I'm trying to figure out which clock to use when reclocking my Naim CD3.5.
Now hold on! Before typing, read further!
There will be other mods done to this player, including moving the built-in toroid offboard into a dedicated psu box. This will power only the digital sections. Another trafo will power the analogue sections (via ALW superregs).
A final transformer (3.2VA toroidal 0-18v, 0-15v or 0-12v depending on requirements) will be used to power the clock. It will be rectified by a regular bridge before entering a capacitance multiplier for smoothing. Finally, another ALW superreg will be used to provide the correct voltage to the clock itself. Here's a diagram of the proposed capacitance multiplier:
The clock, powered from the above circuit + ALWSR will be connected into the holes left by the soon-to-be-removed oscillator. Now comes the question. Which clock will perform best under these conditions? If you have figures to back up opinions, that would be great.
Edit: I'm looking at Trichord 4, Audiocom superclock 2 or 3, Tent XO2 (or the bare module), LCaudio, Kwak... any others?
Let the flame-wars commence...
Now hold on! Before typing, read further!
There will be other mods done to this player, including moving the built-in toroid offboard into a dedicated psu box. This will power only the digital sections. Another trafo will power the analogue sections (via ALW superregs).
A final transformer (3.2VA toroidal 0-18v, 0-15v or 0-12v depending on requirements) will be used to power the clock. It will be rectified by a regular bridge before entering a capacitance multiplier for smoothing. Finally, another ALW superreg will be used to provide the correct voltage to the clock itself. Here's a diagram of the proposed capacitance multiplier:
An externally hosted image should be here but it was not working when we last tested it.
The clock, powered from the above circuit + ALWSR will be connected into the holes left by the soon-to-be-removed oscillator. Now comes the question. Which clock will perform best under these conditions? If you have figures to back up opinions, that would be great.
Edit: I'm looking at Trichord 4, Audiocom superclock 2 or 3, Tent XO2 (or the bare module), LCaudio, Kwak... any others?
Let the flame-wars commence...
Why are there so many out there?
Some people like some and others not.
I decided to go with Elso Kwak's clock V7 but you might choose something different. I like how Elso is doing things so that's why I choose his but it really depends on things like that. I still haven't built it but is should come as soon as I finish that parts list.
Something to do would be to see in previous threads and read clock designers' posts. That's what I've done while I was not searching just for that.
Kwak Clock is told better than LCAudio. IIRC Elso designs with his ears rather than with specs.
Some people like some and others not.
I decided to go with Elso Kwak's clock V7 but you might choose something different. I like how Elso is doing things so that's why I choose his but it really depends on things like that. I still haven't built it but is should come as soon as I finish that parts list.
Something to do would be to see in previous threads and read clock designers' posts. That's what I've done while I was not searching just for that.
Kwak Clock is told better than LCAudio. IIRC Elso designs with his ears rather than with specs.
Well "designing with ears" rather than specs is an interesting proposition. To say something that measures less well (has more distortion, noise, less SNR) is akin to saying you favour a certain amount of added coloration to music.
If something measures 'cleaner' then it should sound better, but there is no accounting for subjective 'taste' in this regard. Randy Slone and Douglas Self make this quite clear and they have made some of the best amplifiers available - if you like ultra-clean and transparent sound. Some other designs add a sense of 3D or some other quality which is artificial but this may bolster a system weak in those areas and so seem to be 'better'. In this case, it is the right component for the job, but in direct comparison a poorer one.
However, since electricity is not a flow of electrons around a circuit as we are all initially taught (see this page: http://www.st-andrews.ac.uk/~www_pa/Scots_Guide/audio/part6/page2.html ) but the transmission of electromagentic energy using free electrons, then there is probably a good chance there are effects which standard measuring equipment such as oscilloscopes could not pick up. A such, I am not totally in the 'measurement' camp or 'subjectivist' camp. I prefer to think that there are still many aspects to electricity we still do not understand or cannot measure. So listening is important. However, I would always start from getting the best measuring components together first as I find this goes further to achieving a realistic sound than doctoring a system which is already coloured.
