External Clocking - is it all that?

PaulBrewer

madtheory wrote: »

How much is the Big Ben anyway? Wouldn't it be better to buy a Lavry or a Prism converter to replace the 192?

It most certainly would and the Apogee's are a big step up too.

However Euro 1400 won't get you too far the Lavry or Prism road where as , as you'll hear it ups the 192s game significantly.

I rather fancy a Lavry Gold 2 chan converter for the mix , however at Euro 6.6k for 2 chans I may have to postpone ...



Actually - are there any Lavrys in Ireland that we know of?

madtheory

One of the guys is using a Benchmark DAC1 on his HD rig, he says it's far better than the 192. I dunno of any Lavrys.

When HD first came out, if I remember rightly, wasn't it rumoured that the 96 better sounding than the 192?

PaulBrewer

madtheory wrote: »

One of the guys is using a Benchmark DAC1 on his HD rig, he says it's far better. I dunno of any Lavrys.

I think Frob uses a Benchmark

PaulBrewer

madtheory wrote: »

One of the guys is using a Benchmark DAC1 on his HD rig, he says it's far better than the 192. I dunno of any Lavrys.

When HD first came out, if I remember rightly, wasn't it rumoured that the 96 better sounding than the 192?

I don't know ....

A well known Converter maker did tell me that when the 192 came out they bought one for a look see - on flipping the lid they noticed what appeared to be a copy of their previous generation clock circuit - they'd considered suing but reckoned if that's where Digi were at leave em to it?

Have you heard the Apogees? I could send down a AD/DA too . Chalk and cheese compared to 192s I reckon but not at the Lavry/Prism pricepoint.

woodsdenis

Afternoon MT

I spent far too long last night trawling through the thread. Very informative.

I am sure Dan Lavry is a fantastic engineer and his products speak for themselves, but he comes across as a complete dick at times. I think Bob Katz's view was more fair as he actually listens to music aswell as having the technical chops. Its a pity it got into an Apogee battle has the subject matter of the thread got diluted. I dont own any Apogee gear so I cant comment on it.

1. I certainly cant get into a technical discussion about what Dan Lavry says as I dont have the technical chops to do so. To put it in a nutshell, he says that any PLL converters will not get any better when externally clocked. This is disputed by some on the thread. Fair enough. It seems a very black and white way of looking at it, as many audio pros and engineers/manufacturers disagree.

however


2. Bob Katz, I think is more realistic. He says if an external clock makes your converter sound better than there is something wrong with the design of your converter. I think this is absolutely correct. If you use Lavry/Prism converters and externally clock them I would suggest there is no difference. Bob tested this with his TC 6000 and it made it marginally worse according to him. If you read this quote from the thread he clearly states that his list of "great converters is very small, indeed" I dont know what brands they are but at a guess I would say pretty high end. My conclusion to all of this
is that just because a converter uses a PLL converter certainly doesn't make it immune to improvement by external clocking, it just means it should have been designed better. This is Bob Katz's view. Dan Lavry takes a more
blinkered one as an engineer/designer might.


Bob Katz quote
"Another way to look at this is "the battle of the marketing against the science", or "can $1800 (retail) of external components really beat $40 of internal"?

For (in a particular test) if an external clock really beats an internal one, then the manufacturer of the converter under test must really have cut some cheap corners, OR at the time of manufacture, his technical chops were not as good as they oughta!It takes a tremendous amount of knowledge of analog, digital, RF and systems design to make a great-performing converter. Which is why my personal list of "great" converters is very small, indeed.

I know a converter designer who is modest and does know what he is doing. He also knows the limitations of his design expertise. So he freely admits that to design a great PLL would take far more time and money and consulting work than he could afford, so instead, he made sure that he had a well-performing internal clock in his converter, and exploit that as an advantage. I thoroughly agree to that philosophy.

I also know another converter designer who built a converter with "no compromise" in mind. It took one man-year to design the clocking circuits, both internal and the superior PLL. How long do you think it will take for them to make up their R&D costs?

Additionally, BOTH of these converter manufacturers have been entirely honest in their marketing statements of the capabilities and limitations of their gear."

Anyway very interesting reading, any one interested in this should read the
whole 14 pages for themselves:eek:

woodsdenis

madtheory wrote: »

One of the guys is using a Benchmark DAC1 on his HD rig, he says it's far better than the 192. I dunno of any Lavrys.

When HD first came out, if I remember rightly, wasn't it rumoured that the 96 better sounding than the 192?

