Sampling jitter is one of the most misunderstood thing about digital audio. It’s very complexed issue without clear solution and happen in everywhere. Today I’m going to teach how sampling jitter happens in digital audio world and why it’s still a problem in highend audio till this day.

First of all, let’s talk about how sample rate works. PCM will send pulse signal containing bit depth and sample rate. For CD format, it has 16 bit depth and 44100 sample rate. In 1 second, there’s 44100 samples and each sample holds 16-bit information to determine amplitude level.

In ideal world, each sample in CD format should arrive at exact 22.675737 microseconds interval to reproduce analogue signal according to design. But in real world with limitations and errors within hardware and software level cause the gap between each sample to be not the exact 22.675737 microseconds like we expect.

 

 

When transmitting data, those samples will arrive with timing variation. It happens randomly with no predictable pattern how long each sample will arrive before or after the expected interval. This is called jitter and it happens on every single individual sample, thus it’s called sampling jitter.

This kind of jitter is just like a noise in digital audio world. You can’t expect magic anti-jitter trick to eradicate them completely. Some technique like phase locked loop or PLL can work as locking phase of signal to have like exact 44100Hz instead of 44150Hz or 44050Hz. Otherwise, you may have incorrect sampling rate locked like some devices due to lack of proper jitter management.

 

 

How can we tackle with sampling jitter issue? Can’t we just reclock every single sample and have perfect zero jitter? Well, there’re two big issues about tool we need but don’t have yet.

1. We need reference clock with exact no jitter at all like 0.0000ppb precision The highest precision clock we can find in audiophile market is around 3-5ppb. For clock in bluetooth and network components, they’re mostly used around 50-100ppm from what I found in stock environment after modding products in my experience. Getting 0ppb won’t be possible anytime soon as a consumer market product.
2. We need powerful processor to handle whole process under very strict time frame. For 44100Hz as reference, you need to finish analyzing jitter and forward each sample with corrected time with in 22 microseconds. There’s no way you can sync to reference clock and align the deviation then forward corrected time alignment sample in that limited time due to hardware limitations. Even if you somehow can do it with 44100Hz, nowadays people use high-res too so for things like 192000Hz sample rate, all process need to be done in 5 microseconds. And if you were to do this on DSD256, you need to finish it in 85 nanoseconds. None seems possible with today’s technology at the moment.

 

 

That’s why there’s still many die hard analogue fans who still prefer vinyl over modern digital audio format. Because some problems can’t be resolved completely and even with the best digital audio equipment that cost fortune to own one today, there’s more to invest on analogue domain for better satisfaction.

Some could say $30k CD player is crazy expensive but there’s a guy who’s willing to spend $30k for one cartridge to use with his $100k turntable and $50k pre phono. And for them, digital doesn’t bring what they can enjoy like they do from their vinyl collection, at least not until digital audio is advanced enough to tackle more serious problems better.

 

 

Some people may introduced silver bullet approached in the past like just buffering it to eliminate jitter. It happened with Async USB before that we have companies claiming using Async USB will eliminate jitter completely with buffering technique but that actually increased output jitter in buffer instead and later on it was proven that you still need high precision clock on Async USB to handle buffering with less jitter. The same goes to ethernet and wi-fi as described in IEEE before by Sharp.

The most effective ways to reduce sampling jitter is to reduce the chance of it happening and scope of range of sampling jitter to happen. For example, if you upgrade clock from precision 100ppm to 10ppm, you’ll likely have less sampling jitter from jittered clock with higher precision. And upgrading components in network circuit can reduce noise that can affect jitter indirectly.

 

 

I’m not sure if this will make sense to you but this principle of sampling jitter will happen in everywhere and will affect signal in digital to analogue process to certain degree. From my experience sampling jitter effect was confirmed from my own observation in USB Audio, USB/SATA storage, ethernet, wireless network, and other outputs in DDCs for computer audio.

The only place where I couldn’t detect audible jitter is when I setup WAN bridge from modem to another router. It seems signal before being modulated won’t be affected by jitter as much as signal after modulation. Or maybe problems with network lies in home environment more than from ISP.

So, the easiest solution for network improvement everyone can consider is to upgrade clock module in hardware equipment with higher precision clock and maybe capacitors too with lower impedance to reduce noise in network circuit. I’ll one of my works as example for reference. Inside red box is clock module being upgraded with noise reduction sheet applied.

 

 

I hope this will help audiophiles understand how digital audio works better based on data from professionals like IEEE and Sharp. I have others but I picked this one as they’re mostly known and respected.