This opens up a can of worms. So I best shut up now.
InifiniteGain
If something measures 'cleaner' then it should sound better, but there is no accounting for subjective 'taste' in this regard. Randy Slone and Douglas Self make this quite clear and they have made some of the best amplifiers available - if you like ultra-clean and transparent sound. Some other designs add a sense of 3D or some other quality which is artificial but this may bolster a system weak in those areas and so seem to be 'better'. In this case, it is the right component for the job, but in direct comparison a poorer one.
However, since electricity is not a flow of electrons around a circuit as we are all initially taught (see this page: http://www.st-andrews.ac.uk/~www_pa/Scots_Guide/audio/part6/page2.html ) but the transmission of electromagentic energy using free electrons, then there is probably a good chance there are effects which standard measuring equipment such as oscilloscopes could not pick up. A such, I am not totally in the 'measurement' camp or 'subjectivist' camp. I prefer to think that there are still many aspects to electricity we still do not understand or cannot measure. So listening is important. However, I would always start from getting the best measuring components together first as I find this goes further to achieving a realistic sound than doctoring a system which is already coloured.
This opens up a can of worms. So I best shut up now.
InifiniteGain
Jitter.
Few people have the equipment to measure it.
..and there's your reason for so many clocks out there!
Anybody can build a clock, listen to it, and say it sounds different!
If you want low jitter clock , find the measurement of interest, and pick the clock with the lowest jitter.
If you want a different sound try every clock you can find, and see how they sound. Then sell the one's you don't want.
Regards,
Ashley.
Hi
How much do you want to spend??
You could have a clock module built by a specialist ( one in the USA but I cannot remember the name ) but the cost for a one-off is astronomical!
The best advice has already been given - buy a few, try them, keep the one YOU like and flog the rest!
Happy listening
Andy
How much do you want to spend??
You could have a clock module built by a specialist ( one in the USA but I cannot remember the name ) but the cost for a one-off is astronomical!
The best advice has already been given - buy a few, try them, keep the one YOU like and flog the rest!
Happy listening
Andy
The problem with designing and measuring clocks is that there don't exist the appropriate measuring tools to measure the jitter in the form we need. (Well sort of.)
Talking about minimising the absolute value of the jitter is of similar use as saying you wish to minimise "distortion." What matters is the distribution of energy in the jitter spectrum. Further, what is even more crucial is the level and nature of correlated products in the jitter spectrum. And this measure only makes sense in terms of the clock's use in a DAC. Indeed it varies depending upon the nature of the DAC, since the DAC topology becomes a convolution of the jitter spectrum. In principle, you could measure and calculate this. It has, to the best of my knowledge, never been done. Desperately difficult, and all you end up doing is simulating the DAC.
The distortion products are also harder to measure, they are not harmonic in nature - so don't show up in a simple FFT based harmonic distortion analysis easily.
So, designing with one's ears is not so silly an idea. But it can be done very carefully, and with an understanding of what is being done. Guido Tent and co describe how they listened to the clock whilst designing their DAC. This is a great way of cutting to the chase and effectively using their ears to measure the auto-correlated jitter components on the clock. It is in some senses a second order measure - as the clock is yet to be convolved into the audio signal in the DAC, but none the less a great way of looking for signal correlated jitter energy, which is the most important thing to be rid of.
The problem with efforts like Doug Self's work is that it is not clear that they have chosen the correct distortion metric to design to. Simple harmonic analysis of the signal is not how our ears work. As I like to point out is is quite simple to construct pathological examples of distortion that will be represented as an almost imperceivable increase in the noise floor in THD+N measures, but will actually be quite audible and disturbing. A DAC is quite capable of creating such distortion. Indeed almost all the nasty distortion products will be of such a nature.
