Do you hear any difference?(master clock)

I've read countless people saying a Grimm clock is a betterizer. I'm tempted to check it out, BUT when I tested out clocking my Lynx Aurora (n) off of the high end clock of the HEDD Quantum, I found that it sounded wider at the expense of losing a bit of the center. But perhaps an external clock is better suited to a multi unit setup like yours.

Phone speaker I liked the first one I heard: aurora internal daisy chained.
Great track, had me wondering how the engineer dialed the mix so good.

the next one on the list, I no longer heard the engineer, instead I noticed some sharpness that was razor clear but kind of bothered me. The rest were similar to this, but seemingly different balances. A couple seemed close to similar feeling of the first.

I played the first again, once again same marvel at the mix feeling.

So I’m crazy and would fail a blind test ;P

e: I probably should checkout high res files to rule out streaming funkiness. Might not get to for a few days

This one definitely wasn't as fun to make as doing a compressor shootout but I hope everyone finds it informative!  Thanks for taking the time to check it out.  

This is a really cool test and absolutely confirms the kinds of differences I've heard between clocks in my studio.

For me, the Mutec as the master clock is the clear winner, but the Aurora/Mutec and Avid/Mutec aren't far behind.

To describe what I'm hearing, I think the Aurora on its own sounds just a hair too crispy up top for my taste — pretty much exactly what I'd expect from my Aurora 8. The Mutec softens the highs and enriches the lower mids on the Aurora. Definitely a "betterizer" in this case for sure.

The Mutec master sounds the most relaxed and natural to me. On the Aurora, by contrast, the transients hit a little too hard and everything feels just a little too stiff. Again, I would expect the same from my Aurora. But it's definitely not bad or anything. If the Aurora was my only point of reference, I'd say it sounds great.

But yeah, the Mutec is a keeper, IMO.

Ok so I downloaded the high res files, used headphones. Great track by the way, excellent tasteful production.

I still liked the aurora internal daisy chained, but also the ones with mutec were great but different.

So I went ahead and did some blind testing, which is never fun, more like hell.

Switching back and forth, like all blind tests I didn’t notice any differences at first. This could have led me to believe there’s nothing there if I would have stopped.

So I decided to focus on two of my favs and just swap between them. auroral internal daisy chained and Mutec internal.

After about 20 minutes of this, I did start to perceive some differences. Reminds me how hard blind testing is, your ears and brain have to kick into a higher gear or there’s no way to pass.

The hard part is staying focused on the part of the sound that has differences without wanting to just tune out and enjoy listening. If I felt confident I perceived the difference, I’d usually pick the right answer. When is lose that difference, I basically was just guessing.

So anyway, the difference I perceived blindly that wouid give me the most right answers: the way the mutec followed the vocal articulation seemed more linear and open. The low mids also felt ever so slightly more wider and open.

I thought I liked the aurora better but after hearing this part of it blind I think the mutec has a little something better going on.

Again, this is just shure headphones and I doubt any scientist would buy any of the math. But inside when I was pretty confident in a choice I was usually right. That’s enough for me

just wondering, if you were to subjectively quantify this difference, are we talking 1%, .01%: what?

I understand the search, but if the difference is somewhat negligible, perhaps listen to a variety of music you might work on, before making a final decision?

After all, this dough could be spent on a major rgo wings and beer post covid blow out party: it’s all about priorities !

Nope, no difference.

Remember, most modern converters have natively oversampling converter ICs. This means they need a much higher clock rate, typically 128-512fs, or 128 to 512 times the desired sampling rate. If you input a Word clock at the sampling rate, it's usually used as a master clock for a PLL to multiply up to the required frequency, which will always ADD jitter.   Sometimes these converters have PLLs internally as well, and sometimes they do fancy things like double sampling..  Most of these fancy tricks are not clear to the user because they're used to give small amounts of performance increases to make the datasheet specs look slightly better than the competition..  But some like double sampling are very sensitive to jitter on the clock.

Oversampling does little to affect jitter.  While oversampling looks like frequency division on paper, since the clocking is the absolute in the equation, jitter remains mostly the same.  Oversampling helps quantization noise greatly though, which is why it's standard in all sigma-delta converters today.

Also, jitter has a frequency response if you were to look at it in frequency-domain with a FFT and a little math, and you'd get Phase Noise.  Phase noise is generally worse the lower in frequency as the randomness and magnitude of the frequency deviation is worse as you get closer to the fundamental frequency of the clock.  These frequencies are also well within the audio band as PN is usually measured from 1Hz to 1MHz.  

