Hi all,

Recently, we decided to set up a test with two identical cd players where one player should have a low jitter clock "upgrade". Well, that will not be the topic of discussion here. Prior to that test, I wanted to perform some other tests. The basis is that the standard crystal is replaced by an HP8640B signal generator with frequency modulation input. This FM input is then modulated by several signals to investigate the effect on the signal that comes out of the cd player.

The cd player-under-test turned out to be a Yamaha CDX-496 and is shown here:


with the following relevant circuitry:


Basically, the crystal is removed as well as the capacitors C43 and C44. The logical ports of the IC are connected with 1M to ground and a center tapped mini-transformer is used to couple the external 16.9344MHz sine wave. A 820Ohm resistor is used to load the transformer characteristically. This is shown here:



The cd player runs if I set the ouput level to 2.1V. The display of the generator is given here:


The FM input of the generator is limited to 1% peak deviation and that results into 160kHz of modulation. For the test I set the modulation to the max. This is a situation which does not resemble practice but it is only meant for demonstration purposes. Furthermore, I have chosen for 20Hz as modulation frequency. That is my lower limit of the LF generator.

This extreme form of jitter can be heard as follows.
Clean track: no jitter
Jittered track: 20Hz FM modulated carrier with 160kHz peak deviation

I have no idea what the jitter is in terms of ps. May be someone here can help me out.

For me an important question was if jitter introduced harmonic distortion or intermodulation distortion. For that purpose, I used a test cd with several sine waves and some two-tones. To be precise:
* 399Hz,
* 3990Hz,
* 8819.5Hz,
* two-tone 15435 and 17640Hz,
* two-tone 250 and 8020Hz,
* two-tone 60 and 7000Hz and
* white noise.

As far as my old laptop can do the job, I made a movie of the recorded response. The sequence is the same: first the clean track and after that the jittered version. Here is the movie. As far as I can see, there is no harmonic distortion nor intermodulation distortion. Sometimes the lower part of the graph is blanking in the movie, but that has to do with my old laptop. Let me zoom in at two cases, the first for the 399Hz case:



and with the jitter:


, where it is obvious that the level of (deterministic) jitter is gigantic.


The two-tone with the two high-frequency signals is plotted with a linear x-axis for clarity:


and with the jittered clock:

, where the same modulation-effect can be seen but no intermodulation. There are no significant components at 2*f1-f2 and 2*f2-f1.

For the ABX enthousiasts, I prepared some fragments of Tool (Vicarious). The first fragment is the clean one: no jitter. The second one is jittered with itself. That is, the output of the cd player is now connected to the input of the FM modulator. The peaks of the musical content is set to 160kHz peak deviation. In other words, the clock is heavily FM modulated with the output signal of the cd player itself. You can find it here: with jitter.

Regards,
Jacco

I'm not sure I fully understand. Please correct me where I am wrong:

You created an artificially high level of jitter and demonstrated it is audible?

The jitter is expressed in picoseconds (ps) because it is very small.

This is an example of very high jitter:

Yes, that is one thing. The other thing is that it does not create harmonic distortion. And the next steps are to investigate where the level detection is and to come up with a rule of thumb to predict the amplitude of the unwanted signals.

Regards,
Jacco

What I see from your experiment is pretty much what I would expect: the signal gets modulated by the jitter. It turns out that for small jitter (frequency modulation), then the effect is actually very similar to amplitude modulation except that the additional signal introduced is 90 degrees out of phase.

So if you have two tones at 1 kHz and 3.5 kHz and you apply FM modulation with a period of 20 Hz. What you would get is additional tones at 980, 1020, 3480 and 3520 Hz. As max jitter deviation becomes larger, you would start seeing other tones at +/- 40, +/- 60, and so on.

Now, when you say you're not seeing inter-modulation distortion, I assume that you mean that there's no "interaction" between the tones of your original signal. That can actually be demonstrated fairly easily. Say that your input signal is x and the output of the modulation is J(x), you can see easily that J(x+y) = J(x)+J(y), i.e. it doesn't matter if you add two signals before or after the jitter is applied. Hence, the effect of the jitter can be separated and there cannot be interactions between x and y.

In terms of perception, the spectrum of the "distortion" caused by the jitter is very similar to that of the original signal. That means that it would tend to be simply masked until either 1) the amplitude of the jitter causes the distortion to go above the masking threshold 2) you start actually noticing the frequency changes 3) the speed of the jitter spreads the harmonics far enough to destroy the masking effect.