& Jonathan S. Abel
Ask the Doctors!
Drs. David P. Berners and Jonathan S. Abel Answer Your Signal Processing Questions.
How are the UA 1176 and LA-2A algorithms different from others on the market already?
|
|
& Jonathan S. Abel |
The 1176 and LA-2A each have specific characteristics which bring about situations in which these goals are not enough to give accurate emulations, and, in some cases, situations in which these goals become ill-defined. For example, both the 1176 and LA-2A have program-dependent compression. One of the things that often confuses people about the 1176 is that if you measure the static compression curves using sine waves at different amplitudes, the compression ratios will appear to be much higher than what is selected on the front panel. However, if you were to look at the 1176's response to a transient-type signal, you would see that the ratios are very close to those indicated by the ratio buttons. The release times and curves are also very program-dependent.
"Eventually, the two models converge, and at that point we know we have it nailed."
This brings up another subtlety in the modeling of a vintage compressor. Simply having the release "time" correct does not guarantee that an emulation will sound anything like the original. The gain will follow a certain trajectory during release, and it is critical to get that entire trajectory right, as opposed to just the time it takes to go through the trajectory. We typically work on our emulations from two directions: going forward from the schematic, and working backwards from the way the hardware compressor behaves. Eventually, the two models converge, and at that point we know we have it nailed. By the end of the process, we can type in actual resistor and capacitor values from the schematic and have our digital model behave in the right way.
Another thing we like to do is work with analog engineers and physicists. John Hinson worked with us in isolating the parts of the compressor circuits which were responsible for different aspects of the compressors' behavior, and we also worked with Brad Plunkett, who had worked on the original 1176 hardware. Both of them had a lot of insight as to the mechanisms which were important to the operation of these compressors. For the el-op behavior in the LA-2A, we also worked with USC professor Ed Maby, who is a world-class semiconductor physicist. With his help we were able to isolate factors which defined how the release of the compressor changes for different types of program material.
For compressors, or for any processors with a lot of nonlinearity or program dependence, it's critical to have a model based on the physics of the unit, rather than just the unit's behavior in specific test circumstances. That's because when something has got nonlinearities or program dependencies, it's almost impossible to predict what that piece of gear will do in every situation; you simply have to understand the processes behind its operation. For the LA-2A, things get even more messy model-wise, because in addition to having the program-dependent release, you also end up with compression ratios and thresholds which are frequency-dependent. It's pretty easy to get a threshold which is frequency-dependent by using a sidechain filter on the signal detector, but in order to get compression ratios which depend on the signal frequency, you need a very special type of nonlinearity, and this is part of what gives the LA-2A its signature sound. So, in this case, you can see that if one of your goals in doing an emulation is to get the "knee" right, then there is no way to do that with one knee, because for the LA-2A you're going to have a different ratio and knee shape at every frequency, even discounting the program-dependent aspects of the compressor.