So the takeaway is that "jitter" rarely explains the frequency component of the noise.  You could have a very small amount of jitter integrated between 1Hz and 10KHz but have a huge spur around 15KHz and your "jitter" number would be small.  You'd have strange artifacts in the audio, but your jitter would be small..  

But there's one thing that I REALLY hate about clocking that nobody talks about..  And it's related to spurious signals mentioned above.

The waveform.  

Clocks are generally square waves..  Mathematically, square waves are an infinite number of harmonically related sinewaves.  Why does this matter?  PLLs are not smart.  They take an input clock (called the reference) and compare it to a divided down version of their output signal.  The difference between the two causes the PLL to push or pull the output frequency until it's the correct multiple of the reference (called being "locked").  

Once the PLL is locked, harmonically related output and input signals will interact with each other.  If you have a reference frequency with harmonics that extend into the frequency of the output signal, then they can add spurious signals back into the phase detector inside the PLL.  These small differences in frequency between the two will cause a mathematical frequency mixing and you'll get spurious signals at intervals of the reference frequency (called reference spurs) as well as internal division frequencies (colloquially called PFD spurs).  PFD spurs shouldn't be a problem here since we're not working with broadband signals that need the PLL to work over a large range of frequency steps.  Engineers who are used to working with PLLs would inherently add lots of filtering to the reference and output signals to reduce the number and magnitude of the harmonics.  PLLs don't care if the signal is square or sine, but the results are drastically different.  I ALWAYS try to reduce the harmonics to less than an octave if possible.  Most audio design engineers don't have the first clue about this stuff.

However, reference spurs can be a huge problem here too. 

If a PLL is locked to a 44100hz Word Clock signal but the PLL loop filter and other internal division settings have allowed it to lock +1KHz offset at 512fs(22.5792MHz), then the output of the PLL will have a spur roughly 1KHz from the incident signal and the output will have a spur at 22.5792MHz as well as the output signal at 22.5802MHz.  That spur will mix down through oversampling into your audio band.

The usual way to fix this with oversampling converters is to have a harmonically UNrelated clock frequency that is outside the band of interest, either lower or higher than the frequencies you're wanting to sample, that way any spurs will be outside the band of interest and can be filtered easily.  With Word Clock as your input, that's impossible to do!

So besides the whole issue with existing jitter on the Word Clock input, any system that uses Word Clock as a direct reference for a PLL risks adding spurious signals to the output as well.

So what do these spurious signals sound like when mixed with audio?  They tend to add higher frequency harmonics (similar to an aural exciter effect) which makes audio sound "more alive".   Half the time folks "hear a difference" from an external clock, they're hearing the distortion caused by harmonics.

I'd be totally suspicious of any external clock that makes this sound "more clear".

This concludes this episode of "Why Internal Clocks Should Always Be Used And External Clocks Are A Waste Of Money".

No difference in discernible sound is not the same as as a null.  Nulling is extremely hard to do as even fractions of a dB of difference in level or phase between tracks can lead to orders of magnitude of difference in nullification.  It's not a 1:1 ratio at all.  

I don't have numbers for audio but for RF I'm generally worried about 0.1dB of gain and 0.1deg of phase making the difference between nulling a signal down to -60dB vs. -30dB.  it's ridiculously sensitive.

So yeah, I don't hear any "discernible" difference, which is actually a good thing here.  It means that the external clock is not destroying your signal.  It's just not making it any better.

Also,

If you're struggling to hear tiny differences, it's probably not making as much difference as you think.

If you're going to spend a ton of money on something and you can't really tell that it's making a difference, then it's probably wishful thinking that you're hearing.  Just sayin'.  This has bitten me so many times when doing my own testing on stuff.

Let me answer you by asking you a question..  

Why do you need to clock multiple interfaces with one clock?  Is there a technical reason they need to be synchronized?

Try it!  

But, I think you'll find that the playback is fine.  If your devices are interfaces, then the clocking is relative to the unit, and audio is usually buffered.  If it's simply a converter (which are rare these days), then it's usually relative to the datastream like AES or ADAT (both are self-clocked and don't require external sync anyway).

However, I found this article just now.  It has some generalized info but it also has some tests and graphs for internal/external clocking performance.  The long story made short is that external clocking made almost every converter worse, and the best case was that it broke even.  That's not a great return on investment.


www.soundonsound.com/techniques/does-your-studio-need-digital-master-clock

Maybe let someone else create a comp of the two mixes and then see if you still hear the edit points?

OK, here's my comp of all the files:


drive.google.com/file/d/119YRykUQRdLGUOfzvya_RdZd6YmUWMFh/view?usp=sharing