Talking about minimising the absolute value of the jitter is of similar use as saying you wish to minimise "distortion." What matters is the distribution of energy in the jitter spectrum. Further, what is even more crucial is the level and nature of correlated products in the jitter spectrum. And this measure only makes sense in terms of the clock's use in a DAC. Indeed it varies depending upon the nature of the DAC, since the DAC topology becomes a convolution of the jitter spectrum. In principle, you could measure and calculate this. It has, to the best of my knowledge, never been done. Desperately difficult, and all you end up doing is simulating the DAC.
The distortion products are also harder to measure, they are not harmonic in nature - so don't show up in a simple FFT based harmonic distortion analysis easily.
So, designing with one's ears is not so silly an idea. But it can be done very carefully, and with an understanding of what is being done. Guido Tent and co describe how they listened to the clock whilst designing their DAC. This is a great way of cutting to the chase and effectively using their ears to measure the auto-correlated jitter components on the clock. It is in some senses a second order measure - as the clock is yet to be convolved into the audio signal in the DAC, but none the less a great way of looking for signal correlated jitter energy, which is the most important thing to be rid of.
The problem with efforts like Doug Self's work is that it is not clear that they have chosen the correct distortion metric to design to. Simple harmonic analysis of the signal is not how our ears work. As I like to point out is is quite simple to construct pathological examples of distortion that will be represented as an almost imperceivable increase in the noise floor in THD+N measures, but will actually be quite audible and disturbing. A DAC is quite capable of creating such distortion. Indeed almost all the nasty distortion products will be of such a nature.
Hi
When all said and done, HiFi is about listening to sound.
If you are happy with the end result - fine!
Who am I to comment - or anyone else for that matter?
As with other changes to your system, you are going to have to try and try again until YOU are happy. But beware, you could end up back at the beginning!
Andy
When all said and done, HiFi is about listening to sound.
If you are happy with the end result - fine!
Who am I to comment - or anyone else for that matter?
As with other changes to your system, you are going to have to try and try again until YOU are happy. But beware, you could end up back at the beginning!
Andy
Hi DragonMaster,
If I may quote myself....
I just want to clarify I was not applying Randy Slone's or Douglas Self's uncompromising attitude towards "listening". I should have merely pointed out good measurements (if you know which ones to take) are IMHO the best place to start. If you look at high end audio kit, which typically does sound better as you go up the price ladder (and I should know having reached a £25000 system) and there is the tendency (that is, generally but not always so) for these systems to measure better than their cheaper equivalents (except with valve/tube equipment where some cost a fortune and measure terribly). From that observation, one must conclude that their is a fairly strong correlation between measurement and sound quality - but it does not mean that it is definitive. Since I do not belive it definitive, some aspect of "listening" MUST come into it.
The reason I said what I did was up until then there were not many postings on this thread and I wanted to stimulate the debate as I am interested in the opinions and science of what makes a good DAC clock and why as much as the thread starter. You can never get all views from all sides as well as you can when you bring up something controversial
InfiniteGain
If I may quote myself....
I just want to clarify I was not applying Randy Slone's or Douglas Self's uncompromising attitude towards "listening". I should have merely pointed out good measurements (if you know which ones to take) are IMHO the best place to start. If you look at high end audio kit, which typically does sound better as you go up the price ladder (and I should know having reached a £25000 system) and there is the tendency (that is, generally but not always so) for these systems to measure better than their cheaper equivalents (except with valve/tube equipment where some cost a fortune and measure terribly). From that observation, one must conclude that their is a fairly strong correlation between measurement and sound quality - but it does not mean that it is definitive. Since I do not belive it definitive, some aspect of "listening" MUST come into it.
The reason I said what I did was up until then there were not many postings on this thread and I wanted to stimulate the debate as I am interested in the opinions and science of what makes a good DAC clock and why as much as the thread starter. You can never get all views from all sides as well as you can when you bring up something controversial
InfiniteGain
When it comes to components and circuitry that deals with amplitude, ie in the analogue domain, listening is everything. For example, the little maroon 47uf/40v electrolytic caps that Naim used in their older amps are not exactly the best cap in the world, however they take some beating in listening tests.
