Yesterday I made a digital loopback test after reading a post on gearspace.com about a possibly wrong ‘placement of recorded audio events’. With my current setup the loopback recording indeed is constantly 22 samples off. Funny side note it’s 22 samples early … so I can travel in time through re-recording 
I use Nuendo 12 with a RME UFX+ interface on a windows 10 workstation.
it’s only 0,4ms when recording with 48kHz and 0,2ms with 96kHz but with a sharp transient you already hear phase issues even with this tiny offset (I re-recorded the digital audio click/beep).
Now my questions are the following
a) why is this / what’s causing the negative offset between DAW and interface driver?
b) it’s constantly -22 samples so I can compensate it easily within the audio system settings (… now that I know ). But turning off ASIO guard makes it worse by 23 samples. Why is this? Is this a bug? My understanding of ASIO guard is that it only affects tracks that are NOT rec enabled. Changing the ASIO guard settings from low to high doesn’t affect or change the -22 samples. only turning it off raises the offset to -45 samples.
c) there is an option within the audio system settings named something like ‘compensate audio latency’ (“Aufnahmelatenz berücksichtigen” in german). I made several tests with this option active as well as deactivated but there was never a difference regarding the -22 samples offset. What is it supposed to do / to compensate?
Thank you guys so much for an answer and be aware to measure your system and enter a possible offset for an accurate recording system.
Tell us about your test method
- To have a visual sharp transient I rendered the click to audio (quarter notes, 100bpm)
- that became my playback track routed to an output of my interface
- rme offers a handy loopback function and I recorded its appropriate input on a new track in my DAW
(when I add the 22 samples offset both tracks null perfectly when the phase is flipped on one channel)
Go analog out and back in again with that transient and see if it lines up. I think Nuendo is compensating for RME’s reported a/d latency and you’re not going a/d and back again.
Good point 
with an DA/AD pass it’s only 1 sample off (late) and of course I can live with this … with or without compensation.
But knowing that digital interfaces have faster RTL I fed a digital channel with the same signal too (ADAT in this case). And guess what this channel is early again (16 samples with 96kHz and 20 samples with 48kHz).
I couldn’t care less whenever there’s an DA/AD pass later on or if something’s traveling through air anyway. I’d just say good to know and to keep in mind in case there’s a parallel path that stays strictly digital.
and what about ASIO guard? Why is it changing and over-compensating my recordings when turned off?
It shouldn’t be. Is it not the same (1 sample) when going D/A - a/D with Asia guard on or off?
Tell us your test method on this one too…
No, it is not the same, when I turn off AG. That’s what I wrote in my first post too. When I turn off AG it ‚overcompensates‘ with -12 extra samples. No matter if I use an analog or a digital patch or do the software loopback, it‘s always the same: turning off AG results in a different, slightly early positioning of the recorded material!
My method: the analog loopback recording is with an balanced audio cable (analog OUT back to analog IN). Digital loopback recording done with an optical ADAT cable connecting ADAT OUT with ADAT IN.
You don’t have a the big orange button “constrain delay compensation” turned on, do you?
2 Likes
no, it’s off (grey)
I know about the constrain delay compensation feature, never ever used / needed it so far.
Can anyone please confirm or disprove that turning off ASIO guard offsets the recordings on her / his system? You can use a software loopback or real patch cables. In any scenario turning off AG changes the positioning by -10 samples at least to the same setup here with AG on.
O samples difference here with Asio guard on or off. Lynx AES 16e. Win 11 Intel i9. N13 latest version
ok, thank you so much 
. I think I’ll open a support ticket then.
in case you still have N12 installed on your system too it would be very, very helpful if you could run a quick test within N12 to exclude the version from the list of possible culprits.
I’m going to repeat the test on my MacBook and alternatively with an UA apollo interface later this week and I will report back if the results change from the current setup (RME UFX+, Win10, Intel Xeon, N12 latest built).
1 Like
I couldn’t wait and did a quick run at home (MacBook + Apollo). With the same nuendo version 12.070 it works like it should and there’s absolutely no difference with AG on or off
Now I have to examine further what causes this behavior on my studio workstation.
1 Like