Pink Paper #002 – The Future of Clocks

I read it, interesting findings (listening to reverb tails etc), however if still subjective.

Sure. Digi brought out the 192, most popular interface/converter around since Moses was a small boy and still quite popular. Many in use, maybe still the most widely in use.
People complained about sound and jitter.
Blamed it on internal clock.
New clocks came out like Big Ben.
External clocks seemed to improve sound
Things synced together better.
Was it really as bad as we thought?
Maybe not.
Digi comes out with own clock, the Sync.

10 MHZ is the magic number.

There are a few of the highlights.

The problem was and still is People spend big $ there for it sounds different and different must be better and the more they spent the better it must sound.
The thing is there are instances where a master clock can help, the problem is 9 out of 10 probably don't need it but the guys selling it make them feel like they need it .
I do run an old micro sync but it's here to sync the analog 8 track !

The Digi HD interfaces can be running on external clock even though the settings appear to say internal. I haven't used one in years but there is a way to set a single HD interface up to actually run on its internal clock and it sounds better that way than with the exotic external clocks.

All external clocks increase jitter. Some designs force the jitter spectrum into something that sounds better. The same clocks make the best converters sound worse. Apogee wanted me to try out a Big Ben until they found out I was using Benchmark converters and backed out.

If you drink the more expensive monkey-poop coffee, your jitteriness would be forced into a spectrum that sounds better.

Son of a BISCUIT, spock! An actual flux capacitor in its natural habitat: The analog synching clock!!

Considering that they explained rubidium clocks incorrectly, I'm not sure I'd even take the rest with a grain of salt...



Also, 10MHz is a standard for most test/measurement equipment as a synchronizing frequency, not a clock source...  



Most products with modern DAC/ ADC chips have internal 256fs and 512fs frequency sources for sample clocks, such as 22.579MHz for 44.1K  and 24.576MHz for 48K sample rates, etc.  Using an extenal source only acts as a new reference for those clock sources, NOT as a replacement for those internal sources.  If your internal 512fs souce is a PLL using a 10MHz clock that happens to be jittery, then an external source with a MUCH lower jitter could still yield a better output from the PLL..  Or just a more harmonically pleasing profile of jitter..

Because 10Mhz is not suitable for direct clocking, it must be used as a reference for a PLL of some type.  "Audio grade" PLLs have always had notoriously poor jitter performance, but up until the last few years, nobody really cared.  It was more about synchronizing multiple devices in systems, rather than providing a more robust clock source.

But PLL wise, a PLL can never provide better performance than the reference clock that drives it.  There are a few PLL tricks though:

1. Higher frequency reference, with PLL set to internally divide reference frequency..  Dividing a clock down ALWAYS divides jitter.  I ALWAYS use as high of a clock frequency as I can, and use the internal PLL dividers.  This tricks the loop into having less jitter than the reference natively exhibits.

2. Use a PLL with significantly higher native output frequency but with output frequency dividers.  Again, jitter divides with output frequency division.

3. Isolated LDO supplies for the PLL and it's various VCC inputs like CP, core, etc, and separate LDOs for reference oscillator, etc.

4. Extreme decoupling.  Largest capacitance values in smallest packages provide lowest ESL/ESR.  Decade decoupling(100pF+0.01uF+0.1uF) is really poor for high speed decoupling but seems to have a loyal but extremely misled following.

5. Clock/transmission line theory..  One that gets VERY little use or understanding.  So what if your fs clock is 0.1ps jitter when your L/R clock into your DAC IC is skewed 1ps from mismatched-length transmission lines..  Now you'll get jitter-like anomalies, but never be able to "fix" it with another clock scheme.



There are probably hundreds of small things that most designers overlook that could help in audio clocking...

But then again, there are numerous studies that suggest people actually like non-random jitter due to the effect it has on creating eve-order harmonics..  (tube like richness, and such..)

For example, I have measurements that show my converter clocks being 20-40% better than another, very similar, group of products..  But the outcome is that some prefer my design, some prefer the other products, but I digress as this really boils down to taste.



So I have to ask, where are the logarithmic phase noise plots that visually corroborate these opinions?  How do we KNOW that what the authors are hearing ISN'T a more even-harmonic jitter content, rather than a CLEANER content?

We don't.  We just have to take it on faith that their opinion is somehow a proven quantity without need for scientific proof..

Let's just say, that if I presented this as a whitepaper for my job/company/industry, I'd immediately be asked to go back and bring proof..  There are no "opinions"..  Only proof, when it comes to quality analysis of a product.

Well, it just so happens that I do this stuff day-in, day-out..  At least let my toiling and drudgery be education for the masses!