ASIO latency much longer than predicted

Discuss music production with Ableton Live.
Post Reply
Envergure
Posts: 21
Joined: Sun Apr 28, 2019 12:37 am

ASIO latency much longer than predicted

Post by Envergure » Sun Apr 04, 2021 8:41 pm

Ableton Live Suite 11
Windows 10 x64
Behringer UMC404HD interface (connected to a USB 3.0 port on a PCIe card; no difference if connected to the motherboard's USB ports instead.)
Behringer UMC ASIO driver (performs better than ASIO4ALL or any other drivers available in Preferences)

Hello, I am playing an acoustic instrument into Ableton and listening back to it through software with some audio effects, and the audio latency is much longer than it's supposed to be.

To measure the real latency I bypassed all my effects, placed a mic directly in my headphones, snapped my fingers nearby, and measured the time between echos of the snap. No underruns occurred during the test. All audio is wired (no Bluetooth), and my headphones were powered off.

Setting: 512 samples (48 kHz audio)
"Overall latency" in Preferences: 27.7 ms
Measured latency: 77 ms

Setting: 64 samples (48 kHz audio)
"Overall latency" in Preferences: 5.33 ms
Measured latency: 55 ms

Setting: 1024 samples (96 kHz audio)
"Overall latency" in Preferences: 27.7 ms
Measured latency: 56 ms

Setting: 128 samples (96 kHz audio)
"Overall latency" in Preferences: 5.33 ms
Measured latency: 30 ms

So, in addition to the latency predicted by the Preferences window, there is an extra latency of about 2400 samples (~50 ms at 48 kHz, and ~25 ms at 96 kHz).

I've heard it claimed that a delay of less than 100 ms is imperceptible, but even 30 ms is very hard to deal with IMO. It would be excellent if I could get it well below 30 ms without the extra CPU load of using 96 kHz audio.

miyaru
Posts: 1267
Joined: Tue Jan 10, 2017 12:08 pm
Location: The Netherlands

Re: ASIO latency much longer than predicted

Post by miyaru » Sun Apr 04, 2021 9:08 pm

FWIW:

The cheaper the interface, the worse is the latency.... sorry.

I don't have the best interface in the bussiness but I can run it at 32 samples. Still I find it unacceptable how the latency is with some or the be honoust most FX.

Therefore I have an old Lexicon MPX1 patched in digitally in the monitor path of my Focusrite Scarlett 18i20 3rd gen.

This way I can listen to an monitor signal with FX, record dry and add FX later on in any DAW. This is the only way to get FX almost latency free, as digital domain always adds latency.

So record and monitor dry, or do as I did........ Good luck!!!!
Greetings from Miyaru.
Prodaw i7-7700, 16Gb Ram, Focusrite Scarlett 18i20 3rd gen, ESI M4U eX, Reason 12, Live Suit 10, Push2, Presonus Eris E8 and Monitor Station V2, Lexicon MPX1,
Korg N1, Yamaha RM1x :mrgreen:

pottering
Posts: 1802
Joined: Sat Dec 06, 2014 4:41 am

Re: ASIO latency much longer than predicted

Post by pottering » Sun Apr 04, 2021 9:27 pm

Bypassing effects doesn't remove latency from plugins in Live, the plugins are only muted and bypassed, not actually removed from the signal chain (this one is the main suspect in your post IMO)

USB adds some latency by itself, could be as much as 5 ms (maybe for just input, so double for both input/output).

Using ASIO or Wasapi in non-exclusive mode basically turns all the shown numbers into pure bullshit, the real latency will have almost no relationship with those buffer-derived numbers displayed in the drivers options.

People using ASIO or Wasapi drivers in non-exclusive mode to listen Youtube while using a DAW are basically deluding themselves.

The number I read that is not perceptible by most people is about 13 ms, 100 ms is a huge latency that anyone can perceive.
♥♥♥

Envergure
Posts: 21
Joined: Sun Apr 28, 2019 12:37 am

Re: ASIO latency much longer than predicted

Post by Envergure » Sun Apr 11, 2021 10:03 pm

Thanks for the replies!

Pottering was correct--delay compensation for the disabled effects was still being applied to the live set. That was the main culprit in this case.

Without the FX chain, the delay is too short to measure by the method I described, and barely perceptible to me.

Post Reply