FREE Custom IRs for Your Guitar and Its Pickup (Acoustic Guitar IR Impulse Response)

[jonfields45]

After you install Octave you want to run the GUI version:



I've circled where you need to set the directory to the one with your guitar sample and my script, jf45ir.m. The "X" marks the spot where you type the command to run the script.



If your recording is mgit.wav then you would type:

jf45ir('mgit')

And hit the carriage return. The three IRs and one frequency plot will be placed in the directory with your sample and my script.

The WAV file guitar sample needs to be 60 seconds long (or longer), pickup left, mic right, no clipping, peaks within 15 dB of clipping (pretty easy to achieve), and the mic and pickup should be roughly the same volume in the WAV file.

[wood nacho]

Is there any reason this wouldn't work with my piezo hammered duclimer pickup? The pickup sounds quite good but is definitely missing some of the 'airy' quality of a microphone.

[63telemaster]

Thanks Jon and Cuki for your work and thanks for those screenshots....they are a big help

I'll attempt to produce an IR from my D28 with Black Angel pickup and load it into my HX Stomp. If it's successful I have a particularly nice sounding 000-28 Martin Shenandoah that I can also sample along with a Sunrise pickup so I'll be able to do some comparisons.

I'll pm any samples over to you for your database, but this could take a few weeks with my current workload. Thanks again for the help.

[jonfields45]

Originally Posted by wood nacho

Is there any reason this wouldn't work with my piezo hammered duclimer pickup? The pickup sounds quite good but is definitely missing some of the 'airy' quality of a microphone.

Give it a try and let us know. There is nothing in the algorithm that assumes something that would make it not work.

I think most other approaches to this problem add some equalization post IR generation or make careful selections of the sample to process based on assumptions about the instrument to be processed. This script is just instrument agnostic math.

[Mcirish]

I stumbled upon this from a post on mandolin cafe. Funny how this is all playing out. Back in early December, I started thinking about creating an IR for each of my instruments. I play 7 instruments in my band and have never been completely happy with the sound I got. So, with my DAW, I started experimenting in creating IRs from a matching EQ based on the pickup and a mic version of each instrument. It ended up being about 20 steps to create the IRs but the results were well worth it. I bought a Mooer Radar first but then realized it would not work live because I need a quick and easy way to change the presets on the fly. I bought a Joyo Cab Box. I just got back from using it on a tour. It worked fantastic. The sounds I got were completely natural. My $10k mandolin finally sounded good cranked through the PA. I can post my instructions if anyone is interested.

The Joyo Cab Box is "almost" perfect. The only issue is that you cannot set the volume of each preset unless you turn on the power amp emulation (which I don't want). So, I just have a couple pedals I use to match the gain of some presets to the others. Not a big thing but Joyo could easily give the option to turn off the power amp emulation and still keep the volume settings. I've emailed them a few times but they seem pretty unresponsive.

I see using IRs for acoustic instruments as the new standard. No other acoustic pedals have come even close to what this can do. Whatever manufacturer finally takes this and fully implements it is going to be successful. The ToneDexter is similar but way too cryptic.

[Mcirish]

Just to keep the discussion going, here are my steps I used to create my IRs using Cubase/Nuendo, Ozone, Wavelab and Voxengo Deconvolver. I'm sure I can improve on the procedure but this did work well.

1) Instrument with mic into preamp - to DAW
2) Instrument into preamp DI - to DAW
3) Get levels somewhat close to each other without any clipping
4) Record both tracks at the same time in various ranges of the instrument
5) Normalize both resulting recordings
6) Export those two recordings - at least you need the one that was mic'd

7) On the DI track, add Ozone to it and do an EQ match to the mic'd recording
8) After you have the match, save the EQ match (you will need it)
9) In Deconvolver, generate a test tone. 6 second sweep 24 bit 44.1k mono
10) On a new track in DAW, add the test tone file
11) Add Ozone as an insert and apply the matched EQ to the sweep
12) Export the sweep with the EQ and save as a wav file

13) Open Deconvolver and add the test tone (the raw one - not EQ'd) in the test tone file line
14) In the file folder line, add the exported test tone with EQ on it
15) Select where to save the output file
16) Bit depth 24 bit, Normalize, low cut at 70hz, hi cut at 14khz
17) Process - the saved wav file is the IR
18) Edit the IR in a wave editor to trim the silence at the beginning and fade the tail -
a. if you don't trim the blank at the beginning, you will have latency
b. IR should be around 25ms - inexpensive IR loaders will truncate to 20-25ms
19) Load the IR into an IR loader and test

I'm looking forward to trying Jon's program. That would sure save a lot of steps.