I think the issue was if you recorded at 192 on a 192 io it sounded worse than 96. Why anybody would record at 192 I dont know. I actually agree with Dan Lavry on that one.:D

madtheory

Yes, Dan Lavry is your typical engineer- poor people skills. So he probably doesn't sell as many converters as the charmers over at Apogee. He should still write a book though!

Personally, I prefer the science, and try my best to understand it As you say Denis, John Watkinson is one of those rare people- an engineer who can explain things very well.

Agreed, 192k is nuts. It's a misunderstanding of the sampling theorem, which is one of those universals that applies to everything from ship building to population surveys.

Anyways, I'm looking forward to testing the Big Ben. I'm gonna test it with my M Audio Firewire 1814 too, that could be interesting!

madtheory

The importance of blind AB testing (and how it's done):
http://seanolive.blogspot.com/2009/04/dishonesty-of-sighted-audio-product.html

Science and subjectivism in audio:
http://www.dself.dsl.pipex.com/ampins/pseudo/subjectv.htm

All links from http://www.ethanwiner.com

woodsdenis

madtheory wrote: »

The importance of blind AB testing (and how it's done):
http://seanolive.blogspot.com/2009/04/dishonesty-of-sighted-audio-product.html

Science and subjectivism in audio:
http://www.dself.dsl.pipex.com/ampins/pseudo/subjectv.htm

All links from http://www.ethanwiner.com

MT How am I going to watch Britain's got Talent tonight with all this studying to do;)

jtsuited

hope I don't derail the thread but it always amazes me the lack of scientific good practice involved in your average audio based a/b blind test.

Which would be fine if the flawed tests weren't put forward as 'scientific fact' by some people.
Btw, if you want to educate yourself on the distortion of scientific data by vested interests, read Bad Science by Ben Goldacre. Best book I've read all year.

madtheory

Sorry Denis!

That sounds like an interesting read jt. I'm having a hard time trying to imagine how a blind AB test could be messed up. Although I have seen several reviews of speakers, in pro audio mags, where they author claims he's doing a blind AB test- but on his own? Very strange

jtsuited

just to add my 2cents to this:

I have heard the 192's clocked with a Big Ben. Very very noticeable difference. Jaysus an epic thread and my only contribution is to agree with Paul. What has become of me?

studiorat

Has anyone considered the sync I/O? Digidesign's own external clock. Is it any different to the clock in the 192 or 96 converters?

I hooked one up to a RADAR a few years ago and everyone in the room agreed that it was an improvement, on what's already a good sounding machine IMO.

I kinda scanned through the threads here, one thing that struck me was the fact that just because something looks good on paper doesn't actually mean it is going to sound good. One only has to look at the frequency response of a tape machine to figure that one out.

Seziertisch

jtsuited wrote: »

just to add my 2cents to this:

I have heard the 192's clocked with a Big Ben. Very very noticeable difference. Jaysus an epic thread and my only contribution is to agree with Paul. What has become of me?

You must understand
Though the touch of your hand
Makes my pulse react
That its only the thrill
Of boy meeting girl
Opposites attract
Its physical
Only logical
You must try to ignore
That it means more than that

Chorus
Whats love got to do got to do with it
Whats love but a second emotion
Whats love got to do got to do with it
Who needs a heart when a heart can be broken

It may seem to you
That Im acting confused
When youre close to me
If I tend to look dazed
I read it some place
Ive got cause to be
Theres a name for it
Theres a phrase fits
But whatever the reason
You do it for me
Repeat chorus

Ive been taking on a new direction
And I have to say
Ive been thinking bout my own protection
It scares me to feel this way

Chorus
Whats love got to do got to do with it
Whats love but a sweet old fashioned notion
Whats love got to do got to do with it
Who needs a heart when a heart can be broken

Aridstarling

Haha! Win post!

teamdresch

Interesting review -

http://farmelo.com/blog/?p=79

madtheory

teamdresch wrote: »

Interesting review -

http://farmelo.com/blog/?p=79

Interesting, but not at all scientific:
We threw up a number of different mixes, and our first impression was: “hey, different clocks really sound different.”

PaulBrewer

Here's Apogee's reply -

"Jitter specs are nefarious at best. Everyone manipulates jitter specs, making this statistic almost meaningless. What I can tell you is that the Rosetta 800 clock performs 2Xs better than the clock in the AD-8000-SE. The C777 clock used in Big Ben is 10Xs better than the Rosetta 800. Now, in the interest of providing a jitter spec for the Big Ben, we tried every test imaginable to get a reliable figure. Unfortunately, all of our jitter measuring equipment has significantly MORE jitter than the C777 clock, making it almost impossible to calculate how low it is. Apogee is not claiming that the C777 has no jitter, as everything has jitter, but the C777 has by far the lowest level of jitter ever seen in a digital audio device.

