POWER Management
Nov 1, 2003 12:00 PM, By Jim Batcho
In an age of music production when computer mixing power is so prevalent that it's almost silly, you can remember all of the classic, brilliant records that were made with just four tracks and try to remind yourself that it's all about the ideas, not the tools. But that doesn't help when your buddy is streaming six stereo channels of the latest software synth into his mix without a hiccup while you're struggling to squeeze in a second reverb. It's a reflection of that rabid id, which forever urges that too much is never good enough.
It's not all greed, though: There is a paradox at work behind the development of desktop-production tools. The arc of chip development is matched at almost every degree of progress by the arc of software development. As computer processors increase in speed, the software you run seems to cancel many of the benefits gained. Modern operating systems (like the elegant but thick Mac OS X) and increasingly feature-rich sequencer applications create enormous demands on the hardware on which they run. Add to this the prevalence of streaming protocols for software synths and samplers, and the whole mechanism becomes pretty much a wash.
Nevertheless, some recent developments on the software front are aimed specifically to help throw the curve in your favor. Many software companies are making efforts to counteract the demands created by their own development to make resource management more, well, manageable. Digidesign was the first to the plate in such endeavors when it built the infinitely handy Deactivate Tracks feature into its flagship Pro Tools TDM systems with version 5.0. Thankfully, starting in version 6.0, it is available on LE-based systems — the very ones that really benefit from it. Recently, in true Leibniz-Newtonian fashion, leading sequencer manufacturers Emagic (Logic Audio), MOTU (Digital Performer) and Steinberg (Cubase SX) almost concurrently released an identically titled feature called Freeze into new versions of their software. Magix's Samplitude/Sequoia also features a Freeze capability. And Syntrillium Cool Edit Pro (now Adobe Audition) has had a track-locking feature available for several years. The idea behind the freezing and deactivating operations is to allow users to free up system bandwidth, resulting in increased mixing power. Clearly, CPU preservation is becoming a top priority for software manufacturers.
This article intends not only to detail some of the ways these products are working to help the CPU-deprived but also to discuss the many options available to mixers for getting the most out of their workstations. But beyond that, the larger scope is to offer a broad methodology for approaching the mixing process on systems that lack processing power. Using a method of linear mixing emphasizes resource preservation to best-utilize the power you have.
MIX MAXIMIZATION
It may seem obvious, but it's important to always maintain an awareness of the big picture when mixing with low system resources. It also requires a bit more patience, more organization and a good deal of preparation. If the project is something that you're recording yourself, you should anticipate the eventual mix before you even begin tracking. The no-brainer engineering attitude is that you should strive to attain the best-quality recording that you can. The less EQ and tweaking you'll need to do later, the more resources (not to mention time) you'll save. If you're remixing someone else's recorded tracks, you'll have to make do with what you have, for better or for worse. But no matter how it was recorded, you'll want to make the mix yours from the start by getting organized. Begin by making sure that all tracks and audio regions are properly labeled. There will probably be times when you'll need to go back to previous raw tracks and settings, so it's good to have an idea of where everything is and what everything is called.
Active resource management begins when you open your mix. As with any mixing endeavor, avoid heading straight for the kick-drum insert to pull down your favorite compressor. Back up a minute, and think about the whole mix. Listen to the song without touching a fader, and take note of what's happening. Then, start thinking about how you're going to build upon the foundation that you have in front of you. Nearly every mix session requires various essentials, including reverb, compression and varying degrees of EQ. With computer-based mixing, plug-in effects are not all created equal. In general, effects that add a great deal of color to a track tend to create more burden on your system. For example, reverb plug-ins are notoriously processor-intensive. Generally, the better the reverb, the more demands it places on your CPU. EQ and compressor plug-ins are usually more economical, but they, too, can vary widely in both quality and efficiency. The more special-effect-type plug-ins — such as delays, choruses, flangers and treatments like analog emulation — are all over the map.
If you had all of the power in the world to mix and process your tracks without ever fearing a CPU-overload error message, I'd propose a progressive mixing approach like the following: Get all of your basic tracks sounding good at the same time using EQ and compression, then move on to streaming software synths and samplers, then go back and apply reverb treatments and special effects like chorus and flanger. But you need to take a linear approach when system resources are low, first to head off any potential troubles and second to take advantage of recent CPU-friendly software development. The linear method goes something like this: First, take on any surgical repairs (fixing tracks). Next, tackle any CPU-intensive needs, such as streaming software synths and samplers. Then, address your basic processing (EQ and compression). After that, turn to CPU-intensive processing: reverbs, additional effects and busing assignments. Lastly, prepare your finished mix for automation and final bounce. Covering each step individually, I'll address some potential headaches along the way.
