Guitar Cab Impulse Responses by Transient & Sine-Sweep Excitement

by rievezahl in Circuits > Audio

2875 Views, 12 Favorites, 0 Comments

Guitar Cab Impulse Responses by Transient & Sine-Sweep Excitement

Instructable_Main.png
cabs.jpg
mikes.jpg

Hello everyone interested in (home) recording electric guitars, today I've got some great news for you! You can not only have my free guitar cabinet impulse responses (IRs), but also learn how to make your own ones in quite some depth to get you prep'd for creating the IRs of your dreams!

EDIT: New IRs have been processed and added for download in step 2! A better picture documentation (than the first session) of the cab microphonations is given in step 11

EDIT: This Instructables also serves a research purpose, hence I will give a clear formulation of the research questions:

  1. How different do transiently excited IRs sound from sine swept and deconvolved ones? How close to 'reality' is each of those?
  2. What about practicality and effort / value of each approach?
  3. How do transient IRs sound when excited by 'alternative' excitation waveforms? What purpose could these serve?
  4. Does polarity inversion (speaker push vs. pull) have an audible influence on the IR?


Beware when watching the videos in the Step 10: Validation, they seem to answer more than they actually do! Please read descriptions and/or this instructable to the very end before forming an opinion! Detailed answeres to the question based on the information gathered herein are given in Step 12: Conclusions & Outlook

This is a very specialized 'ible, but the content still escalated far beyond what I had planned. I have to assume you are familiar with the concept of IRs, what they are and how they work. In case you aren't, herw are a few resources to check:

Ultimate Guide to IRs - guitargearfinder.com,

Create and Record IRs - soundwoofer.se

Create Your Own Impulses - GREBZ

My apologies, the rest of the intro section has become a little bit messy.

What to expect:

I will show you how to record guitar cab impulse responses using transient excitement as well as sine-sweep methods, plus some cabinet microphonation tricks. I will go into some detail of transient excitement waveforms, comparisons, strengths and weaknesses in contrast to each other as well as to the sine sweep. For download I will offer:

  • a variety of 15 transient excitement waveforms (8 if you don't count the polarity inverted ones, but pushing and pulling a speaker might not do the same thing!)
  • also an all-in-one file, so you can not only record all transient and sine sweep IRs in one go, but also have one metal and one rock guitar sample pre-amp track , recoded with the Marshlands JCM900 Pocket (one of my other instructables!) to conveniently make null-tests with.
  • a total number of 221 (Dec. 2020) + 199 (July 2021) = 420 guitar cab IRs recorded from my own set of cabinets and microphones.

I hope the '.tar.gz' format of the archives is ok, as this site won't let me upload '.zip's. If thats the point, please leave a comment and we will find a solution!

High-end IRs by sine-sweep excitement are said to be more accurate, but require deconvolution of the sweep with a specialized piece of software, a few more clicks and a little bit on knowledge, all of which you will acquire here for free.

While definitely less accurate, transiently excited IRs have the charm of their incredibly simple making process, as well as being something different, yet still highly useable in certain ways.

Skills needed:

  • I assume you know your DAW and/or you look up how to do the things this Instructable makes you do in your specific DAW for yourself - most probably I can't answer DAW-specific questions unless it is about Ardour or Garageband. If you're totally stuck you might find somebody to help you in the comments, but I cannot guarantee.
  • some mikes need phantom power and some can be damaged if phantom power is applied with lack of care. Be sure you know your mikes and how to deal with phantom power!
  • your room, your cables, your mikes, your audio controller: I assume you know how to hook everything up to match the schematics and descriptions provided. Feel free to comment if something is unclear, though.
  • your mixer and devices: I assume you know a thing or two about pre-amps, level sliders and indicators.
  • wave form manipulation and audio editing: possibly there is an easier way to trim recorded audio into an exact desired length and export short stretches than from your DAW. I can recommend Audacity or Sound Forge for these tasks. I assume you have some experience or willingness to look things up on how to do the tasks in these applications.

Some background:

In music production, there is a decline of expensive professional recording studios and audio engineers due to the rise of home studios and bedroom productions, as for the availability of cheaper decent hardware gear, but most importantly even free software that is now very capable of producing good results. This trend has both, advantages, like music production being accessible to a greater number of creative heads, but also a prize, like a lot of equipment is being (ab)used without knowledge how to handle it properly or professionally, or professional producers having less and less time to care about the quality of their decisions for the rising need of productivity to be worth their money. In bedrooms and studios alike, IRs are applied instead of playing through real cabs (at real volume) for their ease of use, reproducibility, price (just think of acquiring a cab, mikes, storing them, going through the hassle of miking them every single time for anew...) and for the reason that they are almost not distinguishable from real cab recordings if done right. You can think of IRs as deep-frozen cab microphonations. The convolver is your microwave, put in, ready, go. The downside being that standardization is not exclusively a good thing in creative fields. Everybody starts to sound the same if everybody uses the same sets of IRs.

The sound influence of guitar cabinets, and most importantly the speakers on the overall guitar sound is getting more and more attention (and let me tell you: its getting it for a reason! It's huge!). Just bear in mind, these are the only parts of the instrument meant to create audible sound (waves).

