RECORDING July 2022 - 40

INSIDE
THE BOX
Record and Play Buffers
A DAW system includes both input
(recording) and output (playback) RAM
buffers. Only the playback buffer is
involved for playback, but when recording
and monitored through the DAW,
that live signal passes through both before
the performer hears it in the headphones.
This is where latency might become
noticeable.
How Much is Too Much?
Most players subconsciously adapt
to several milliseconds of latency in the
monitored audio signal-in fact, latency
less than 10-12 milliseconds is usually
not a problem. However, some performers
are more sensitive to it than others.
To put system latency in perspective,
sound in the air travels about a foot/millisecond,
so monitoring a signal that's
delayed a few milliseconds might be
equivalent to an electric guitarist standing
a few feet away from the amp. As
long as the user-adjustable RAM buffer
setting keeps the latency down in this
range, it shouldn't be an issue.
Buffer Size
The size of the RAM buffer is measured
in samples. Typical options are 32, 64,
128, 256, 512 or 1,024 samples. Lower
settings result in less latency-better
for the performer-but put more strain
on the computer, potentially resulting
in dropouts, clicks and pops. Higher
settings put less strain on the computer,
allowing for larger sessions and more
internal processing, but at the expense
of higher monitoring latency.
The specific amount of latency for a
particular buffer setting is related to the
sample rate. For example, at 44.1k, a
128-sample buffer setting might result in
about 3 milliseconds latency for the record
buffer and another 3 milliseconds
latency for the playback buffer. Add
to this another 2 milliseconds of AD/
DA converter latency, and a performer
is experiencing around 8 milliseconds
of round-trip latency. Sometimes this is
40 RECORDING July 2022
helpfully displayed in the DAW's Audio Preferences panel (though sometimes without
the additional converter latency added in).
With that same system, reducing the buffer size to 64 samples cuts the round-trip
latency in half (3 milliseconds +ADDA), doubles the buffer size to 256 samples and
the round-trip latency doubles as well.
The Edge of Perceptibility
Generally, 256 is the highest setting to use for recording. It is right at the edge of
perceptibility. It might bother some performers (like me) and not others. Generally, I've
found that 128 is usually fine for recording in most cases.
Even lower buffer settings of 64 or 32 samples may be usable with powerful modern
computers and USB 3.1+ or Thunderbolt interfaces. Still, on slightly older computers,
128 samples is a good compromise between system overhead and suitably low latency.
Test your system before a critical tracking session to determine what it can handle. Think
maximum track counts, plugin and virtual instrument usage to make sure there are no
surprises. Also, keep in mind that high sample rates (like 96k) may offer less latency at
a given buffer size, but they also put more strain on the computer, so test that too.
Recording Versus Playback
For many users, it is common practice to switch the buffer size between recording
and playback. Setting it as low as possible during recording, then increasing it when
mixing will allow for CPU-hungry plugins and processing. Some users like to find a
median setting that serves them for both.
Zero Latency Monitoring
So far, everything I've talked about assumes that the live audio is being monitored
through the DAW. But there is another way to approach monitoring that avoids the
issue of latency on live signals, and that's to monitor live/recorded audio before that
signal passes through the DAW.
Many interfaces offer an option to take any incoming live signals and split them
off, sending them straight to the output (speakers or headphones) before they pass
through the digital converters and the DAW/computer RAM buffers-this is referred

RECORDING July 2022

Table of Contents for the Digital Edition of RECORDING July 2022