[Doug Young]

Originally Posted by Mcirish

Just to keep the discussion going, here are my steps I used to create my IRs using Cubase/Nuendo, Ozone, Wavelab and Voxengo Deconvolver.

I've done something very similar using Logic's SpaceDesigner/IR Utility, which tends to make a process a bit shorter. It produces good results, tho not as mic-like as ToneDexter or Cuki and Jon's algorithms. The EQ IRs are a bit more direct sounding to me, and work well in cases where the other approaches are too lively. It was interesting to create these, but I generally just use the ToneDexter IRs.

I use the CabM pedal from TwoNotes, which supports up to 200ms, and have it loaded with my EQ-based IRs, ToneDexter IRs, and a few from Jon and Cuki. The CabM does support per-patch volume adjustments, so I can make all the IRs have the same volume. It also supports an additional 5 band parametric EQ, 100 patches, and text names, which is a nice touch when you have that many IRs.

I assume your comment about ToneDexter being "cryptic" is just because it has numbered patches. I solve that pretty easily with a strip of paper with the guitar/pickups taped to the front. Its not much different than most pedals with presets - most don't have programable text displays, tho the display on the CabM certainly is a nice feature when you have a lot of IRs to keep track of.

[GuitarLuva]

Originally Posted by Mcirish

Just to keep the discussion going, here are my steps I used to create my IRs using Cubase/Nuendo, Ozone, Wavelab and Voxengo Deconvolver. I'm sure I can improve on the procedure but this did work well.

1) Instrument with mic into preamp - to DAW
2) Instrument into preamp DI - to DAW
3) Get levels somewhat close to each other without any clipping
4) Record both tracks at the same time in various ranges of the instrument
5) Normalize both resulting recordings
6) Export those two recordings - at least you need the one that was mic'd

7) On the DI track, add Ozone to it and do an EQ match to the mic'd recording
8) After you have the match, save the EQ match (you will need it)
9) In Deconvolver, generate a test tone. 6 second sweep 24 bit 44.1k mono
10) On a new track in DAW, add the test tone file
11) Add Ozone as an insert and apply the matched EQ to the sweep
12) Export the sweep with the EQ and save as a wav file

13) Open Deconvolver and add the test tone (the raw one - not EQ'd) in the test tone file line
14) In the file folder line, add the exported test tone with EQ on it
15) Select where to save the output file
16) Bit depth 24 bit, Normalize, low cut at 70hz, hi cut at 14khz
17) Process - the saved wav file is the IR
18) Edit the IR in a wave editor to trim the silence at the beginning and fade the tail -
a. if you don't trim the blank at the beginning, you will have latency
b. IR should be around 25ms - inexpensive IR loaders will truncate to 20-25ms
19) Load the IR into an IR loader and test

I'm looking forward to trying Jon's program. That would sure save a lot of steps.

I like your idea on utilizing Ozone. I haven't used that plugin for a while and kinda forgot its capabilities. The EQ matching is something I will try so thanks for sharing. I've used Jon's IR procedure with great success so far.

[jonfields45]

Originally Posted by Mcirish

Just to keep the discussion going, here are my steps I used to create my IRs using Cubase/Nuendo, Ozone, Wavelab and Voxengo Deconvolver. I'm sure I can improve on the procedure but this did work well.

Hi Mcirish,

I'm looking forward to hear how you make out. Don't hesitate to email me directly if you have any questions.

If you look at my code you will see if is just brute force (huge windows, FFTs, and IFFTs in double precision floating point) deconvolution of two correlated LTI systems (guitar to pickup, and guitar to mic in this case) with a little extra math that I've not seen published but seems like a no brainer to me. There's no assumptions being made about guitars or pickups or mics. I'm sure the results can be improved with EQ to taste and some ambience/reverb, but it seems to me that should be the end users' judgement and not cooked into the IR generator.

In my own usage I have a 9kHz high cut and a 20 Hz low cut as my first stage (HFNs can be a bit brittle sounding and subsonic filtering is something I do without too much thought), then 60% IR 40% pickup, then my foot switch assigned compressor for single note solo boost, and finally a touch of plate reverb.

Thanks!
Jon

[Cuki79]

Originally Posted by GuitarLuva

I like your idea on utilizing Ozone. I haven't used that plugin for a while and kinda forgot its capabilities. The EQ matching is something I will try so thanks for sharing. I've used Jon's IR procedure with great success so far.

Hi everyone,

thanks Mcirish for this contribution,I think Yamaha had a patent for that technic but I honestly can not put my hand on it... They have published so many patents afterwards.