SURGICAL REPAIRS
Imagine and take note of those elements of the mix that need to be repaired or that need uniform processing across an entire track. Think first about any tracks that need to be repaired by editing, tuning or resampling. Go ahead and tighten any drum or percussion performances, resample any snare hits and tune vocals (if that kind of thing doesn't go against your studio religion). With these types of operations, go ahead and commit without looking back. Make sure that all of your crossfades are good, that there are no audible problems and so forth, and then consolidate the new tracks or bounce your Auto-Tuning to disk and delete any unused regions. Incidentally, although I am a proponent of committing to editing and arranging decisions and moving forward by deleting old files, I also realize that mixing on slow computers often requires moving back a few steps in time if some previous operation isn't working with the rest of the mix. With editing, resampling and vocal tuning, you shouldn't need to go back, because those operations shouldn't affect the sonic structure of the mix. However, this is the only stage at which I recommend deleting old files when mixing on slower computers. From here on, avoid deleting anything until you've reached the CPU-intensive processing state, as mentioned previously.
Next, separate in your mind which effects you want to automate and which effects can be uniform across the track throughout a tune. Automation of real-time effects will have to wait until the end whereas uniform operations — known as file-based, or non-real-time, effects — can be tackled at the beginning. The benefit of file-based operations is that they cause no system drain. Because file-based processing prints a static result across a whole track, you need to feel pretty confident about the effect it will have. Still, you can always go back to the original tracks if you've practiced good file management (and haven't deleted audio regions), so try not to be too cautious. Some good candidates for file-based effects processing include surgical highpass filtering to remove rumble, amp modeling on dry guitars or adding a particular vital effect that defines the character of the track. Remember, don't delete the original, unprocessed tracks until later.
CPU-INTENSIVE NEEDS
Once the basic editing and file-based processing is done, you can begin thinking about the more creative aspects of the mix — specifically, software synthesis and sampling. The ability to stream software-based synthesizers, samplers and drum machines into sequencer applications was one of the major developments in the industry during the late 1990s. Soft synths have since surged in popularity as processor speeds have increased. Today, nearly every sequencer app supports soft-synth integration using VST, ReWire, RTAS, HTDM, MAS, DXi, Audio Units or DirectConnect protocols. Regardless of the format type, each operation functions under the same basic concept: Within a particular software sequencer, a user can stream additional host or third-party soft synths into a track on the software application's mixer channels. In the computer world, this creates a huge burden on a computer. Because it's an overdubbing process and a mixing endeavor, the process is best initiated at this stage of the mix. This is also the point at which freezing and deactivating tracks comes into play, so it's important to take a closer look at these new developments.
The Deactivate Tracks feature in Pro Tools is simple enough to use. Say you have a channel strip with a few plug-ins and synths, some automation and a couple of busing assignments. This can create a severe drain on your system. To free up some muscle, hold down the Control+Command (Alt+Control) keys, and click on the audio or auxiliary icon at the bottom of the channel strip. The entire strip will appear faded on the screen and be rendered inactive. Deactivation solves two problems: First, it alleviates system drain for you to instantiate plug-ins on other tracks, and second, deactivated tracks don't work against your voice count, meaning that it makes more voices available for building more tracks. The drawback of it is that the audio on the track is no longer audible, so you have to bounce the contents of the track and then deactivate if you want to hear it alongside the rest of the mix. But the instantaneous flexibility it offers is a tremendous help, and you'll use it a lot more as you progress forward in the mix.
The latest versions of Logic Audio, Digital Performer and Samplitude/Sequoia take things a step further with their Freeze Tracks functions. Freeze is different in that it conducts an instant offline bounce of a chosen track. All operations that are unique to the track — such as plug-in instantiations, panning, automation and so forth — are temporarily printed (or frozen) to the track. The great benefit of freezing is that the track, with all its previously cumbersome data, is still audible. The limitation is that you can't make changes to frozen tracks. Still, you can always unfreeze them and make changes before freezing them again. Now that you're unburdened, you can go on to other processor-intensive actions. As you do, keep in mind that freezing and unfreezing is a balancing act. In a weighty mix, once you unfreeze something, you'll have to freeze something else to clear the pipeline. In other words, it's unlikely on slower computers that you'll be able to have a completely open and unfrozen mix.
With your streamed data arranged to your liking, freeze the track. If you're working with Pro Tools, bounce the arrangements at the same resolution as your session and import the resulting mono or dual-mono file into your project. You should name the bounce something similar to the track itself, but designate it as audio data. For example, if the track is named “Arp Synth,” name the bounce “Arp Synth Bounce.” Create a new track called something like “Arp Synth 2,” pull the bounced audio into the track, and deactivate the original one that was weighing heavily on your computer. Do not delete the original or remove it from the session, because you will probably want to go back and tweak and rebounce it later. If you want to temporarily get it out of the way on your mixer, deselect it in the Show/Hide Tracks list.
BASIC PROCESSING
Needless to say, the art and science of EQ is an extremely deep topic that goes far beyond the scope of this article. But even with low system resources, there are some ways to approach EQ that are different from standard mixing practice and will spare your computer from unnecessary burden. First of all, whereas most mixes are best done by listening to them in their entirety, with low resources, you'll have to take the linear approach: Take on each element of the mix one by one. With this approach, it's best to work from high complexity to low complexity. In a normal mix, it's often wise to get vocals sounding good from the start. But on slower systems, I like to start with rhythmic elements because they usually involve a lot of tracks and there are a lot of frequencies on multiple instruments to negotiate. They're also better candidates to bounce and import, thus lessening the burden on your computer. Resist the temptation to instantiate an EQ on every drum channel, and instead use your ears to identify areas that need the most help. Also, try to use transparent, simple EQ plug-ins, as they tend to be more CPU-friendly.
