While “better clock” may mean to some more outputs, easier user interface, better reliability and so on, the question here, thus my reply is about clock performance.
From a technical standpoint, a better clock means first and foremost lower jitter.
I have not looked at all the clocks on the market, and this forum is not about gear recommendations, nor is it about reports about listening to specific gear. This is a technical forum.
The most important clock specification is jitter. Given a choice for one clock specification number, it would be the jitter spec. I have yet to see the jitter specification on any of the clocks that I looked at. That makes clock comparison very difficult.
Of course the question of which clock is better should be viewed from the perspective of how good should it be before one notices no difference (the point of diminishing returns). So here is some guidelines: A clock yielding say 1psec of random jitter will perform no better then a clock with 20psec, because by the time you get to the converter, the jitter level will be much higher then either.
There are various sources of jitter, and the final overall jitter is not the sum of the individual sources. Random jitter accumulates with accordance to RSS rules (root mean square).
Say you have 2 sources of jitter, one is 10psec, and the other is 50psec. The overall accumulated outcome is sqrt(10*10+50*50) = 50.99psec
Say you have 2 sources of jitter, one is 1psec, the other is 50psec. The overall accumulated outcome is sqrt(1*1+50*50) = 50.01psec
Therefore, when the clock is driving a system with 50psc of jitter from sources other then the clock itself, the difference between using a 1psec clock and a 10psec clock has no practical advantage (less then 1 psec).
My example demonstrates that one is best to be aware of the practical range of the jitter levels we deal with in the real world. I can assure you that by the time you get the clock signal through a cable, some IC for receiving the signal in a different chassis, and then followed by a phase lock loop circuit, your starting point is already so high that any of the commercial clocks I looked at will yield very similar results.
Of course my statements above are contrary to typical audio marketing , where the goal of selling gear is often a source for much hype and misinformation. Given that most (if not all) of the clock makers do not even publish the basic specification – clock time jitter, the conversation turns into how things sound, which can be very subjective and self serving, thus not suitable to this forum.
It would be good to keep in mind that a word clock is a fundamentally a very simple signal. It goes up, then down, then up then down… The best word clock is the signal that does its repetitive task precisely, where the timing of each “go up then down” cycle is as identical to any other cycle. The whole concept of digital conversion is based on fixed cycle time for each sample, which is in fact no jitter, to some practical point.
A “better clock” is a clock with less jitter, to a practical point of diminishing returns.
There are those that claim that some clock or another can “do something” to improve sound. That is marketing at it’s low point:
First, the clock box does not have any knowledge of what the converter does. The clock sends the same “up then down” signal to a converter, be it a near perfect converter, a poor converter, converter with power off or what not.
Second, the clock box does not have any clue regarding the signal (the music) that the converter is processing. The converter may be playing loud rock music, soft vocals or it may be playing no sound at all (between music cuts).
Third, different converters react differently to clock timing variations. Differences in clock recovery circuits (PLL’s) and converter architecture simply shoot down the idea that different converters would react the same way to a variation in clock signal.
Any of the above 3 points is sufficient to end any claims about clock improving sound.
Such claims are particularly disturbing.
My comment to people that like to listen to gear: the claims regarding listening to clocks is NOT analogous to say someone liking a tube sound (or a certain mic, speaker and so on). Contrary to the clock claims, the tube “does know” about the signal, it “processes” the signal (the music), it does not have to guess what is playing elsewhere. A tube does not “operate” on some other chassis it has no idea about. It is in series with the signal.
If an engineer were able to improve sound of a converter by altering the clock timing, they would not do it at an external clock box. The engineer would be able to overcome the differences between transmission links and AD converters by applying their “technique” directly to a known AD chip of their choice. So why do we (converter hardware designers and IC designers) stick to a fixed minimal jitter concept? The answer is simple: we do not know how to improve music by forcing some timing alterations. I can alter predictably the sound of say one tone, but music is varying all the time, and the sound alterations due to a proposed clock alteration is simply UNPREDICTABLE in the context of music. As a rule, modulating the clock (any deviation from ideal no jitter performance) will introduce uncontrollable distortions and or increased noise.
I can not argue with tastes. But claims regarding sound improvements based on a single (or limited) listening tests to given clock and AD model(s) should not be automatically extended to become general statements.
The sonic alteration of a tube or a transformer is PREDICTABLE. We pass the signal through the tube, and it is the INTERACTION between the music signal and the tube known characteristics that make the sound.
In the case of the clock, the clock box is 100% disconnected from the music, and the proposed improvement is done with 0% knowledge of both the hardware and the sound that goes through it. In other words, we are down to a fundamental claim that a clock other then a “perfect metronome” can be better. Such claims do not stand up to professional scrutiny.
Having said the above, there are many cases requiring external clock sources and clock distribution. With no published data about jitter, and with the knowledge regarding the other factors influencing system jitter, my own personal tendency is to look at feature and costs.