New type of convolver

Dell · 19 Sept 2010

Nice!

I'm going to download the videos right now.

Cheers!

McBarGig · 20 Sept 2010

Thanks Dell

I want to post some other information about this new take on old technology...and...a little background.

For years people have been trying to get a true representation of a Cabinet from an IR file. IMO, You cannot represent 4 minutes of guitar playing with a 4 ms recording of a cabinet, that is the length of most cabinet files(give or take). It just does not explain everything the speaker is doing. So, why not record an IR that represents the entire song, every break, every start every nuance, the whole damn thing. Is there a rule that says you have to use a 4 ms IR file? Of course not..Problem is, try loading a 4 minute IR file into any convolver and we'll talk again in 6 months...

So, the idea is to record 2 tracks..One with your mic'd cabinet from the floor, and a DI track both at the same time, so that they are identical, the only difference being that one was recorded from a mic'd cabinet, the other was just direct.

You take these 2 tracks and deconvolve. What you are left with is one big IR file. That IR file represents your cabinet on the floor as it was being played for the entire song. You also have a DI Track of this performance. You take the Newly created IR File and place it into your daw.

The rest is almost like a standard chain on your DI track.

TS(Whatever)>Amp sim>Cabinet(In this case its IR-A-LIVE)

Then the only difference is, instead of loading a little 4 ms IR file, you route the track containing the de-convolved cabinet to IR-A-LIVE. Now you have a complete tonal representation of what your cabinet really did throughout the entire performance. IMO that just cannot be truly represented in a little 4 ms IR file...

What about non-linearities? We have no way of representing these yet, but, what you can do, is since your IR is on a whole other track, there is nothing to stop you from doing some further processing to it, add a compressor, eq, etc..

We have some people trying to explain more in depth what we can do about the non-linearities using this process. I mean in terms of Volume swells and in general, dynamics, all the information is for sure preserved in the original recording of the Cabinet. Since the Cabinet file and the IR file are both identical in terms of performance, we were thinking about reading the original information from the Cabinet recording and replicating the dynamics information into the convolution process, well after it. If the volume goes up by 2 db, then we increase the volume by 2 db...

Its still really early in this project's development, and I am sure as we move forward, things that are not apparent to us now will become apparent.

Disco_Steve · 20 Sept 2010

errm, I dont think you quite understand the physics of what your trying to do, and I think your deconvolving is not deconvolving! sorry, my deconvolver uses so much CPU that it wont work in real time. and you want to continually deconvolve and follow the convolution path, store that out to recolvove with it, but you are only useing one file. you need a multi dymential array to do that, a differnt IR for every single sample in your audio, to convolve with

sorry but you are not remotely doing it 'properly' I think your just adding the differnece and subtracing the difference, thats what it sounds like from your text anyhow. and that is not deconvolving, adding and convolution are VERY different things!

I have no doubt its a cool effect, and I havn't listened to it yet. so I may be talking compleat poo!

also I dont see the point in having a perfect 'IR' as you call it, becasue you have made it from a recording taken from a real amp, which will only replace the lost 'ambience' of exactly the original recording with that amp on that daty presure temperatiue mic preamp cables EM interference etc. so it will be useless on any other material or even the same track but played throgh with just one sampe wrong would ruin everything! so I can't see the point. if you have a recording why do you need to convolve the track into the recording, and you have worked out the convolution by comparing with the blanck anyhow, so it hasn't gained you anything!?

I am very tired today! so sorry if this comes across a bit harsh. i'm just triying to undestand what you are doing as compared to what you think you are doing, and why it will be usefull.

if you are actually doing real 4 min convolution and made it work in SM then WOW!

McBarGig · 26 Sept 2010

OK Steve, to clarify...haha

The deconvolution... In the above statement where I said to use the Mic'd cab, you would actually need 2 files...One of a Line level output of the head, and one with the cabinet and head..Then you can deconvolve that and end up with just the cabinet. The technique I mentioned above is flawed, but most people that record are not recording with a DI from their head

Now in terms of convolutions, IR-A-LIVE has that capability to bring in live sources, not just pre-recorded, so theoretically, there is no limit to the size of a deconvolved(or not) file that comes in. 20 minutes, half an hour if you want...🙂

"sorry but you are not remotely doing it 'properly' " I assure you, its proper convolution...Deconvolving, we do none of that..., Yet...