Lets take a look at the complete chain of influences on the sound of an electric guitar for a moment:

  1. player's playing
  2. strings type and condition
  3. guitar pickups
  4. guitar physical resonance influence
  5. guitar electronics and poti/switch settings
  6. guitar cable
  7. amp model and settings, including tube types, pedals & effects if applicable
  8. amp-to-cab cable
  9. speaker type
  10. cab geometry & materials, including damping, open sections &/ bass resonance tunnels
  11. room acoustics
  12. miking: microphone types, positions, settings and characteristics

Now, I'm not here to talk about the effect of cables or whether the instruments' neck wood has an audible influence (I mean I'm sure I couldn't hear it, but...). At least not more than to tell you: in most components of that chain it is easy to mess up the whole sound, while it takes experience and knowledge, and often enough also fresh (or perfectly pre-worn) wear parts to get it right. Much rather I want to stress once more that the cab and specifically the speaker, as well as the miking setup can hugely affect the tone. And with IRs, their tonal characteristics can be transferred to the digital world - fast, easy and silently. Just load a <1s .wav file to your plugin and your guitar track sounds like played through the exact speaker/cab/mike/room combination the IR was recorded with.

Ready to roast yourself some freshly baked IRs? Lets get going!

Supplies

Supplies, Case 1: End Product Download

You're what I call the "end consumer" and want to download my IRs and rock out your tone with them. In this case you need:

  • guitar (or DI-track* when re-amping) ...*Direct-Injection (DI) means you record your guitar directly, without any amplification or distortion, so you can use software amps or re-amp that signal later.
  • DI-box (for the start, your audio controller / sound card might do, but a really, really high impedance, low noise input would be better. You might see an Instructable for something like that here in the future! ;)
  • PC
  • Digital Audio Workstation (DAW), like Logic, Cubase, Reaper, Ableton, Ardour, Garageband... or other real-time plugin host of your choice
  • IR convolution plugin of your choice, like Poulin LeCab , PULSE, NadIR or others
  • sound output device (headphones, speakers, ...)

Supplies, Case 2: Making Your Own IRs

You wonder about how to record your own cab IRs quickly, easily and most importantly the way that best suits your needs, or you want to get knowledge &/ experimental data on cab impulse excitement and response? Here you go. This needlist is not my invention, but I honestly cant re-find the source I borrowed it from. Well of course, much of it is common sense, but let me clearly state the technique is not mine.

  • guitar + player or DI-track + playing device. Good cell phones are ok, just make sure no EQ is applied!
  • guitar amp of your preferred sound
  • hi-fi amp (probably a good clean channel also does it, but the flatter the frequency response the more accurate the results)
  • guitar cabinet
  • microphones: depending on your liking: SM57 or similar will be great, also a large diaphragm condenser & further dynamic mikes will be highly useful. Maybe you want to try a small diaphragm condenser or ribbon mike too.
  • mike stands
  • cables
  • PC + audiocontroller &/ mixing table (the point is: you have to be able to center/mono-blend at least 2 tracks so the signals overlap and phase cancellation can be heard live, while the 2 channels should be able to be recorded separately. plus: in case of condenser mikes, phantom voltage supply is needed)
  • pair of good, flat-freq'ed, analytical headphones or control room monitors
  • another pair of closed-back headphones, or some that are otherwise designed to well shield the listener from the outside environment. And they shouldn't sound pleasant. Much rather they should honestly tell you any problem with the sound! That's your true friend in the studio.
  • patience and willingness to fiddle with things is a must, experience with miking guit amps definitely helps, but as always, anywhere is a great way to start and gain your experience on the way.

Ready-To-Go IRs

CRATE2xV30.jpg
Technics_Woof+Tweet.jpg
Austrovox120G.jpg

For those who are after some special free ready-to-go IRs, here they are! (Go all the way down to find even more)

This is what you need to do:

  • download my ready-to-go IRs directly from this section
  • open your favorite DAW or other (realtime) plugin host
  • set the line in signal or DI-track as input to you current signal chain.
  • mount your IR-host in that signal chain and make sure you have a
  • pre-amp, distortion effect or software amp right before the IR host.
  • load an actual IR into the IR host, then
  • feed a DI or live guitar playing signal to that track to hear it being played through your software amp and my combinations of cabinets and microphones.

I have to appologize for the quality in advance, as everyone was sleeping in the house when I finally got the chance to record these IRs, Thus I had to keep the volume low, like, really low. That is not optimal for capturing the essence of the speakers, especially as I had to algorithmically remove a lot of humming and other background noise..

NOTE: I always recommend using a (sub)sample-delay when blending IRs, so you can play with phasing to make the most for your purpose. However, if you are going to use my transient IRs when blending, you will OBLIGATORILY NEED to phase-align them, as some transients have a gentle pre-swing that naturally delays the main impulse beyond what I wanted to compensate those IRs with instant attack.

Thats it, have fun playing!

'THE RIG':

Cabinets & Speakers:

  • Austrovox 3x12" (Greenback-similars)
  • CRATE bass cab repurposed for guitar with 2x12" Vintage 30 speakers
  • Technics active speaker with a woofer and a tweeter cone

Microphones as of 2020-12-25:

  • Behringer B2 Pro large diaphragm condenser, with switchable polar patterns: cardioid / 8 / sphere. Please don't judge this mike before you worked with it, as it is one of those lucky shots behringer really got right in my opinion.
  • AKG C1000 small diaphragm condenser
  • AKG D22 dynamic
  • SHURE SM58 dynamic

Microphones as of 2021-07-10 (see step 11):

  • Behringer B2 Pro large diaphragm condenser
  • t.bone RB100 ribbon mike
  • t.bone DB300 large diaphragm dynamic mike
  • SHURE SM58 dynamic
  • SHURE SM57 dynamic
  • Sennheiser MD-421 dynamic

Download-Links:

  • The most listenable single mike and some blended IRs:

https://drive.google.com/drive/folders/1oYm7ntEwQpMpxZtLNYMTTDow2fIDDarE

  • 2020-12-15 transient IRs:

https://drive.google.com/file/d/194li0XH6Bfsj9eb51...

  • 2020-12-25 sine-swept IRs:

https://drive.google.com/file/d/1ltC1n4ihdEBDKQAs-...

  • 2021-07-10 transient IRs:

https://drive.google.com/file/d/1lDcC9h5fqAFgQzC-f...