Another advantage the C777 has over most clocking devices is that the statistical distribution of the C777’s minimal jitter error is a Gaussian or "bell" shaped distribution (hence the name, Big Ben). We found that in looking at this spec on other clocking devices on the market, very few of them exhibited the deterministic, Gaussian distribution that is present in the C777. It is well known that Jitter with this characteristic is much less audible than any other sort of distribution. This in combination with the exceptionally low jitter performance on Big Ben is why the unit makes such a noticeable audible improvement when connected to your DAW.

Jitter

One of the big problems with clocks is that it’s very easy to make statements about measurements without knowing what you are seeing or really listening to. As we at Apogee know very well, it’s so important to always keep questioning what results you are getting and why. Ask yourself when testing, am I seeing what I want to measure or not? With clocking tests, it is very often the case that you are not.

As stated earlier, jitter performance is only important when doing A/D or D/A conversion. For digital transfers its only relevant when the jitter is utterly extreme, so extreme that bits get corrupted which would result in glitches (very nasty and audible obviously, depending on your luck and the kind of jitter).

Think of jitter like an FM modulation of your audio (indeed that's what it does). If you apply exactly the same FM modulation to both channels at the same moment (which is what you did here) then you could have all of the jitter in the world and the digital signals would still cancel out (pure math). I mean if you apply nutty pitch bend on a synth and send it out to 2 channels then phase invert those and add them you'll get nothing, and yet that doesn't mean that the pitch bent sound is identical to the sound with no pitch bend applied. While this is an exaggerated comparison, in the end it's exactly the same, under a magnifying glass...

In the end, what this test tells you is that your math teachers were right; sin(a(t+dt) - sin(a(t+dt) =0; obviously whatever your dt (jitter) is it will cancel out.

That doesn't mean it's insignificant with real audio. You really don't want to be FM modulating your audio. If you look at harmonics in a real sound (as white noise is kind of a bad signal obviously) and apply even slight FM modulation to it they all turn from nice clean harmonics into wider bands, sort of smearing out the spectrum. It might be an exaggerated comparison, but the effect of jitter does to the spectrum something similar as blur does in graphics. And just because you are blurring you are also messing with the intricate phase relationships of the spectrum of the original signal. And our ears are very sensitive to phase as this is what gives us among other things, important clues our brains use to determine sound location, what we like to think of as stereo image.

Understanding that most folks do not have access to the equipment necessary to carry through a mathematical/scientific test that would allow you to prove this outright, our strongest recommendation is to just listen. After all, that is why we care about all of this in the first place, to make everything sound better.?

madtheory

PaulBrewer wrote: »

Understanding that most folks do not have access to the equipment necessary to carry through a mathematical/scientific test that would allow you to prove this outright, our strongest recommendation is to just listen. After all, that is why we care about all of this in the first place, to make everything sound better.?

It's possible to do a scientific test just by listening, i.e. a blind A/B test. S'funny they didn't mention that...

madtheory

PaulBrewer wrote: »

If you apply exactly the same FM modulation to both channels at the same moment (which is what you did here) then you could have all of the jitter in the world and the digital signals would still cancel out (pure math).

What are they referring to? Are we to assume that the hypothrtical A to D and D to A are in the same box? That's the only way to get the cancellation they're talking about. Some links to measurement and test data would greatly improve the authority of this reply from Apogee. Is this an extract from a bigger document?

PaulBrewer

madtheory wrote: »

Is this an extract from a bigger document?

No - it's just an Email to me.

Are we to assume that the hypothrtical A to D and D to A are in the same box?

I believe this section is referring to jitter having less importance in digital transfers i.e. no A2D or D2A .

madtheory

A sceptic questions audibility of jitter:

http://www.avguide.com/forums/jitter-audibility-robert-harley-and-keith-johnson-comment

Again, he's questioning the assumption because there's no scientific data, just some "golden ears" claiming he can hear it.

PaulBrewer

madtheory wrote: »

A sceptic questions audibility of jitter:

http://www.avguide.com/forums/jitter-audibility-robert-harley-and-keith-johnson-comment

Again, he's questioning the assumption because there's no scientific data, just some "golden ears" claiming he can hear it.

A sceptic questions everything !