"I dont see the point in having a perfect 'IR' " OK well, this is a multi-stage answer..🙂 In the guitar world, one of the things that people always complain about in terms of convolution cabinets is their lack of realism. I submit that the only way to mimic that realism is by having the entire performance of the cabinet, not just 4 -8 ms. So, one must remember that this is performance based, currently there is no way to punch in a track, unless you are recording the cabinet and the di at the same time. Then when finished, you can take the cabinet recording outside of the DAW and deconvolve it, bring it back in and send the deconvolved track to IR-A-LIVE for convolution. This is some extra steps I know, but if we can manage to get a real time deconvolver running, then that will eliminate that step.

"even the same track but played throgh with just one sampe wrong would ruin everything" Correct, thats why when recording you have to record more than one track of the same performance. One DI and one Cabinet..

"so sorry if this comes across a bit harsh." No it didn't, no worries.. Watch the videos, and tell me if its a convolver....Now in this video we looped a standard IR file, but that does not have to be so, it could be a long IR file like I said, and it would make no difference. http://dl.dropbox.com/u/923791/IRALIVE/Reaper2.wmv

edit: "and I think your deconvolving is not deconvolving! " It should not be deconvolving...haha Its a convolver! If I said it was a deconvolver somewhere I must have made a mistake. I'll check my post and edit it if I did. If I did say that, sorry about the confusion..

*edit again I just read through my posts Steve and I guess I can see how one would assume its a real time deconvolver, but I was a little sloppy in my description...It is a convolver...not deconvolver.

Nu Audio Science · 28 Sept 2010

You keep saying that the cab file needs to be deconvolved but i haven't found an app to deconvolve more than a few seconds yet, any hints on what you are using to do the task ?

Also no disrespect meant here but i don't think this is very effective at all from my tests so far There are so many artifacts that the cons severely outweigh the pros at this point Using an IR that works great in kefir or such, sounds terrible using this (For what its worth Reaper has been able to do this for a long time and truthfully does it much better than this)

It is a nice idea and i am sure in the future this will be done (However lets be honest here, if it is going to be done then it will have the deconvolver built in on one ch and the convolver on the other, because recording to two chs, L DI and R micced cab is just so common)

I don't really want to put you off here because as i said this is a nice idea and i'm sure at some point in the future will be very common, however i doubt very much it will ever be built in Synthmaker, please feel free to prove me wrong 😉

I really wish you would stop messing around wasting your time on SM Ken and get on with some native (C++) goodness You have the knowledge and the fanbase to make a real go of it

NAS

McBarGig · 30 Sept 2010

Well as I said many times, stay the course man...Its coming, but in its time. This whole deal was put together as a plan and deviation from that plan at this point is not possible...

In terms of Reaper doping this, yes it does(which I only just found that out from Justin), Live Input? I don't know, and Stereo , no..

Thanks for your input...🙂

Nu Audio Science · 1 Oct 2010

But you didn't answer my most important question What are you using to deconvolve the files I can't test this fully till i have some way of doing that Voxengo keeps dissing my ass with a 50 second guitar file

NAS

McBarGig · 2 Oct 2010

Well I'm not...Thats why I said at the reaper forum, here, KVR, GAM, everywhere, we have not fully tested it yet, and we need a real time deconvolver. But, it does convolve anything you send to it, now in terms of that, if you send it a straight source, of course you're gonna end up with a really bad sounding vocoder-ish type of thing. But of you loop a de convolved IR then you'll get a continuous cabinet. In this configuration, there really is only one benefit to it, that is the fact that you can process the IR by placing effects on that track..

Nu Audio Science · 3 Oct 2010

Without the ability to deconvolve or a deconvolver that will do entire tracks available somewhere it seems like a nice experiment but nothing else

NAS