Re-opening the Nebula for cabs discussion

[Ermz]

Some of you may remember the fairly large set of threads we had here in the past regarding the use of Nebula for cabinet simulation. A large chunk of us were very adamant and hopeful - so much so that we chipped in to buy Marcus (Metaltastic) a copy of Nat2 and Nebula 3.

The tests didn't fare quite as well as most of us had hoped and the idea was eventually abandoned after all the threads crawled to a halt with no new developments.

So, coincidentally whilst browsing a convolution reverb thread on Gearslutz I came across one of the developers at Acusticaudio. As the discussion developed I made mention of Nebula and our inability to get convincing cabinet simulations with it in the past.

Here is what the developer had to say in regards to my recital of the experience:

Something is wrong in what you did.

The sampler:
If you limit to 1 kernel/no dynamic you get an impulse response. Nat calculates it @64 bit, which in general is not possible using other deconvolvers (64 bit deconvolution is not supported by all other developers). We support a long testtone, till 60 seconds or even more. So you should get a better result and a better S/N ratio.
Papers said that this testtone is the best one even for sampling outdoor.

The player:
Nebula engine is in general better than the one provided by other developers (convolvers) for several reason: for example the early tail could be played by direct convolution, which has a greater precision than FFT. FFT is calculated @64 bit with best accuracy (no speed optimizations), several algorithms are provided.

If you add harmonics you'll get harmonics. If you add the dynamic variable you add this dimension and so on.

Do any of you guys that tried capturing your cabs with Nat have anything to add here?

I've invited zaphod to this thread to hopefully share some more insights with us and understand what went wrong and where.

 

[AeternusEternus]

Okay..

I went ahead and bought a copy at the same time everyone bought Metaltastic his copy. I opened my own thread, which is now linked in my signature, and just started sampling stuff relentlessly. I've done a lot of tests. Many test tone lengths, many Acustica-created NAT sessions with varying kernels.

I've got impulses and programs that match, and the sound result is about the exact same, but Nebula was always determined to be a little bit more unpleasant to the ears. The unpleasant sounds everyone described were the higher frequencies. Even the non-dynamic one-kernel "impulses" that I created with Nebula sounded harsher than their impulse equivalent.

I maintain that Nebula sounds great for cabs, but I stopped making them because no matter what session files I tried, no matter what gear or mics I sampled, people just weren't getting into it. I think that part of the falloff had to do with the fact that it's ME making the programs and impulses. I have no Fredman-style end-all programs or impulses under my belt. It's just, why keep making Nebula programs if impulses sound better? Nobody would get on board, especially if the latency issue existed in any capacity greater than that of something like KeFIR.

So, I dunno. I use it now for the BBE 882i that I sampled and for the Doc Fear EQ. I'm always interested in seeing where it will go.

If zaphod comes in here, I documented everything I could have documented VERY thoroughly with all of my cabs. If I did something wrong, do let me know, and maybe I can try to correct it. It was uncharted territory at the time.

 

[Metaltastic]

Fuck, I barely remember - I think the only thing I varied was the amount of kernels and then a couple of other templates, all with pretty similar results...

 

[Ermz]

Right, thanks for chipping in so quickly guys.

From what I recall the one thing consistent across all of them was that harshness in the upper registers. Not only that but it seems the mids didn't fill out as well. I'm speaking of course from a purely subjective point of view, but I did reach a consensus in that regard with a few other people. Even with only 1 kernel and what was essentially just a standard IR, the plain IRs sounded more pleasant than the nebula programs. The great thing here is that we have a reference point. We can for instance record a DI track and reamp that through the real cab, alongside the IRs and programs made from that exact same cab, mic, positioning etc. So we can solidly say, without a doubt that there was something awry. Tonally, the IRs were closer.

@Aeternus: I think it's commendable that you kept it going for so long. The inconvenience of the whole operation no doubt contributed to everyone jumping ship. I recall being quite overwhelmed by the entire thing, and then after all of the work to have the IRs sound closer was a bit of a punch in the face. I'd love nothing more than to get back in line with this and see if there is anything that could be done to have Nebula create convincing cabinet simulations. Even if we could just track with IRs, then throw Neb on for the final bounce, that would be fine.

I have a feeling that the only people that can help sus this out are the Acusticaudio developers, or at least some folks who are very mathematically inclined.

 

[giancarlo]

hi,
I think your troubles are a consequence of lacking of proper documentation from our side. I'm sorry because I can't be everywhere and I can't help everyone properly. If you are interested in sampling gear using nat/nebula and sharing that contact me, I'll try to add you to our beta-testing group.
Betatesters have a little advantage: nat generates plain files. So you are able to test them using different convolvers and compare the impulse reponse with the impulse response created using other products.
Being a betatester is not simple though
1) you should not be a developer
2) there is a long list of candidates

If you are going to generate commercial libraries for our products we'll try to be faster with you.


