Lag in audio

gpoorman

Hi all. I'm new to the forum and have only recently returned to Mobius. Quite some time back I went with a full computer setup for live playing. Not happy with any software loopers, I kept my EDP and used the Mainstage I/O plugin to continue using the hardware EDP for looping. With a 64-bit Mobius, I'm trying to go back to software. I had everything setup but something was just off. Playing around today, I narrowed things down into a case I think I can describe. I'll record a loop playing steady quarters notes ... say ... around 140bpm. I record 16 notes. Then I jump right into overdubbing and, again, play 16 notes this time a 5th up. As I'm playing the overdub, my timing is dead on. Once the overdub is finished and the loop plays back, it appears to start off well and then as I get into beats 8 to 16, I start to hear a noticeable "flam" (for lack of a better word). Like the two layers go off. I'm not sure what to look at this point as far as settings go. I'm using a UA Apollo X4 audio interface. I've got the audio setting such that Mainstage is showing a round trip of 0.9ms. There is no audible latency. Looking for some advice as to what to try to tighten this up. Thanks, Glenn

jeff

Usually this is latency compensation. From the Help menu select Environment and look around 4 lines down for this:

Audio block size: 256

Make note of that number, it should match the hardware block size being used by Mainstage but hosts don't always report that accurately. I don't have Mainstage handy, but bring up whatever it uses for audio device configuration and see if they report a block size rather than just a millisecond time. .9ms is about right for 256.

If the number you see in Mainstage is not the same as the number you see in Moibus' Help->Environment, then this needs to be adjusted. From the Session menu select Edit Session, then the Globals tab. In the tree on the left select Latency. There will be two fields in the form labeled Input Latency and Output Latency. Usually these are left at zero which means to use whatever the host is reporting as the block size (which is what you saw in Help->Environment). You can type in numbers here to fine tune compensation. If the numbers in Help->Environment were wrong, enter the number that Mainstage uses for both of them.

gpoorman

Thanks Jeff. So I misspoke a little. The 0.9ms was the output latency. I was started to hear some other artifacts though so I think I might have been overloading my computer a bit. So for accuracy (I'm looking at everything now), my Mainstage settings are:
Sample Rate: 88.2KHz
I/O Buffer Size: 32 samples
Resulting Latency: 4.8ms roundtrip, 1.1ms output

In Mobius, the environment shows Audio block size: 1024

I don't know if the I/O buffer size in Mainstage is what I should be looking at. I don't see the equivalent of the audio block size (unless that's it).

jeff

Well, for starters that's a very low block size, especially at that sample rate. That will cause a lot of CPU stress if you're using anything other plugins besides Mobius. It might work, but it's definitely in the danger zone. When you go under 128 or 64 for the block size, the host may decide to give the plugins larger blocks that it actually uses with the audio interface. That's why you see Mobius saying the block size is 1024. Mainstage may actually be doing another level of buffering above the audio interface, or it may just be combining 32 32-bit blocks into one when it calls the plugins, it's hard to say. You could start by going to those Input Latency and Output Latency overrides I described earlier and type in 32 for both.

I'd recommend that you increase your block size to 128 though, especially if you're using other plugins. Again, if Help->Environment still shows 1024 then you'll need to put in 128 for the Mobius latency overrides. Basically, for Mobius to do latency compensation correctly it needs to know exactly what the block size is being used between the audio interface and the host. But the host is free to change that when it passes blocks to the plugin. All I can see is the block size the host gives me, 1024, which is not accurate, so that's why the two latency overrides exist for you to type in the right values. It's not something most plugins care about, but loopers do.

Typically, really low block sizes are used when you're trying to monitor "through" the computer to reduce your instrument-to-ears latency. Most people use "direct monitoring" instead where you connect your monitor speakers directly to the audio interface and bypass the computer for monitoring. I strongly recommend this, it will save lots of headaches if you can set it up that way. This is fine for looping because Mobius will compensate for latencies and keep things lined up with what you are hearing. That does mean however that you can't use amp modeler plugins if that's what you're using to process your sound. If you need to do that and monitor through the computer, then there are some other latency adjustments that will need to be made to make overdubs line up with the sound you hear in the monitors.