Next, it's best to work with the bass so that it sits well with the kit, particularly the kick drum. With the rhythm section out of the way, you can then move on to the all-important vocal tracks. EQ and compress lead and harmony vocals so that they sound good as a unit and in tandem with the bass and drums. From there, you can work the other melodic instruments by equalizing them around the vocals and the rest of the mix. (Note: There are no set rules. If you're dealing with a primary melodic instrument with a broad frequency range, such as a piano, you may wish to rethink the above workflow approach. You should still use a linear approach but perhaps in a different direction.) Hopefully, you'll have enough CPU cycles to get the drum kit sounding the way you like. If you run into problems at any stage afterward, freeze the drum tracks or bounce and import the drums as they are and then deactivate the multitrack channels. Be sure to name the drum bounce something that you'll remember as a temporary stage in the process: “Drum Sub Temp 1.” If you still need to free more resources, continue freezing and bouncing in the same linear progression that they were equalized. For example, the next time you hit a brick wall after you've worked the drums, freeze or bounce the bass track complete with EQ and compression and then move on.
CPU-INTENSIVE PROCESSING
Now, turn your attention to that all-important CPU hog: reverb. For good reason, many mix engineers tend to use outboard gear for reverb processing. Not only is the quality better, but premium software reverbs, like Audio Ease Altiverb, suck up a great deal of processor power. If you lack outboard and are using plug-ins, you'll need to prepare for some juggling. As a general rule, it's good to have different reverb settings on different instruments. A vocal usually requires a unique set of parameters whereas a drum kit demands something different. For this reason, you should set up two sets of auxiliary channels, one called Drum Verb and one called Vox Verb. Bus the drum-kit channels to the Drum Verb aux, and bus the vocal or the primary melodic instrument to the Vox Verb aux. Pull down your reverb of choice on both. If the second creates problems, deactivate or freeze the first. It's not vitally important to be able to hear both at the same time, so deactivation isn't an issue at this stage. If your system protests even one reverb, you should start thinking about freezing or deactivating everything except the drums and work on them in isolation, then bounce the result. After you've set appropriate reverb levels on the different drum channels, listen to the result in relation to the rest of the mix.
When you're fairly satisfied with the kit and you think you're going to run into future resource-management issues, go ahead and bounce a potential final drum submix with all EQ, compression and reverb intact. Name it something like “Drum Sub.” (The reason you don't want to simply bounce a nonreverbed drum sub and then apply 'verb to the sub is that different elements of the kit require varying amounts of reverb. For example, the kick is usually best kept completely dry, but snare, toms and overheads should be seasoned to taste.) Now that you have a final drum submix complete with reverb, you have much more freedom to move on to the other instruments. With the kit out of the way, hopefully, you'll have enough system resources to treat the other instruments.
Concentrate next on the vocal track. Craft a suitable reverb sound on the Vox Reverb auxiliary channel. If you want additional effects such as delay on the vocal, create a new aux channel or throw a delay plug-in before the reverb. Delays are not quite as processor-intensive as reverbs. They also don't need to be as robust as reverbs. Usually, a sequencer's stock delays do the trick just fine. As you move on to other instruments in a mix and begin adding special effects, like flanger or chorus, you may need to find ways of balancing your mix with additional freezing or deactivation. Just remember to name any bounces and never remove any deactivated channel strips from your session until the end.
FINAL MIX
As you near the completion of your mix using the linear method described here, you can always go back and rework things if you've maintained an organized process. Say you're nearing the end of your mix, and the bass isn't sounding right anymore. If you're working in Logic Audio or Digital Performer, you can simply unfreeze the track and retweak it. If you're working in Pro Tools, you'll need to reactivate the original bass channel strip, change the EQ or compression parameters and bounce it again. Name the new mono track “Bass Sub 2” (your second bounced effort), import it into your playlist, and check your mix again. If you properly name all of your bounced audio, it's a simple process to go back to your deactivated tracks to fix them, rebounce and rename, and always be able to see where you've come from. The same goes for any other instruments that you've submixed. In essence, this process creates “mix stems” that are easily manageable as you move toward a final mix. The more manageable your mix, the easier it is to automate volumes, pans, plug-ins and other elements that constitute the final mixing stage. As for the bounce, you probably won't have power left to put anything on the master fader. It might be best to bounce a nondithered, full-resolution stereo file and pull it into a new session for mastering.
As you can see, new features like deactivating and freezing tracks offer tremendous value to those with low system resources. You can fairly easily reactivate or unfreeze the multitracks and dig back in section by section to fix what's needed. Moreover, if you move your mix to another location, such as another project studio with more power or a mixing room in a major studio, you can just reactivate everything, and all of the plug-in settings will be there, ready for you to mix the whole thing at once, if desired, on a more robust system. A lack of power may make you a potential Freudian case study, but it doesn't have to stop you from making great music.
| Want to use this article? Click here for options! |  |
