking for better explanation of the specsheet figures, because they appear to be in contradiction to the real nature of the product.

Daniel Fuchs wrote:

The jitter value quoted by Mytek is only called "internal jitter", and does not refer to external sync and actual conversion. No other (comparable) figures are provided.

Hi Daniel,
This is why I’m asking RME for their comparable figure for “internal clock jitter” if that’s how you prefer to define it.  So I as a customer can make a fair comparison.  How does it compare to the 0.82 picoseconds of Mytek Brooklyn, or the calculated 0.88 picosconds of Antelope Pure2.
I understand that a few inches further down the copper tracks on the circuit board it may measure worse, and by the time it’s passed through buffers and a consumer S/PDIF optical interface, it will be degraded further, but we’ve got to be able to compare on a level playing field to start with, so let’s start with the internal crystal clock jitter.  Do you have this figure anywhere?

Re: ADI-2 DAC question about "Steadyclock FS" femtosecond jitter spec

propianist wrote:

This is why I’m asking RME for their comparable figure for “internal clock jitter” if that’s how you prefer to define it.  So I as a customer can make a fair comparison.  How does it compare to the 0.82 picoseconds of Mytek Brooklyn, or the calculated 0.88 picosconds of Antelope Pure2.
I understand that a few inches further down the copper tracks on the circuit board it may measure worse, and by the time it’s passed through buffers and a consumer S/PDIF optical interface, it will be degraded further, but we’ve got to be able to compare on a level playing field to start with, so let’s start with the internal crystal clock jitter.  Do you have this figure anywhere?

No, and I'll refer you to Matthias' statement on that matter. As he pointed out, "the only relevant jitter is the one that you finally see in the AD and DA conversion". You seem to have no issues with Mytek only publishing that somewhat elusive "internal jitter", instead of figures that would be more comparable to the figures we do publish, and which are more relevant. However, none of this is available even in the manual.

Have you read our tech info on Steadyclock, by the way?

Regards
Daniel Fuchs
RME

6

Re: ADI-2 DAC question about "Steadyclock FS" femtosecond jitter spec

You seem to have too much time on your hands - I don't. I just will give one last reply:

> If you’re only measuring to approx 1 nanosecond resolution at the outputs, how can you quantify subtle improvements to the design like the “SteadyClock FS” enhancements?

I wrote that already -  the only relevant jitter is the one that you finally see in the AD and DA conversion. For example by standard jitter measurements based on the JD signal (fs/4 with and without modulation). And we will publish those when we have the time to present them.

Regards
Matthias Carstens
RME