It's not hard to hear a difference - it appears to be difficult to explain it though ...

madtheory

Here's a good explanation of how three DAs deal with jitter when clocked externally:

http://www.soundonsound.com/sos/jan09/articles/daconverters.htm

Getting The Jitters
Each of these units can operate over a wide range of sample rates, covering all the standard frequencies from 44.1 to 192 kHz, but the methods employed vary significantly. When you select 'External Clock' in a cheap and cheerful digital audio interface, the internal clock's oscillator is simply disconnected, and the converters blindly follow the incoming clock frequency, complete with any jitter from the external clock, plus any more that's picked up en route via cables, power-supply noise, and so on — so you could end up with more jitter than you started with!
A better approach is to employ a low-jitter internal clock, which is physically situated very close to the DAC (to minimise the added jitter of cabling), and whose frequency can be tweaked. You then force this internal clock to follow the incoming clock frequency. Typically, PLL (Phase Locked Loop) circuitry is used for this function. This generates an error-correction signal in response to any frequency difference between the internal and external clocks, and uses this to correct the internal one, but the error signal is low-pass filtered, leaving it fast enough to follow any longer-term changes in the incoming clock rate, but slow enough to exclude jitter noise. Effectively, the D-A converter is running on its own internal clock, but the PLL is used to synchronise this to the external clock.
Apogee, as do various other digital audio manufacturers, take this approach further in the MiniDAC by employing a dual-stage clock (essentially the same one that's used in its more expensive stablemates). The first is optimised to track timing variations in the incoming data and store it in a digital buffer, whereas the second further attenuates jitter, as described above.
The Benchmark DAC1 has a rather different design, and uses technology the company have named 'Ultralock'. This uses a very high-quality, asynchronous sample-rate converter that oversamples the incoming data (at any sample rate from 32 to 192 kHz) by a huge factor, and then down-samples it again to an unusual fixed sample-rate of 110kHz, clocked out by an extremely stable, internal crystal oscillator. This, of course, means that the emerging data has been re-computed rather than being 'bit transparent', and also that at sample rates higher than 100kHz data is actually discarded, but Benchmark claim that even 192kHz audio sounds better through the DAC1.
The Lavry DA10 offers three clocking modes: Wide, Narrow and Crystal. Wide is intended for non-standard sample rates, and where they may vary significantly, and it will lock to any sample rate between 30 and 200 kHz, using a sample-rate converter similar to that of the Benchmark DAC1. The Narrow option uses a more traditional two-stage PLL design like Apogee's, with an initial coarse lock followed by a very fine one, and provides bit-transparency, which makes it the most suitable option if you need to run multiple units in parallel (for use in a surround setup, for instance).
However, it's the Crystal mode that will appeal to the biggest proportion of stereo listeners, since this results in the lowest jitter levels. It's an elegant design, whose error-control signal is itself a 12-bit D-A converter with 4096 steps. Incoming audio data is stored in a buffer and clocked out by an extremely low-jitter internal clock, whose frequency is only updated about once every 10 seconds. Each update either increases or lowers the control DAC value by just one step, and this control signal is then filtered to smooth out the tiny steps and remove any remaining interference. The DA10 only supports rates up to 96kHz in the Narrow and Crystal modes — but then designer Dan Lavry doesn't believe that 192kHz sample rates offer any benefit over 96kHz, and there are arguments to support him in this view.

madtheory

So myself and a friend tested the Big Ben, and could not detect a difference. It was a HD3 rig, Adam s3A's, acoustically treated room, two sound engineers, wide selection of good recordings from CD playing back in Pro Tools. We each sat and listened while the other changed clock source, so it was blind A/B. At one point I thought I heard a miniscule improvement, but it was on internal at the time! Call it a statistical anomaly. There was no difference in sound with my Firewire 1814 either.

We had a plan to record the results by miking up a speaker and switching clock from Big Ben to internal, but didn't get around to it, for a personal reason I won't go into here. We did hit on a quick and inaudible way to switch though- disconnect the wordclock cable! PT switches seamlessly that way.

Thanks for the opportunity Paul!

Bluebirdstudios

Don't want to get into a tech debate which few people will have any interest in or understand but do want to point out my observations.

I use big ben to master clock PT 192s, Lynx auroras ,bricasti , Pcm 96 , eventide eclipse and other outboard. I have noticed a difference especially on reverb tails and low end definition. When I had reason to disconnect the big ben even one of my clients had remarked things got a little cloudy !! A real blind test he had no idea of what clock was doing what.
SO is the big ben or other master clock essential for all studios the surprise answer is NO not in some circumstances. Let me explain.

