Compare clock jitter specs on Firewire interfaces (Traveler, Firestudio, etc.)

[hoagie]

I’ve read this forum off and on over the years, but I’m pretty sure this is my first post here.

I’m putting together a mobile rig for tracking 12-20 channels. For the core interface, I’ve fairly narrowed down the choice to a Motu Traveler, Presonus Firestudio, or a *used* RME Fireface 800. (Whichever device, it will be augmented by outboard preamps via ADAT and SPDIF).

I’m also hoping that this interface can do double-duty in my studio, providing better clock and converters to my existing setup based on a Motu PCI-424 (1224, 308, 2408). The 1224 is getting long in the tooth; I’m thinking that one of these newer interfaces will have improved specs.

Here’s where my brain starts melting: The Firestudio makes big claims for its ultra-low jitter clock. And I’ve *heard* that it’s clock is better than the Traveler’s. Yet the Firestudio’s touted jitter spec of “<300” pico seconds looks high to me. Compare that to clocks like Lucid and Rosendahl from ten years ago, where the jitter specs were in the range of 10-20 ps, or the Big Ben at 40-50 ps. But then, RME shows its “SteadyClock” technology as capable of 700 ps(!) Clearly, there’s some disconnnect here. I would think that jitter could be more objectively measured than, say, THD or dynamic range.

And does anyone know what the Traveler’s jitter spec is?!

Of course I know the smart talk about how jitter is just one variable among many, etc. But if want to be able to use this live-rig interface also as an improved clock (and likely converter too) in the studio, then jitter really is a key factor.

So which of the three boxes above would be best to replace (or improve) my Motu 1224? Or does anyone want to pitch out a wildcard, like the TC Konnekt 24D? :-)

My brain clock jitter is now up in the high milliseconds...

Hoagie

[theblue1]

First thing you need to realize is that supplying an external clock source to a converter does NOT eliminate that converter's internal clock from the picture and does not -- in normal circumstances with modern gear -- "improve" internal timing but actually tends to increase jitter as that converter's internal clock must constantly attempt to "correct" itself to stay in synch with the external clock source.

Obviously, if you want to use more than one converter simultaneously, you have to synch them. In a complex, multi-converter rig, it may be more convenient to use an external clock source (like a Big Ben or other) with a star topology but in general, a properly terminated daisy chain using the internal clock of the first converter in the chain will tend to cause the least jitter in the synched converters.

Unfortunately, there are a number of people who -- for whatever reasons, some of them are celebrity endorsers and others are simply people who bought the Kool-Aid and apparently figured they then had to drink it -- claim that a "superior" external clock will "improve" the accuracy of a converter's sampling.

The science does not support this.

You can read all about it in what is undoubtedly the epic thread on the topic -- and you can read some amusing rejoinders from reps of a certain external clock company, too: PSW Recording Forums: Dan Lavry => Proper word clock implementation


NOW... if someone says they LIKE the sound better with external clocking -- no argument possible there... it's only if they claim it improves something measurable like accuracy, that they're crossing the line.

And both Dan Lavry (in the thread above) and Digidesign (in their whitepaper on clocking issues) have both suggested that those who LIKE the sound of external clocking are entitled to their subjective judgement but that an implicit conclusion from such a claim is that they may simply LIKE the sound of increased jitter.


Anyhow... this ish is, like, my bête noire...

And, amusingly, I used to run a rig with an external clock going to two converters, but it was an old ADAT system that was designed that way [BRC supplying dedicated clock via BNC].

I got caught up in it when a certain external clock maker and their celebrity endorsers started promoting this "improve your converter with external clocking" rigamarole. That didn't sound right to me (from my limited understanding of the issues at the time) and I went looking for answers. Eventually, that led me to Dan Lavry and the Digidesign whitepapter and... well... now it's my pet peeve for reasons that probably have much to do with personal pathology, I'm sure. (FWIW, I use a MOTU 828mkII, which went from everyone's darlin' a few years ago to can't-get-no-respect status -- and that seems to provoke SOME folks to want to mod it or try to improve it in various ways. Me... I figure it is EXACTLY the same unit it was when everybody thought it was so groovy -- so THEY must have changed. And I looked into the modding issue, too, and did not much like what I found. If I was that unhappy with it -- I'd abandon it rather than trying to magically "fix" it somehow.)

[hoagie]

theBlue1:

I do indeed realize that supplying an external clock source to a converter doesn’t eliminate that converter's internal clock from the picture. I’ve read the material you point to. Unfortunately, my post wasn’t clear about how I’d actually be using a new “mobile box” with a better clock and converters in my studio. (My writing wasn’t clear because I was trying not to go into too much boring detail.)

