Sean Costello (Valhalla DSP) on reverb design, March 2019

play slideshow there we go all right now I know what I'm talking about here all right so just quick start what is reverberation I don't know if any of you need to have that answered but it's like I would say like like Wikipedia talks about is like the persistence of sound after a sound is produced which i think is good because it's like the idea is like it doesn't say it's echoes it doesn't say it's like the whole bunch of reflections in a room is just based on what we hear which is that it you don't say why it's persisting but it persists and I also like this reverb developer from the 70s and 80s Christopher Moore came with this phrase spectral plasma and I really like that it's a good way of kind of showing that like the the reverberation has all the sounds that you put into it but it just sort of jumbles them up it takes out any sort of temporal coherence and just gives you sort of a smear of sound and that's like I get into that like I think that like reverbs and echoes extend a sound sign but they're perceived differently for example I could just show you you know I'm sure you all understand but like this is an echo like [Music] you hear discrete repeats in there whereas a reverb it's gonna be much more like there's no real discrete read echoes and there can be and reverb but that's not what most people are used to with that all right and I don't remember how it just got to that slideshow sorry about this and play play thank you so much so a reverb can consist of thousands of echoes like spaced closely enough together so you don't really hear any gaps in the signal but also can just be randomizing the sine phases of the individual components in the signal like a really good smeary reverb does take a giant FFT of a signal and then randomize the phases it sounds really reverberate so even though there's no echoes in it at all so one of the terms I use are T 60 it's the what use standard measurement of time time takes for the signal decay to one one thousandth of the original strength and they often do it in different frequency bands so it's like but like usually when they measure like the are T 60 is like in the mid-range or something like that so early reverbs one of the things that you had in like churches and cathedrals as probably one of the first structures where the reverb was not only sort of designed in but was a major function like major kind of like factor in what was done in those spaces because they had massive reverb times six to nine seconds common sometimes exceed 12 seconds just to give you an idea of what that sounds like let's just bring this up and let's put it in like this a six or let's put a nine second reverb with a short decay [Music] that lasts a long time and some of the huge cathedrals it really gets longer than that so and if you think about it priests like couldn't really talk in a space like that because if you heard someone talking one it's a giant space it's hard to be heard and everything smears together so like all the you know all the speech just turns into this big wash so the idea of Gregorian chant it might have been a way to communicate ideas like in the area where you really can't hear the individual phonemes or anything of speech and the other thing is that fast musical passages like if you're doing something it's just a blur so if you're doing like a like you can hear that but it's like I saw van Halen at the Kingdome in 1988 and he was doing his low eruption solo and just sound like ambient music it was really beautiful in that way so but like that's the idea it's like monophonic phrases can turn into polyphony just because they hang out in that space so music and the cathedral needed to be slow and really deliberate and kind of design about the fast thing you know just just be used for like your last several notes are gonna be hanging around there for quite some time chamber music is like in big rooms but not as big as concert halls and I don't know the names for those sort of rooms because I don't really know classical music that well and it's like it's a shorter reverb range like 1 to 1.4 seconds and it kind of increases the spatial perception increases sort of the richness of it but it doesn't blur things together so something like that would be let's turn the attack down so it's like [Music] so you know you could you can see where you can actually have more complicated passages in that and you'd be able to hear like all the notes concert halls were designed to like have a lot more people in them and then also a lot bigger orchestras and so symphonic works sort of evolved with that like they have a longer reverb time and it turns out that when people talk about the best concert halls like they talk about Boston Symphony Hall or they talk about the one in Vienna they usually have an art mid-range art e60 between 1.9 and 2.0 seconds and the longer but there's lower frequencies you have a much longer reverb time than that and that gives you a higher perception of spaciousness and development which I don't think we're going to be able to hear in these speakers tonight but you can just get a little bit of the idea of what a reverb like that sounds like and in this case it's already kind of set up to do something like that [Music] which sounds short compared to cathedral but it's still a fairly long reverb for what you what you'd have like you know for like orchestral music so this is where I get into more of my strengths I'm not a classical guy but I know more about artificial reverberation and so one of the first artificial reverbs came from the ideas I mean this actually goes back a couple of thousand years in some ways they had echo chambers in some Greek theaters or like they had kind of like jars that would resonate this like but in the world electricity is a simple idea he sent to take part your signal you send it to a speaker that speakers in a room it's a dedicated room for that you put a microphone somewhere else you recorded it's coming out of the microphone you mix some of that the original signal and then there's ways of like you know you probably want the speaker aimed away from your microphone like the microphone