Here a short answer: if you limit kernels to 1 and you avoid dynamic programs and disable "repeats" parameter you HAVE the impulse response.
It's generated using a sine sweep. It can't fail, this is the standard method for creating an impulse response.

First step.
For example start using session/template
Clean Pre (1 kernel) - smooth2.xml

In such case you would have a number of impulse responses specified in the expert page. Just limit the number to "1". Disable REPEATS (set to 1).
This is your starting point, the impulse response. If it sounds differently than using other products/method contact us in our forum, we'll try to analyze your matter.

Second step.
If you add a dynamic behaviour you should get a closer result, because you'll take in account a different behaviour of your device at different gain levels.
Dynamic behaviour is heavy on CPU and could lead to crackles if you are going to process low frequency content, because the IR switching. In any case you could solve it using the LIQUIDITY control. You loose a bit of the dynamic behaviour but you will not hear crackles any more.

Third step.
Now you could change session and deconvolve your previous work increasing the number of kernels (harmonics). In general there is a trouble with S/N. It means that suddenly if you use an high number of harmonics you could get a completely wrong result (exactly what you are describing, an high-frequencies boost). This happens frequently if you try to sample noisy gear. Infact high order harmonics damage heavily your resulting session, because they are multiplied for a big factor (their value is multiplyed even for 1000).
So I suggest to go carefully, increasing the number carefully. You start from 1, than 3, than 5, analyzing the resulting value in your analyzer (nat has a built in analyzer). When you get a completely different frequency response you stop.

Nebula has several advantages, though. For example if you have knobs in your gear you are able to sample in different positions and so on.


Here a session useful for creating a simple impulse response, this is your starting point:
http://www.acusticaudio.com/uploads/9/0/impulse response.rar

------------------
www.acusticaudio.net

 

[xturbox]

can you reference any good paper to read a theory behind dynamic convolution that you coded?

 

[AeternusEternus]

hi,
I think your troubles are a consequence of lacking of proper documentation from our side. I'm sorry because I can't be everywhere and I can't help everyone properly. If you are interested in sampling gear using nat/nebula and sharing that contact me, I'll try to add you to our beta-testing group.
Betatesters have a little advantage: nat generates plain files. So you are able to test them using different convolvers and compare the impulse reponse with the impulse response created using other products.
Being a betatester is not simple though
1) you should not be a developer
2) there is a long list of candidates[/url]

Wait, so you mean to say that if I'm a betatester, I can generate files with NAT that could be used in other convolution plugins? So I could create an impulse and load it into KeFIR? Because if that's what you're saying, then I need to betatest. Because if a Nebula-sampled impulse and a regular impulse sound different in KeFIR and are made from the same equipment and stuff, we got problems that are out of our hands as samplers and users. If they sound the same, then we know it's just tweak city to get a more realistic sound.

 

[Ermz]

^+1

Thank you for the response, giancarlo. I think it may be a good idea to have at least one beta-tester that's using your software primarily for cabinet emulation, and I can't think of anyone better than Aeternus. He seems to have the will to push it to where it needs to go.

 

[Jarkko Mattheiszen]

Yeah, the whole Nebula episode was a bit of a turn-off for me. The end results didn't sound as good as I had hoped, and to be honest, the user interface of Nebula is pretty horrible. And it's processor-heavy as fvkk. So after fiddling with a couple of Nebula "impulses" and doing some tests, I just went back to firing up Boogex and loading up a decent impulse.

 

[cobhc]

^+1

Thank you for the response, giancarlo. I think it may be a good idea to have at least one beta-tester that's using your software primarily for cabinet emulation, and I can't think of anyone better than Aeternus. He seems to have the will to push it to where it needs to go.

+1 Aeternus is the man for the job! The only reason I never got involved in Nebula was having to speficially use Nebula for impulses.

 

[abyssofdreams]

In risk of asking too much, I think the best thing would be if you, giancarlo, or anybody else from the Nebula/NAT-team would do a dynamic program with a recorded cab itself.
This way we could exclude any "noob" errors and the likes. It doesn't have to be an awesome tone but just so we get an idea as to how close we could get with Nebula programs compared to the real deal. Ideally we would have a short clip or shootout with different presets/settings. Would that be possible?
I remember that Francesco wanted to do exactly this but in the end he was very busy and probably forgot it or something...

We aren't looking for a static impulse but a dynamic one that captures all the details a static one naturally isn't capable of.

Anyways, thank you for all your help but I have to admit that I find it really hard to understand the tech stuff behind all this and that's coming from a programmer

I wish I could be of any help!

 

[Sickan]

You can check my mix where I use AE's 1960AV 57 CDC3 ("modified" tho).

 

[giancarlo]

Wait, so you mean to say that if I'm a betatester, I can generate files with NAT that could be used in other convolution plugins? So I could create an impulse and load it into KeFIR? Because if that's what you're saying, then I need to betatest. Because if a Nebula-sampled impulse and a regular impulse sound different in KeFIR and are made from the same equipment and stuff, we got problems that are out of our hands as samplers and users. If they sound the same, then we know it's just tweak city to get a more realistic sound.

yes you could. They sound the same, believe me. Most of initial programs like avalons were deconvolved using voxengo deconvolver. We compared different convolution engines, nebula sounds very very good even as simple convolver. Betatesters often compare the output coming from nebula with the output coming from the real gear using frequency and harmonic distortion analyzers. Ok for betatesting, contact me.

