Lavry and Katz comment on external clocks (converters)

While hunting around for something else, I stumbled across these two quotes, which I found interesting, and thought some of you would as well:

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

bobkatz wrote on Sat, 02 October 2004 21:40

An external clock is a bandaid for a "cure" which can only be done properly within a good converter design. In fact, any converter which does not perform equally as good or BETTER on internal clock than external is also defective.

~~~

danlavry wrote on Sat, 02 October 2004 22:54

A mediocre internal crystal implementation is going to outdo even a good external clock implementation.

From whence the quotes came...

I read those earlier today.

Certainly puts me back at square one with my clock questions that no-one's bothered to answer, here and at unicorn nation.

Yeah... the science seems pretty straightforward on this. Digidesign has a white paper on it that basically says what Lavry says only does it a little more diplomatically.

http://recforums.prosoundweb.com/index.php/t/13628/0/

You'll see the link to the PDF in the second or third post and a discussion of the paper. The paper's worth at least skimming.

There's a long and long running thread in Lavry's PSW forum that features a number of appearances from an engineer at a certain converter and external clock maker. I find some of the assertions by that guy very amusing. Almost new agey. It's hard when you're arguing technology and the science isn't on your side.

You see, grasshopper, in this world there is good jitter and there is bad jitter...

Hey, but what do I know? I would have been an English major if I'd ever declared. (Ah, for those halcyon days when you could go to university for four years without declaring a major and nobody even shrugged.)

I'll let the guy from the company with the big ad budget argue with the living legend who helped design their first converters -- and with the much much bigger company... It's sort of David Vs. Goliath only the bully who used to beat up Goliath just came to help David...

I just hope no giants fall on me.

The answer (to any question about a master word clock) is this: It's necessary to synchronize the data in a system that has multiple digital audio devices that have their audio interconnected. That's the only purpose of an external word clock source - SYNCHRONIAZTION.

If you have crummy devices, they might sound better with an external clock source. If you have good devices, there is no good reason why they should sound better. The external clock doesn't replace the internal clock, all it replaces is the reference frequency source (the crystal) for the internal clocking system. And a good design will effectively filter out any jitter or other phase noise injected by the reference source. There have been reports of "even my wife, in the other room" hearing something better when using a different clock source, however, but nothing conclusive (at least not to me).

If you need a master clock, then you need it and it doesn't hurt to get one that claims to be good. If you don't need to synchronize multiple digital audio devices, then you don't need a master clock. And if you have a device that's improved when connected to an external clock (for other than synchronization) then you're probalby better off in the long run by eplacing that device than by clocking it externally.

The assertion that you don't need an external clock if you use only one good quality convertor has always made sense to me. However, I have heard so much testimony on the benefits of high quality external clocks that I was even pondering purchasing one. Has anyone done blind listening tests?

I've never run into anyone who has recounted a properly rigorous blind listening test -- but there are jillions of anecdotal and subjective endorsements of the practice floating around from folks who've sunk a couple thou or so into external clocking.

Lavry has suggested -- not necessarily sarcastically (although THAT may be subjective, too ) -- that some people may actually prefer the sound of some jitter.

The conclusion section of the Digidesign paper suggests much the same thing, a bit more diplomatically.

FWIW, the engineer from that prosperous converter/external clock manufacturer seems to fall back on a "good jitter/bad jitter" rationale (that sounds highly suspect to this would-have-been English major) and a reliance on subjective /anecodtal reports of improvement.

An interesting topic, no doubt.

BTW, there may be times when an external clock in a star topology is more convenient but for simple multi-converter rigs, the most convincing info I've seen seems to suggest that for simpler multi-converter systems, a daisy-chain with the lead converter supplying clock down the chain will provide the least jitter.

__________________

Lavry's writing -- though often technically detailed and fairly challenging -- on this general topic and others is very interesting. (Bob Katz and Jon Hodson, on Lavry's board, GS, and others, are another couple guys who seem to have his slide-rules glued on straight.)

But Lavry is definitely someone who does not suffer fools gladly.

I myself was corrected by Mr. Lavry, firmly but relatively gently, on a couple of fine points I rounded off (to the point of factual inaccuracy, apparently) in an analogy I was trying to make. (Mind you, he agreed with my practical advice. He just thought how I got there was poorly thought out and potentially created greater misunderstanding.)