  • 2021-07-10 sine-swept IRs:

https://drive.google.com/file/d/1P3sa9V6cH89Gusyx6...

Finding Your Tone

RecordingIRs_schematic.png
DSC01442.JPG
DSC01322.JPG
DSC01443.JPG
DSC00128.JPG
DSC05822.JPG
DSCN4684.JPG
DSCN4689.JPG

Now heres the hardest part for any audio chef who wants to cook their own special sauced IRs. Feel free to skip it if you're just here for reading about the excitement-response-experiment.

Although the outcome might be just a mike placement, there is so much you need to get right and so much that can go wrong with that setup. It really is unbelievable to the beginner, so step down your hunger for immediate results and grab a good portion of patience, because the heavy lifting has to be done now.

That said, you should now set everything up to sound just the way you want it to. Grab the guitar player of your trust or prepare some DI-tracks of your own playing and feed them to your favourite guitar amp and drive your favourite cab using the best tone settings that sound just so right.

OK, now it's time to grab some mikes and decide for a miking technique.

  • A simple and very wide spread technique involves a single Shure SM57 or similar / equivalent kind of mike. Prepare a mike stand to hold it in place in front of the cab, mount the mike and hook it up to your recording system. The problem you have now: you need to set the mike the way it sounds good, but standing there, the cab itĺself blasts at you. So a really really well acoustically separating yet faithfully sounding pair of headphones make your quest a lot easier. Options are to have someone else place the mike to your instructions while you listen from another room and watch via webcam. Also, mike-placement robots are available for that job, but probably not cheap. Always record some sample tracks and listen to them without the cab blasting. compare some and try to reproduce the placement of the best sounding one!
  • 2-Mike Technique - playing with the big boys.
    • 1 decent condenser mike, like a large diaphragm Rode NT1 or Neumann if you can afford those, but a lot of others will do as well. A small diaphragm condensers might also work, but the one I tried had a little less bright high end compared to the large diaphragm mikes. Also ribbon mikes might bring in some flavor here.
    • 1 dynamic mike, like a shure SM57, SM58, also snare, tom or even kick mikes might be worth checking out. Maybe you even find small diaphragm condensers suitable for that job, but again, I felt like the one I tried lacked some of the lower mid range punch compared to the dynamic mikes.

    With the 2-mike technique, you put your main mike (in case of guitar cab IRs most probably the condenser, but might be the dynamic one in case of bass cab IRs!) to the place where it captures the tonal characteristic best. Condensers sound bright and transparent in general, thus providing bright, clear and present high end for your tone. The dynamic mike is there to provide the sound with a solid low-end and all that 'chuggy', 'punchy' mid range.
    However when 2 mikes are placed to capture the same source and mixed together, phasing will be a serious issue. With all the frequency range of a guitar cab and all the notes a guitar can play, the mikes can theoretically never be perfectly in phase for everything. And they never are. However your ears will tell you, there are some few 'sweet spot' placements that sound better than most of the others. At these settings, the mikes are said to be 'in phase' - despite science saying they really aren't.
    Whatsoever, your ears being your ears and guitar sound being a VERY subjective topic, what sounds good to you might not sound good to others, or even you a few hours later. what I want to say is: there is no perfect microphone setting, but its all about mood and feeling, just like styles of playing are. The better you know what kind of sound you want to achieve, the less time you will waste on checking the infinitely countless possibilities of placements, but feel free to start wherever you have to and take your time getting used to the abrupt changes that even small position changes can have.

    In the special case of the 'Fredman Technique', 2 SM57-like mikes are put to a speaker at different angles, hence capturing different characteristics despite being the same mike model at the same speaker.

Be creative, yet critical and you might invent your own technique for your own sound. It's a game worth playing after all!

Note: The 'tone finding' pictures have nothing to do with my recorded IRs, they are merely there to show you a variety of cabs and mike positions I came up with when recording local bands.

Miking Tip 1 - center panning (critical!): Make sure you do listen to a center-mix of both mikes (do not full-left one mike and full-right the other), as the phase interference is the biggest topic to deal with, thus, you really need to hear that and you only can hear it the way you should if you center-pan both and level them accordingly. It's also called the "mono" button on some interfaces

Miking Tip 2 - leveling: To hear the phasing effect best, the levels of both mikes should match, otherwise the signal from one mike dominates and phasing effects are under-represented. Not hearing what is actually going on is a problem, so make sure you boost the dynamic mike more than the condenser, play with the levels until you are sure to hear the phasing as clearly as possible. Also make sure you update the leveling every time you do major changes to the positioning, as you might not always notice how much the actual level of a mike changes, as the phasing can do weird things to the percieved loudness of the mixed signal.

Miking Tip 3 - phase inversion vs. feeling: One approach to find sweet spots in both, positioning and level is to invert the signal of one of the mikes and search for both, position and level where the cancellation is at its maximum, hence the least signal is heard. This ensures that the least cancellation will take place without signal inversion, hence you get the fullest tone possible.
If you don't have access to (real-time, latency-free!) signal inversion or you just want a more intuitive, less instrumental approach, you can just find a positioning and leveling, where you hear and feel the phasing sounds best for your liking. However, phasing issues can be tackled also later, as the better IR-loaders allow you to introduce sample-wise delay. If yours does not, sample-delay plugins shouldn't be much of a trouble to get, if not even included in your DAW. Although I highly encourage you to play around with phasing in the IR-loader (its lots of fun, it really is!), I would also highly recommend trying to get the phasing right in the first place.

