Automation Timing Accuracy

[rbrown46]

Something that has bothered me for a while is the accuracy of automation timing in Live. Certain uses of automation (e.g turning a band-pass filter on and off for a breakdown) depend on having accurate timing. A filter getting turned on 40ms before or after a breakdown can be very audible. Accurate automation is even more important when doing detailed sound-design type stuff. I ran some tests between different DAWs to see which has the best automation timing. The technical details of the tests are at the end… I'll spare you if you aren't interested and cut to the conclusions first.

There are two major factors to consider: if automation timing depends on buffer size, and if automation timing is corrected for as part of plugin delay compensation (PDC).

Live's automation for plugin parameters is dependent on buffer size. If you have a huge buffer (e.g 2048 samples), then any plugin automation may be up to 46.4ms off (I'm doing everything here assuming a 44100khz sample rate). Automation for Live instruments, effects and mixer parameters is not dependent on buffer size. It will be sample accurate regardless.

There is no PDC for automation in Live. If you stick a plugin like Ozone with crazy latency on a track and try to automate anything that occurs in the signal chain after the delay-inducing plugin (including mixer parameters), then that automation will be off.

As far as other DAWs, Logic's timing is the best of the DAWs I tested. Automation is considered for PDC, and audio unit automation is updated every 32 samples regardless of the buffer size. This is great. The problem is that automation values in Logic get quantized to the equivalent of MIDI CC messages (127 discrete values). You hear stepping on everything. This makes it unusable for me.

Cubase is worse than Live as far as timing… it doesn't do PDC for automation and all automation is dependent on buffer size, regardless of if it is a Cubase parameter or a plugin parameter being automated. Reaper does PDC for automation but, like Cubase, everything is dependent on buffer size (both Reaper parameters and plugin parameters).

Reaper comes out on top for me. While automation is dependent on buffer size, this isn't critical when using small buffers, and there is a preference for setting the buffer size that is used when bouncing takes. Reaper also supports tension curves. Whether or not Reaper's UI is tolerable is another issue...

Test Details:

To do the tests I created a 10 second test signal where all sample values are 1.0 (all high).

First, I tested the interpolation each DAW does for volume automation. I automated the volume of the test signal from 0 db to -? db in a single step after one second, recorded this to another track, and measured the amount of time it takes to go from high to low (1.0 to 0.0).

Second, I used iZotope's Ozone 4 with everything on to test automation PDC (this creates 48.5ms of latency). With the same volume automation as before I measured the time between when the signal should have been muted and when it was.

Third, I tested the influence of buffer size (sometimes called block size) on automation timing. I tested the original volume automation with Ozone removed at 2048 and 64 sample buffer sizes, and measured how far off the mute occurred for each. Then I re-instantiated Ozone and automated it's output gain parameter, and measured the timing there.

Tests were done with Live 8.1.4b3, Logic 9.1.1, Cubase 5.5.0 and Reaper 3.63, on OS X 10.6.4, with an Apogee Duet, on a Mac Pro.

- Live has a ramp time of 2.7ms from high to low
- Live does not do PDC on automation.
- Live's automation timing for AUs is dependent on the buffer size. The larger the buffer the more off it will be. Non-plugin parameters are sample accurate regardless of buffer size (volume, panning, built in Live instruments, etc).

- Logic has a ramp time of 5.0ms from high to low
- Logic does PDC on automation
- Logic's automation timing for AUs is not dependent on the buffer size (at least for Ozone 4). Seems to have a constant 32 sample update rate.
- Automation range is limited to 127 values, for nearly all automation types (mixer volume, plugin parameters).

- Cubase has a ramp time dependent on buffer size. 5.0ms for 2048 sample buffer, 0.7 for 32 sample buffer.
- Cubase does not do PDC on VST automation. Automation for other cubase parameters (volume, etc) is compensated.
- Cubase's automation timing for VSTs is dependent on the buffer size… although many trials were necessary b.c. a diff. result was seen each time.

- Reaper has a 1 sample ramp time from high to low.
- Reaper does PDC on automation, both for plugins and Reaper parameters.
- Reaper's automation timing for VSTs and Reaper parameters (volume, panning, etc) is dependent on the buffer size.
- Freezing a take has a configurable buffer size, which will increase automation accuracy.

[tw1nstates]

Why does reaper have so many functions that live still doesn't have.

I am seriously starting to think about going over to the Reaper for a bit to check it out properly.

[Beto Larski]

As i do lots of automation the lack of plugin delay compesantion really bogs me off. I hope that will be implemented soon.

[leedsquietman]

Draw in the automation. This is what I do and it avoids these issues.

[rbrown46]

Draw in the automation. This is what I do and it avoids these issues.

No... this issue is for drawn automation. You may have not noticed this issue if you work with low buffer sizes or plugins that don't have much latency.

Recorded automation is correct only in that what is heard during recording is what you hear on replay... but visually the timing will be wrong.

[leedsquietman]

I typically use low buffers for tracking and increase them for mixing and mostly use zero latency plugins. Sometimes I do the automation early while tracking at low buffers if I already have a idea of what I want.

I have Cubase, Reaper, Live and Renoise at home and can't say I've noticed any differences to be fair, I haven't ever done any scientific testing on it though.