Nebula is the cab sim holy grail.

[AeternusEternus]

Hey guys, I just wanted to chime in here real quick. I know that using Nebula as a cab sim was brought up in the past, and it kinda died out because I really couldn't get the sound perfect. Well, since then, my sampling and reamplification skills have gotten a little better, so when Ermz brought it up in a new thread, I gave it another shot.

At this point, I'm very happy to say: Impulses can take a hike. Any kind of static cab simulation, while it's been good to me in the past, is old news now. Over the past few days, I've been testing and have finally come up with a sort of proof of concept because I've got a cab simulation that sounds exactly like the real thing, dynamics and all. It's not a perfected art just yet, but if I did it once, I can do it one thousand times more.

Follow along in the thread, it's only 5 or 6 pages right now.

http://www.ultimatemetal.com/forum/andy ... ssion.html

Woo!

Edit: Sorry about the link, hahaha. I suck.

[Agreed]

That link doesn't go anywhere, but I am very interested to see what you have found. I haven't worked with Nebula in the past but the technology is very impressive and if it could be applied to cabinets to capture their response under a variety of conditions but with the frequency and phase accuracy of an IR I would love to use that.

[AeternusEternus]

Well, hm. Looks like I need to hit five posts before the link becomes clickable. Anyone seen any good movies lately?

[AeternusEternus]

That link doesn't go anywhere, but I am very interested to see what you have found. I haven't worked with Nebula in the past but the technology is very impressive and if it could be applied to cabinets to capture their response under a variety of conditions but with the frequency and phase accuracy of an IR I would love to use that.

Honestly, this is way more accurate than IRs. Like, they don't even enter into the equation for me anymore. Giancarlo from Acustica Audio allowed me to be a beta tester, and a lot of my tests have been comparing IRs to Nebula programs to see if they match. But I quickly learned that I'm using like 5% of Nebula's capability by doing that. The integrity of an IR varies and depends on the convolver, the editing, and in some cases, the test tone. I find that Nebula's a much more stable way of sampling now, though it IS a bit of a workout to get it going. It only took me a year or so

[Agreed]

What exactly does Nebula offer for cab simulation that IRs don't?

[AeternusEternus]

Well, I've heard many people call IRs inconsistent and inaccurate. I know that when I've got real cab vs impulse comparisons, the impulses come up short every time, but they're the closest to the pin, and you can't discredit them for that.

But Nebula offers dynamics. Impulses are static and can't really offer much beyond tonal similarity. Nebula's got some movement to it. The sound is more accurate, but when you're playing on it, you really do feel like you're playing in front of an amplifier. Plus, you kind of have to be careful, because Nebula can model small amounts of distortion, so if you have any speaker or tube breakup involved while you're sampling, you're going to get some of that in the final product. It's got a very high resolution of 64bits, too.

To be short, it's an entirely new way of sampling. I don't understand much about what goes on under the hood, but the mathematics of it are much more accurate, and the end results are much more gratifying.

Edit: When I create more programs, I'll see what I can do about creating a program or two for Nebula 3 Free so people can see what this really can do.

[Agreed]

So how does the sampling process work that you're able to capture the dynamics that an IR doesn't have? Could an amplifier be modeled with Nebula?

[AeternusEternus]

So how does the sampling process work that you're able to capture the dynamics that an IR doesn't have? Could an amplifier be modeled with Nebula?

It's because Nebula is not a convolver, or even a dynamic convolver, which would already be an upgrade to standard IRs. Instead, Nebula uses kernels to model the things that are sampled. This is why the power of Nebula is so absurd. It can model Avalon preamps PERFECTLY. I've sampled a BBE 882i Sonic Maximizer myself. It can do hardware and real reverbs alike, which, given that it can sample anything, makes it the best for reverb on the market. It can sample equalizers, time-based variants like choruses (which is completely impossible with IR), small distortions (nothing like a pedal or a head yet, but more like for tapes and such), and so on. Even now, it can do compressors!

It's a multi-FX sampler, and is the only one in the world. The focus is not on usability or the interface or the distribution, but rather the power and versatility, which gets stronger daily. Cab modeling was not something that was in the original game plan for Acustica Audio, but it worked like crazy.

The scary thing is that the answer to your question is YES, it theoretically CAN model an amplifier. But at this time, Nebula only models up to 10 kernels, and even now, the CPU usage is enormous. So it's impossible right now, because it would require more kernels and CPU usage than is really possible right now.

[Agreed]

Any published work on the topic, or is it totally proprietary for the time being?

