I have to do a more precise (and correct) statement. Formerly the FW driver had a Safety Offset of 64 samples. As this one is for record and playback it's 128 samples. Now we have this one reduced to 32 (=64), but added an additional playback Safety Buffer (like in Windows) of 32 samples. So its 96 instead of 128 before, [large]the total improvement is 32 samples less delay[/large]. I will change the readme text to make this more clear.
I've discovered an interesting development concerning "the total improvement is 32 samples less delay".
With driver versions 2.59 through 2.75, I consistently measured a +2 sample recording delay in Logic 7.2.3 through 9.1.1.
I measured by recording an output physically patched (with a cable) to an input. Then pan the source track hard L and the resulting patched track hard R and bounce to a stereo file. Open in the Sample Editor and count the difference in samples between L & R. I check it every time I update the Fireface 400 drivers and it's always been 2 samples late. So, in Logic's Preferences, I set the Recording Delay setting to -2 samples to compensate. Re-test and it's dead on perfect.
I just updated to driver version 2.83 with firmware 1.70 and checked the recording delay.
Now, it's 34 samples late.
It seems like 32 samples has been added to the original 2 samples that I've measured for years.
And, when I set the Recording Delay setting to -34 samples to compensate and re-test, it turns out dead on perfect.
Is this the result of the "additional playback Safety Buffer (like in Windows) of 32 samples"?
Shouldn't the driver report those 32 extra samples to Logic/Core Audio?
If not, everyone should check their recording delay and adjust if necessary.
It would be great if there was no need for the user to have to test this and perform manual compensation.
Macmini '19 3.2 GHz 6-Core i7, 32GB RAM, MacOS 10.15.2, LogicProX 10.4.8, FF400, UAD2 Satellite Octo