Support Report: UAD-1 Plug-in Instances in Native DAWs: How Many and Why
By Tom Freeman and Casey Schultheis
A question that we get regularly in tech support is, "How many plug-ins can I run on a UAD-1 card?" Many factors affect how many plug-in instances you will be able to run on a UAD-1 card, but the most significant are the sample rate you are using, the size of the plug-in algorithm, and whether the plug-in is being used in a mono or stereo situation. The first two factors are fairly straightforward. Generally, the higher the sample rate, the more processing power is required from the UAD-1 card. If you double the sample rate ( by going from 44.1 to 88.2 kHz, for example), you will effectively halve the number of instances you can run (for plug-ins that do not do internal upsampling).
“A plug-in like the Fairchild, with its amazing sound and very complex design, takes more DSP cycles than a plug-in like the 1176SE, which still sounds great but is a much simpler algorithm.”
The "size of the algorithm" refers to the fact that some of our plug-ins are simply larger and take up more of the card's processing power. The instance chart in Fig. 1 gives an overview of how our plug-ins differ in terms of size and the number of instances. A plug-in like the Fairchild, with its amazing sound and very complex design, takes more DSP cycles than a plug-in like the 1176SE, which still sounds great but is a much simpler algorithm. That is why you can run more 1176SEs on a UAD-1 card than Fairchilds.
 |
|
Figure 1
The UAD-1 card can run quite a few instances of some plug-ins at the 44.1 kHz sample rate, but your choice of sample rate and mono versus stereo aren’t the whole story. |
The third main factor takes a little more explanation but is still relatively simple to understand. Some of our plug-ins have higher instance counts as mono plug-ins than as stereo plug-ins; other plug-ins use the same number of cycles in mono and in stereo. Why? The main reason is the way different types of plug-ins work. This is an oversimplification, because there are many other related variables, but a comparison of EQs and compressors will illustrate how plug-in type affects your number of instances. EQ first: a mono EQ uses less DSP than a stereo EQ, because a stereo EQ must process each channel independently (see Fig. 2). In the simplest case, a stereo EQ plug-in should use about twice as much DSP as a mono EQ plug-in. The Cambridge EQ is an example: you get double the number of instances when you use mono versions instead of stereo.
 |
|
Figure 2:
This EQ demands the most DSP in the Filter processes, which must be duplicated for each channel. |
However, in the case of the 1176LN compressor, the number of instances you can achieve is the same for stereo or mono routing. Why is this? Like their analog counterparts, many compressor plug-ins take a stereo signal and determine a peak or RMS level estimate for the signal (see Fig. 3). The estimate of the signal level is itself a mono signal formed either from summed stereo audio or from one channel of the stereo audio, depending on how the compressor is designed. The amount of gain reduction the compressor applies is based on this signal estimate and the settings you have chosen, and the compressor reduces the gain on both sides of the signal by that amount. By applying the same amount of gain reduction to each stereo channel, the compressor preserves the signal's stereo image. The processing is still stereo in/stereo out, but the gain reduction is determined by a single signal source. Because the compressor does not need to process two completely independent signals, instead using a single level estimate to carry out gain reduction even on stereo audio, a compressor plug-in will typically not have twice the number of instances available when treating a mono signal.
 |
|
Figure 3:
The bulk of a compressor’s processing is devoted to the “Level Estimate” function, which needs to be done only once, whether the signal is mono or stereo. The gain is then applied to all channels, using a minimum of DSP cycles. |
We've described the main principle that determines how many instances you can get based on the routing of the plug-in, but there are other factors at work as well. For example, the LA-2A algorithm does some stereo processing first because it models the hardware unit's behavior; it then proceeds with the more conventional method of determining gain reduction. This is one reason the LA-2A has a unique sound, and it is also the reason that the instance numbers for the LA-2A are slightly higher when using mono versions of the plug-in, but not double the number of instances.
So how do you take advantage of the extra instances when working on mono tracks? Any application other than Logic 5.5.1 or below will automatically determine what type of plug-in to use based on where the plug-in is inserted-if you place the plug-in on a stereo track, you will get a stereo instance. Users of Logic 5.5.1 or below need to place a mono plug-in on a mono track and a stereo plug-in on a stereo track. For users of this application, we have included mono versions of our plug-ins, and there is an option to choose mono versions when you install any recent version of our software. If you use any application other than Logic 5.5.1 or below, you don't need to choose this option. However, James Cigler has some interesting tips on using mono plug-ins on stereo tracks for Nuendo and Cubase SX users in his interview in this issue of the WebZine.
Questions or comments on this article?