might be aimed there speaker aimed there you shut the doors it's really nice if you have like if they worked well in bathrooms like the first ones like there was a who was at Bill Putnam senior just recorded something for the harmonic hats and he used a bathroom echo chamber and then concrete stairwells were really popular for this stuff like Columbia Records they had a seven story stairwell speaker on the top then a microphone on each level which and that's like a lot of amazing records were recorded with that reverb now and like the same thing like at Abbey Road like you had like a storage room that was repurposed to be a reverb chamber the problem is it's like you know with those repurposed rooms there's isolation if it's a bathroom you can't have anyone using the bathroom or running water or anything like that that was something you know so you had to kind of like take that room out of commission the least during the recording so they end up of designing reverb chambers and they designed for optimal sound like they tend to have non parallel walls kind of like a wedge like this where you'd have like the short part of the wedge let's see if I have a picture you have the speakers and then you have the microphone aimed away from that and like you wanted hard reflective surfaces concrete is good they often do like multiple layers of drywall covered with plaster and the characteristics at really high early echo density echo density is like some people talk about the grain of a reverb and how I tend to perceive that is kind of like what the initial attack sounds like do you hear any sort of like is it just a little of sound or is it kind of like that before it builds up and in a chamber it's like any other physical room the echo density bills with the square of time so it like if you have a certain density just give it a little bit and it's going to build up and also builds by the square of frequency again that's true of any like 3d physical space and then the RT 60 like the reverb time above two killers it starts to fall really quick because of the air absorbs e of air so it's like a 10 kilohertz you can't have a reverb time higher than 1.25 seconds or so and you can have really long RT 60s at low frequencies and that's not really that big a feature like you know if I do is example here like let's turn off all the high frequency damping it's like you can hear like I could let's put it in like a long sound [Music] like that's something that's a very digital sound we used to breathe like dit nowadays we used to digital reverb sounding like that in the olden days though it really you had so much high frequency absorbency that you would never have liked like a long reverb would always sound darker like that and that makes a big difference when you have speech going through it or like singing because it's like you know every time has like sip this is sill bit sibilants elephants in sibilance thank you those sounds in a digital reverb we'll just like like can be like go on for a long time it's a really 80s sound and that's part of its like that was part of the sound the 80s like these reverbs acting in ways that physical rooms just can't act whereas in reality it's like any physical room is almost like a natural de-esser the bigger it is the less you're going to hear those like any sort of consonants ringing back and if they do ring back it's only going to be for about a second to a second and a half at the most so that makes it kind of like you know like the chambers are really good for like more like you know acoustic music singing kind of like it's stuff where you want to have this like I don't know they talk about warmth and realism and all that if that's something you're looking for drawbacks of it it's a lot of room to have an echo chamber and like you know it's like one those things it's not just like building you it's not just that you have to have a speaker in there and a nice microphone you have to have the space you can't buy one you have to have a design but you have to have the actual physical space in it it has to be built it's not the sort of thing you want to build in a rental it's it has to be like this skills you have to have people that know how to apply plaster well and all that and you can't really adjust the reverb size unless you have like some moveable wall in it which a power from you like a ceiling that lowers up and down and there is a reverb chamber in Blackbird studios in Nashville that does that so that I mean it's a beautiful sound but it's it's expensive and you can't you can't adjust it much so in the 50s emt electro mess Technic released the emt 140 plate reverb Raider and that was really a big innovation all it is is like it's a sheet of thin rolled metal and it's about like maybe this big across by about this high and they're in a plywood box yeah that's the size of a twin mattress it weighs like 400 pounds and it's got a driver on the like it's suspended in the steel frame it's got a driver to put the sound in one place and a couple of like some sort of pickups that kind of pick up on the bending waves and it sounds really beautiful there's a picture of one but they don't look like that when you see them in real life when you see in real life they look like a plywood box they're very not photogenic and then they have like you can't see it here but they have a damping like plate that comes the closer it is the more it kind of cuts the reverb time but never actually touches it it's do the air and it's like it's probably asbestos or something like that I mean it's it's like a lobbyist things like you just horribly carcinogenic materials we used in the making of these things so one of the things that you hear into spirt like in a plate reverb is what's called dispersion and you hear this kind of like whip crack or ray-guns sound it's hard to hear that in like actual reverb it's always easier to synthesize it it's like it's a kind of classic [Music] like that sort of resonance sweep is what you'd hear like if you do an impulse response as a plate reverb you'll hear that in there is a little Raygun sound that's I don't need okay then and what's cool is that so that's something is like what that is caused by is