My first thought was who is this boutique bozo to criticize ME? (I was, after all, almost an ENGLISH MAJOR. )

Uh, then I did a little bio research on Lavry and realized he is really the real deal. And 3/4's...

In this industry?

Guess I don't look so stoopid now for saying I couldn't hear a difference wtih the PreSonus FS clock vs. the clock in my existing system.

But I do recall when I first got an Ensoniq PARIS system, it sounded way better than what was out there at the time. I asked their engineers about this, and they said they thought it was due to the fact that they'd paid a lot of attention to creating something with very low jitter. So I think excessive jitter could be audible, but I wonder -- what with the level of today's technology -- whether devices are still being made that have excessive jitter.

Whenver I see those angels-on-the-head-of-a-pin discussions, I always just want to say "You want your system to sound better? Write a better bridge! (Said only semi-seriously.)

That's because there are so many crappy converters out there. The point stands that if an external clock sounds better to you, and you have everything synced to the internal clock on your best converter, then your best converter's clock still sucks. Which up until quite recently was the case fairly often. And I find it funny that Digidesign published a paper about this because they're one of the worst offenders - loads of people reported improvements when using external clocks with Digi gear and that says more about Digi than it does about the "benefit" of external clocks.

It's one of those things that you'd have to test with your own gear to determine if you like it better or not. I wouldn't buy anything that you can't send back. And if I found I liked my converters better with an external clock I'd look into better converters before shelling out the dough for the external clock, unless I had a ton of money already invested in a slew of crap converters maybe.

Best post of the year so far...

I know you're kidding, Lee, but just for the kids: putting an external clock on a crappy converter tends to increase the overall jitter.

With regard to "using your ears": subjectivity is obviously important in the musical field.

But SOME people have tried to claim that feeding an external clock source to a standalone converter will improve its accuracy.

That's a WHOLE different thing than someone saying their new Big Ben (er, just to pick a well-known external clock out of thin air) sounds better to their individual ears.

No one can argue about the latter but there's LOADS of hard science -- and practical test data -- to argue against the former.

I'm gonna go with this one...

I disagree with the semi-serious part, though.

P.S. What's a blind listening test?

When I say blind listening test I mean having some sophisticated listeners audition the difference between internal clocks on several devices and several brands of external clocks to see if there is an audible difference and if so, which sounds best. The blind part is that they should not know which product(s) they are listening to.

It takes a good bit of effort to minimize the variables to create an unbiased test, but since people spend many thousands on these products, I think it would be worthwhile. The question is; who will volunteer to do it and/or who will pay for it?

If I get to keep a set of Lavry converters, I'll volunteer to be a subject. Not, however, to pay for it.

BTW, I'm pretty sure Stranger was continuing the joke Craig started, but I'm a big fan of scientifically rigorous testing and I know there are a lot of folks who are really unclear on the notion, so a little extra explication never hurt no one.

It can be a big PITA to do proper blind testing but -- when doing subjective testing -- it's one of the most important ways to filter out factors that might subconsciously color the subject's perceptions.

That said, my guess is that your subjects might very likely find -- depending on the set up being tested -- noticeable differences and be able to identify one from the other at least part of the time.

And it's certainly within the realm of possibility that a number of them might well prefer the externally clocked rig.

That doesn't mean, however, that that rig is more accurate in transcribing an analog signal into digital.

Obviously, we intentionally introduce other subtle and not so subtle forms of distortion into our tracking and mixing -- and often pay big bucks to do it.

As I intimated earlier, jitter can be measured, but one's subjective response to increases or decreases in jitter is just that -- subjective.

If you say you like one sound more than another, who can argue with that?

But, coming back around full circle (a rarity for me in a digressive mood), it falls to blindfold testing (if you're feelin' kinky you can use real blindfolds ) to help people sort out what they actually subjectively like from what they really, really want to like.

In general I'm a big fan of scientific rigor too, and have even participated in quite a few blind tests. But for a lot of these types of questions I really think it's kind of silly (in case you couldn't tell from my last post ), because as soon as you start trying to control the factors that would affect one's perception, it becomes "not a real world scenario" so the results don't mean a whole lot. And we are, after all, talking about a subjective thing as you say.

Blind tests are useful up to a point but I think at least as many people draw erroneous conclusions after a blind test as before one. If I add a piece of gear to my setup and it sounds noticably better, I'll keep it, and I don't really care if somebody else's blind test (which probably used a different setup from mine) says differently. If I have to dance on the head of a pin and listen through $10,000 monitors to tell the difference... then it ain't worth it. Seems pretty straightforward to me.

