On Topic: Modeling Software of Analog Devices

[Saint Johnny B]

OK, we've seen over the last few years that software "models" claim to emulate real world analog devices. Amps, mics, pre-amps, 'verbs, compressors, and whatnot. Some claim that they replicate the orginal, often vintage, analog unit in a very accurate way. And some of this stuff seems to sound really very good. So who knows how they actually "model" a real world device in computer software. That does not sound like a trival task, but how would get software to know how to, say bump a few db of the 10k frequency but only under certain weird circumstances. I don't know how they can really, precisely, and accurately claim to have "modeled" the exact same behavior of an analog unit. Some techies and software types could perhaps offer some explanations of the process used to write this kind of software. Others of course are free to chime in with your own experience with these models and emulations Good, bad, close, not quite ready for prime time. Like 'em. Hate 'em.

[Chuck Moore]

The bomb factory emulations for PT, while being usefull for their own sounds, don't really sound like their real world counterparts. The emulations on the UAD card however are a different story. They do a great job of re-creating the sounds of their namesakes. I use a Nuendo system with a UAD, recently A-B'd the 1176 against a real 1176 on tracks split from the nuendo system into an SSL 4000E, one track had the UAD comp into one SSL channell, the other had an unprocessed track into a channell with the real thing as an insert. With levels matched, and through the same signal path it was impossible to reliably pick the emulation in all but the most extreme settings. I've found there to be more differences between individual 1176s than between a nice example and the UAD version. Don't have access to a pultec, but I'm told it's equally hard to tell the diff.

[aeon]

modeling is typically done by means of: 1. black box transfer functions, in that you input a range of values and see what you get as output...and from those numbers you code a function curve such that the behavior of the code based on a given input is as close as possible to that of the phyiscal device. 2. component-level modeling, similar to #1 above, but instead of building code for the entire process, you code for every component in the device, recreating the circuit piece-by-piece...at a much greater computational cost, but with the possibility of much greater levels of emulative accuracy. 3. convolution with an impulse response of the device being emulated...works great for time-invariant devices like a room, or things that work only in the spectral and dynamics domain...things that function in the temporal domain generally cannot be modeled via convolution. of course, much modeling has been a combo of #1 and #2 in that the maker does not model on a component level, but does model subsystems which are integrated into a common whole, e.g., a guitar modeler which has models for the amp pre, power stage, cabinet, mic and pedals, if any... something between #1 and #2 also has to be used if a control or parameter in a given device influences the behavior of other parts of the device, e.g., changing the mid control on a guitar amp will influence the function of the bass control that came before it and the treble control that comes after. so in the end, to model something well means you have a many-factorial equation that will account for every significant variable...frequency, level, the action of every part of the device whether internal or an external control, the changes that occur in components when they age, get hot, etc. cheers, aeon

[Saint Johnny B]

Thanks Chuck. The UAD stuff was done by Bill Putnam Jr's group and modeleed after his Dad's orginal stuff. Bill Putnam Sr was the man. Anyway, I believe the 1176 stuff was modeled on a pristine unit they found at Stanford University. I think that one of the reaosns that there may be some variance between real world units is that there were diffences in the manufacturing runs, but maybe more importantly, the components like pots, resistors, capactors, and so on age differently and theire values can drift, thus, one unit can sound different from another of the same brand. Anyway, I didn't know that BF stuff was not quite the same, but I'm glad you are getting good results from the other stuff. I know about so-called expert systems and artifical intelligence systems and thing like that on a general level. Like they will try to capture the knowledge base of some doctors and then use a computer programe to help diagnose a condition. Pretty heavy stuff. But I'm still unclear as to how you'd go about putting together a computer programe that models tubes, transistors, ics, transformers, pots and resistors. What? Would you try to come up with every possible setting, input levels, frequencies, etc? It sounds like it would be extremely difficult to do well in software and not a trivial task. Maybe I'm missing something. Perhaps, some computer jocks or really smart people can enlighten us on how this computer modeling is done. Maybe Craig knows?

[KenElevenShadows]

I've never used the UAD or the real stuff being emulated, but I would also agree that the BF stuff is quite useful. The Mooger Fooger Analog Delay - I don't know whether it sounds like the real pedal or not, but to my ear, it actually sounds like a decent analog delay of some sort.

[Chuck Moore]

Yeah the BF comps are definately usable, sound better than the ones that ship with PT for sure. But your not gonna fool anyone into thinking you had a real LA-2A or 1176. Having the UAD allows me, for example , if I'm mixing out of Nuendo on an SSL, to use both the real 1176 and the UAD one on different tracks, without the UAD if you want the 1176 on three acoustic guitar tracks and a bass track you were in for a 'lil track bouncing unless you were lucky enough to have several of them at your disposal. And for quit a bit under a grand, its the bargain of the century. The first one has paid for itself many times over.... considering adding a second card in my system.

[bond1]

Bargain of the century, yeah i'll definitely go along with that (regarding the UAD-1). I can't tell the difference between my hardware 1176 and the UAD-1 card. Especially considering you can throw one instance of an LA2A or 1176 accross a STEREO track which would require 2 hardware units, and STILL have enough power left over to run 7 or 8 more instances is pretty amazing. At 600 bucks its a no-brainer and definitely the smartest upgrade I've made in a long time.

[Saint Johnny B]

Aeon, thanks for the explanation. I kind of get a little of what may be involved, but, I'm one of those guys who is always in the slow class. :) Sounds like Chuck and Ken like the BF stuff even if it ain't quite like the real deal, it still works and sounds good. Also, sounds like the UAD is getting raves from Chuck and bond1 as being really like the real deal. I'd guess that BF and UAD took a different approach to modeling the originals. Maybe someone can tell us more about the different approachs these 2 companies used in trying to do modeling. In fact, maybe some would like to talk about how the other modelers do it, Line 6, etc and so on. Thanks. :)

[KenElevenShadows]

[quote]Originally posted by Johnny B: [b]In fact, maybe some would like to talk about how the other modelers do it, Line 6, etc and so on. Thanks. :) [/b][/quote]The Pod stuff - I don't know how they do it - but for the money, it's really excellent. To my ear, it doesn't quite sound like the sound of a great tube amp in a room. But it sounds really great. And it's very useful and flexible. Very impressive for what it is.

[Saint Johnny B]

I was over at a producers place a while ago and he was doing some smooth jazz, aka Kenny G. (yuch, no life there), some rap, and some alt/punk stuff. I really liked the way his bass sounded esp on the alt stuff he did with this girl who's going to the Berklee School of Music, anyway, it sounded good...massive like he was using an SVT turned up to 11, so I said, how'd you do that? He smiled and replied, "POD. I did all those songs with a POD." It changed my mind about those POD things. And it got me wondering about all this modeling technology, it keeps getting better and better. And yet...some of it does not seem quite like the real thing, very close, but not quite there yet. But some seem to be saying the UAD is there. So I guess it can be done today. So I guess this is now a discussion of both 1) "How Modeling is Done," and, 2) "How Good or Bad Certain Company's Implementations of Modeling is." Thanks everyone for your thoughts and contributions. :)

[EXAGON]

And what about soft synths? IMO Pro 53 sounds nothing like the original, while Moog Modular V sounds very similar to the original. So anyway it's very important to check the demos before believing any ad.