that higher frequencies get less delay than lower frequencies so you get this sort of natural curve if you look at like the frequencies it's like this kind of like the lower frequency if I vlog don't group delay and it goes down to very short for the high frequencies and what's nice is because you've got the left and right pickups at different distances they have different delays at different frequencies and so the high frequencies for left and right in the plate might be a millisecond apart low frequencies it's like 15 or more milliseconds apart and what that does it gives you this really complicated Haas effect because it's like you know what the haast effect is basically if you have a delay between left and right channels whatever channel has a shorter delay you're gonna hear it coming from that channel so it does that but it's really complicated and then it's like so it's going to be like over different frequencies and then so like the echoes build linearly with time because it's two-dimensional but also builds literally the frequency so it's like it's something that interesting enough that's something that digital reverb czar good at too but they're weird echoes because every one of those echoes it doesn't delay all the frequencies so you get a chirp and then that runs for an echo which trips that chirp and then that chirp is trip so what that means is that like if you've got like a 1 second millisecond delay in the high frequency it's in 15 milliseconds in the low run that through like reflections a few times and it's like the by the time you hear it echoes like in a low frequency the high frequencies have to hit like just bounce back and forth dozens of times and what we end up hearing that as it's just like instant diffuse reverb it's not like chatter chatter chatter it's just I wrote this when I was like working on Valhalla plates so I was really thinking a lot about plates if people wanted me to skip through this it's not but yeah I go into like again the arty 60s so it's a lot like like the high frequencies decay a lot quicker than the low frequencies just naturally essentially you're like for a plate reverb it's like you have kind of like fixed high frequency decay and you're just turning up the low frequency decay on it that's the only way you're really adjusting yeah oh yeah did that really thank you for asking so basically it's like you know reverb time to get to decay to 1 minus 60 DB of the original signal it's a very convenient measure and one that they have like digital stuff for so and but they often talk about just in the mid-range or something like that you usually just kind of pick a mid-range frequency and go with that like around a thousand Hertz is where they commonly measure it because that's often we're like that's and then it's like rolling off below that they'll talk about like you know but they might even say it like they often just happen in several frequency bands and that's actually a way that's like you know some reverb Raiders have decay that you can adjust in several frequency bands and they could see much it does give you a more natural sound like that same thing where it's like the high freq you know like like continents decay away really quickly and so that's a sound like I mean like for like a kind of organic reverb sound like something that sounds like reality that's important and that there were some like they use stainless steel plates in the late 70s and they actually had like a longer high frequency decay alright now we get into the stuff I like digital reverb and this is I called the mainframe early years because this stuff like Manfred Schroeder he published a couple of papers in 1961 natural sounding artificial reverberation colorless artificial reverberation and they show it had to do it to digitize signals and so you can see here he's doing like those things on the left our comb filters on the right that's a weird all pass filter and we still use these now I use these every day when I'm making Lieber it's like comb filters and all pass filters and what they call delay lines which is just that little tau signal means delaying it that's used to this day now we need shorter publish these papers in 1961 he said that the quality was indistinguishable from the natural reverberation of real rooms which isn't true like I remember hearing this in the late 90s heard this counting composition from the early 70s I think it was Terrenas and there is a sound in it that I was just bees it was just a sound of bees and I realized that's the reverb that was like he sent stuff to the reverb and the reverb was just sounds like bees when you make something like the Schroeder algorithms just they do not sound they sound really metallic but part of that is like you got to cut the guy some slack it's like literally he would have to like it would take hours of time on a mainframe computer that you're at I'm cheating with other people to render a minute of sound and then they put that on magnetic tape and they drive 30 miles like I think like Bell Labs was in Princeton New Jersey or something and then they had to drive 30 miles to where there was actual a d/a-converter in the 60s so right now if I have to wait two minutes for my thing to compile I get kind of angry it's like oh my gosh I'd like you know to like be able to listen to it so it's like like versus like hours plus 30 miles drive and then presumably 30 miles back so it's like it's hard to tune stuff that way so I like in the 70s they started making digital like dedicated digital reverb hardware and stuff from out of the gate sounded way better and the people that developed that stuff very blessed her and David Rees enjoy the really big guys did very blessed her designed a reverb for EMT David grease injury got his stuff basically he was acquired by lexicon they implemented the Schroder stuff they were really excited to find these papers on this and then they just like this doesn't sound that good but the reason why they were able to figure out but they improved on it because they designed the hardware realizes stuff they designed programmable Hardware they're over here it in real time and then they got on whatever their vt1 hundreds or whatever they