Creating my own IRs and I will share in this thread soon

[Desertdweller]

Not sure if this belongs here or in the recording section, but it's digital so why not?

I took it upon myself to start shooting my own IRs recently. I have a fantastic old 90s Mesa slant/straight 4x12 cab that I bought new back in 1999, and though I love impulse responses from so many manufacturers, none of them really sound to my ears like my cab. I've got years of recording guitar under my belt, and I have enough gear to get it done, so this is my challenge for myself. The end result I'm shooting for is that I can make a set of IRs that I will be happy with and can share with others. I have ZERO ambition right now of selling anything, I do intent to share a free IR pack with everyone as soon as I'm happy with it.

Gear I'm using:

• Mesa 4x12
• SM57 dynamic mic
• sE Voodoo VR1 Ribbon mic
• Various Mesa amp heads to get the sound and mic position right
• Synergy 5050 power amp to shoot the IRs
• ART MPA II tube mic pre in low plate voltage mode to receive the mic signal
• Reaper as my DAW and white noise generator

So here's my quick hit findings:

1. The white noise blip way of creating the IR is fantastic. I see no need for the obnoxious & loud sine sweep.
2. As always, SM57s are so touchy. You can find a perfect spot but accidentally nudge it a mm or two and it's a completely different sound
3. Mixing 2 SM57s together produces superior results
4. Ribbon mics are amazing - they pick up what your ear actually hears
5. Shooting your own IRs is not hard to learn, but much like learning an instrument it's difficult to master
6. I still love doing things the way I'm not supposed to - I used a tube Syn5050 power amp instead of a SS amp to shoot IRs and they are coming out fantastic.
7. Visually aligning and gluing in Reaper is a great way to phase align IRs.
8. My Torpedo Captor X produces a squashed bass response compared to other load boxes. I've been using it to demo my IRs as I create them, and I find that running IRs through this device does not accurately produce the bass response of using the CabClone IR in my Badlander, the Cab Block in my FM9, or a VST IR loader in Reaper. I did a factory reset and wipe of the Captor and it is still the same result. Quite interesting.
9. I'm having a ton of fun doing this super nerd shit
   

More to come as I work through the kinks, I'm hoping to post a clip or video and then share at least one or two IRs or maybe even a whole pack soon.

 

[James Freeman]

Great!
Can't wait to try!

white noise blip

Is that a single sample click?

 

[MirrorProfiles]

Looking forward to hearing them. I have a 1999 Mesa Slant OS that sounds killer, great cab.

Ultimate test for me is comparing the sound of the mic’d cab vs the amp+IR (with either the same load or a load box with similar characteristics). If they sound indistinguishable then it’s job done.

 

[MirrorProfiles]

btw, may be cool to test the IR’s using the cab as the load and either take the line out from the amp, or put the captor between head and cab. You can avoid the weird Two Notes impedance curve that way, which Mesa amps can be very sensitive to

 

[Desertdweller]

Is that a single sample click?

That's it!

I take Reaper's white noise generator:

Record it to a track and get it down to a very small sample:

 

[Desertdweller]

When phase aligning, I'm trying to get the first peak dead-on in my IRs, and then the second peak as close as possible. This has produced great results as I can mix any two of the IRs I've created and do not have phase problems. The red line is where I'm trying to align to:

 

[James Freeman]

Why not just use a single sample dirac instead?

Homepage noisetime - How to create impulse responses

Homepage noisetime - How to

www.noisetime.com

 

[Desertdweller]

Why not just use a single sample dirac instead?

Homepage noisetime - How to create impulse responses

Homepage noisetime - How to

www.noisetime.com

Because I've never heard of it
Now I will have to look into that too!

 

[AlbertA]

1. The white noise blip way of creating the IR is fantastic. I see no need for the obnoxious & loud sine sweep.

Sorry but that's a terrible stimulus signal.

There's reasons why we don't use white noise blips. First is the crest factor, you want to spread energy over a period of time so that you don't overload the DUT chain (increased SNR)

A logarithmically swept sine also has useful properties:
1. All non-linear products are pushed to "negative" time
2. If your sweep ensures that every octave starts and ends in sine phase, then the distortion products in negative time will be in ascending harmonic order as you go into negative time - ideally though, you will set a level where these products are minimal in level.
3. A useful thing about ending at sine phase with a sweep that extends to the nyquist limit is you avoid having to window the end, which in turns avoids pre-echo and ringing after deconvolution.
4. If you start the sweep at 10 Hz, you can use the first octave (10-20 Hz) to window the sweep to avoid low frequency ringing during deconvolution.
5. Note the length of the log sweep is unrelated to final IR length.

You can use Room EQ wizard, which already has all the tools to create IRs: https://www.roomeqwizard.com/

 

[AlbertA]

Why not just use a single sample dirac instead?

Homepage noisetime - How to create impulse responses

Homepage noisetime - How to

www.noisetime.com

That only works in the digital domain.

For actual capture of physical guitar cabinets that would be a terrible signal, because it's a full level signal concentrated in just one sample (think 1/48000 of a second) - you would get terrible signal to noise ratio.

 

[Desertdweller]

@AlbertA I've actually very different opinions on this, and I appreciate the input! Some feel the white noise blip is superior to the sine wave and gets a more accurate IR. My understanding is sine waves are better for tuning home stereo equipment, but recently I've seen more videos and forum posts of guys using the white noise method.