I'd also recommend using 44.1 or 48 for your sample rate, again to reduce the possibility of CPU overload. If the goal of all this is live performance, then the audience isn't going to be able to tell the difference.

jeff

"I don't know if the I/O buffer size in Mainstage is what I should be looking at. I don't see the equivalent of the audio block size (unless that's it)."

Yes. Hosts have different ways to talk about this, what I call block size is usually the same as "buffer size". I haven't used Mainstage in several years, but I'm pretty sure that's it.

gpoorman

This is great info!! I'll spend some time with this today.

I have seen the same recommendations before regarding sampling rate and buffer size but when I do that (monitoring through the software) I always feel a disconnect between what I'm playing and what I'm hearing. As far as monitoring through the audio interface, I do use interface plugins for my pre-amping but I use AU plugins in Mainstage for all of my additional effects. I was under the impression that if I monitor through the audio interface, I would lose all that. Is that correct? Wouldn't I even lose the Mobius output?

jeff

Yes, if you need to use plugins to process your live sound then you would need to monitor through. The lower the buffer size the less lag there will be between what you play and what you hear. 128 seems to be about the upper limit for what is tolerable for most people. You can go lower, but then it starts adding stress and risks audio dropouts of you use demanding plugins. I would say mobius is in the low-to-average range for CPU demands compared to amp modelers, or anything else that does a large amount of signal processing. Mobius doesn't really do any DSP, it just copies what comes in and throws it back out with a delay. But when you start combining multiple tracks and consuming a lot of memory it can add up. With lower buffer sizes you should also use a 44.1 sample rate.

When you monitor through the computer, you are already doing a form of mental latency compensation. Since there is that slight lag, you tend to adjust your play style to compensate, playing a little early, much like you would if the monitor speakers were a long distance away and had to compensate for "air latency". In that case the latency compensation that Mobius does is not necessary since you're already doing that. It seems counter-intuitive but what you want in this case is to turn Mobius output latency compensation off. To do that go to the Output Latency field and enter the number 1. You can't put zero there, because zero means "whatever the Host tells me the block size is". In your case since the host is saying 1024, Mobius will by default add a very large amount of compensation which you don't want. 1 isn't technically none but it's close enough.

Input latency may or may not still apply, this depends on the degree to which you are "playing ahead" to adjust for the monitoring lag. Start by setting that to 1 as well. If it feels like Overdubs are being recorded a bit to early, recording content before where you want it to start, then increase that number up to the hardware buffer size until it sounds about right.

jeff

Addendum for the UAD...I'm sure you know this, but the Apollo is effectively a hardware plugin host that can can host it's own amp modelers and effects. If you find what the Apollo provides for your sound processing is acceptable, you can move that over there and use hardware direct monitoring, then just use Mainstage for Mobius and other things that aren't part of your instrument sound.

This is what I usually do, hardware for the guitar sounds and the computer for looping, delays, and backing tracks. You'll get the best "feel" from the instrument since it doesn't have to take the tortured path all the way through the computer. But that obviously doesn't work for everyone.

Looping is actually one of the easiest things to offload to the computer since everything it does is delayed. As long as it can do latency compensation correctly, overdubs should line up just fine no matter how large the buffer size is.

gpoorman

Hi Jeff. Ok ... so monitoring aside. I left my sample rate high but changed the buffer size to 128. This didn't produce any noticeable quality difference to my ears so that's good. I manually changed the latency numbers in Mobius to 128 like you suggested and ... low and behold ... it was like a fog lifted!! Just fantastic! I just sat and played for an hour straight before coming up for air.

So three other questions (one related and two not so much).

The "Mute Mode". I tend to always set my quantize to "cycle" (note I'm a long time EDP user). On the EDP if I set my mute mode to "start", unmuting will start at the beginning of the loop immediately with no regard for what quantize is set to. In Mobius I'm seeing that mute mode is following the quantize setting and waiting for the start of the next cycle to unmute and then starting from the beginning of the loop. Can I get the unmute to restart immediately without turning quantize off?
Is there a way to program a long press of "Multiply" to do a global reset?
Back to the monitoring. I understand what you're saying as far as the effects in Mainstage vs the UA plugins. I guess I'm not exactly sure how to accomplish that. So would you essentially turn off routing of the original signal back to the audio interface then? Basically monitor the original signal from the interface but the only thing coming from the computer is the output of delays, Mobius, maybe reverbs?

