Question on Latency

[rlared]

I have a question about latency. I know this is a big topic and is quite complex, but I figure there's some experts out there amongst us. This question is spawned from the MSoundFactory discussion where some of the built-in effects introduce latency.

I understand if you have a latency-inducing effect in your VST instrument, there's going to be a delay from when you press your keyboard key to when you hear the output of MSF. I believe the DAW will record when you actually pressed the key, but the delay in hearing output will screw up your timing.

What I'm not sure about is, let's say you input MIDI notes into the piano roll and MSF is adding 1 second (to be extreme) of latency. Then you go to record vocals. Let's assume you're monitoring your vocals through your audio interface at zero latency to make that part simple. Will the DAW compensate your vocal recording knowing that there's a latency between when the MIDI piano roll note triggers MSF and when it actually plays out through your speakers? So basically would it move your vocals backwards by a second ?

If it's DAW specific then I'm curious about Reaper and Cubase.

[reggie1979]

Cubase and Reaper have delay compensation. The most important thing when recording is the latency of your ASIO buffer (or core audio on apple)

For example, my buffer is set to 64 giving me about 5-7 ms of latency. It's hardly noticeable. However, the lower your buffer, the bigger CPU hit.

[jancivil]

First, in Cubase Preferences: Record/MIDI tick ASIO Latency Compensation Active by Default
This obviates some if not most, if not all of this hypothetical issue.

Per the assumption the MIDI will be recorded with no latency, the DAW 'records when you hit the key', no, the latency exists here first.

Depending on compensated or not, or settings in "Adjust for Record Latency", the quality of that compensation and so forth, you can end up with things recorded all over the place as seen in Key Editor. So once you have the MIDI part, where your timings are are known to you; there may well be additional latency on playback and it might not be as predictable (I have no idea of that product), but timing of your notes is already subject to the latency of your system and the buffering you're using.

If the FX seem so late as to be a big issue, you can record the vocals via ASIO Direct Monitoring, where the computer's latency is bypassed and it's just the negligible ("0") latency of the audio interface. If the FX are part of the instrument's audio coming back to Cubase, ie., as known to the Instrument Channel in Cubase, I don't know what to tell you except to commit to and print that MSoundFactory track as compensated and record vocals using Direct Monitoring anyway.

In a typical record vocals situation, the FX are more of outboard paradigm in a Cue Mix scenario, and basically to hear FX in that external sort of cfg, you need the proper audio interface providing a Cue Mix scenario.

[rlared]

OK well I know delay compensation accounts for the various plugin latencies and basically takes the "worst" track latency and adjusts everything else to play back a little later accordingly so it's all lined up. But does that also work when recording audio for example? I guess I'm trying to figure out, with the exception of an instrument track having latency where you hear it after you play it (which is clearly a problem), why is it bad if other tracks introduce latency?

[jancivil]

Recording audio itself is not subject to the vagaries of MIDI to a virtual instrument, no. There's latency, yes.
I record (instrumental or vocal) as audio when there's nothing going to soft instruments, it's either a blank slate or I've committed the parts, printed to audio so there's not so much latency. If the computer exists between you and a recording there is latency. Direct Monitoring removes that from the equation. This gets complicated at this stage, and I would recommending printing the MSoundFactory result before doing vocals.

Steinberg, however has an interface for the "Cue Mix" scenario (first I heard of that was MOTU) where you can use the Control Room and ASIO Direct Monitoring simultaneously. I haven't done, so I can't tell you.
When Steinberg launched the MR816, however, they also adapted Cubase so that, with their own hardware, the Control Room and ASIO Direct Monitoring could be used simultaneously.

https://www.soundonsound.com/reviews/steinberg-ur824

[rlared]

First, in Cubase Preferences: Record/MIDI tick ASIO Latency Compensation Active by Default
This obviates some if not most, if not all of this hypothetical issue.

Per the assumption the MIDI will be recorded with no latency, the DAW 'records when you hit the key', no, the latency exists here first.

Depending on compensated or not, or settings in record compensation, the quality of that compensation and so forth, you can end up with things recorded all over the place as seen in Key Editor. So once you have the MIDI part, where your timings are are known to you; there may well be additional latency and it might not be as predictable, but timing of your notes is already subject to the latency of your system and the buffering you're using.

