# Leslie Horn emulation

There are many factors to consider when emulating the rotating horn. The simple things like and 800 hz HPf is basic. The doppler can be done by delaying the horn and modulating the delay. The pan is a basic procedure too. The harder more complicated items are the flare and frequency response of the horn. Unfortunately the most difficult emulation of all is the standing waves produced inside of the cabinet. This seems a near impossible thing to do unless you break the rotation down to fewer degrees of resonant circuits. No you can't take a resonant filter and change it resonant frequency as the horn spins. Because the resonant tones in the cabinet have to die down on their own. So as you can see it takes a lot of simple delays with feedback to pull that off. The bass part of a leslie cabinet with the rotating hole is another chore but like the horn. In fact, many players don't even like the bass section as it makes it difficult to play a smooth base line. So, just food for thought there. The drawbars are the easiest of all,

Tagged:
«1

the standing waves of the cabinet should be achievable with convolution by an impulse response, no? and if not, its usually pretty easy to do with a parametric EQ

An impulse response is needed for every phase angle of the horn as it rotates. I guess it could be done with several convolution verbs. Probably the more the merrier. Don't think parametrics will work, it would sound like a wa wa pedal. lol

if you had one you could measure the impulse response at every phase angle... but as far as recreating it you wouldn't want to use FIR convolution, as those are not filters you can really 'modulate'

you could however try to approximate the behavior by using something closer to parametric or a filterbank, if you you had an understanding of how the impulse response/resonant qualities processed through phase angle and how to map that to modulation behaviors of the filters

The solution is actually pretty easy but it would take a lot of processing power. Basically what we hear is a comb filter with resonance for each phase angle. Sine the horn has a dispersion pattern it means more comb filters as it sits still. When it moves the input to the comb filters needs be like a swept input to a string of comb filters each with its own pan setting. I guess you pay a premium for a nice colorful sound. Probably why leslie simulators don't sound like the real thing. It might be done with a non real time plugin with a single comb filter with many passes. A comb filter is nothing but a delay with feedback. The feedback is just the reflections inside the cab. So ya, that could be done with a single computer. I guess to do it in reaktor would require the entire audio clip be loaded as a file and it would have to be play the file over and over and sum the outputs of the comb filter, something along those lines. Not sure if reaktor can do that on its own but maybe it can with the song position pointer and the midi start and stop. Seems a software package could pull it off a lot faster though; not being restricted to some fixed sample rate.

im having trouble visualizing the setup you're describing but intrigued. i can definitely see how there'd be comb resonances at play in an enclosed cabinet like that. there is a core library comb filter that's pretty good, maybe it'd be worth the trouble of throwing something together to see what you get?

Yeh, I tried before, used 18 of them, each one was 20 degrees apart. Think of it as a circle like the horn spinning. I basically crossfaded one into the next in succession and started over at 360 degrees. I only crossfaded the input to them, not the output. The outputs were panned and summed. It was working somewhat but it sounded choppy and I was out of cpu. At the time I didn't think to drop the sample rate though on the comb filters. A comb filter basically increases tones an octave apart and turns of the tones in the middle. In the cabinet it's the reflections that produce standing waves if the horn is stable. As the horn moves the reflections die down. That's the damping which in reaktor is nothing but the amount of feedback in the delay.

have you tried just convolving a leslie simulation with a cab impulse response? i thought about this for a bit more and realized that i think that the modulated delay of most leslie simulations is probably a pretty good physical analogue for what is actually happening, and that the cabinet is almost certainly able to be accounted for with convolution

my reason for saying this is that most leslie simulations that i've seen have no measures taken for the cab simulation part, and usually cab simulation is a part of a linear time invariant system

I totally get what you are saying. Yes, I've tried it with a convolution impulse of a leslie horn cabinet. The overall sound is way better because the horn is continuously changing its volume and filtering as it rotates. So yah, the sound of the cab does follow the amplitude and filter changes being fed into it. It's the best it can sound and in reality isn't too bad. In fact it's actually a pretty decent sound for a back fill instrument. It's got a better bass sound than the leslie versions with that rotating bottom end. So you can at least play a bass line. And what a powerful knock your socks off bass it can be. So all in all, for rare uses it's pretty fun. ... So getting back to the cabinet drawback, a real leslie changes its intenal reflections for a full 360 degrees. The one thing a single convolution reverb can't do. At least the ones I've seen that load in a fixed impulse response. So you see the problem. It would take a lot of separate cab simulators to follow that horn. You could set up a reverb and modulate the size with the horn cab but that in turn causes a doppler effect. It would be pretty wild with too much modulation but may sound pretty cool too if set up properly.

