Printed from:
Les Brockmann Music Engineering . Writing ON MUSIC & ENGINEERINGOn Music & Engineering

Wednesday, November 29, 2006

Beat Monitoring Latency with MOTU Cuemix -- part 1

Fig. 1

Supported equipment

First of all, MOTU's "Cuemix Console" software is available for both Mac and Windows (supplied free with the hardware installer discs or online), and works ONLY with the following computer audio interfaces from Mark of the Unicorn:

All old and new PCI interfaces can use Cuemix with the 424 card (this applies both versions of MOTU's PCI-424 card, PCI-424 and PCI-424e): 2408, 2408MKII, 2408mk3, 1224, 308, 1296, 24i, 24i/o, HD192.

The PCI-324 card does not support Cuemix on OSX. (However, if you are still using a 324 card, you can contact MOTU, (617) 576-2760, for a reasonably-priced upgrade exchange.) The 24i/o, 2408mk3, HD192 do not work with a 324 card on any OS.

All FireWire interfaces (and the 828mkII USB 2.0) support Cuemix except the original 828 and 896.

So, if you use any of these audio interfaces, read on. It doesn't matter which music/audio/MIDI software you use, i.e. Digital Performer, Logic, etc. But if you use any other brand of audio interface (or any Pro Tools hardware or software), this doesn't apply to you.

Latency — where does it come from?

Let's review: Latency delay is a problem which is characteristic of audio interfaces which do not have onboard audio DSP, but instead use the computer's processor for all audio routing and manipulation, often referred to as "native" signal processing. (There is some latency in all digital audio devices, but it's generally not considered a problem with devices offering less than 1 or 2 ms. throughput delay.) When a "native"-hardware-equipped computer system receives an incoming audio signal from an external source such as a synth/sampler or microphone, there can be, depending on the buffer setting in the software, fairly substantial delay, up to several tenths of a second.

Why is this? Analog audio comes in through one or more of the inputs, is converted to a digital signal by the interface box, and then passes into the computer, goes through its buffers, memory, and processor, is routed into your audio software and goes through whatever routing and processing it may do, and then, if you are set to monitor that signal, it must turn around and go all the way back out, and then be routed to your speakers or headphones.

In order to do this the computer needs a number of processing cycles; this delay is usually measured in number of samples. Your audio software will have a setting, somewhere, that allows the user to set how many samples the computer is allowed to use as it routes and processes the audio. This will probably be from 64 to 1024.

Why would you change this setting? If you have set an audio track to record sound from some outside source, and have the software set so that you can monitor "input" in the standard old-fashioned way, you will probably notice that the larger this number, the more delay you will hear, and as the number is set smaller, the delay is decreased. If you, or another musician, is playing and monitoring the live sound through headphones, you will find that any setting much greater than 64 or 128 samples leads to an unacceptable delay between the live performance and the monitored sound. And if you put up your "Processor Monitor" window in your software, you will see that the computer is working harder when the number is set smaller.

Here is an imaginary scene that might help to visualize what is going on in your computer: Suppose you are trying to transfer sugar from one barrel into another, and you need to move one cup of sugar every ten seconds. First of all, start with a coffee mug; it should be fairly easy to scoop in the first barrel and then dump it into the second barrel once every ten seconds (this corresponds to your computer with a buffer setting of a large number, such as 1024). Now, instead, use a tablespoon — you will have to move much more quickly in order to move the same amount of sugar in the same amount of time (this corresponds to a smaller buffer setting). What if you used an even smaller measure, such as a teaspoon? At some point you will no longer be able to keep up, and either fall behind or spill the sugar. Either way: bitter tea!

Your computer has the same situation going on, pouring audio digits through its circuits and in and out of your hard drive. If you are using plugins, soft synths, or video, that takes processor power as well. If the buffer number is set so small that the computer can't keep up, the audio you are trying to record will end up with pops, crunches, or distortion. In extreme cases the software will bog down or crash.

What if your computer simply doesn't have enough power to keep up with all you are asking it to do, to run at a low-latency buffer setting, to monitor incoming audio with a manageably short amount of delay, at the same time — now what do you do?

Cuemix Console to the rescue

Suppose you could monitor incoming audio sources with practically no delay at all, and at the same time let your computer breathe easy with a large (1024) buffer setting. With Cuemix Console, you can.

On the MOTU PCI-424 card, and in their Firewire and USB2 audio interfaces, there is a special "mixer" chip especially intended for this type of monitor mixing. This can be set to take any incoming signal and immediately send it back out an output, without it having to take the time for the journey all the way through the computer. Cuemix Console is simply a control panel for this mixing capability that lives in the hardware.