But it's the problem of determining whether it really does sound noticeably better to you.

Don't get me wrong: I do not go around blind testing every piece of gear I want to buy. (I do kind of wish I had in a couople of instances, though. ) It's simply not practical.

But we also know that there are a lot of factors that can go into our perceptions and that those perceptions are highly mutable, contextually.

Ethan Winer has a very interesting informal write-up on some testing he did in treated and untreated rooms with very minor differences in mic placement and found very large frequency response variations due to reflection/comb filtering issues. While his tests in one of his own treated rooms (and that, of course, is his business) showed considerably less comb filtering, it was still a significant factor in unevenness of frequency measurement at positions only a few inches apart.

Winer posited that such wide variations in even treated environments may make much subjective testing so variable as to be inconclusive over anything but a very long battery of tests. (I'm probably mangling that. It was a while ago that I read his write-up.)

Yeah I remember that, and I agree. I make my own determination of whether something sounds definitively "better" after I've either heard it in a variety of environments, or can reliably identify it (for better or worse) just by listening over ordinary consumer gear. Otherwise, I assume that I like something better for my own setup but my preference may not translate to other gear chains or other rooms.

I've had this experience quite often with things like microphones. For example, in my own home studio, it turns out the good old SM57 is always the mic I go back to when it comes to miking guitar cabs. Occasionally I use something else but in most applications, if I compare several mics I prefer the SM57. I found this funny at the time I "discovered" it because in other studios where I've worked, that's almost never been the case.

Just to cite a recent example, when my band recorded at Han's studio in Holland, Han put up several mics on the guitar cab and said, "I'm going to try the Beyer M88 on your guitar. I bet you'll love it."

"I dunno," I said, "I have M88's and they're great, but most of the time I still prefer the SM57 for guitar, at least in my room."

Han seemed kind of surprised, but he went ahead and put up a 57 for comparison. But when we listened back to the recorded tracks, there was no comparison at all - the M88 clearly killed the 57.

I think this is because Han knows his own room, his own setup and his own mics, and has learned through years of experimentation what sounds best. And I've done the same in my room. But interestingly enough, it wasn't hard at all to agree on what sounded "best." I'll bet if he visited me and we recorded something here, he'd think the 57 sounded better. I'm sure the differences can be accounted for by a zillion different factors including what Ethan was talking about. That's why I think a given blind test in a given room through a given chain is kind of silly, and everybody just needs to learn by experimenting what sounds better for them.

I can go you one better than that. I plugged my vintage 1967 Wavetek function generator into the external word clock input of my Mackie HDR24/96 to experiement with variable speed and it didn't sound any worse for the jitter that I'm sure was present in this "master word clock." So, yes, today's converter chips do a pretty good job of ignoring jitter.

But there's more to an A/D or D/A converter than a chip. People like Dan Lavry and the Lynx folks work really hard at surrounding them with whatever it takes to keep the the clock ticking accurately. People like M-Audio and Presonus pretty much work from the application notes and do a pretty good job because it's hard to make a really bad converter with today's components.

It's possible to make a better one, however, and if you have the money and the justification to own it, you can get it. And like everything else in this industry, there will be some people who will tell you how it blows away everything else, and others who will tell you how badly a pretty good ordinary converter sucks.

Not if the converter itself is bad enough. But it's hard to buy a new one today that's that bad. The good news is that it probalby won't make it sound worse either, unless you're really, really listening hard (in which case you should have a better converter anyway).

An external clock serves a purpose and it's not to make converters sound better, it's to make a system work together. And for that, unless you're using digital devices with asychronous inputs (which, by the way, are getting pretty darn good) you gotta have a master clock. So you might as well have a pretty good one.

Originally posted by Lee Flier In general I'm a big fan of scientific rigor too, and have even participated in quite a few blind tests. But for a lot of these types of questions I really think it's kind of silly (in case you couldn't tell from my last post ), because as soon as you start trying to control the factors that would affect one's perception, it becomes "not a real world scenario" so the results don't mean a whole lot. And we are, after all, talking about a subjective thing as you say. Blind tests are useful up to a point but I think at least as many people draw erroneous conclusions after a blind test as before one. If I add a piece of gear to my setup and it sounds noticably better, I'll keep it, and I don't really care if somebody else's blind test (which probably used a different setup from mine) says differently. If I have to dance on the head of a pin and listen through $10,000 monitors to tell the difference... then it ain't worth it. Seems pretty straightforward to me.