were using at the time you know like and put in the custom assembly language like you know or machine language or whatever they had for those boxes and we're able to you know it took a while to program but they were able to hear stuff and get the quality up a lot quicker so that's the emt 250 I talked about it's about this high they call it r2d2 it cost twenty thousand dollars in 1976 I think the first house I lived in would have cost about that back then as a kid I didn't own the house in 1976 here's lexicon 224 it's a lot smaller only about this wide by this tall sounds great to this day cheaper is only $8,500 in 1979 now one of the things about these boxes is that they have really low sampling rates like eight like maybe like a 20 killer sampling rate and they have these like filters to get things out so the bandwidth slike around 8 kilohertz to 10 kilohertz it's really dark you don't hear the high frequencies but it's kind of a ways like it turns out as I was talking about with plates and chambers you don't really hear those high frequencies hanging out very long and in natural reverb anyway so just not having that sound there is kind of a way of getting around that because otherwise you just hear this high-energy his sort of stuff so one of the things this is one thing that makes me happy because I love the sounds like they didn't have much delay memory line available and delay memory lines like the total amount of memory allocated to like the delays in the reverb and like you know because the idea is that you have multiple delays running a parallel or in series and that's how you get the complexity AMT 250 only had 300 milliseconds worth and I was like 1k chips and there's like 80 of them like still out to 24 is one second I don't know it sort of chips it is it but that's you're going to hear patterns in that so I talked wow this is a really tricky description of delay line modulation but it's like it's kind of like the idea is like if you think about it like you hear almost these resonances like metallic sounds tend to be coming from you hear things rigging and they're ringing close enough to each other that it kind of bugs you but not so close that they sound like the same thing you'll hear a beating between them like boom boom boom and it's like it's unpleasant and so modulation it's just kind of like moving those around over time and imagine you're taking pictures of it so it's sort of blurs a space in between them over time that's the idea it can sound great or it can sound really like seasick and I talked about linear interpolation in delay lines that's really TWiki but I can give you an example of that so let's take like a short get something kind of like the original lexicon 224 algorithm let's get a sustained sawtooth in there because that really single one tends to really bring up I need to get the decay on that am i sending to the right send let's see if that's actually working oh I turned it off that's why so let's say like we're gonna have like a fairly long you know middle range long and decay and let's turn up those high frequencies and let's put in the 80s mode or 70s mode so if you hear it like that might be really hard to hear that's good I don't know if you can hear it here but there's like definite sort of patterns in the sound but if you start moving the length of the delays okay I'm just there you go yeah so here we go like here it is without any modulation and let's do one that actually has really short decay legs you can hear the metal that way better [Applause] here that kind of that pattern it sort of brings out but you turn up some modulation and you don't hear the pattern but you do hear this sort of like lush sound and then it's that you know [Applause] it's a really nice sound that same thing turn off the modulation [Music] so that's what the modulation is doing it isn't designed it wasn't intended originally to I couldn't even show you an algorithm based on the original EMT 250 let's shrink that size very ringy [Applause] very lush but getting kind of seasick it's like so you have to sort of dial in the right amount of modulation on these old boxes and on things that emulated [Applause] [Music] but its way of cheating that's like this is using about 300 400 milliseconds of delay as I and you can hear the ringing in there but like you turn on the modulation that goes away and so I think I talked about this next we'll get some noise coming out you wouldn't be able to hear it over these speakers but just trust me if you do it's subtle but it's like but it's one of those subtle things I think people like so 80s digital reverb czar everywhere everywhere turned up gated reverb they were stopped trying to do anything natural with it it's just totally unnatural sounds and you put them on everything and I don't it's not really my thing but but if people like it then I'm happy to provide the tools for them so like this is like lexicon 224 they had rich chamber rich plate which were new algorithms it didn't have rich hall if you know that joke you're a certain age for eight lexicon for DL it's kind of like the 224 but it does some randomization in there and so those don't have the coursing I can actually play an example of that like so let's go to like something that's more like the something more like a 224 type outer then I'll put it in the 80s mode get the GUI all 80s out and then you can hear that [Music] let's make it a little bit not absurd you got that lush coursing for some reason in the mid-80s David Greeson sure decided that chorusing was bad so he did a different sort of modulation which changes things over time but it doesn't really give you the pitch change like I'll turn that off you can hear it that doesn't sound that different but to some people that's the sound they really love to hear I personally prefer the the pitch modulation and it still had 18 bit integer processing those two bits make a big difference but not that big a difference the converters made a big difference Quantic it's a cool picture I've never seen one of these in my life I've never encountered one of these in person people rave about them as like being really realistic they use a lot more delay memory than the than the