MAXIMISING YOUR VIRTUAL INSTRUMENTS
Maximising Your Virtual Instruments
Technique : Computers
These essential survival tips will help to prevent your appetite for virtual instruments from exceeding your computer's ability to run them.
Paul White
Over the past couple of years I've become a big fan of virtual instruments, especially those that conform to the VST plug-in standard. However, despite the power of current desktop Macs and PCs, some of these instruments place heavy demands on your available CPU capacity. Many people seem to confuse CPU power with RAM capacity and assume they can solve speed-related problems by fitting more memory, but this is not the case. While it is true that you will have problems if you have insufficient RAM to run your music software properly, fitting more than you need will make no difference at all to the performance of your system. While I don't claim to have all the answers, I'd like to pass on a few tips I've discovered for making the most of virtual instruments in practice.
Rules Of Thumb
If you want to maximise the number of virtual instruments you can run, the first thing to do is make sure you avoid inessential CPU demands such as screen redraws. Close unnecessary windows and scale your arrange window so that it fits the complete song. This will avoid a hit on your CPU overhead as the screen redraws after the song cursor has run off the end.
Low latency settings make virtual instruments easier to play, but also make your system more likely to fall over when pushed close to its limit. You need low latency while recording so, to reduce pres
sure on your system, mute any virtual instrument tracks that are not essential to the performance of your overdub. Note that you may need to remove the virtual instrument from the track altogether to completely shed the CPU load, so keep your system performance meter open to see if simply muting the track produces the saving you need. Also, be careful not to take the CPU load meter too literally. If it often shows more than 75 percent load, you're probably closer to meltdown than you imagine.
Surrogate Sounds
An alternative to muting a track altogether is to use the MIDI data of your existing virtual instrument parts to play back 'guide' sounds from a soundcard synth chip or external synth module while you're overdubbing your new virtual instrument part. Once all the parts have been recorded, you can reinstate the original virtual instruments and increase the latency and buffer settings so that you get glitch-free playback. Latency doesn't matter when you're mixing, only when you're recording.
Power-hungry effects such as reverb can dramatically reduce the CPU capacity you have left for your virtual instruments. Less sophisticated (and less demanding) reverb plug-ins sound just as good as the more sophisticated alternatives on some instruments , so try them out and see what happens. As a rule, reverb quality shows up most on percussive sounds and vocals. Furthermore, you'll make more economical use of reverb if you configure a single reverb plug-in as a send effect rather than using multiple reverb plug-ins in channel inserts. Another valid alternative is to use a DSP card such as TC Works PowerCore or the Some plug-in instruments (such as Native Instruments' Dynamo and Reaktor) that can be set to work at different sample rates take less CPU capacity at lower bandwidths, so experiment with the settings and see how low you can go before the sound quality suffers. In this case, the virtual instrument sample rate is independent of the sample rate you have set for your audio interface. You'd be surprised how many synth sounds work OK at a 32kHz sample rate -- unfortunately, only a few virtual instruments offer you the choice.
Universal Audio UAD1 card to provide high-quality plug-in effects that run largely independently of the host processor. This is a good option if you need very high-quality reverb and you don't have the means to integrate an external hardware reverb into your mix signal path. The more routine processing you can hand over to these cards, the more host power you have left for your instruments.
Adjusting Sample Rates
Running your audio interface at high sample rates will reduce the number of virtual instruments that you can use at one time. Unless you have a good reason for working at higher sample rates, I suggest sticking to 44.1kHz as that's what you'll eventually need for CD production anyway. Working at 24-bit resolution makes sense though, as it allows you to maintain quality while leaving a useful amount of safety headroom between your nominal working level and clipping.
The Lite Option
Even after applying all the economies described so far, you may still find that you can't run as many virtual instruments as you want to, especially if some of them are power hogs such as some of the modular soft synth packages. The simplest solution is to take your most power-hungry virtual instrument tracks and 'bounce' them to a regular audio track. Once recorded as audio, there will be no hit on your CPU, just the disk access demands of one more mono or stereo audio track. This option is good if your performance meter shows you are nearly maxed out on CPU but still have plenty of disk access capacity left. You lose the ability to add filter effects from your synth, but there are plenty of filter plug-in effects that you can press into service if you need to. Saving as audio does take away your ability to change the sound later in the process, but if you retain the original MIDI data track, you can always go back and bounce the track again if it doesn't work out when you come to mix.
Another possible answer is to transfer the sounds of more power-hungry instrument plug-ins to your software sampler, as this will take significantly less CPU muscle than most synths. A complex Reaktor loo
p, for example, can be saved as a single sample and will require almost no CPU power to play back. If the loop contains a dynamic control change element, then sample several bars of the loop -- RAM is cheap! Using a sampler is a good option if you still want to apply dynamic filter settings to the loop during playback, as you can use the sampler's own filter.
Simple Synth Tweaks For Reducing CPU Demands
As a rule, the more polyphony you allow a virtual instrument to have, the more CPU power it will use, and where the sounds have long release times, even a simple melody line can eat up a huge amount of polyphony. That being the case, it makes sense to set the lowest polyphony you can get away with without compromising the musical performance. You may also find that switching off the filter section of some virtual instruments reduces the CPU overhead, so don't leave a filter in circuit if it isn't needed.
Creating multisamples out of regular synth sounds is a little more time-consuming, but using a virtual sampler such as Steinberg's HALion or Emagic's EXS24 it shouldn't take more than 10 to 15 minutes to create a good-quality multisample with crossfaded loops. This can be a real power saver, especially if there are some sounds that you find yourself using a lot. Admittedly, sampling takes away much of the real-time control element of the sound, but if you're honest with yourself, I'll bet there are lots of sounds that you use just as they are, with no real-time control other than modulation and pitch-bend. A beneficial side-effect of working this way is that, once sampled, it's very easy to change the overall envelope of a sound and to apply filter sweeps -- it's often much easier than doing this on the original virtual instrument plug-in.
Never-ending Story
No matter how much computing power we get, the software designers will always find a way of using more of it, so I don't anticipate the pressures on CPU power diminishing in the near future. Using additional computers (an option Steinberg is pursuing in a commendably imaginative way) offers a partial solution, but my guess is that the user will still need to apply workarounds, like the ones described here, for a long time to come. 
Home |
Search |
News |
Current Issue |
Digital Editions |
Articles |
Forum |
Subscribe |
Shop |
Readers Ads
Advertise |
Information |
Links |
Privacy Policy |
Support
Sound On Sound : Est. 1985