In a previous post we had an encounter with jitter in audio. The talk about jitter is hard to miss these days. There are plenty of products on the market claiming to have low jitter or to reduce jitter.
There are also many pages on the Internet that do a pretty good job of explaining what jitter is. If you are not familiar with the concept, a quick Google search will bring you up to speed. However, if you want to find out what it sounds like, that's not so easy to find. The top result for “What jitter sounds like” was this video - 5 minutes that I never get back.
So I thought “What if I could simulate jitter, listen to it, and judge for myself?”
This post is the result of that thought.
If you want to know how my simulation works, read on. Otherwise, you can skip this section where you can do some fun jittery listening. To simulate jitter, I used handy software called Matlab. Matlab is great for doing all sorts of mathematical things. I wrote a little script that does this:
Opens a music WAV file
Upsamples the file by a factor of 100
Downsamples the file by a factor of 100
Saves the result in a new WAV file
The magic happens during the downsampling part. Normally if you want to downsample by a factor of 100, you would take a sample jump over 100 samples, take another sample, then jump over another 100 samples … you get where I'm going.
But here, instead of jumping 100 samples, I jump 100 plus (or minus) a number (lets call that number x). Doing this simulates a situation where your clock is ticking a little earlier or later than it is supposed to.
There are various causes of jitter in audio, so there are various types of jitter. Here I experimented with two:
Noise jitter is when the ticks of the clock are off by a random length of time. That means x is a random number. I used a normally distributed random number, which is what we are more likely to have in the real world.
Periodic jitter is when x oscillates in time.
This jitter can be caused by the 60 Hz (or its harmonics) from your power supply leaking into your clock generation circuitry.
Every circuit is unique, so in the real world periodic jitter can come in different frequencies. However, for the sake of simplicity, in this experiment, I used 60Hz periodic jitter.
I ran the simulation on 44.1 kHz (CD quality) sample WAV files adding different types and levels of jitter to each sample.
Now the fun part
There are seven versions of each sample.
Three have periodic jitter, three have random jitter, and one is the original - untouched.
In the samples we have 2µs (microseconds), 4µs, 8µs, 16µs RMS jitter.
It is unlikely in the real world for a digital-to-analog converter to have a jitter higher than 2µs RMS.
I threw in the 4µs, 8µs, and 16µs samples to see what some extreme cases sound like.
As you scroll down you can listen to them and make your best guess of which is which.
Then you can click the sample's title (i.e. 1kHz v1) to see if your guesses are right.
Before you move on, here is something to keep in mind. The clock interval of a 44.1 kHz signal is 22.7 µs, meaning, there is a sample every 22.7 µs. Jitter is expressed in RMS microseconds. So when we say there is a 2µs RMS jitter, it means the clock was, on average, off by 2µs.
Let's start with a 1kHz sinusoidal wave.
Let's see what the spectrum of the 1kHz tone looks like after jitter is added
Now, let's listen to some real music. I hope you like Pink Floyd!
Next, “The World is Round” by Day One
Now an excerpt from a favorite artist of mine, Vanish Ark. We are going to do something different with this sample. I've heard a really slow periodic jitter, sounds like a WOW that you expect from an LP record. LPs have an RPM of 331⁄3 which mean the WOW should have a frequency of 0.55Hz. So I added a gigantic 1600µs RMS periodic jitter at 0.55Hz. Now, it's totally unrealistic for a digital system to have jitter like this, but let's see what it sounds like.
This wraps up our mini experiment. Of course, the system that you used to listen to the samples also adds its own jitter on top of what was added by me. However, these samples should still give you a good idea about the sound of jitter. I hope this post has shed some light on this illusive phenomenon. How did jitter affect your experience? Did it effect the music the way you imagined? Do you see a flaw in the experiment?
Access early bird offers and updates on our new and exciting product, coming soon on Kickstarter.