If the FX seem so late as to be a big issue, you can record the vocals via ASIO Direct Monitoring, where the computer's latency is bypassed and it's just the negligible latency of the audio interface. If the FX are part of the instrument's audio coming back to Cubase, ie., as known to the Instrument Channel in Cubase, I don't know what to tell you except to commit to and print that MSoundFactory track as compensated and record vocals using Direct Monitoring anyway.

In a typical record vocals situation, the FX are more of outboard paradigm in a Cue Mix scenario, and basically to hear FX in that external sort of cfg, you need the proper audio interface providing a Cue Mix scenario.

Regarding MIDI, can you elaborate on what latency there is when recording a MIDI note input from a keyboard for example? (assuming you're recording MIDI in the DAW)

Where you mention ASIO Direct Monitoring above, that's where my question is. The playback that you hear and sing along with is going to have some sort of latency, right? For example, on an instrument track, it would route a "D" note into the synth, which produces audio, which goes into an effect which requires a bunch of CPU processing, which causes latency, which goes through your DAC to your speakers, which also adds latency. The DAW would then say "OK the instrument and plugin cause 50 ms of latency, so I'm going to play the other zero latency tracks 50 ms later." So then you sing along, and the time between when the "D" note was triggered to when your vocals were received is 50 ms + DAC latency + speaker latency + ADC latency + processing latency. So does your DAW account for any of that by moving the vocal "backwards" after recording?

[jancivil]

"The playback that you hear and sing along with is going to have some sort of latency, right?"
No, the part is on time as playback, the ability to record on time using the soft instrument as effected by high buffering is still there, but now playback is predictable.

External FX, let's go with the most typical case, reverb, is taken out of the equation, IE: sends, inserts *in Cubase* won't be heard. So people with recording studios clamor for a classy sounding Cue Mix and DAW developers came up with hardware with "Cue Mix" capability, in Cubase's case you use Control Room with certain buttons pressed.

[rlared]

OK I don't quite understand the buffering aspect in terms of how it affects playback latency. . . I will have to do some reading. It's a tricky topic. Thanks for the explanations though.

[jancivil]

The DAW would then say "OK the instrument and plugin cause 50 ms of latency, so I'm going to play the other zero latency tracks 50 ms later." So then you sing along, and the time between when the "D" note was triggered to when your vocals were received is 50 ms + DAC latency + speaker latency + ADC latency + processing latency. So does your DAW account for any of that by moving the vocal "backwards" after recording?

First, that soft instrument latency is not predictable like that. I don't see how it would, itself do such a thing.

What you *can* do is go into "Device Setup", now known as 'Studio Setup', VST Audio System, and tick Adjust For Record Latency with a guess as to how many samples late you're seeing, or actually in many cases for me the compensation has things land early in the time line.

Since you don't seem to need to record the MSound MIDI at the same time as the vocals, I recommend committing the part(s) to audio before dealing with the vocal recording and set your latency really low.

You're working from an untrue premise as to late playback in the first place though.

[jancivil]

There is no playback latency.

[rlared]

The DAW would then say "OK the instrument and plugin cause 50 ms of latency, so I'm going to play the other zero latency tracks 50 ms later." So then you sing along, and the time between when the "D" note was triggered to when your vocals were received is 50 ms + DAC latency + speaker latency + ADC latency + processing latency. So does your DAW account for any of that by moving the vocal "backwards" after recording?

First, that soft instrument latency is not predictable like that. I don't see how it would, itself do such a thing.

What you *can* do is go into "Device Setup", now known as 'Studio Setup', VST Audio System, and tick Adjust For Record Latency with a guess as to how many samples late you're seeing, or actually in many cases for me the compensation has things land early in the time line.

Since you don't seem to need to record the MSound MIDI at the same time as the vocals, I recommend committing the part(s) to audio before dealing with the vocal recording and set your latency really low.

You're working from an untrue premise as to late playback in the first place though.

Well it's all just a hypothetical, I'm just trying to understand how latency works, and why people are so concerned about it. Other than the case of playing a live VST instrument, I was trying to figure out why people care about latency so much.

[jancivil]

Well, recording the vocal with FX (or with the instrumentalists behind her playing soft instruments, or using DAW FX, etc at that time, hypothetically) is something people care about and you're dealing in record latency; the issue isn't playback latency, recording the vocals with FX means similar quality latency as with soft instruments. ASIO Direct Monitoring means that's out of the equation, but so are the FX. Except you can do with such an interface and teh Cue Mix. Which is all hypothetical to me, I never did.