Unless you are recording with Pro Tools HDX or the new Universal Audio LUNA Recording System we all have to deal with latency. There is no getting around it. In this article the team at UK interface and console manufacturer Audient explain all you need to know about latency and how to make the best of your audio system when recording, tracking and mixing.

Latency - What Is It?

When recording using a computer, one thing you’ll want to understand is latency and how it can affect your sessions. Modern computers are wonders of technology and can process audio at incredibly fast speeds. However, despite this super-speedy processing, you’ll still get a few milliseconds latency as the audio travels from your interface, through your computer and then sent back to the audio interface to be played out of your computer. Our ears can be quite sensitive to this latency, typically anything around 6ms will be audible as a delay, meaning it can be difficult to play to a backing track if you’re a few milliseconds out. Therefore, understanding how this latency occurs and how we can manage it can mean you get better recordings and happier artists when tracking.

To understand latency in audio systems today we are going to use the analogy of the postal system (snail mail!) Let's first imagine the ideal situation:

The ideal situation when someone sends a letter is that this letter instantly arrives at the recipient's address, not even a second after the letter is posted. However, this isn’t possible as a physical journey has to take place in order for the letter to arrive with the recipient.

Audio systems are the same, while for monitoring and recording the ideal is that the sound is instantaneously returned back to the listener, in most modern systems the audio has to travel through multiple processors and wires to get to the listener's ears and this delay is what’s known as latency.

While most of the time latency is unnoticeable and if it remains consistent it’s easily fixable in post-recording, this article is going to discuss the different causes of latency and the things you can do to fix them.

Direct Monitoring

Going back to the post office analogy the quickest way to get a letter to someone would be to deliver it yourself, which in audio terms would be called direct monitoring.

This is when the input is sent straight to the headphones greatly reducing the latency. On the iD range, for example, this can be done through either the monitor mix control or using the iD mixer.

Occasionally, direct monitoring is not possible so the journey through the signal chain has to be done.

Processing

The other adjustable causes of latency would all be found within the processing of the audio either in the converters or the plugins. Continuing with our postal service theme each processor would be a post-distribution centre where the post is processed and sent on to its next stop at either another centre or the recipient’s house.

Like with audio systems and processors the more centres a letter has to go through the longer it is going to take for this letter to reach its destination. So one step that can be taken to reduce latency is to decrease the number of active processors/plugins (centres) being used in a system either by waiting til mix down to apply them or by rendering the audio tracks with effects on so the processing power is reduced.

The final thing we can do to improve our postal service is to improve the efficiency of each centre. For audio systems, we have two different things to help with this, buffer size and sample rate.

Buffer Sizes

Computers like to do work with large amounts of data at a time every once in a while rather than small amounts of data in an almost constant stream as this means no other processing (such as graphics) can be done at the same time.

This, for audio systems, causes a problem as audio requires a constant stream of samples due to being a continuous waveform. To do this, audio systems have what’s known as a buffer.

For our postal analogy, this will be our delivery trucks. Now with buffers they release a certain size packet of audio every certain number of milliseconds and we are going to imagine this is the case with the postal service too. Let says a truck has to deliver 100,000 letters every 10 hours.

Now where the buffer size would come in is it would be the size of the delivery truck used. If each truck could fit 50,000 letters, a truck would only need to be released every 5 hours giving the centre a reasonable amount of time to do other things in between and get the truck fully loaded with the letters.

Now if the truck could only fit 5,000 letters then a truck would have to go every 30 minutes meaning the centre has to spend most of their time loading the trucks putting a lot of strain on the centre to get other things done like management etc. This is often a cause of audio system crashes (because the buffer is too small and the computer can’t handle it).

So, just pick the largest possible buffer right? Wrong, if we now bring latency into the analogy, if we choose a truck that could fit 100,000 letters in it we would only need to send one truck every 10 hours. However, this then adds 10 hours to the time it takes our letter to get to the recipient when it could have been only half an hour.

So the buffer size is one that often has to be compromised on and things such as number of plugins and tracks being recorded can greatly reduce the buffer size needed to stop the computer crashing thus reducing your latency.

For those enjoying this postal analogy, the clicks and pops often heard when the buffer isn’t full before releasing this packet of audio are effectively empty space in a truck that wasn’t full, so the next centre is left waiting for the next truck to arrive.

Sample Rate

Finally, sample rate can also be used to increase or decrease latency. The sample rate is effectively the speed the letters arrive and the number that need to be processed. To explain this we are going to imagine the difference between 48kHz and 96kHz (double 48kHz). 96kHz for our postal service would effectively be like our postal service running at double the speed meaning you halve your latency (almost!).

However at 96kHz instead of having to deliver 100,000 letters every 10 hours, it would be 200,000 letters every 10 hours greatly increasing the strain on the centre (processors), but the letters would arrive in half the time. So if you are at the smallest or close to the smallest buffer size and your computers not struggling to run at all this can be a good way to decrease latency further by increasing the sample rate.

Conclusion

Understanding why latency occurs will help manage the latency you and your artists hear and help you get the best from your tracking sessions. Experimenting with your buffer size and sample rate can give you great improvements.

Furthermore, the new v4 drivers for the iD range and Sono mean you can use lower buffer sizes with higher stability in larger sessions.

Sometimes, settings on your PC can stop you getting the best performance, so we’ve written a handy guide to help you optimise your PC into an audio processing beast. You can find that guide here.