In this series’ episode around d/a converters, we’ll look at jitter. What exactly is jitter? What types of jitter are there and, not unimportantly, what are the consequences of jitter when it comes to music reproduction?
At Alpha Audio, we are quite regularly measuring jitter (and phase noise). The reason is that we can fairly easily see the impact of external factors based on increased or decreased jitter values. For this we use both the Wavecrest SIA3000 and the Aeroflex PN9000 (phase noise). Both devices have advantages and disadvantages, but together we get pretty good insight into what is happening.
What is jitter?
If we simplify jitter, it is nothing more than a deviation in a (clock) signal that is supposed to be constant.
So jitter is especially applicable to signals that time a digital circuit. Think of providing a timing signal for a dac chip. If this signal – or better: clock signal – should operate at 11.2 MHz, for example, but there is a small variation in it, the dac chip will sometimes perform a conversion too early and sometimes too late. And that causes distortion of the analog audio signal.
Jitter is usually measured in picoseconds or sometimes even femtoseconds. These are very, very small deviations. However, we now know that it is audible. We can hear deviations of only a few picoseconds.
Now, in specification sheets we often see only one value for jitter. It says for example: 10ps jitter. It’s very rare, but sometimes we also see the phase-noise described. This is done in dBc/Hz. We will explain the difference for you later.
The point is: jitter can be broken down into multiple segments. It cannot simply be expressed in one value. So if it doesn’t say which type of jitter it is, it’s not much use to us. In many cases it is the 1-Sigma value, but well: that is our assumption after countless measurements and comparisons with the specification sheets.
Metric values for jitter
Absolute jitter
The absolute jitter indicates the deviation of the position of the edge from the ideal. With the Wavecrest, this is where we see the max values. The SIA-3000 also indicates in the histrogram and avg-jitter values what the average deviation is.
Cycle Jitter (Period jitter)
Cycle jitter indicates the variation in duration of successive clock cycles. Thus, it measures the duration of a cycle and indicates the differences from the mean or ideal.
Cycle to Cycle
This is the variation in duration of a cycle relative to each other (i.e., consecutive cycles).
Types of Jitter
Random Jitter
Random jitter – the name actually says it all – is an unpredictable component within jitter. It is also known as Gaussian jitter. Random jitter can be caused by things like thermal noise.
Deterministic Jitter
Deterministic jitter is predictable and reproducible. It is caused by things we can control. Think interference, clean power supply and crosstalk, for example.
Total Jitter
Total jitter is the sum of random jitter and deterministic jitter.
Phase noise
Phase noise is related to jitter. You can best see that phase noise visualizes jitter in the frequency domain. Jitter specifies clock stability in time. Phase noise describes the variations (noise) in the phase of a signal as a function of frequency. Hence the unit dBc/Hz. This value therefore comes at a certain “offset” from the carrier (carrier wave). For example -110 dBc/Hz at 10 Hz. By this we mean that the ‘noise’ is -110 dBc/Hz at 10 Hz of the carrier (carrier frequency). For example, 20 MHz.
From the phase noise, jitter can be calculated. The Aeroflex PN9000 has certain tools for that. The rms jitter we calculate is usually pretty close to what we see on the Wavecrest.
So why would we want to measure both jitter and phase noise? They are two ways of looking at something. Phase noise plots by frequency. Jitter plots by time.
How is jitter audible?
Now that you have some understanding of what types of jitter there are and what the similarities and differences are between phase noise and jitter, we can take a look at the effect.
How jitter manifests itself varies. In fact, it also depends on what a manufacturer has done to suppress jitter. There are various “anti-jitter” circuits. Think buffers, phase locked loops, or other ‘dejitter’ techniques. Think reclocking, advanced oversampling, etc….
Manufacturers stick all kinds of fancy names on them, but the fact is: it is often a combination of a phase-locked loop and an elastic buffer that temporarily stores the signal and then reclocks it. And that works. But it doesn’t solve everything. Low-frequency jitter is often still present. And well: that is precisely the jitter that affects audio….
In many cases jitter is audible in sharpness (especially t’s and s’s), loss of detail and flow in the reproduction and a flat and unfocused stereo image. In short: everything does get affected. However, what usually stands out immediately is sharpness.
The annoying thing is that many people associate these characteristics with digital sources. Analog would be better because it does not have these properties. However, a digital source does not have to have these properties either. If a manufacturer simply pays attention to a decent clock circuit, then at least the basics are in order!
Next Episode
In the next episode, we will take a look at up-and oversampling. What is it? What are the differences? And when does it make sense?