Mike,

What are you using to measure the RTL (round-trip-latency) figures you posted above?  I'm assuming these are @ 44.1KHz.

The fact that you can't use the lowest buffer could be a system problem, but it is too early to tell unless other UFX owners chim in.

By the way, Demonic, your latency figures are only the reported ONE-trip buffer size and NOT the entire RTL figure as reported by Mike.  Just so you know, they're only half the story :-)

Take care!

Have you measured the roundtrip latency with a loopback (output to input) connection and aligned the source and the recorded tracks in Logic to see the real world roundtrip latency including all hidden safety buffers and the latency introduced by the AD/DA converters or do you compare the latency given from Logics driver readout?

best regards
Knut

@Admin Knut :

I compared latency given by Logic readouts. I didn't perform real world test. I'll do it as soon as possible because I have some doubt about Logic's measure reliability

To my ears, with software monitoring on and Totalmix monitoring disabled, with the same sample rate and buffer size, I have more "slapback" effect with 002rack than with UFX when recording my own voice. Regarding the numbers, should be the opposite.

@Jeff :

I am a Logic beginner not expert at all. So thanks for this info. Meanwhile, I cannot find this option in Logic 9 and the article seem to be Logic 8 related. I compared differents soundcards with the same settings so whatever IO safety buffer is engaged or not, we can "see" the differences.

But I will double check about this option before performing any "real world" measures.

Yes I agree, values about Digi002 rack and built-in ins/outs are weird and difficult to trust.

That's wrong. I would prefer saying that those latencies numbers are simply wrong and unrealistic. But that isn't one trip latency either ;-).

Once again I only gave what Logic displayed in audio prefs, and Logic gives 2 values, the ones I posted.

BTW, I tried making measures but it is not as simple as you think. Why? Because Logic (and also PT9) automatically compensate for the buffer in/out value (and yes, Automatic delay compensation is disabled). Whatever buffer size chosen, source and recorded track are already phase aligned.

So I have to record both source and "roundtrip" track at the same time and need to do this for each buffer size, fore each audio interface, with and without software monitoring  (or Low Latency Monitoring) engaged and with PT9 and Logic 9 in order to be sure what I am exactly measuring (and to be sure that no additionnal latency is introduced somewhere in the audio chain).

I will post results when I sort all of this out.

Thanks.

Ok. After real life measure :

- UFX USB and FW gives me the same values above if, in Totalmix, I route the software output to hardware output and software monitoring is enabled in Logic9 (didn't find any IO safety buffer option but I have a french version).

If I disable Software Monitoring, and directly route input to hardware out with Totalmix, latency is fixed whatever buffer size used and give me 56 samples@44,1k  (1,27 ms), 51 samples@96k (0,53ms) and 45 samples@192k (0,23ms), which is close to Gerhard oscilloscope measures.

I made the same measure with PT9 and have same results (except @32 samples which is unfortunately not available).

- With Digi002R, the story is different :

With Logic9 / PT9 (LLM Off) / PT9 (LLM On) @ 44,1k
304 (240+64) samples (6,9ms) / 272 (240+32) samples (6,2ms) / 88 samples (2ms) @32
466 (338+128) samples (10,6ms) / 402 (338+64) samples (9,1ms) / 88 samples (2ms) @64
735 (479+256) samples (16,7ms) / 604 (476+128) samples (13,7ms) / 88 samples (2ms) @128
1376 (864+512) samples (31,2ms) / 1120 (864+256) samples (25,4ms) / 88 samples (2ms) @256

Conclusion : numbers given by Logic are pure fantasy and cannot be trusted depending on the interface used.

Measuring the roundtrip this way will in my experience always result in lower than 'real life' latency. A signal sent to the output and back to the input - if that's what you mean by (output to input) doesn't travel through the same latency inducing steps that a recording situation in Software Monitoring mode does.

IMHO Logic is to blame, as it is aware of all the buffers used and could calculate and display the real/total buffer size = latency easily. This way you would no longer see stupid values designed to fool the costumer.

I wonder if Logic is not using this 'total' size internally anyway? Otherwise a 002 (and many others) would not be able to record its own playback correctly, as the second track in Logic would immediately reveal a time offset (first track: playback, second track: record).

Install "AU Lab" with the XCode developer tools on your OS X disc. This allows to see all numbers if the driver reports them correctly. Look into the manual to get the AD/DA latency numbers if you want to know the exact driver buffer size.