 

[Desertdweller]

Here's my first test clip to share. This is a 2 mic mix, SM57 and VR1 together. I'm running this with no other effects on the rhythm guitar, and I'm using the Recto1 Red amp model in the FM9.

 

[AlbertA]

@AlbertA I've actually very different opinions on this, and I appreciate the input! Some feel the white noise blip is superior to the sine wave and gets a more accurate IR. My understanding is sine waves are better for tuning home stereo equipment, but recently I've seen more videos and forum posts of guys using the white noise method.

The great thing about math is that it's not up to feel :)

If the goal is accuracy, a logarithmically swept sine is a superior stimulus signal to a white noise blip stimulus. Assuming they deconvolved the input white noise stimulus, you end up with residual noise in the IR (i.e. reducing the accuracy) - the other issue is that it's not inmune to non-linearities in the chain corrupting the recovered impulse response.

The tradeoff is this - in order to reduce residual noise with a white noise blip, you have to increase the level of the white noise blip to the cabinet - doing so can put any of the devices involved in the capture in the non-linear region (something that wouldn't occur normally when you are just recording your playing) - this is why the crest factor is one of the measures of comparison for various stimulus signals.

In contrast, you can spread that energy in a logarithmically swept sine stimulus signal over time, so that get much better SNR in your capture and therefore in the recovered impulse response. And as mentioned above, it has the property of rejecting all distortion products out of the impulse response - another win.

If they don't even deconvolve the stimulus, that's like the equivalent a applying a bit of random EQ to the IR.

And for completenes, because many times, people confuse sonically pleasing with accuracy, I consider those two completely separate concerns.

For example, I could capture a guitar cabinet that to me sounds really bad, but I could do so in an accurate way - in other words, the resulting IR would accurately represent the guitar cabinet that is not sonically pleasing to me.

By the same token, you could make a non-accurate Impulse response that's sonically pleasing.

 

[Desertdweller]

If the goal is accuracy

Just to be clear, the goal is crushing skulls

Seriously though I'm always so happy to do things the wrong way and get good results from it. I'm not supposed to use a tube amp either but I think a very balanced one like the Synergy is working out really well.

I do 100% appreciate your input on this and find what you've posted about sine waves fascinating. Everything I've read recently runs opposite to that, so that tells me I need to do some sine wave experimenting too. As a first-timer and absolute amateur at making IRs, I appreciate your thorough responses!

 

[JiveTurkey]

Sorry but that's a terrible stimulus signal.

There's reasons why we don't use white noise blips. First is the crest factor, you want to spread energy over a period of time so that you don't overload the DUT chain (increased SNR)

A logarithmically swept sine also has useful properties:
1. All non-linear products are pushed to "negative" time
2. If your sweep ensures that every octave starts and ends in sine phase, then the distortion products in negative time will be in ascending harmonic order as you go into negative time - ideally though, you will set a level where these products are minimal in level.
3. A useful thing about ending at sine phase with a sweep that extends to the nyquist limit is you avoid having to window the end, which in turns avoids pre-echo and ringing after deconvolution.
4. If you start the sweep at 10 Hz, you can use the first octave (10-20 Hz) to window the sweep to avoid low frequency ringing during deconvolution.
5. Note the length of the log sweep is unrelated to final IR length.

You can use Room EQ wizard, which already has all the tools to create IRs: https://www.roomeqwizard.com/

Hey remind me. Are you the creator of the Antcarrier FFIRs?

 

[AlbertA]

Hey remind me. Are you the creator of the Antcarrier FFIRs?

Hah no, that's an awesome australian dude - I'm in Texas.

 

[paisleywookiee]

Sorry but that's a terrible stimulus signal.

There's reasons why we don't use white noise blips. First is the crest factor, you want to spread energy over a period of time so that you don't overload the DUT chain (increased SNR)

A logarithmically swept sine also has useful properties:
1. All non-linear products are pushed to "negative" time
2. If your sweep ensures that every octave starts and ends in sine phase, then the distortion products in negative time will be in ascending harmonic order as you go into negative time - ideally though, you will set a level where these products are minimal in level.
3. A useful thing about ending at sine phase with a sweep that extends to the nyquist limit is you avoid having to window the end, which in turns avoids pre-echo and ringing after deconvolution.
4. If you start the sweep at 10 Hz, you can use the first octave (10-20 Hz) to window the sweep to avoid low frequency ringing during deconvolution.
5. Note the length of the log sweep is unrelated to final IR length.

You can use Room EQ wizard, which already has all the tools to create IRs: https://www.roomeqwizard.com/

It's too bad it uses Java. I won't touch that, even if it's a private runtime.

 

[-BOOGIEMAN-]

Here's my first test clip to share. This is a 2 mic mix, SM57 and VR1 together. I'm running this with no other effects on the rhythm guitar, and I'm using the Recto1 Red amp model in the FM9.

Let us know when they are ready

 

[Baba]

1. Ribbon mics are amazing - they pick up what your ear actually hears

I could not care any less about IRs, but I agree with this wholeheartedly, which is why I love them.

 

[Desertdweller]

Let us know when they are ready

Thanks I will! I’m thinking next week. I think they’re just about ready, but I want to test with one more load box that should arrive early next week. I’m seeing quite a bit of variability in the Captor X in the bass response compared to other load boxes and I want to be sure of what I’m hearing before I publish anything.