Latency of 3rd party interfaces compared to HD Native

[jacko]

I did some round trip latency tests of different interfaces. I will be updating this post once I get more devices tested.

Pro Tools 12.5.2 HD.
Everything unchecked in playback settings.

Sample rate 44.1KHz, buffer size 32:

1. Apogee Ensemble Thunderbolt: 2.86ms
2. Focusrite Red4 Pre: 2.9ms
3. Avid HD Native Thunderbolt + HD I/O: 3.35ms
4. UAD Apollo 16 silver: 4.4ms

Sample rate 44.1KHz, buffer size 64:

1. Apogee Ensemble Thunderbolt: 4.3ms
2. Focusrite Red4 Pre: 4.35ms
3. Avid HD Native Thunderbolt + HD I/O: 4.8ms
4. UAD Apollo 16 silver: 5.85ms

Sample rate 88.2KHz, buffer size 64:

1. Avid HD Native Thunderbolt + HD I/O: 1.93ms
2. Apogee Ensemble Thunderbolt: 2.15ms (this needs to be re-tested)
3. Focusrite Red4 Pre 2.18ms
4. UAD Apollo 16 silver: 2.92ms

Sample rate 96KHz, buffer size 64:

1-2. Avid HD Native Thunderbolt + HD I/O: 1.78ms
1-2. Apogee Ensemble Thunderbolt: 1.78ms
3. Focusrite Red4 Pre: 2ms
4. UAD Apoolo 16 silver: 2.97ms

[Darryl Ramm]

Since there are sweet spots at 96 kHz and lots of folks like to work there it would be great to also measure latency there.

And more details would be nice, like exact Pro Tools version, playback engibe settings, OS version, driver version, firmware versions,

[jacko]

Tomorrow I should be able to test Focusrite Red4 Pre connected via Thunderbolt.

[jacko]

I updated the first post with Focusrite Red4 Pre latency values.

[propower]

Thanks for posting these! The Ensemble between 88.2 and 96k looks a little odd (too much change for such a small sample rate delta).

[jacko]

Originally Posted by propower

Thanks for posting these! The Ensemble between 88.2 and 96k looks a little odd (too much change for such a small sample rate delta).

The latency in samples for 44.1KHz and 88.2KHz was 2x[buffer_size]+62.
For 96KHz it was 2x[buffer_size]+43. Hence the difference is bigger than proportionally to sample rate.

For the Omni the formula was the same for both 88.2 and 96: 2x[buffer_size]+43.

[propower]

Originally Posted by milaszewski

The latency in samples for 44.1KHz and 88.2KHz was 2x[buffer_size]+62.
For 96KHz it was 2x[buffer_size]+43. Hence the difference is bigger than proportionally to sample rate.

For the Omni the formula was the same for both 88.2 and 96: 2x[buffer_size]+43.

I don't know what formula you are referring to. There is 2X the buffer in all Native cases: time_DAWbuffer = 2*(Buffer/Sample_rate) in seconds

Then there is A/D, D/A plus any driver delay to get data on or off the data bus.

At 96kHz - 64 buffer is 2*64/96000 = 1.33ms
Measured RTL at 96kHz (Ensemble in this example) = 1.78ms
A/D D/A and data bus delay ~ 0.45ms

For 88.2kHz I expect this to change by
~1.78ms*(96/88.2) = 1.94ms
I just measured this (using Logic as the DAW today) and the data is 1.93ms (as expected)

All measurements are with scope and function generator analog to analog. I trust no other method

[jacko]

Originally Posted by propower

I don't know what formula you are referring to. There is 2X the buffer in all Native cases: time_DAWbuffer = 2*(Buffer/Sample_rate) in seconds

Then there is A/D, D/A plus any driver delay to get data on or off the data bus.

At 96kHz - 64 buffer is 2*64/96000 = 1.33ms
Measured RTL at 96kHz (Ensemble in this example) = 1.78ms
A/D D/A and data bus delay ~ 0.45ms

For 88.2kHz I expect this to change by
~1.78ms*(96/88.2) = 1.94ms
I just measured this (using Logic as the DAW today) and the data is 1.93ms (as expected)

All measurements are with scope and function generator analog to analog. I trust no other method

The formula is something that I worked out myself by measuring those latencies. It is the same as you're describing. Something you describe as "A/D D/A and data bus delay" varies from interface to interface and for a single interface it may vary depending on the sample rate. In case of Apogee Ensemble Thunderbolt it is 43 samples for 96KHz and 62 samples for 88.2KHz.

For HD Omni it was 43 samples in both 88.2KHz and 96KHz.

For Apollo 16 MK1 this value is 130 samples for all sample rates.

Focusrite Red4 Pre has this value at 64 samples for all sample rates.

[jacko]

Originally Posted by propower

For 88.2kHz I expect this to change by
~1.78ms*(96/88.2) = 1.94ms
I just measured this (using Logic as the DAW today) and the data is 1.93ms (as expected)

Will you get a chance to make the 88.2KHz measurement in Pro Tools as well? I will get an access to Apogee again in about two weeks time.

[Terry Wetzel]

Pardon my unscientific thinking on the issue of latency. In my humble opinion latency is an undeniable fact of digital,(and analog) recording. It's a fact and it's not going away. There are ways to deal with it and most, if not all the manufacturers of Interfaces have developed ways to record without it being an issue. As a musician I've been involved in sessions that predate digital recording and I can't recall latency even being mentioned. The particular sessions I am referencing required vocal overdubs of both main vocal and background. As long as there were no headphones issues as far as levels were concerned, latency never reared it's ugly head! I don't believe the average musician or vocalist would be bothered or could even discern 1 or 2 ms. of latency! My advice. Deal with it and move on!