 

[Ermz]

In the mean time is there anything you could tell us here to keep us going. Any good starting points for Metaltastic, who also has a copy of NAT and a good cabinet to sample with it? Our main issue was the harshness in the upper frequencies, but for some reason that occurred even when using only 1 kernel. The main ambition was to have cabinet IRs that are actually dynamic and model speaker distortion, amongst all those other little things that separate current IRs from the real world.

 

[giancarlo]

Yeah, the whole Nebula episode was a bit of a turn-off for me. The end results didn't sound as good as I had hoped, and to be honest, the user interface of Nebula is pretty horrible. And it's processor-heavy as fvkk. So after fiddling with a couple of Nebula "impulses" and doing some tests, I just went back to firing up Boogex and loading up a decent impulse.

nebula3 pro is light on cpu.
Here a screenshot of nebula3 pro / nat3. As you can see on a cheap core2 duo an eq is around 3-4% for a single core, so less than 2% of the whole cpu power. Here an emulation of an expensive tube equalizer.

 

[GarethSE]

nebula3 pro is light on cpu.
Here a screenshot of nebula3 pro / nat3. As you can see on a cheap core2 duo an eq is around 3-4% for a single core, so less than 2% of the whole cpu power. Here an emulation of an expensive tube equalizer.

See, that's the thing, I'd jump on board and try and help (if i had any money that is, this is just ifs and buts is all), but in all honesty I don't think I could handle working with that.
I look at that and my brain just goes HOLY SHIT WHAT THE FUCK WHAT IS GOING ON HELP PLZ OMG OMGOMGHELPZPLZPLZPLZ AHHHH and then it explodes in a shower of giblets, coating any passers by with a meaty concoction of confusion.
The interface needs work because at the moment figuring it out feels like half the battle, and it's just taking away from the focus of trying to get realistic sounding cab simulations.

 

[abyssofdreams]

See, that's the thing, I'd jump on board and try and help (if i had any money that is, this is just ifs and buts is all), but in all honesty I don't think I could handle working with that.
I look at that and my brain just goes HOLY SHIT WHAT THE FUCK WHAT IS GOING ON HELP PLZ OMG OMGOMGHELPZPLZPLZPLZ AHHHH and then it explodes in a shower of giblets, coating any passers by with a meaty concoction of confusion.
The interface needs work because at the moment figuring it out feels like half the battle, and it's just taking away from the focus of trying to get realistic sounding cab simulations.

well put and big plus uno

 

[giancarlo]

the interface will be simpler.
But believe me, the interface is simple..... let me explain: you have

1) a lcd display, and it's showing the preset name
2) 8 big knobs with a label. In this example you have control for 100Hz, 500Hz and 12Khz. It's an eq. There is also a DRIVE knob (distortion). Forget the last control (liquidity), it's for experts.
3) input and output gain

Imho a software or hardware sampler is way more complex.
I hope it helps.

 

[GarethSE]

the interface will be simpler.
But believe me, the interface is simple..... let me explain: you have

1) a lcd display, and it's showing the preset name
2) 8 big knobs with a label. In this example you have control for 100Hz, 500Hz and 12Khz. It's an eq. There is also a DRIVE knob (distortion). Forget the last control (liquidity), it's for experts.
3) input and output gain

Imho a software or hardware sampler is way more complex.
I hope it helps.

if that's all there is to it, then whats all the other crap doing there?

and also, "its for experts"? this is a forum full of people who I'd say are full of fucking expertise. don't partronise us.
i'm sure we all figured out the parts you outlined in points 1 through 3, but the rest of it just doesn't need to be there.
it's bad design, own up to it and make some attempt to fix it.
it'll open up your demographic to no end, and people like myself who quite simply don't have the drive to deal with a shoddy interface will be all over it like flies over fecaes.

sorry if i come across as somewhat vitriolic, but that doesn't quite cut it i'm afraid.

 

[sash poask]

if that's all there is to it, then whats all the other crap doing there?

and also, "its for experts"? this is a forum full of people who I'd say are full of fucking expertise. don't partronise us.
i'm sure we all figured out the parts you outlined in points 1 through 3, but the rest of it just doesn't need to be there.
it's bad design, own up to it and make some attempt to fix it.
it'll open up your demographic to no end, and people like myself who quite simply don't have the drive to deal with a shoddy interface will be all over it like flies over fecaes.

sorry if i come across as somewhat vitriolic, but that doesn't quite cut it i'm afraid.

+1 i agree with you man, while i would in no way even think of classing myself as one, there are plenty of experts on this forum, so saying things like that does come across as patronising, but im sure that was not the intent...right?