To use it, first change the setting in your audio/MIDI software so that it does NOT monitor the input signal from any record-enabled track at all. (When playing back a recorded track it should monitor output as usual, whether or not the track is record-armed.) Instead, we are going to use Cuemix Console to derive a mix of the live sources.

Notice in each page of the Cuemix Console, there is a fader, pan pot, and mute button for every input channel of your hardware. Notice also, on the right-hand side above the "master" fader, there is a pop-up which lets you select any of your hardware's outputs in stereo pairs. (See figure 2.) You can create a completely separate mix of the inputs, routed to any output pair.

Fig. 2

Start by selecting the output pair that you want to use for monitoring. If you are working without a mixing console, that would be the pair that routes directly to your speakers and/or headphones.

Next find the fader or faders that corresponds with the input source(s) you want to monitor. This would be the signal from a microphone or similar device. Notice the faders are labeled according to the inputs on your MOTU converter. If you want to see names that are more logical, you can change the names in "MOTU PCI Audio Console" (a separate application that is also provided with your MOTU interface) and then those names will appear in every application that accesses the interface. (If you always keep things plugged in the same place, try names such as "Mic1", "GtrPod", etc..)

That's all there is to it. Raise the fader and set the pan and you will hear the source. Make sure the "master" fader is up too, and the "mute" buttons aren't muted.

Caution zone

There are two important limitations to understand about monitoring through Cuemix Console. First, the sound that you hear in your speakers is now not the same as that reaching your mix bus in Digital Performer or other audio app. That is, DP doesn't "hear" the live mic or other source being monitored. So, this won't work for mixing, such as monitoring external synths or other instruments with lower latency. You'll still have to record them onto tracks or use an aux channel to monitor them for mixing.

Secondly, after recording, don't forget to turn off (mute) the channels in Cuemix Console before you quit that application, or the source sound will still be "live" in your speakers. Remember, Cuemix Console is just a "control panel" for the monitor mixing capabilities in your audio interface — it has to be told to no longer route those channels for monitoring or it will continue to do so.

How dry it is

One other little problem with monitoring a source with Cuemix Console is that the sound will be "dry", no reverb. Since you are not monitoring the source using your main audio app such as DP, you can't just turn up an aux send and use a plug-in 'verb. When you play back the track it will be wet but when you are recording it will be dry, unless you can add some 'verb after the output, such as if you have a mixing console.

There is a work-around for this. Have you got an old hardware reverb box around that you're not using? Lots of folks do (how about that old LXP-1?). It can have digital or just analog inputs and outputs. Patch it directly into your audio interface with two channels in and out. (This is another case where it might be convenient to label those ins and outs in the software.) Leave it there and let it be dedicated for just this purpose.

Now, in Cuemix Console, use the pop-up on the right to select the output pair that feeds your monitor speakers or headphones, and find the two channel faders that represent the input to the interface from the output of the reverb; unmute, raise, and pan them left and right. Then go to the output pair that is feeding the reverb inputs; find the channel fader that corresponds to your microphone or other source and raise the fader a bit. This is your "send" to the reverb. That's all there is to it. Season to taste by adjusting the "send" fader. Being a different reverb device for the source than for the track output, it won't be quite the same sound as when you play back the recorded track, but you should be able to get it close enough.

One thing that can make it a little easier to work with CueMix mixing features: most major DAW packages (DP, Cubase, Logic, etc.) have a way to control CueMix monitoring (enable/disable, level, pan, etc.) from within their mixing window. This is covered in the manuals for the 828mkII, Traveler, UltraLite, etc., and is cool because it means the user doesn't have to switch to CueMix Console to manage the input monitoring. Read the manuals for details on how to do this with each app (DP, Cubase, etc.).

A useful compromise

That's all there is to it. Cuemix Console is a useful tool that helps to solve the latency problem under certain conditions. Is it as effortless and convenient to use as working with ultra-low-latency computer audio hardware such as Pro Tools? No, but if it's worth it to you to save the substantial difference in money between that and the MOTU audio interfaces, then it's nice that it is available.


I'm going to go up on a limb and make a prediction: that our current problems with latency delay, using "native" computer audio hardware, isn't going to be a permanent thing. For example, I've just been reading about a new interface from Apogee that lowers latency to 32 samples or 1.6 milliseconds of latency at 192 KHz sample rate (which beats Pro Tools under certain circumstances); my sources at MOTU report they are also working with the latest fast computers and hardware with very low latency. Count on this to ignite a competition among all the manufacturers.

As computers and digital hardware get faster and more powerful, someday soon we're going to look back at this whole latency hassle as "old news". When that happens, remember you read it here first.

Thanks to Jim Cooper and Matt LaPoint of MOTU for their help in researching this article.

This is not intended as an advertisement for any product. All products mentioned are trademarks of their various manufacturers. ©2006 by Les Brockmann.

Previous Posts