Miking Tip 4 - a place to start if you can't find a good spot: This one following tip is from my experience, I haven't heard anyone tell this to anyone yet, now I pass it on to you: The sound from a condenser mike is mono, but otherwise pretty similar to what your ears hear. Now I warn you (!) turn down the amp until you are sure no harm will be done to your hearing, because one way to find a good spot for the condenser (or single) mike is by silencing one of your ears with a finger or ear plug and holding the other ear right to the cab. Move and turn your head across the speaker cone(s) and find the spot and direction that sounds best this way. That's probably where you want your condenser or single mike placed! You get a more immediate, quicker and clearer Idea of the sound profile around the speaker than if you're constantly distracted by the need to manage the degrees of freedom of mike stands. Especially true if miking cabs that feature a tweeter (see next one on the list).

Miking Tip 5 - tweeters - pain or gain? Some bass amps, but also experimental cabs I used for extreme metal guitar tones feature a tweeter and a woofer. now you have got 2 sound sources that overlap in spectrum at least at some mid-frequencies, which means: phasing is already an issue when using only 1 mike, and starts to become the mike positioner's nightmare come true when adding a second mike. Now you have a total of 4 phases to struggle with, but this also opens up a vast multidimensional ocean of possibilities to play with all the positions, angles, phases and tones. Most will be disgustingly awkward, but some got me effects that I could never have dreamed of getting without a tweeter cab.

Miking Tip 6 - closeness vs. distance: To capture pure essence, people tend to want to place their mikes as close to the speaker as possible, which makes them even want to remove the cloth that usually protects the cabs and put the mikes INTO the cone. Well, feel free to do that, but if you cant take that cloth off non-destructively, better think first before you do something you'll regret later. If you have such a protective barrier, better leave some cm or an inch of clearance when you place the main mike, so you have space to move the other one closer if phase interference makes you need to. You can iteratively move both closer when you got the hang of that phasing. With closeness to one speaker, you also minimize the signals picked up from other ones, hence multi-cone-phasing is thereby minimized. On the other hand, also further-away placements can have their benefits, capturing more of the room-ambience and a more distant version of the primary signal. The dynamic mike might still need to be somewhat close to a speaker to decently capture the low frequencies it's supposed to. Try and listen before making a judgement and go with what suits your ears best!

Miking Tip 7 - open back cabs: If you are dealing with cabs that hare open at the back (so you can see the back of the speaker cones), you can also mike them from behind. Don't forget to invert the polarity of the signal, at least for the beginning and think about the phasing in terms of absolute distance from the speaker membrane on either side. Feel free to play with polarity and distance to your liking as you progress.

Cab IR Recording Setup & Excitation File Download

RecordingIRs_schematic.png
20201225_041414.jpg
20201225_041337.jpg
20201225_033914.jpg
20201225_033936.jpg

OK the hard part is done now, and most of the recording setup is already in place anyways. Now it's time to change the guitar amp for a hi-fi one. If that's not feasable for you, you can at least change channels to clean, however, the flatter the frequency response of your amp, the more accurately the recorded IRs will represent the actual cab sound.

Oh, and don't forget to download the excitation waveforms provided with this step! Unless you make them yourself of course.

Hook up a device capable of playing excitation waveforms. A good cell phone does it; if you know what you are doing, you can also use the PC and audio controller you want to record the IRs with. Just make sure you are not feeding the monitor mix of the recorded responses into the excitation signal that feeds the hi-fi amp!

Now you have a setup / signal chain like in the schematic depictured for this step:

  • [playback device] --(cable)-->
  • [hi-fi amp] --(cable)-->
  • [speaker/cab] --(air mediated pressure transduction)-->
  • [microphone(s)] --(cable(s))-->
  • [audio controller] -->
  • [PC]

Recording Tip 1 - playback noise:

EDIT: there's more to be said

(a) It is good practice to have the playback device be on battery power when playing into the amp. Power line transformers and rectifiers as they are built into laptop and phone chargers tend to produce odd noises. If you are using an external audio controller with its own power supply, thats fine these are pristinely conditioned for low-noise characteristics.

(b) surrounding noise is less of an issue than I expected, that is, unless you have excessive surrounding noises or you are forced to record at really low volumes.

(c) the most stubborn noise to get rid of is amplifier hum or buzz. This hums&/buzzes make for almost all of my need to do algorithmic noise removal and occurred with instrument amps as well as with a PA-amp. To admit, those weren't the most expensive ones out there, but still very disappointing. The only 'clean' amp I found was a Hi-Fi amp meant for 'audiophile' music listening. All Amps that weren't designed to be connected to a guitar cab did sound really really dark. To achieve pleasing results, treble controls needed to be cranked, but there the 'linearity-thing' starts to raise questions.

(d) before you throw out a good amp, also check the cables and connections from the playback device to the power amp! Check a few cables to see whether there are differences and opt for the quietest, of course.

Recording Tip 2 - percieved flatness:

EDIT: Obsolete: Make sure no EQ is applied on the playback of the waveform files! this would change the overall waveform.

Why this might not be the right approach? Interfering with Recording Tip 1 - Guitar amp outputs are VERY bright signals (rich in high frequencies), thus, amps with a "flat" frequency curve sound way darker than one would expect. Probably it is the best practice to play a pre-amp signal through your "flat" amp, like those you find at the end of my excitation wav file and tweak settings on your playback device until the played back signal sounds just right through your cab.

Recording Tip 3 - levels: There are 3 places where levels do matter:

  • cab-speaker amplitude
  • mike diaphragm amplitude
  • audio input levels of both channels in your audio controller &/ mixer

While you might or might not be interested in IRs captured from cab speakers that were driven into nonlinear (saturation) zone*, you definitely by all means want to avoid clipping of the mike and audio levels! So activating any -20dB or similar switches on your mike is a good idea here, as a lot of headroom (peak-wise) on the audio controller will save you from setting everything up in vain, since clipping really destroys the fidelity of the IR! If you lack some loudness then, you can still gradually deactivate loudness barriers and bring up the level, but better do it bottom-up than top-down. Be mindful with cab loudness, especially with very old and/or custom modded/repaired speakers, as once I have blown the coil of a custom repaired 60's Vintage30 speaker during band practice, just by playing it loud. now I use the remaining other speaker of the same kind only for solo practice or recording at mild amplitudes.