The above was in conditions of regular power supply i.e from ESB Now I've upgraded all my equipment and have on-line UPS systems supplying the studio equipment. This time when disconnecting Big Ben the difference nearly dissapeared.
So in conclusion this master clocks systems will have more effect and better improve things when the whole sysytem is suffering from poor power supply - improve your power supply and the benifits seem to become NEARLY negligible. The quality of power supply is rarely discussed in forums even this thread had no mention of it and all the tech specs you read about low jitter are based in perfect lab conditions where power supply in regulated and clean.

IF YOU HAVE 1 - 1.5 K TO SPEND / SPEND IT ON A HIGH END UPS SYSTEM. ALL YOUR EQUIPMENT WILL THANK YOU AND SO WILL YOUR CLIENTS AND THEN CONSIDER MASTER CLOCKS FOR THAT LITTLE EDGE THAT SOME PEOPLE HEAR AND SOME DON'T !!

kfoltman

PaulBrewer wrote: »

That doesn't mean it's insignificant with real audio. You really don't want to be FM modulating your audio. If you look at harmonics in a real sound (as white noise is kind of a bad signal obviously) and apply even slight FM modulation to it they all turn from nice clean harmonics into wider bands, sort of smearing out the spectrum. It might be an exaggerated comparison, but the effect of jitter does to the spectrum something similar as blur does in graphics.

Anyone wants to do a test?

1. Connect a good sine wave generator to the A/D converter (by "good" I mean no oscillator drift, low noise and precise tuning)
2. Look at the realtime STFT graph of the signal (say, 2048 points)
3. Pick a frequency corresponding to the first harmonic (sampling frequency divided by... 1024 I think)
4. Try to set the sine wave generator so that STFT shows a single peak at first harmonic (not counting negative spectrum and inaudible noise) - it might not be *exactly* sampling frequency divided by 1024 because the clock is never 100% precise, but should be close enough
5. Repeat steps 3-4 for different harmonics

kfoltman

Ok, another test:

- test environment: one PC, two decent quality soundcards (not clock-synced), output of first soundcard fed into input of second soundcard; by "decent quality" I mean anything that has dedicated ASIO drivers (not really high bar to be honest)

- some software tone generator playing a sine wave through first soundcard (the level must be high enough to ensure good S/N ratio and not high enough to cause clipping)

- some software narrow-band notch filter cutting out everything at the frequency of the original sinewave (+/- clock drift effects), with output sent to speakers or some kind of spectrum analyser

If the DAC in the first soundcard or the ADC in the second one is really causing any frequency modulation to occur, it should be visible as peaks in the spectrum, or audible in the speakers (the original tone has already been cut out, so any residue other than noise should be caused by jitter, intermodulation distortion of soundcard's inputs or some other unwanted effect).

krd

I think I know what this question is about.

A good sound card - will have hardware on the card doing the MIDI clocking - a crap sound card will generate the MIDI clock by software - I'm using Ableton on my laptop - my laptop is not crap - it just has a very crap sound card - it doesn't take much for the midi clock to glitch - Because windos is not a Real Time operating system - it can just make decisions to give the midi clock a low priority - and even though the MIDI stream is 36kbps - the laptop can't always handle it.

The reason i say this - I remember years back someone who was hunting down Atari STs - purely to use them to clock their midi - as the ST had a hardware clock on it's sound card.

I don't know if Audio phase arguments come into it - simply because my Roland stuff can't handle that much MIDI information - and there's always a slight delay - (under ten miliseconds) in the Roland stuff playing back - if I send a busy stream to my Roland things they panic. I don't know if anyone could guarantee their devices will clock in some way that the Phase will be perfect.

I know for certain - that if i record a baseline on to my computer from my Roland - then record the same basline - they'll be out of phase. I don't know about softsyths - but I imagine they wouldn't be able to be in phase either, due to the maths that generates the tones.

There's SMTPE timecodes - I've never seen anyone apart from people working on video use it - and they were using these expensive SONY hardware magic boxes for the timecoding. (Video timecode has to be far more precise then Audio - but I suppose with digital editing it's not as much of a headache as the old analog signal)

kfoltman

krd wrote: »

A good sound card - will have hardware on the card doing the MIDI clocking - a crap sound card will generate the MIDI clock by software

MIDI has very little to do with this particular topic.

I mean, jitter exists both with audio and MIDI data, but the scale of the problem is much worse with MIDI. And it's not limited to soundcards - many hardware synthesizers use slow microcontrollers that can't keep up with processing events. Some MIDI events are processed later than they're supposed to, and the delay isn't constant - almost a textbook example of jitter. And those delays may be in millisecond or above range, comparing to audio converter jitter, which is in nanosecond/microsecond range.

On the other hand, some software synthesizers brag about sample-accurate event processing, because events are timestamped using sample position within a buffer. But that's a completely different story...