Jitter Buffer Calculator for Network Audio

If you’re listening to a song and the digital link drops out at a crucial point, you’ll probably hear a sudden granular crackle instead of nothing. Then there will be an embarrassing pause as everyone waits for music to catch up. In most cases this isn’t due to bad luck but a buffer problem.

Packets of audio travel across a network. They can follows various routes via different switches and routers. One packet might go slow and get caught up behind other data, like a video update on another device. The other packets goes by while it waits. At their destination, they are captured in what’s called the jitter buffer, where they wait before being played out. If the buffer is too big, you’ll experience your own voice delayed noticeable in your headphones. Too small, and you hear glitches.

How to Calculate the Right Audio Buffer Size

After entering your network measurements into the calculator on this page it does the rest for you. No more trial and error by clicking buttons until it sounds about right. You know what the inputs mean so now you can dial-in the system rather then poking around trying to get it right.

First, look at packet interval to see how much audio goes in each packet. Big packets reduces overall traffic overhead but introduce delay before the other end gets the data. Little packets introduces less delay but more overhead that can slow down a congested network. It’s always a balancing act.

Now turn to the jitter measurement. Because average jitter can be deceptive, the tool wants ninety-five percentile value of jitter. It could be true that the average is good, but then there are those occasional spikes and performance is bad. You don’t want to only protect yourself from the middle-of-the-road stuff; you want to protect yourself from worst case, especially when that worst case happens regularly.

Most folks report this on the send side of things. But what you really need to know is how to measure it on the other side of the net where packets arrive. That’s when all the network chaos takes place. So if you see a spike burst in your logs, don’t just shrug it off. Enter it into the spike field so math will take into account that rare-but-severe event of late arrivals.

There’s also the matter of clock drift. Even high quality audio interfaces has an internal clock which will drift from other clocks over time. A small amount one way or the other can add up to lost samples over time. That’s where the drift allowance comes into play with the calculator for how far out of sync you think they could be when you press “resync.” You’ll require more headroom if you’re doing a longer performance with no common clock. You only remember when audio begins skipping towards the end of the show.

Here’s a handy reference table that outlines typical situations. Like satellite link vs. Studio LAN, and shows you what size of buffer you need based off distance: When the path is short and stable, like your local gigabit switch, not much cushioning is necessary. Once you’re cellular backhauling or going over public internet, though, that all changes. Jitter increases. Increase the buffer to account for increased jitter.

In short, don’t be afraid to turn up latency as long as it doesn’t drop below one hundred milliseconds. For live musicians, they’d prefer zero latency but they value reliable latency above fast latency. If a latency of a hundred millisecond fluctuates between 80 and 120ms, then we can compensate by adjusting our time. If our latency drops frequently and averages thirty milliseconds, it completely ruins the performance.

If you feel your network isn’t stable, use the safety margin slider. This will add a certain percentage on top of what’s calculated so you have some leeway when there is unexpected congestion.

Lastly, be aware of memory needed. The receiving device has to have enough RAM for its large buffers. That adds up fast if you’re pushing lots of channel over a hardware interface or maybe even in your laptop. Knowing the raw PCM size will help you plan out your system resources before hitting record. It’s all about finding the balance between the buffer depth and the clock speed. When you gets it right, the glitches dissapears.