As audio controllers and recording software possibly state audio levels in RMS values, better do some test recordings and check the waveform!

Recording Tip 4 - acoustic separation: Also a good practice thing is to have playback and recording devices (basicly the operator and all his interfaces) in another room than the one with the cab and mikes. This means a lot of cables have to go through a door or wall (optimally with a stage box). This might raise the risk of faulty cables problems, but lowers the risk of introducing unwanted noises, such as breathing or sliding of cloth fabric even at slight body movements.

Recording Tip 5 - multiple take alignment reference: Whether it is the lack of recording channels, lack of space for mikes in front of the speaker(s), or if you want to have your mike set in multiple polar patterns recorded from the same position, there are numerous reasons to split the IR-recording into multiple takes. If you decide on doing so, it is good practice to leave one reference mike always in place, optimally at a location that leaves you place to work on the other mikes' positions without worry. This way you can conveniently align your recordings later, only seeking the best overlap of the reference tracks. However, if you fear knocking that mike when repositioning other mikes for another take, or if you decide to not care about the phasing of mikes before the recording so you have to do manual realignment later anyway, then it could be the better choice to omit such a reference-mike. If recording channels are your bottleneck, reference miking obviously has the downside of eating up one channel each time. NOTE: switching polar patterns of a mike changes its response to air pressure profiles, thus the same mike in different polar patterns does not count as a reference mike!

*is there even a linear zone in electrical vs. air pressure amplitude from a guitar speaker? How linear is this actually? And what's the nonlinearity shaped like? Questions over questions that might require quite some literature research. Any way, here is a highly (!) interesting interview with Jensen guitar speaker manufacturer representative Ignazio Vagnone. Please watch it if you ever had questions about speakers, it's the greatest and most direct source of real information on the topic I came across in a long time.

EDIT: Download link for excitation files: https://drive.google.com/file/d/1Jj8_VYm7mcOPVO-CN...

Post Processing of Transient IRs

cropTransient.png
presCrop.png
IRlength.png

To extract those shiny IR-gems from your raw recording, you need to do some post processing:

  1. Normalization: You did leave some security headroom to avoid clipping. That might cause some of the tracks to appear rather silent. Do track-wise normalization to something around -6 or even -3dB. This should bring the silent ones up a bit.
  2. Alignment: You might have recorded your IRs in multiple takes, In this case you first need to get all the channels of all the recordings you did in one project of the audio editor of your choice. If you left a reference mike in place every time, you can align all the tracks of each take for best overlap of the reference track with that of the previous take. Another option to realign the tracks is by listening to the validation tracks. Hit the solo button of 2 tracks you want to align, and listen to a mono-mix of their validation samples. You can now drag one of the tracks (sample-wise) for the best sounding alignment, just as you did with the mike positions in step 2 - Finding Your Tone, but in the software-realm this time. It might seem easier to you to only do it this way - and it might truly be, BUT you only have discretely sampled data now, and you can't change the angle, just the offset. Just as back then, don't forget to pay attention to the leveling! You can help yourself with the manual alignment by means of signal inversion of one of the tracks and searching for the thinnest, lowest volume tone, again analogously to step 2.
  3. Trimming: Select all the tracks from shortly before a transient (leave some security space!) to about 0.5s after it. Copy all the tracks to a new place and take a look at the start of the transient. Especially look for the first sign of a pre-swing, or in other words: the first place, where you see a signal that might be part of the actual transient. The track with the earliest one is the one that matters. go to a few samples earlier than the one you spotted and discard any audio data before that sample, making sure you do the same thing to all the channels to keep the alignment. Now you take care of the post swinging section.
  4. Normalization: Did that already didn't we? Well, the other one was for the whole track, so we could better see the transient waveforms. The individual transients vary greatly in loudness, thus normalization of each transient in each track is important to get comparable results, so you don't need so vastly varying amplification when actually using the IRs.
    Note: This step of individual IR-normalization should also be applied to the readily deconvolved IRs from sine-sweep excitement.
  5. Noise Removal (worst case only): If you're forced to record your IRs at low volumes like me or you have excessive unwanted noises polluting your recordings, then you might have to take noise reduction measures. in the easiest case you can apply notch filters to get rid of individual frequencies. However, this will greatly affect you primary IR-tone. Audacity has a stunningly well working 'noise reduction' feature, that lets you teach the algorithm what your unwanted noise sounds like (select a silent region, far off any transient on your track for the learning part) and apply a certain amount of reduction of that exact kind of noise. I'm not sure what this algorithm does in detail, but I'm pretty convinced it's way more sophisticated than learning a noise spectrum and applying its inverse as a filter. This would surely destroy the primary tone, while.. I mean you hear the primary tone is being affected, but way more slightly than I'd expect. However, best practice is always to provide good signal-to-noise-ratios by playing the transients at moderate to higher volumes in the first place so you don't need dirty clean-up tricks afterwards.
  6. Export as .wav: Make sure you hit the solo-button of the track you want to export, making sure its the only solo button active, or mute all others. Hit the export button and give it a name that carries some sort of recognition value.The output file counts legit'ly as your transient IR!

