Question about loopback tests

I'm currently doing some loopback testing to measure audio latency in my system.

My current I/O buffer size is set to 256 samples, and my project sample rate is currently set to 48k. At these settings, Logic is reporting a 15ms roundtrip latency for my device.

For the loopback test, I'm running a cable from a physical out on to a physical in on my focusrite.

When I run a ping test using the I/O plugin, I'm getting a value of -19 samples. I verified this with a recording loopback test, and indeed logic is displaying the recorded audio exactly 19 samples early.

From what I understand, this means that Logic isn't reporting the correct latency for my device, and that I should use the Recording Delay setting in the Device Preferences window to compensate for these 19 samples by entering a value of +19 for this particular buffer size and sample rate setting.

Which brings me to my question. How is it possible that logic is reporting that the recording is coming 19 samples early? From what I understand, my device may be doing some additional processing that might add to the sample delay which logic doesn't report, but isn't it impossible that it comes early like this? All the examples I've seen of people doing loopback tests have resulted in additions to the sample delay, not subtractions. Am I missing something here?

EDIT: It might be worth mentioning that I'm not using any internal ADAT/digital routing in my scarlett mix control (at least not that I'm aware of). I understand that if there is internal routing which keeps the signal in the digital domain, logic might be "overcompensating" with the auto delay compensation which causes the recording to appear as if it was recorded earlier than it was. On my focusrite, I'm currently running the loopback from my headphone out to my mic line in, so the signal should be passing through the physical domain as expected, but I'm not exactly sure how to make sure there's no digital routing in scarlett... I'm trying to wrap my head around that right now.

.R

It's the other way round.

The driver is misreporting its roundtrip latency via core audio to Logic.

I would approach Focusrite and ask why it is that way?

Reporting back in, just in case anybody else searches for a solution in the future.

I got in contact with the Focusrite support team. Apparently, it's perfectly normal that audio devices could report the incorrect roundtrip latency to Logic in both -(negative) and +(positive) values. So it looks like my test result of -19 sample offset @ 48k is not odd.

Solved?

Hmm...

Normal or not is not the real issue.

The issue is why it's not reporting the correct roundtrip delay through core audio.

Are they using a generic driver that they can not control?

The interface I'm using is always reporting dead on.

Hi Eric,

I'm not sure why my particular device is reporting an incorrect roundtrip latency.

Per the Focusrite team, it's "normal".

This also aligns with some information I got from Tom Rice's Macprovideo tutorial (Logic Latency Toolbox, which Jordi recommended I check out), in which Tom explains that some audio interfaces will not report the correct roundtrip. Unfortunately, Tom doesn't go into the reason why, so I'm just as curious as you are. He does mention that this is exactly what the recording offset slider in logics preferences was designed to do (compensate for incorrectly reported roundtrip latency).

So yeah, I'm just as curious as you are, but so far I've heard from two separate sources that this is "normal".

I'll continue to search why this happens and report back if I find anything.

.R

... in which Tom explains that some audio interfaces will not report the correct roundtrip. Unfortunately, Tom doesn't go into the reason why,

Because the interface manufacturers are too lazy to test the exact latency of the drivers and hardware they use.

Or they're trying to hide something, but I don't think it applies in this case.

Of course, their drivers could just be so crap that they couldn't reliably test the roundtrip across multiple platforms.