Any interested in an external clock source project?

Just out of curiosity.. why would you even bother with designing an external clock if you don't believe in external clocking?
Anyway, I think the only time when someone would actually benefit from having an external clocking is when there are multiple devices running in parallel. Even then it's a compromise, not an improvement as someone mentioned..

...nah, I'll stick with the Svartbox.

True. If there is a demand and you can do a better job than what's currently available in the market then it makes a perfect sense to seize the opportunity. Just thought it was a bit strange for someone to say that they don't believe in a product they are about to design.. btw I absolutely DESPISE BLA and 'almost' all of their offerings (they do have their own designs and I respect those but nothing else) so in a way I am hoping that you would come up with a product that can compete with theirs which in my opinion are being sold for a ridiculous price. 

Didn't offend me at all.  I think these things are tools for people to use.  Sometimes a tool can make a big difference, sometimes no difference, sometimes it just gets in the way.

I think the lower the quality of the converter system (I.E., more consumer grade), the more it'll be helped by clocking.  The better, the converter system (I.E., true professional) then the less something will help, to the point of hurting.

I think learned people who talk about internal clocking being better than external clocking are correct for the most part, but you have to assume that since these same folks are professional designers, they are speaking of professional grade gear..  Not prosumer or commercial gear.

This is generally what I believe, that a professional level converter system will have little to no difference with external clocking, if it's not slightly hurt by it that is. I still believe that a lower end box will still have some performance increase from better clocking.

But you also have to be aware that the design and implementation of a piece of gear is critical if you are going to analyze the performance possibilities.

If you have a converter with good quality conversion chips, but poor WC buffering/up-conversion, then you'll never realize the potential of the converter ICs when using external clocking, and you'll almost definitely have either equal or worse performance.

However, if you have so-so converter ICs, but excellent WC buffering/upconversion, then you might have something that results in much better results.

Interested




cheers

Wiz

So to start the discussion, there are many ways to get clock sources. I will discuss a few..

PLL/VCO: An oscillator of a common value is used as a phase reference for a PLL who's job it is to control the frequency of oscillation for a VCO. The PLL does this by using a complex divider/multiplier network to match the phase between the output of the VCO and the reference oscillator. The jitter of the oscillator adds to the jitter of the PLL/VCO system. These are good at suppressing close-in phase noise, but the loop filter of the PLL can allow mid-jitter to be high.

Summary: PLL/VCO is not the best, but certainly good enough when done correctly.  With these, clean/proper design, layout and parts choice is key for performance.  These are plentiful for high frequency clocking, but get increasingly hard to find good audio frequency products.



Clock oscillator/D-type Flipflop:  You can cascade D-type flipflops with their /Q (NOT Q) outputs fed back to the input, to form chains of divide-by-two.  You can take a common audio clock of 22.5792MHz and divide it with 8 of these sections to create 88.2KHz.  9 sections make 44.1KHz, etc.  These work well, and are commonly used in the world.  The CMOS flip flops don't add much "jitter" to the clock during division, but they can add random noise depending on the decoupling/layout.

Summary: Decent performance and cheap.  Can be sensitive to noise.  Some people swear by these as the best option for audio clocking.



Oscillator/ripple counter:  You can use old counter IC's to "divide" down the clock.  Essentially the clock is just a gate that switches bits out of the counter IC.  These are, for the most part, extinct.  Most of the other options like PLLs, DPLLs, frequency synthesizers, DDSs, and all kinds of other programmable oscillators and custom devices have taken over as their availability grows and costs drop.  You could make an asynchronous ripple counter from JK flipflops, but they are slow as each stage must flip before the next.  Again, not a big deal to a master clock though.

Summary: Nah, no reason to even look at these.


Oscillator/ASIC: Now we get to the good stuff.  Custom frequency divider IC's.  They take all the best stuff and package it.  Most of these are unknown types of dividers due to proprietary designs.  Smaller packages mean less noise, less power, less parasitic issues, for cleaner signals and low jitter.  They cost more, but deliver more, down to single picosecond jitter addition.  The only tradeoff is they typically are only available in high speed interfaces such as LVPECL and CML.  This makes adapting them to external signals hard.

Summary: This is likely the path I'll choose to investigate since I'm aware of RF divider ICs that have some of the best jitter specs, and are relatively cheap enough to use.

Now to discuss features..

I'd reuse the LCD and a lot of the firmware from the converters..

WC outputs.. How many? 2? 4? Do people even use more than one of these?

SPDIF/AES? I honestly don't know why anyone would use these just for clocking..  The AES stream is a function of the audio and the system clock..  So i'd have to generate a dummy stream of some kind for it to work properly.. I hope the other guys do this..

What else is even needed for a clock??

I would go at least 4....perhaps 6.

Don't forget a good PSU:)
You thinking internal or some sort of external switching supply?

Perhaps, I misunderstood. I just meant 4 or 6 bnc connectors.

I was not a part of the test, but I was told that the head engineer did a sync test and was able to hear 3 WC jumps through devices. They also did loop tests after and found that it was not sample accurate. Now this was a few years ago. The studio had one clock distribution, and was using a mix of spidif, aes, and ADAT I/O outboard, set up as inserts in PT HD. So they did not loop the WC through the outboard.

most of the people that are gonna buy this (basically forum members here) have one interface with 8, maybe 16 channels. they don't have 96 channels of conversion that need clocking.