Thoughts on the Transient Excitement Waveform

So welcome to the heartpiece of this instructable, the thing that bugged me until I designed this study. When I went to research how to make IRs, I was confronted with this surprisingly strange instruction: "generate 1 sample of white noise". I thought I knew white noise as being random generated sample values, probably fourier synthesized to fulfill the requirement of even frequency distribution over the whole audible range. I thought, why not pencil point one sample to +max?
The 2nd thought was already stated in the comments: this way you only either push OR pull on the cabinet speaker, so you will never experience the full range of where the speaker can (and will) move when playing the guitar into it. Thus, it is very unlikely for the resulting IR to be an accurate representation of the real sound.
The 3rd thought is elaborated on in the variety of excitation waveforms I came up with in the subsequent steps, derived from a variety of sources. Their characteristics and background are given in their respective steps' sections.
The 4th thought, however got me thinking again: my knowledge on what an IR host / convolution processing plugin actually does is so poor that again I'm unsure whether or not I can even judge a waveform on a theoretical level. Thus, unknowledgeable people like me need to smash a few things at the wall and see for themselves what sticks. Maybe even the more knowledgeable might profit from experimental data as there might still be contributing factors they hadn't seen coming. Appreciate any criticism and suggestions for future improvements!

The Delta Dirac / Kronecker Delta / Single Sample Transient

01.png

So this is the easiest made, just put one sample to max in either direction, to make the speaker either push or pull. However never both at once. Still in one respective it is also the most elegant transient form: it has a truly flat frequency spectrum.
Note: The Delta Dirac function is a theoretical construct describing an infinitely narrow, more or less Gaussian bell shaped peak with an area of 1, thus needs to be infinitely high. In discrete electronic data, the equivalent is called Kronecker Delta, which is one sample of a data set having the maximum possible value, while everything around it is at zero. Usually, these terms are applied to strictly positive scales, thus their usage here may be argued.

RESULTS: (1, 6)* This is definitely the most faithful of all transient excitement methods. The result lacks a little bit of lower mid punch and feels a bit scratchy on the high end. Highs feel an ever so slight bit boosted, compared to the original, so the result is a little bit thinner, but not way too thin. We might call it the 'poor man's instant IR'

Note: All frequency plots of the response recordings are given from the case of 2x12 Vintage 30 cab, with Behringer B2 Pro microphone in 'cardioid' polar pattern setting. Workload and data would explode if I gave all the spectra of all cabs and mikes in all settings.

*I try to reference the numbering given to the transient waveforms in the text for transparency. Note-on-note: the numbers 6,7,8,9,10 are the inverted waveforms of 1,2,3,4,5, as are 12 and 14 for 11 and 13. The reason being that 'pushing' and 'pulling' the speaker might not be trivially the same. I mean, they are the same to my ears, but I do not consider myself an audiophile hardliner. You will find only one polarity of each waveform as pictures in these steps, however of course you are provided everything in the downloads.

The (Pseudo) Step

02.png
03.png
04.png
05.png

Another well-established way of exciting systems to measure their response is not differentially, as with the Delta Dirac, but cumulatively after stepping the excitation signal from zero to one. Audio amplifiers however don't work this way, their output always creeps towards 0 when DC voltage is put in. Thus a real step would not make sense, but you can see how I tried to get close to one in the waveform pictures. Their frequency spectrum again is interestingly straight*, but inclined towards the bass region.

*excluding the first and last ones, obviously

RESULTS: the "smoothing" of the "step" has a tremendous effect on the low-end content of the IR.

(2, 7) The 1st attempt with the "diamond-shaped" waveform has no low end and is very mid-pronounced. Sounds very thin, but you might want to have this as an effect without having to apply extreme filtering.

(3, 4, 8, 9) The other 2 have very bass-pronounced tones, which makes for great IRs to use on "dark mike" tracks, the characteristics your dynamic mike is supposed to capture. Love to use them as such, depending on how much low end I really want to put on there. the clipped excitation waveform IR has audibly "boosted" high end - I mean it still sounds very dark and boomy, but still there is also a little bit on top.

(5, 10) These are a special case, with only one sample full on one side, the next full on the other and everything else silent. The fact that on neither side there is much area beneath the 'peaks' and they sum up to zero means that there can't be much bass information. To really make the mass of the speaker cone swing, there would have to be more (amplitude x time) area on either side. Although surprisingly not apparent in the spectrum, the sound does not have any bass but is just one really bright hiss. Maybe suitable for effects, or for the bright mike of a bass track.

* This makes sense, because the low DC resistance of speaker coils would drain a lot of energy, causing them and/or the amp to get excessively hot and possibly even burn through. A DC-block / high-pass can easily be achieved putting a capacitor in series.

Multiple Zero-Crossings

11.png
12.png
15.png

Small steps of imagination lead me to introduce more single sample extremes to get something of both: single sample elegance AND full push-and-pull range of the speaker in one excitement transient. As the waveform gets more complex, so does its frequency spectrum, but it seems like more is lost than gained in the excitation spectra.

RESULTS: A solo'd track with only these IRs sound quite as heavily notch-filtered as the spectra obviously are. However the change in tonal color range between the individual ones is quite respectable.
(11, 12) The excitation waveforms containing 2 zero-crossings can be regarded as a pseudo 'full pull' or 'full push', as if it was an effort to extend the range of the 'delta dirac' impulse. However the speaker picks up a lot of momentum from the 'gentle' pulls (lots of peak-area on one side) that swings on and gives the IR quite some low end. The 'disturbing transient' gives it some mids on top, with the notch in between. Very viable as a 'dark-mike' sound.
(13, 14) The one with 3 zero-crossings reminds a little bit of the 'pseudo step' shape, but with 2 additional transients in between. The results are as mid-pronounced as the 'pseudo step' with only short gentle pre and post pulling; overall these 2 sound very similar, with their usage again rather limited to mid-pronounced effects.
(15) The 'chip' as I like to call the excitation waveform with rising and falling alternating transient samples, because thats the noise it makes when listening directly through headphones. It is very focused on the high mids, very bright and lacking most of the low end information - which makes sense, as there is no larger area on either side.