The majority of the tracking I do in my small studio involves 4 or fewer tracks at a time. So I’d be using the new box as my main AD and DA most of the time. I did actually say “to replace (or improve) my Motu 1224.” There, “improve” was a poor word choice: I didn’t mean to imply a reclocking scenario, but simply that I wouldn’t actually get rid of the 1224. I meant to say “and improve upon” (the function it serves in the studio).

So, sorry to have woken up your bête noire!

But before you tell your bête to go back to sleep, you point to a situation where I would have to face reclocking issues -- for which I’d have to do some experimenting. On the less frequent occasions when I’d want to track through the 1224 at the same time as with whatever new box, presumably I’d want to use the new box’s better clock as master. Ideally, I’d want to slave the Motu PCI-424 directly. But the only dedicated sync input on the PCI-424 is a 9-pin ADAT connector. So I might want to find a little box (M-Audio maybe?) that does nothing but format conversion from Word Clock to ADAT sync.

Of course, I could connect the new box directly to the 1224’s Word Clock input; but then the PCI-424 itself would be receiving 2nd-generation clock (new box to 1224 to PCI-424) to distribute elsewhere (e.g., to my 308 and 2408, on which I have digital devices connected, not using any analog converters there). Messy.

So I just might find that in those studio situations where I want to track more at a time than the single new box can input, it might be better to keep the PCI-424 as master... hopefully not degrading the converters in the new box too much by having them clocked to the Motu system. (There’s your bête noire.)

BUT, at the moment that scenario is a bit of a digression. My main interest is in sorting the disparities in clock jitter specs.

1) How is it that a supposedly better clock, such as in the Fireface 800, can tout a jitter spec of “<700 ps,” the Firestudio can claim ultra-low jitter at “<300 ps,” and a ten-year-old Apogee can claim only 20 ps? Where’s the line between meaningful and meaningless with respect to jitter specs?

2) Among the Traveler, Firestudio and Fireface 800, is any one considered to have a *significantly* better clock than the others?

Hoagie

[theblue1]

Why didn't me and my bête noire keep our big gaping maws shut?




OK... I should stipulate that I didn't put your gearlist together in my head before I wrote my post -- and I had to do some careful research (using my favorite methodology: reading eBay postings quoting sales lit) just to educate myself about the PCI-424 and it still didn't include detailed synch info (damn that eBay seller guy! You'd think all he cared about is selling his damn gear!)

But I can see the complexity of this issue here... it starts getting Chinatown-esque (in my fuzzy mind, anyhow) at the point where the 424 comes into play... even though there seem to be some parallels to my old ADAT/BRC rig (and I feel like the fact I currently have an 828mkII means I SHOULD know something about the 424 just by affinity -- but I don't and more than a little hazy about the implicit w/c ish with it and its interconnected devices.)


Anyhow, me and my bête noire are convinced that you're no rube and that I should just get the heck out of the way and let you get to the important job of sussing out your jitter spec issues without further digression or explanation -- since in that area we've returned to one element of certainty: I have no expertise when it comes to actually interpretting jitter specs or reconciling seemingly disparate specs...


Next time someone waves a red flag in front of this particular bête noire, I think I'll just have a nice picnic by the sidelines and watch... instead of throwing myself under its cruel and uncaring hooves...

[tINY]

There is RMS jitter, cycle to cycle jitter, average jitter and peak jitter....

Unless you know at least RMS and Peak jitter you can't really evaluate much of anything. It's like other specs, the methods and numbers get manipulated by marketing and it always stays a little foggy.




-tINY

[hoagie]

Originally Posted by tINY

... It's like other specs, the methods and numbers get manipulated by marketing and it always stays a little foggy.

Yar, figures. So can anyone point to any rigorous, real-world A/B tests of the effects of clock jitter? I've seen and heard comparisons of converters (e.g., the Mytek website), but haven't seen or heard a test with all variables controlled for except clock.

-- Hoagie

[Kees de Visser]

Originally Posted by hoagie

Yar, figures. So can anyone point to any rigorous, real-world A/B tests of the effects of clock jitter? I've seen and heard comparisons of converters (e.g., the Mytek website), but haven't seen or heard a test with all variables controlled for except clock.

A small group of audio experts is working on a jitter test, including a jitter simulation application.
No deadline, no results yet, but I'll try to post news about the progress in this forum.

[ricknaqvi]

Firestudio’s touted jitter spec of “<300” pico seconds

Just a correction. The Firestudio's jitter spec is <20 pico seconds.

Thanks,