I'd love to try it myself but apparently you can't use custom programs with the free version, so unless they're looking for another beta tester I'm going to have to take your word for it and not know what I'm missing.

[xtp]

Good to see you are back at it dude, was a bit worried you might give it away there for a while.

[bmanic]

Any published work on the topic, or is it totally proprietary for the time being?

I'd love to try it myself but apparently you can't use custom programs with the free version, so unless they're looking for another beta tester I'm going to have to take your word for it and not know what I'm missing.

You can look up "vectorial volterra kernels" or something like that and you should get a ton of papers on the theory.

Cheers!
bManic

[Agreed]

Thanks man, I'll see what I can find

[AeternusEternus]

Any published work on the topic, or is it totally proprietary for the time being?

I'd love to try it myself but apparently you can't use custom programs with the free version, so unless they're looking for another beta tester I'm going to have to take your word for it and not know what I'm missing.

Eventually I'll be able to draw up a program or two for the free version of the VST. At least, with Giancarlo's permission, I will.

[Agreed]

That would be very cool, let me know if you get a chance to do that.

I am going to try some of the free programs out and I might end up paying for Nebula, I've been reviewing the literature on volterra kernals and it is a very interesting method, it's actually quite old (developed in the late 1880s, it turns out) but this is a novel application of it. I wish that I understood some of the calculus involved, I regret very much that in my studies I never pursued math very far. I'll see if I can't correct that in grad school, but for now I can at least marvel at the impressive application of old ideas to new problems where they become a novel and exciting solution.

Your clips sound very good, by the way. I can't tell from the clips that it is definitely heads and tails about a well-made IR, but your criticisms of IR are well made, IRs are difficult to work with in many ways that this might not be (if efficiency can be improved or the process simplified - I imagine a cabinet simulation shouldn't be as processor intensive as modeling something with really significant nonlinearities, compared to a compressor a cab is standing still).

[mangel1234]

I've created this link to help with some overview info on the Nebula process:

http://clubvst.forumco.com/topic~TOPIC_ID~3.asp

Nebula has been explained as 'the world's first fx sampler', in that it is able to recognize, sample, and replay very complex multiple forms of information simultaneously, and to an extent automatically. Getting a perfect program is not easy, but once the process is perfected with proper levels, excellent converters, and an understanding of what makes for a good sampling session- the results are light years better than traditional IR's.

What makes the process better than dynamic convolution is that the D.C. process is essentially the same as static IR's done at such a huge number of instances, that when all individual settings are patched together in sequence, that they seem to represent the whole picture- not unlike digital bits make a staircase-shaped representation of audio waves, where the more bits lead to more realistic results. With Nebula's process, the information is not taken one snapshot at a time, but sampled over and over again at multiple volumes in order to compare layer after layer of difference in results. What a preamp or cabinet does at -10dB with a variety of tone sweeps is very different from its reaction at -12.5dB with the same source tones. Also, frequency information gathered is very different from that of limiting/compression/leveling, regular dynamic changes, and harmonic distortion changes.

Each kernel is a live, constantly adapting and updating stream of data replaying what was sampled, lined up with other streams of data. For instance, one order of harmonic distortion receives its own kernel of data played along with the next order etc. Each one replays what was sampled. The brain and ears pick up on subtle differences in waveforms much faster and with much more detail than most analysis software will reflect. The difference in dimensions of information and stair-steps in between even the best dynamic sampling is what tells our ears that something sounds more real.

Ironically, the trickle-down effect is much like good mastering of recordings. If the best process has been accomplished, you actually can create more simplified programs with fewer parameters or fewer levels of dynamic samplings within Nebula, and the snapshots of what has been sampled will sound more real than static IR's. This is similar to capturing a digital image with a great camera with millions of colors, and then reducing it to a 72dpi image for web. If you just shot it at 72dpi it would not have captured complex enough info to begin with to be believable. But, reducing an excellent image down provides a decent representation of the range of color and detail that was originally there. The difference in this example is that if a program in Nebula is simplified for fewer samples, the resolution does not decrease, just the complexity of changes measured.

Another aspect of this is that static IR's have a place in time when absolutes are desired- like absolute filtering or brickwall settings. There's nothing wrong with static when it is replicating something unchanging and stable.

Another significant aspect of the Nebula process is that it allows keyboard sampler-like control of the real-time programs that have been sampled- allowing a program to be played back, or processed, with more distortion or less, more eq or less, more compression or less, etc. All of these factors having been sampled from the equipment for real-not imagined settings, but a full range of data actually collected and streamed with accuracy.