The Sine Sweep & Its Deconvolution

sweeps.png
Voxengo_how2.png
sineSweeps.png

RESULTS: Best and most accurate, just as they told me, as much as I'd like to tell you otherwise. None of the transient IRs can really be compared to how the original cab sound is reproduced by the deconvolved sine sweeps. I have to check my nulling techniques, because my first try subtracting the control recording from the IR-ed pre-amp track yielded a low volume, but awful, stressed sounding noise. The second try was way better, but somehow the tempos differed slightly, so the phase kept changing all the time.

How can this be better?

What transient excitement actually does is try to excite the cabinet at all audible frequencies in a single sample (or a few ones). This comes with all the limitations discussed in the previous steps and heard in the next Step: Validation. However, there is a way to vastly improve the fidelity of the response at each frequency. By sweeping a sine wave over the full spectrum, each frequency gets excited in absence of any other frequency for a short time, so we have a temporal resolution of the freqencies' excitement. The response, obviously still temporally resoluted must then undergo a mathematical process called 'deconvolution' to yield a single short transient that can be called an impulse response and loaded into an IR convolver.

OK, now we have temporal resolution (qualitatively), but how good is that resolution and can we increase it (quantitatively)? Yes - temporal resolution is more or less the time spent on each single frequency. With the finite number of samples of digital audio and theoretically infinite number of frequencies within the audible range from 20Hz and 20kHz, we are forced to skip some of the frequencies. The shorter the sweep, the faster we go over the whole frequency range, hence, the more frequencies will be skipped. Thus, longer sweeps will capture the responses to a more detailed range of frequencies and will therefore be more accurate. And there have been studies confirming that longer sweeps result in more accurate IRs. My sweep is 10s long, but I would be interested in reading whether or not someone could hear the benefit of an even longer sweep! Feel free to experiment.

How to do the Deconvolution:

To make this additional step happen, you have to install a piece of software called Voxengo Deconvolver, available as a 32bit windows application here. They might want you to buy the full version, thus you can't batch process folders full of sweeps and you have to close and re-open the application every 3 files you process. Thus, all the functionality needed to deconvolve IRs is free! There are also rumors about linux packages with similar functionality, but IMO wine and Voxengo are the least trouble for the other OS's.

To actually deconvolute a sine sweep, make sure you have your excitation sweep at hand as well as your recorded one.

!! Be sure the recorded sample is 0.2 - 0.5 s longer than the excitation file, as the length-difference will be the length of the deconvoluted IR !!

Now fire up the deconvolver, select your excitation file (= "Test Tone") and recorded FIle (just select your single file, the 'folder' thing is just to trick you into buying the full version).

Have a look over the settings; I recommend leaving the 'dc Suffix' (do not check this), as the suffix is supposed to distinguish the deconvolved IR from the original file. The output path is equal to the input path by default, I'm not sure if you can change it, but IMO it's not too bad practice this way.

If you're sure you like the settings (and you have backed up the original, just in case), hit the 'Process' button and enjoy your deconvolved IR! Its yours!

I appended the excitation file as well as a reference recording already trimmed to a good length, so you can practice deconvolution before you do any recording.

Validation

allSpecs.png
Guitar cab IRs - various excitation waveforms (read description!)
Guitar cab IRs - real cab vs. sine swept vs. transient IR null tests (read description!)
Guitar cab IRs - transient IR polarity inversion null tests (read description!)

EDIT: The videos from the later session's recordings feature an SM57 in front of a Vintage30 speaker in the CRATE bass cab. Please refer to the video descriptions and/or explanations given in Step 12: Conclusions & Outlook before forming opinions, especially for the null-tests!

Finally, here I provide a kind of hard rock pre-amp sample track in the form of its original recording, as well as the IR-processed counterparts, using all the excitation methods discussed before. Have fun listening and comparing, and I hope it helps someone to decide on whether they should try the transients or just go with the sine sweep. I would'nt recommend doing only transient IRs, as the effort to deconvolute a sine sweep really isn't that much of a trouble and the quality of the results is just way on another level.

NOTE: A pityful situation forced me to record the IRs at low volumes, which made some post-processing inevitable, most notably algorithmic noise removal. This is a bad thing, as it has affected the overall sound of some transient IRs differently than that of others. For you to be able to see and feel the influence of this factor, I also included the whole set of validation tracks with the same IRs, but where the noise removal step had been omitted. These basically have a lot more noise, but also a little difference in the coloring of the primary tone can be heard. Most obvious at the lowest volume transients (1, 6, 5, 10, 15)

NOTE: The sine sweep IRs are in the best phase in a traditional sense. I found myself unable to phase-correct all the transient IRs, not least because some gently starting waveforms just have more delay and I didn't want to delay the others more than needed. Hence, phase must be adjusted individually for the transient IRs.

Thanks a lot for reading this far!

Any questions, suggestions for improvement or extension? I can't guarantee anything, but I'm looking forward to reading your comments!

The New IRs Recorded in July 2021

wholeRoom.jpg
austrovox_perpendicularView.jpg
austrovox_sideView.jpg
Austrovox_topView.jpg
CRATE_V30_room.jpg
CRATE_perpendicularView.jpg
CRATE_sideView.jpg
Crate_sideView2.jpg
Crate_topView1.jpg
CRATE_topView2.jpg
bassAmp_perpendicularView.jpg
bassAmp_sideView.jpg
bassAmp_topView.jpg

[under construction] So, here I present to you the IRs transient and sine-swept, of the newer recordings, along with a way better photo-documentation of their positions.

Sorry "Bad Request" error won't let me upload archives any more, hope to resolve this issue soon!