edited July 2022

ok, not that i didn't understand before but now after that explanation i fully see what you're talking about. i had to picture it in my head. like.....the impulse response of your room will be different if you start moving the furniture around. haha, doyeeeeeee

my first thought after that was maybe the most direct way would be a physical model, some kind of 2d waveguide or resonant chamber? modelling such a complex shape would probably be conceptually very challenging... i guess @Chet Singer singer would know about that if anyone does

the one source i found said a newer approach is 'time variable FIR filter', which made me think: an FFT could probably do it, so like if you convoluted with an impulse response that was a single frame (so, no reverb effect) and you could actually sample the impulse response a bunch of times then if you could modulate through them fast enough, that might actually work pretty good (??)

you might have to sacrifice with a fairly low frequency resolution or frame size so that it could be fast enough.. whats the high setting on a leslie , like 15, 20 hz? 🤔

Yah, now you see the problem. In the cab the reflections are different for every phase angle. You have to keep in mind that those reflections need to die down after the horn moves. So in order to do it with a high detail you need a string of individual cabinet responses. The dispersion of the horn might be 30 degrees so if you had 360 separate cab sims then 30 of them would be working at any given time. So as the horn rotates it moves into the next and drops the last. Yet the last is still reverberating until it dies down. So basically you need a lot of delays and filters running all of the time. It's next to impossible without a super computer. If you skimp down and place one every 10 degrees then you have 36 running all of the time. So it starts to get choppy sounding. The horn can spin any speed you want. I set this one up to work with a fixed tempo or the midi tempo. I usually keep it fixed on a tempo of 79 and the slow speed is like 1 revolution per 4 beats and the fast is like 19. If I step on the sustain pedal it speeds up and slows down on the release. It's not a linear speed change though, It starts out with more energy and as it approaches the end speed it gradually settles in. At least I got that right, I don't know, maybe someday reaktor will use all cores but for now it's just too much. It's not too bad as it is though, the cab res gets covered up in a mix if I use it. Just can't get that real sound like the organ in Whiter Shade of Pale. That was a masterpiece of a recording. Talk later.

edited July 2022

ok, ok ok... i was thinking convolving by a single frame might be an acceptable hack, since the reflections would have a fairly short decay inside of the cabinet, but you're right: there would be a decay which made me realize there is one reaktor module that would be your best bet for pulling this off if its possible (and i don't know why i didn't think of it earlier):

have you considered the modal bank? you may know about the sine bank, how it can be used instead of an inverse FFT to resynthesize FFT output, well.. the modal bank can be controlled in the exact same way but instead controls a bank of filters with adjustable decay time

you'd still need to have enough 'impulse responses' for each phase angle such that it wouldn't be choppy like you say, (so probably a frame size or filter bank size of 128 or so), but if that sufficed you'd actually have a pretty low CPU solution. also whether this would work would depend on whether or not the modal bank is constructed so that a decaying filter would continue ringing even after its settings had been updated (i would think so, for musical reasons??) [but then again maybe not now i'm thinking about it, you might only get ringing out until the filter setting are updated or else the module probably wouldn't be as CPU friendly as it is]

i agree that some sounds are just better as the real thing. i might have to DIY a real leslie someday, im definitely going to make a plate reverb at some point!!!

I kinda got a better sound by putting three cabinet delays in series. It's like having three reflections at one time, each one has some feedback that varies with the phase angle then they sum together. It sort of diffuses the single standing wave that wave making a wow sound. I know it's hard to picture but it works like this. The cabinet is a rectangle and as the horn spins the reflections are longer in the corners than they are when the horn faces the sides. Since it's only one delay with feedback it sounded like a wah wah pedal. The reflections from the feedback have a doppler sound because the delay time is changing. It sounded pretty funky with one but by putting three in series it spread the wow sound out into three of them.So when the horn rotates it's like having a trail of reflections. It sorta drowns out the dominant wow tone with all the other reflections blending in. It's the kind of thing where you pick up on the wow sound without the other reflection, but with all of them going and each one is producing a different reflection behind the first one and they are all three resonating it blends the wow frequency with other tones and it's not so noticeable. Most definately sounds like a reverberant cabinet now. Pretty simple to do and so far no complaints. A lot simpler than using banks but in reality it kinda is a bank of delays. Sounds much better, it's like the cabinet reflections are full of doppler just like the straight swinging horn so you think nothing of it, like it's normal or something. I'll try and upload a sound file in a few days, maybe something from elp. later