gpoorman

@jeff I'd really like to pick your brain a little bit on the monitoring side of this but I don't want to eat up forum space (or your time) with a topic that really isn't Mobius related at all. Is there an email or something where I could hit you with a couple of added questions? Or perhaps just start another topic and mark it as "off-topic"?

jeff

gpoorman Sorry, I've been traveling the last two weeks and haven't been very active here. Let's continue the discussion here since it may be useful to everyone, but if you ever want to contact me directly I'm at

  seriouslyjefflarson@gmail.com

gpoorman

Cool! So I think this is where I get a little tripped up. I understand mixing the monitoring. I use UA plugins for my instrument preamp and use Mainstage to apply delays, reverb, looping. So based on our discussion last week, I tried unmuting my channel in the UA console which gives my hardware monitoring of my instrument signal (and UA preamp plugin). Then I started with a patch in Mainstage that just uses reverb. In order to get the monitoring you described, I sent the signal in my instrument channel to an aux/bus channel in Mainstage where I applied reverb mixed to 100% wet. I sent that bus output to my main output and turned off the output of the original instrument channel. Voila! This was exactly what you described. So far so good.

But now I want some delay also in Mainstage. So now I need to output the reverb of the orginal signal plus delay output both clean and with reverb. I can do that by also using an aux/bus channel for the delay. Output the delay bus back to the main output and also send it to the aux/bus I was already using for reverb. The original channel still has no output. I'm still good but now I'm using an instrument channel along with two aux/bus channels just to apply some delay and reverb.

Add to that now that I have a couple dozen patches in Mainstage with various delay and reverb settings and I feel like this can get super complicated really.

So this has me wondering if I'm missing something obvious that would simplify this? Or maybe this is more of a Mainstage question? Or maybe I just can't make that setup work?

Hopefully you followed all of that ;-)

gpoorman

So in the interest of Jeff's comment where he said other people might be interested in this topic, I've spun many cycles on this and I think I essentialy have it though it's not real straightforward. I play both acoustic fingerstyle guitar and also the Chapman Stick. The Stick is two audio signals so I started with my guitar setup just to keep things simple. Using the Apolllo X4 and Mainstage, this is how things are setup currently. Again, the trick is to have the entirety of the original signal monitored through the Apolllo and the effected (and looped) signals monitoring through Mainstage.

So basically I have:
guitar -> instrument channel -> no output
send 1 -> patch level delay bus
send 2 -> set level reverb bus
send 3 -> set level looper bus

delay -> patch level aux channel -> output back to Apollo
send 1 -> set level reverb bus
send 2 -> set level looper bus

reverb -> set level aux channel -> output back to Apollo
send 1 -> set level looper bus

looper -> set level aux channel -> output back to Apollo

delay, reverb, and looper are all set to 100% wet.

Note that I'm able to use set level busses for reverb as I really only have a couple of different reverbs I use. The delays, however, required a bus for each patch as the delay settings are different.

So for just the audio, this worked and the guitar sounds noticeably better. I ran into a snag, however, when I realized that my MIDI foot controller with an expression pedal set to control the overall guitar volume no longer worked. Since it's mapped to Mainstage it only controlled the volume of the effected sounds. Out of the box, Apollo devices are not setup to allow MIDI control of various parameters. As it turns out, a 3rd party gentleman named Radu Varga created the UA MIDI Control App several years back which turns out to be exactly what I needed. This app allows you to map incoming MIDI messages to all manner of parameters in the Apollo. So using the app, I mapped my MIDI foot controller to not only continue to control the volume in Mainstage, but also the volume of the guitar input channel in the Apollo.

At this point I've redone the entirety of my guitar setup and love it.

Next is the Stick (which I'll likely start playing with today). For the most part, nothing is really any different except that I have to account for the instrument having two inputs instead of one. I do use other effects though that I don't use on guitar. Mainly distortion that I need to be able to turn on/off with my foot controller. First I have to decide if I want to use one of the muriad of UA plugins for my distortion or continue to use my Mainstage plugins. Then it's just a matter of reprogramming the foot controller to turn the various things on/off to make the switch from one tone to another.

Anyway ... that's it. Hope somebody gets something from all this.