Conclusions & Outlook

Let us take a look at the research questions, as there was quite some content in between, wasnt it?

(1) How different do transiently excited IRs sound from sine swept and deconvolved ones? How close to 'reality' is each of those?

There is definitely an audible difference between the "single sample transient" and the sine swept IRs, most notably less low end in the transient one. This could be caused by algorithmic noise removal that was applied on the transiently excited response recording. The "single sample transient" is one of the quieter responses, meaning: it needs more amplification to meet the set standard amplitude. Thereby, amp-hum noises are amplified along, most notably the 50Hz of our power supply system, which is quite low-end-y. Removing this might take out valuable frequencies of the actual response, leading to distorted results.

Despite this problem, both sine swept and transiently excited IRs have been shown to produce quite quiet null tracks when subtracted from a real cab recording of the same source signal. Note when watching the video: a sample timer clock inconsistency issue must have occurred between the 'real cab' recording and the IR convoluted tracks, as the null tracks keep on phasing. However this turns out to be a very lucky event, as simgle-sample-delays are way too coarse to adjust for correct phase alignment. Now we can actually hear (a short moment of) perfectly aligned phases with almost-silent cancellation results. Here you can watch the video: Guitar cab IRs - real cab vs. sine swept vs. transient IR null tests

(2) What about practicality and effort / value of each approach?

As stated before, the creation of sine sweeps and the deconvolution of sine sweep response recordings are not much more of a trouble than the pre and post processing of the transients. Actually, with all the transient design before and noise removal after recording, I would even say sine sweeps are easier to do and they provide more fidelity, just as suggested by theory. Thus, if looking for best possible fidelity, sine sweep IRs should be the go-to option. Now, that's what we have been told, of course, but at least I provide some more data to verify it.

(3) How do transient IRs sound when excited by 'alternative' excitation waveforms? What purpose could these serve?

The enhanced low-end of the larger "step" transients and other heavily filtered sounding transients might serve a purpose, as they achieve these effects without the application of actual filters on the primary recording (remember all those people ranting how bad EQs are when your problem can be solved further upstream in the signal chain). However, as the transient design has to happen before the recording, these designs remain static after the recording and cannot be adjusted to user needs. To sum it up, for special effects and experimental reasons, I really enjoy fiddling around with those alternative waveforms, and I hope others will do as well! Watch this video to see what my imagination yielded in terms of sounds from transient IR excitation: Guitar cab IRs - various excitation waveforms

(4) Does polarity inversion (speaker push vs. pull) have an audible influence on the IR?

The short answer is I can't tell. The more accurate answer is: I can't tell with the recording setup I had access to:

Null testing is actually a more complicated topic than I excpected, as a waaaayyy sub-sample delay is imperative for proper phase alignment. I do have an idea how this could be technically possible, but I didn't manage to dive into that so far. Maybe there is an easy solution I just dont know of. Please let me know if you do! So, fact #1 why my research cannot answer this question is: my tracks are never in phase, thus cannot cancel out properly.

Fact #2 is once more be the tiresome subject of the need for algorithmic noise removal. I could rule this out by systematically applying the same amount of it to the polarity inverted versions of each waveform. But this is much easier said than done, and even if I did, fact #1 would need to be solved too to access actual results. I might find time and motivation to do that at some point in the future.

Nevertheless, the video presenting polarity inversion null tests: Guitar cab IRs - transient IR polarity inversion null tests

Future Outlook:

I have been really surprised nobody seems to have done research like this before, or maybe companies have, but keep it their well-protected secret. I have also been annoyed for people repeating that 'sine sweeps are better than transients', even though the theory clearly says so, but if you want to know how they get to this conclusion, actual data is really rare. So out of my own interest, and to provide the world with at least half-decent information and data on the topic, I did it myself.

I do not expect alternative-transients-excited IRs to become any sort of 'thing', as they are probably too static, too noisy and of too little use for anybody who takes themselves seriously. In science, you have to run against a wall to see it is there. Still, I hope people care about the new concepts, have their own opinions and share some comments and thoughts.

As stated before, I can see larger 'step' waveforms serving purposes for dark mike and bass cab tones, as they provide their own elegant form of bass enhancement.

I'm not too confident about the usefulness of the 'band filtered' sounding transient IRs. they are probably too static as compared to actual filtering that can be tweaked to.

So, thanks for reading, I hope I can help somebody and I appreciate any comments you might have!
See you in the future, bye!

Appendix: Index of Downloadables

So you easily keep track of what to download where:

  • transientIRs.tar.gz - Step 1: Ready-To-Go IRs
  • sineSweepIRs.tar.gz - Step 1: Ready-To-Go IRs
  • excitationWAVs.tar.gz - Step 3: Cab IR Recording Setup & Excitation File Download

  • cabSineSweep.tar.gz - Step 8: The Sine Sweep & Its Deconvolution

  • IRvalidation.tar.gz - Step 9: Validation

  • IRval_noNoiseReduction.tar.gz - Step 9: Validation

I hope .tar.gz is an OK format for everybody, as this site won't let me upload .zip files. suggest me any other supported format and I will do my best to swap or add that.

...and special thanks for the Instructables editor for messing up the distancing of my list without any hope of fixing it.

Edit Log

[under construction] I feel like I am creating a monstrous wall of Text with this 'ible so here's how to find the additions (EDITions?)

Step 1, Introduction & IR download: Added research questions formulation, videos & editing anouncement,download links, updated microphone setup including the "new ones"

Step 3, Recording Tips 1 & 2: Added Insights on playback noise topics, amp choice and settings, added download links for excitation waveform files

Step 10: Validation: videos have been added to this step.

Step 11: Added / Rearranged so new IRs from a new session are included, along with a useful photo documentary of mike positions and room treatment

Step 12: Added Conclusions & Outlook section