These rules don't apply in the digital world of time. Properly implemented, accurate clocks must be better than sloppy inaccurate clocks. If these highly accurate clocks sound bad then it could be that they are showing up a DAC deficiency or weakness in another area such as circuit layout, grounding etc.
So, perhaps my question should have been: what is the most accurate clock available?
These rules don't apply in the digital world of time. Properly implemented, accurate clocks must be better than sloppy inaccurate clocks. If these highly accurate clocks sound bad then it could be that they are showing up a DAC deficiency or weakness in another area such as circuit layout, grounding etc.
So, perhaps my question should have been: what is the most accurate clock available?
HI.
WHAT IS THE MOST ACCURATE CLOCK?
The 2 factors which have to be defined here are :-
1. Frequency Accuracy i.e if I want 1Mhz, what difference would a frequency of 1.0001Mhz make?
2. Frequency Stability i.e. 1Mhz +- 0.0001Mhz measured both as short term (jitter) and long term (drift)?
I used to work with High stability time standards which used Xtal ovens to minimise drift - this practice seems to have disappeared.
The Xtal was also mounted in a 'vibration-proof' mount.
Andy
WHAT IS THE MOST ACCURATE CLOCK?
The 2 factors which have to be defined here are :-
1. Frequency Accuracy i.e if I want 1Mhz, what difference would a frequency of 1.0001Mhz make?
2. Frequency Stability i.e. 1Mhz +- 0.0001Mhz measured both as short term (jitter) and long term (drift)?
I used to work with High stability time standards which used Xtal ovens to minimise drift - this practice seems to have disappeared.
The Xtal was also mounted in a 'vibration-proof' mount.
Andy
Indeed. It is a corollary of what I believe should be called Galbraith's Law. Once any metric becomes well known it becomes useless. He wrote in terms of metrics of the stock market.
OK, back to the plot. Clock accuracy. Exactly the problem. What do you define as accurate? It is well understood that long term frequency accuracy (where long term might only mean minutes) is of no interest in music. It just doesn't matter. Certainly obtaining a high stability crystal is of no value. Clocks used in communications systems need to have very good performance in a manner of greater interest to us. The term "phase noise" is another term for timing jitter. As the phase noise increases it degrades the ability of the channel to carry encoded information. So the comms guys take an interest.
But for audio we have it harder. The absolute amplitude of the phase noise is less important than its spectral distribution, and even that is less important than the presence of significant auto-correlated noise, and if that auto-correlated noise happens to be correlated with the audio signal the DAC is reproducing it gets grim. Such correlation may happen though a number of mechanisms including airborne vibrations coupling into the crystal, oscillator power supply modulation, comparator PS interference, logic gate issues - especially involving gates carrying serially encoded audio data, and so it goes.
OK, back to the plot. Clock accuracy. Exactly the problem. What do you define as accurate? It is well understood that long term frequency accuracy (where long term might only mean minutes) is of no interest in music. It just doesn't matter. Certainly obtaining a high stability crystal is of no value. Clocks used in communications systems need to have very good performance in a manner of greater interest to us. The term "phase noise" is another term for timing jitter. As the phase noise increases it degrades the ability of the channel to carry encoded information. So the comms guys take an interest.
But for audio we have it harder. The absolute amplitude of the phase noise is less important than its spectral distribution, and even that is less important than the presence of significant auto-correlated noise, and if that auto-correlated noise happens to be correlated with the audio signal the DAC is reproducing it gets grim. Such correlation may happen though a number of mechanisms including airborne vibrations coupling into the crystal, oscillator power supply modulation, comparator PS interference, logic gate issues - especially involving gates carrying serially encoded audio data, and so it goes.
rbroer said:Reminds me of all this 6-sigma stuff, measuring "Quality".
A well known joke is:
"Tell us how you measure, and we will perform accordingly"
Quality is very hard to define and very relative
Don't knock all this "Quality" stuff. I've spent many pleasant afternoons sleeping through seminars on it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.