ExternalFX Latency adjust

I’m trying to do some parallel work with the external fx so I need the latency sample accurate. In Cubase if I do the latency detector I usually get a value of around 2ms which is from the converters themselves (although it’s usually 1 sample too much so I have to enter 1 sample less for it to be perfect).

In W10 I’m getting around 28ms which is the converter plus the ASIO buffer (my ASIO setting is 25ms). Doesn’t W10 already account for the ASIO buffer though? When I try this value I get a tight delay (similar to Haas effect timing). If I manually enter in 2ms manually then it seems ok. So it looks like the latency is being reported too much.

Are we supposed to subtract the ASIO buffer from this value?

Also I noticed bypassing an External fx seems to mess with the latency as it goes off then.

Need to test a little more but just curious how it’s meant to work in this situation.


There is always at least one ASIO buffer of latency. WaveLab indicates this, unlike Cubase. This is just a different convention.

Thanks PG. The issue is that when that value is there the latency correction doesn’t seem to be right. When I manually put 2ms it sounds fine, when I use the detected 28ms it’s off. Is that normal? Do we ignore the detection value then?

I have a question here? To the OP, I will assume that you are running your device at 32 samples so the extended sample rates would be 32,64, 128, etc. What device are you running here? Obviously to run at such a low sample rate you have a very serious machine there, kudos. But I do wonder, now focused on my setup, if you did run your device at say, 256 samples, with what PG has described and what you are experiencing, even with a higher sample rate (256) and with ASIO Gard engaged could you simply offset your sample rate 2 samples and get the same results?

Hi, it’s running at 25ms so I think that’s 1024 samples.

The problem I have is the detect seems to add more latency adjust than what’s needed. The 2ms I correct for in Cubase is on top of the ASIO 25ms. In Cubase you just see the converter latency, not the ASIO driver (they are two separate values).

I can eventually figure out the correct value (didn’t spend time on it yet) but was just wondering why the detect seemed to over correct.

So just to be clear my total latency though the system is around 27ms, 25ms of that is the ASIO driver (set to 1024 samples) and 2ms is the converters themselves (which take a small amount of time to measure the signal before passing to the driver).


Ah, thanks for the info.