Actually, subjective perceptual differences can be quantified. That's how the Bell Labs guys came up with the Fletcher-Munson curve - - they tested several hundred people, and ran all the data through statistics to determine the hearing response curve of people in the middle of the statistical 'bell curve', more or less. But it IS a big endeavor, and a true PITA to do correctly.

Moreover, any individual finesse and ability in listening skills would be buried in the data - - you'd get a picture of what 'average' people would perceive. Which, of course, is the purpose of using statistics, to quantify and normalize out all the variances in the sample population.

I suppose you could do the same thing with an individual, by doing repeated tests (like maybe once every two days) over the course of, say, a year or two. But, naturally, then the results only apply to that individual... And by the end of the two years, all the equipment tested would be obsolete

Yes, and not everything is as quantifiable as that.

Thanks for the important qualification, Mike!

I used to always be careful to add some CYA phrase like "with modern, professional grade converters" ... I'm getting sloppy in my old age. Pretty soon I'll be too sloppy to make accurate comments about subjects I don't really know anything about. That day is no doubt coming...

Originally posted by philbo Actually, subjective perceptual differences can be quantified. That's how the Bell Labs guys came up with the Fletcher-Munson curve - - they tested several hundred people, and ran all the data through statistics to determine the hearing response curve of people in the middle of the statistical 'bell curve', more or less. But it IS a big endeavor, and a true PITA to do correctly. Moreover, any individual finesse and ability in listening skills would be buried in the data - - you'd get a picture of what 'average' people would perceive. Which, of course, is the purpose of using statistics, to quantify and normalize out all the variances in the sample population. I suppose you could do the same thing with an individual, by doing repeated tests (like maybe once every two days) over the course of, say, a year or two. But, naturally, then the results only apply to that individual... And by the end of the two years, all the equipment tested would be obsolete

But -- of course -- the knowledge we have derived scientifically is still -- by definition and practice -- that which has been found by accumulated testing and observation to be most likely true.

So, in order to avoid being trapped forever in a solipsistic prison of subjectivity, it's appropriate to fall back on the body of well vetted knowledge derived from our scientific best practices.

OTOH... it's only rock and roll.

Which is dead, anyway. And has been for years. Long live Rock and Roll!

Not to worry. By then you won't be able to hear differences anyway, so you can just say "It works just fine for me."

That's about where I am now, not because I'm going deaf, but because I realistically balance my priorities of how I allocate my relatively fixed income. If a new device in the studio isn't going to pay for itself, I have to consider it to be a toy. I like to play as much as the next engineer, but I like to eat and keep my house warm in the winter and cool in the summer, too.

A $2,500 D/A converter isn't going to bring in a $2.500 project (or five $500 projects) that I wouldn't have had without it.

More realistically, the $2.500 converter plus a new set of monitors ($3K) plus an acoustic tune-up of the room ($10K) so I can hear the benefit of the converter might actually bring in a $15K project that would have gone elsewhere (maybe, if I also had ProTools, which I don't). But that would only work if this was a full time business, which it isn't.

My experience comparing a Yammy AW4416 with internal versus a Genx xternal clock ( for both DA and AD on each given track) was that the sonics were slightly different. No one was clealy better.

Frankly, i jumped into this partly cause syncing makes my workflow more efficient and partly case of input on the old Musiplayer boards from a well known O'Moderator around here claiming that imaging and depth imprved with the genx driving the AW.

In my experiments on challengin sounds such as perc, the jangling key test and a few other common instrument the resulting sound was just very subltly differernt. Depening on the the source I might like one better than the other. It was a horse race. FYI-My monitors and ears are good, my room not so good (esp. at that time).

I now use external clocking for the improved workflow when passing digi signals around to multiple devices. I can do so comfortable knowing that Im not provoking a compromise in the sound.

Possble explanations:

1)Both clocks are good so they yielded just slightly different sonics.

2)The differences are somewhat larger than i think casue of imperfedt monitoring.

OT:Speaking gf clocks.. I'm in Tokyo japan at the moment. Anyone know of the "Meet me at the Roppongi clock" clock? This place is so chaming. It continues to evolve at an alarming rate.