MTRX IO Delays

[creativecontrol]

TB3 is a completely different animal than HDX as it uses coreaudio and therefore you have to choose a buffer size which directly affects the latency values. Yes the AD DA values specified and in the manual are valid.


Jeff

[shawnhonsberger]

Originally Posted by creativecontrol

TB3 is a completely different animal than HDX as it uses coreaudio and therefore you have to choose a buffer size which directly affects the latency values. Yes the AD DA values specified and in the manual are valid.


Jeff

Thanks, Jeff! Understood, that core audio is certainly a different animal. It's just that the table in this thread is great! There is just nothing like the table in this thread in the manual, and TB3 would be a nice addition as that is a primary connection for many folks.

I would love to have a reference list like this for TB3 with a sample buffer of 64 let's say, at each sample rate; like it shows for HDX. The table is such a great reference so that folks with HDX can have a numerical reference and don't have to do the math. I would like to have a reference for TB3 at a 64 sample buffer so that I can always have that handy as a numerical reference.

The only reference I can find in the manual is the following: https://resources.avid.com/SupportFi...uide_5.4.1.pdf

Appendix C: I/O Delays
MTRX Studio has a general system delay of 7 samples on the digital matrix and SPQ (regardless of whether or not SPQ is enabled) path
ways. The digital interfaces add one more sample of delay for a total of 8 samples of latency (166 microseconds at 48 kHz) for digital
I/O.
There is an additional delay with analog I/O—AD (analog-to-digital) and DA (digital-to-analog)—due to the conversion process. There
is approximately 650 microseconds of delay with the analog-to-digital conversion, and approximately 250 microseconds of delay with
the digital-to-analog conversion. These delays are not dependent on the sample rate. Once digitized, the signal is time stamped and buff-
ered the same as with digital inputs.

This involves lots of math and conversions that I'm not sure I can accurately do. Regardless, I think it would be lovely if we had an objective numerical reference for TB3 at a 64 sample buffer added to this list. It would be so helpful for me, but also for everyone with a TB3 connection.

Thank you!

[shawnhonsberger]

Originally Posted by creativecontrol

TB3 is a completely different animal than HDX as it uses coreaudio and therefore you have to choose a buffer size which directly affects the latency values. Yes the AD DA values specified and in the manual are valid.


Jeff

Hi Jeff,

Here are initial figures for 48kHz and 96kHz TB3 (which I mostly work at), but it would be nice to have it all documented in the table for future reference so those of us on TB3 don't always have to calculate it and/or remember it.

Latency in ms for 48kHz 32 sample buffer + 7 samples + 900 microseconds:

• 32 sample buffer at 48kHz: Latency (ms) = (Buffer Size / Sample Rate) * 1000
• Buffer latency: At a 48kHz sample rate, each sample takes approximately 0.02083 ms (1 / 48000 samples/sec * 1000 ms/sec). A 32-sample buffer at 48kHz is equivalent to (32 / 48000) * 1000 ms = 0.67 ms
• 7 samples latency: 7 samples at 48kHz is (7 / 48000) * 1000 ms = 0.15 ms
• Convert 900 microseconds to milliseconds: 900 microseconds is equal to 0.9 ms (900 microseconds / 1000 microseconds/ms)

Total Latency 48kHz:
0.67 ms + 0.15 ms + 0.9 ms = 1.72 ms

Latency in ms for 96kHz 32 sample buffer + 7 samples + 900 microseconds:

• 32 sample buffer at 96kHz: Latency (ms) = (Buffer Size / Sample Rate) * 1000
• Buffer latency: At a 96kHz sample rate, each sample takes approximately 0.333 ms (1 / 96000 samples/sec * 1000 ms/sec). A 32-sample buffer at 96kHz is equivalent to (32 / 96000) * 1000 ms = 0.333 ms
• 7 samples latency: 7 samples at 96kHz is (7 samples / 96000 Hz) * 1000 = 0.073 ms
• Convert 900 microseconds to milliseconds: 900 microseconds is equal to 0.9 ms (900 microseconds / 1000 microseconds/ms)

Total Latency 96kHz:
0.333 ms + 0.073 ms + 0.9 ms = 1.306 ms

Hopefully this math is correct. Do these look correct to you? Let me know your thoughts on adding TB3 values e.g. with a 32 sample buffer to the table/list.

Thank you!

[Darryl Ramm]

Sean

You should be measuring the RTL with Oblique Audio's RTL Utility and comparing that to your calculation. It should be ~< 1 sample accurate. With input and output I/O and any driver overhead and two passes though the I/O buffer.

RTL Utility also shows all the individual Core Audio latency components that the driver is reporting. See the RTL Utility measurement log file as a starting point or you can configure those to be shown in the main results page.

And this should match what Pro Tools experiences for a H/W insert RTL latency (e.g. measure in Pro Tools with ADC disabled).

[shawnhonsberger]

Thanks, Darryl. I appreciate that. Yes, that is true. However, I am looking for a listing of the official "authoritative" numbers from Avid listed out in all sample rates for TB3 - like it is for HDX - at the beginning of this thread.

More as a reference for all, than anything else.