Posted - 08/19/2009 : 14:08:09
So, recently there has been new debate about a long-running topic about emulation vs. IR-based technology.
The new debate once again raises interesting points, but shows that there are still the same misconceptions about the technology being used with Nebula. One comment in a new thread reads as the following:
"but it's never going to be as close as a good emulation to hardware with multiple variance paths built in, because you'd have to sample every possibility."
I have to respond out of a sense of obligation to this, with respect to the intention of those involved.
First, we are specifically dealing with a comparison of UAD2's newest "Fatso" release compared to my new R2R tape collection release. The initial thing to address is that Fatso is a hardware device developed by Empirical Labs for use as an actual compressor, and also as a harmonic saturator. There are many things that it IS, and IS NOT. It is a compression device. It is not a tape device in any form. It does emulate some of the useful characteristics of tape, but it does not accurately replace them. It compresses with much of the control of a standard compressor, allowing one to dial in a very limited, very different type of compression than tape creates. Some instances more closely resembling tape than others, but overall, Fatso is made to be a great creative tool to be used where one may have used tape, but with a great deal of different flexibility and different uses.
In the same manner that the Distressor is its own deal despite having the ability to emulate other popular hardware compressors, Fatso is not a 1:1 comparison of any single tape device.
So, to compare UAD2's excellent emulation in software form, of a hardware device that is emulating another different construction of hardware, minus the complex non-linearity that actually involves tape, and what we have is a software emulation 4 times removed from the comparison being made to my new Nebula collection. 1: Fatso=different kind of hardware all together. 2: UAD2 emulation is a different technology being used to copy the behavior of a hardware device that does something different than R2R. 3: Fatso and 'randomizing'? 4: The plausibility of comparing something that is not even being emulated the same way as the argument describes.
Here's the real deal specifically between these two devices.
Fatso: It is a great hardware and comes with a very high price tag. UAD2 is fairly priced for an 'official' software version that I am sure is very close to the hardware for $249. Good deal! It is a compressor and a harmonic generator with several knobs to turn.
R2R: It is a collection of accurate, nearly indistinguishable from the actual hardware, programs sampled directly from several actual tape machine devices using a technology advanced enough to do so without being static, without being 'random', without being emulated. This includes the changes to frequencu, the changes to volume as it relates to frequency, the changes in harmonics in relation to frequency, volume, and variations in all 3 factors. The only 'static' items are the ability to raise and lower the range of harmonic content sampled within a given program. From a cost per collection perspective, it is not priced less because it is a lesser technology or less accurate. There is nothing in the process or sampling chain that can be improved upon. The absolute best analog and digital processed the sound. The machines are all calibrated and tested at every imaginable combination, and this includes having to run actual tape formulas around in circles in order to produce the REAL harmonic results. This is not only better than an emulation, because it is a recording of the actual device, but it is a direct recording OF tape, not a hardware processor that allows another option for tape. Not only are actual machine(s) sampled, but there are 7 of them! The machines compared represent a monetary equivalent of about 20 times the cost of a Fatso hardware unit, and the software is less than 40% the cost of the UAD2 emulation of an emulator.
Returning to the quote: emulation can do a fine job of recreating a device. The UAD concept of digitally interpretting each part for its purpose and reconstructing these parts inside an emulation is clever and has served well for years. Part of what has been mentioned in the thread is the required increase in cpu for Nebula to include more scenarios in a single program. Um... UAD requires a new card just to run Fatso, so we are not talking about an efficient process comparison, just a good sounding one.
Nebula does use sampling, and it can be compared to shooting a film vs. emulation. But, it is better than any kind of randomization and in this case much better than correctly emulating an emulator:
Do not mistake the sampling process for being a singular operation.
Each element of sound is being individually processed... accurately.
If I were to, for instance, sample the setting of an eq with Nebula compared to emulating this (which I am not certain how much of the UAD process involves actual convolution as well- it is a huge assumption that emulating does not include the IR data collected!). If I sampled the eq as briefly as 10ms and only could sample the frequency, it might sound nice and give me a simple version of that eq filter. But, if I sample it for 10 seconds, and I sample any microscopic changes in volume, timing, harmonics added, etc., it is going to tell my ear and brain a lot more info, even if almost undetectible. Even as long as 45 seconds or more into this process, Nebula can provide real time data of what is happening when I set this eq device. If it is a good device, there is only so much information that is going to fluctuate in 45 seconds. Hopefully, my phase, eq, volume will be somewhat stable! Can you imagine if your eq changes in your DAW were constantly 'randomizing' themselves? What are they randomizing? What did they change? Can I undo it? No, seriously, where there is a need to create algorithms to make random assumptions about a range of duties, there is a less technology than taking a 1:1 absolute sample of the changes the exact device made and replaying them inside your DAW. There is no question that anyone that says different simply does not understand the technology. A simple IR is like taking one decent still photo. Dynamic convolution is like taking a series of 3-dimensional pictures and putting them together for a realistic movie. Nebula's VVKT technology is like improving the camera, film, and number of dimensions in the dynamic process and shooting straight to digital real time without the need to reconstruct multiple thousands of images. It is a live, streaming, true to original, replaying of what the original equipment did. To suggest that randomizing this concept is better, with the assumption that Nebula is like adding multiple static changes together, is like saying that an emulation is good at colorizing film and Nebula is a high quality black and white. It is simply not accurate. A more correct comparison would be to say that emulation in coding as it stands, with the variables taken into consideration, is like doing its best with high quality rgb, where Nebula has made its way to CMYK + a few more in between separations all layered together.
So, more about comparison apples to oranges. Fatso and compression. Nebula and R2R.
High quality harmonic additions to music content can add fullness, character, and even perceived volume to sound. As the amount of correct harmonics are increased in amount, this can add actual measurable volume to the frequencies that correspond to the harmonics.
When working to find a way to accurately sample tape without running into technical problems, we assumed that the compression characteristics would be completely absent from the process. Since the attack/release/ratio/threshold are all elements of hardware compressors that Nebula samples as a separate process, naturally we assumed tape compression would be its own thing.
Well, what I have since discovered is major, major stuff. It is not easy to understand and I am going to place this info more thoroughly in its own technology thread under R2R, but when you are able to capture what eq and harmonic changes are happening to a program at different volumes, real time, at perfect digital transfer, tape reacts in a unique manner from other devices. The compression character that we assume is actual compression is largely made up of the sonic gap-filling process happening between eq and harmonics. The perceived effect of these variables on sound include what we would think is compression.
When comparing actual tape saturation and its compression effect to the resulting Nebula programs that are not measuring the compressing of a signal in a way that the Nebula engine allows independent control for, this information is still being factored to the extent of how the waveforms sampled are being operated upon, and the results are almost identical until you reach a larger amount of compression (more than 1-2dB). At this point, it would require an actual envelope-following process to measure and replay the full-on levelling taking place. So, to compare Fatso as a hardware compressor to a collection that is not attempting to measure compression is strange. Not to mention that Fatso acts like a good actual compressor, and not a natural tape compressing, but Nebula is actually allowing some of the aspects of natural tape compression to be factored in without having even knowing that it could presently be done!
You must be logged in to see this link.
Software, Mastering, Tutorials