Low latency recording With VSTs?

  • Thread starter Thread starter KyleScott
  • Start date Start date
KyleScott

KyleScott

New member
I have a Behringer UMC404 at 3 ms-6 ms; actually, I can get it to 2.4 ms, but it loses quality. As a classical pianist and beatboxer, it is still too high for me. What should I buy? Does the MIDI controller matter at all in the round-trip travel time for playing? My monitor is 16.667 ms at 1080p, which is different, but I can still feel a difference between it and a higher-end 1 ms input latency monitor. So I am sure audio playing is the same, if not worse, because I am an extreme auditory learner. Can we get audio interfaces with billionths of a second latency or nanoseconds? or am I best off playing real instruments? I know multitamberal workflows for instrumentals—not voice—have no latency as they use the hardware synth as the sound engine and not the computer. They also render the final output using the multitamberal synth. Some of them have a sequencer too and skip the computer part completely, which allows you to record voice directly into it. Multitamberal 16-track midi synths often sound plastic compared to VSTs because they use ROM. What is the closest we can get for 0MS playing or multitamberal type workflows but still with VST sound?

Maybe a soundblaster pciex16 setup is better for low latency with a multitamber setup, maybe also with a VST setup?
 
Can you tell the difference in the sound vs the visual of a clap from 6 ft. That's a 6ms delay in air at 20C. The speed of sound is 1125ft/sec or 1.125ft/ms. At a grand piano, you are probably about 3-4 ft from the strings, if you measure from the middle of the string to your ear. Then you have the reflections from the sound board and lid. Does that delay bother you?

With a USB interface, you will never get below the calculated latency which is Buffer/Sample Rate. 128 samples/44100 samples/sec is 0.0029 seconds. You can double the sample rate to cut it in half. 128/88200 is 0.0014 seconds. That's one way. You need to double it for two way communication. Then you need to add in any processing if you are using VSTs. That will depend on the VST, of course. I don't know if 16bit is faster than 24 bit (you are processing 50% more data points).

With microphones or electronic keyboards, you can direct monitor, which eliminates any delay since the signal is routed directly from the input to the headphone output.

A Thunderbolt interface can theoretically cut the latency significantly since it has direct control of the CPU. A Quantum 2626 interface with 32 sample buffer on Thunderbolt was tested at 0.9ms input and 1.0ms output. RME interfaces are supposed to have some of the most efficient processing, yielding low latency. From the numbers I've seen, 2ms round trip is a very common value. Most of these are on Mac systems, since Thunderbolt has been standard on them for years.
 
Rich makes the point well - if you have a delay between hitting the key and hearing it, it is probably not a latency issue with the interface, but latency inside the computer elsewhere. My concert pianist friend swapped computers and couldn't play. The latency value (I think cubase and his modest interface showed 6ms or so was unplayable. My system at the time was 11ms and he played that fine. It didn't mess with his head. On his system, the problem was Cubase and Sibelius 'fighting'. Somehow with Sibelius open, Cubase, despite the displayed latency value was delayed. We worked out that the 6ms was actually around 18ms. We really struggled, and in the end, we deleted both apps, did a windows update - which only updated a few things it reported, then reinstalled the two apps and the driver for the audio interface that we had also deleted. Cubase still showed the same latency, but this time, there was no delay.

We discovered his ability to deal with the delay was about 9-12ms Above that something made him play less well and get agitated. My system at 11 also had a non-weighted keyboard. Swapping it for a bigger 88 note weighted one was worse. Could this be an additional snag for you. The weighted keys add a physical small delay before the note on is generated - common when playing slowly. Could this, plus the computer delay be the straw breaking the camel's back, so to speak?
 
There has probably been more space devoted to "latency" in the audio press and forums than most topics barring perhaps "why are my electric guitar recordings ***t"!

I personally doubt a Behringer 404 is delivering 'just' 3mS of latency? Behringer are not known for superb drivers for their budget interfaces.
That reported 3mS probably has "hidden" buffers in it if you can hear a delay.

A better interface and most importantly, better ASIO drivers should help. I can cite two that are very good but not RME cost. The first is the Native Instruments Komplete Audio Six (the Two might be OK, never had one) Later and better IMO is the MOTU M4. MOTU have long had a reputation for excellent drivers, second only perhaps to RME?
My M4 will play a 25track 'demo' recording containing VSTis and plug ins at 64 samples (1.45mS) without glitching on this i7 Lenovo T510 lappy, so, NOT "Deep Thought" !

So, either something, probably the computer, is putting in extra delay or you are particularly sensitive to latency. If the latter, do not take up the church organ.

I am FAR from any guru here but, doubling sample rate will, in theory halve the delay but there is no guarantee the CPU can cope with the extra data load and clicks and stuttering will happen anyway. BTW. Are you running the MIDI via USB or the DINs on the 404? MIDI DINs can sometimes be faster.

Dave.
 
Back
Top