They actually patented the Image Casting thing (how to make your guitar sounds like another one)

Patent link

Here is another patent from them illustrating IR making algorithm
Patent link 2

[jonfields45]

Originally Posted by Cuki79

Hi everyone,

thanks Mcirish for this contribution,I think Yamaha had a patent for that technic but I honestly can not put my hand on it... They have published so many patents afterwards.

They actually patented the Image Casting thing (how to make your guitar sounds like another one)

Patent link

Here is another patent from them illustrating IR making algorithm
Patent link 2

I took a quick look at one of the patents and immediately recognized the LMS (least mean square) filter adaptation techniques we used in my working days on hard disk drive read channels and ethernet transceivers in the 1990s. I see no reason why this would be particularly hard to program in Octave...

Maybe some more time killing programming for me :~) !!!

What I like about this approach, if I can get it to work (converge), is no window functions, discrete (fast) Fourier transforms, or inverse Fourier transforms.

The expected value is the mic, and the error is the difference between the mic and the pickup. You assign all that error to one FIR (IR) coefficient for a while, and then move on to the next one. You keep going round and round until it (hopefully) converges.

[hiddenmickey]

Originally Posted by Doug Young

I use the CabM pedal from TwoNotes, which supports up to 200ms, and have it loaded with my EQ-based IRs, ToneDexter IRs, and a few from Jon and Cuki. The CabM does support per-patch volume adjustments, so I can make all the IRs have the same volume. It also supports an additional 5 band parametric EQ, 100 patches, and text names, which is a nice touch when you have that many IRs.

I assume your comment about ToneDexter being "cryptic" is just because it has numbered patches. I solve that pretty easily with a strip of paper with the guitar/pickups taped to the front. Its not much different than most pedals with presets - most don't have programable text displays, tho the display on the CabM certainly is a nice feature when you have a lot of IRs to keep track of.

Doug - how do you control the blend between pickup and IR in the CabM for live applications? Let's say I want 60% UST pickup and 40% IR. How would that be accomplished in the CABM?

[Doug Young]

Originally Posted by hiddenmickey

Doug - how do you control the blend between pickup and IR in the CabM for live applications? Let's say I want 60% UST pickup and 40% IR. How would that be accomplished in the CABM?

There's no blend in the UI, but there are lots of ways to create a blended IR and load it. They have software that can do all kinds of things with IRs, including blending up to 8 of them (and an "IR" can be a no-op, or a straight pass-thru), EQing, adjusting phase, length, offset, etc, then loading that manipulated sound into the CabM.

In addition, for me, I'm taking IRs from ToneDexter (among other things), so I just set up ToneDexter with the desired blend before I capture the IR. Right now, I have loaded 40%, 80%, CH0, CH1, CH2 for most guitars. In reality, I generally use CH0 or CH2 for most guitars. There's also no "bypass", so I have blank patch setup that I can switch to if I want no IR at all.

Here's the Two Notes software: https://www.two-notes.com/blendir

This is a pretty nifty tool for capturing IRs from hardware as well. Handles the sweep, deconvolution, etc, that GuitarLuva posted about - just push the button and wait a few seconds while it does its thing - assuming you can attach whatever device you're trying to capture to your audio interfaces in/out. This is a free download, so anyone can play with it. Unfortunately, it creates a proprietary IR format that only works in their pedals (tho the pedals can also load normal IRs, like from Jon or Cuki...)

[hiddenmickey]

Thanks Doug!

[kevinplarson]

Originally Posted by jonfields45

One of the reasons I'm giving this away and making sure doing it yourself is easy is I hate getting stuck in a vender trap.

Once you've downloaded my Octave script you are independent from me too!

The joy of my HX Stomp is it is literally the only pedal on the floor and it goes directly into the 3.5mm TRS input of my CP8. No mess on the floor and no mixer.

However, new guitars and new pickups seem to be a problem I have and I want to be fully independent with IRs for whatever I end up playing next.

The other thing I asked Doug Young to help me with is bootstrapping slightly more capable IR loader pedals. To drive demand for IR loaders that have the IR length to excel at acoustic guitar application first needs the "egg" or IR (wavemap/image) source.

My dream pedal would be 2048 sample IR capable, have mix and volume knobs, a bypass switch, run off standard Tube Screamer compatible 9 volt negative tip power, have that tiniest of pedal form factors, and a $100 MAP. This would be the ideal addition to your existing acoustic guitar preamp that covers everything else.

The next step is to turn my script running on Octave into Windows and MAC programs (that do not require Octave installation). I think this is a reasonable problem for someone with the right skills and development software.

That price point and feature set is not a challenge IMO speaking as a retired EE.

Could you explain more about your stomp setup?