lexicons and the emt but again don't know where you find one rms 16 they did their own algorithms but one of the things about like the old old reverb hardware is that nothing in that was copy-protected that code you could just sit there and read a ROM and then spend next year trying to figure out what it did and so AMS they probably like people say that like I heard from some anecdote someone like delivered a lexicon to the AMS folks and then pretty soon they started getting their stuff together so they probably sort of figured out what lexicon was doing I I hear it doesn't sound the same but it's like it's the basic idea they have a bunch of all passes in a dis series with a feedback around it nonlinear reverb it's that was used they based it on this phil collins drum sound and it was huge people loved that thing so this is where it gets interesting it's like there's a lot of 83 verbs they just not trying to do anything realistic at all like the nonlinear reverbs that's like it's reverb and then they just cut off the decay reverse reverbs it's like it's like what you get like they used to do these in the 60s by like recording recording something reversing the tape running that through a reverb recording that result and then flipping the tape again so the reverb comes before the sound so there's ways of doing that works like you get a reverb that builds up over time super modulated reverbs and then there's shimmer which is like it's a reverb in a feedback loop with a pitch shifter and that let's see I can that's easy to show I have a product that he does that so without the feedback it's just a really long reverb [Music] very modulated you can hear the kind of seasickness and then you turn up the feedback you can control the pitch shifting [Music] and it sounds like a planetarium a planetarium it's kind of instant planetarium like those techniques so 1990s to today like there's a lot more you can do I didn't really talk about convolution here convolution is really awesome it's like I don't love the sound of reverb done with it but you can do any like all sorts of really cool things with convolution and like it's free and most DAWs now so like but for the most part like you know because computers get more powerful they're trying to go for more realistic sound like more early reflections lots of memory for the late tail like TC Electronics and brickowski there's often trying to go for a sound that doesn't sound like there's much reverb in there so they're putting a lot more work into having something that's so good you barely hear it and it kind of glues things together so this is where I got really I get really tweaking and just talked about how these things were done I'm not sure write out those symbols it might be rude off caught Rudolf Koch the book of signs that's a really neat little book so this is somebody's like you looks like Randy you like you came from most like mostly academic or is it I don't know how you call is like academic but cycling 74 like was it like that's like yeah property but they're pretty open about this right yeah but but it seems but they're very like they they there's a lot of like whoa stuff I'm like there's heavy links between like you know max/msp and the academic world and stuff like that so like one of the things about like you know there's academic public speakers but like physical reverbs there's never a mystery it's like how does a reverb chamber work it's like go into the room just check it out you know how's the cathedral work it's like well spend a hundred years and several generations of craftsmen and like most of the world I don't know how they did but you know how it's done it's like so first software reverbs were published in journals but ever since 1975 they've kept a lot of the stuff sort of secret and software inside like it's not easy to reverse engineer and you know you can reverse engines not easy and if they patented it it's like well you're just telling people what to do and then you just change a few things at and you've got your own variation on it or you can do that if in like something that's patented and who's going to prove that you did that so it was like there's not much incentive to patent these sort of things so there's this point where and this is a little bit less true than I was when I wrote this in 2015 it's like there's reverb development all sorts of academic places at karma MIT Media labs are calm and they often do a lot of work around feedback delay networks convolution and physical malls this is a feedback tool a network where it shows a bunch of delays at Z to the minus M going into a matrix and they're great but commercial reverbs sometimes use that a lot of times they don't but but like like F DNS have been used since they were published by stouter and Puckett in 1982 but there's a lot of good reverb Aldrin's that happen before that and so this is like it's kind of like archaeology figuring you know I also I was an archeology anthropology major so I'm just proud gonna say it's like archaeology just like I other people might say it's like gardening or cooking or whatever but that's that's the metaphor that I live with so you can kind of like figure out what's going on like archaeology fear outs going on but the people to do it or either dead like archeology or they've signed NDA so they don't want to talk about or they're still making money or about so you can guess but you might get it wrong but sometimes if you guess wrong and it sounds good whatever a lot of them were like these were like these are the people that Chris Moore Anthony and jello Wolfgang Boop Lightner and then what's that David Rees injure they had a lot of those very long beards going on I know what bull when I first started like doing reverbs it's like wow they all were bearded I was clean-shaven and everyone I knew was clean-shaven and and now yeah it just it just looks better with the beers I just I'm just used to it so I just like I've read everything like you know there's a lot of papers on the subject this is just talking about like this was someone gave in a class so it's like you know four people actually interested in the reverb design so I'm like what I'm things I like I like to trigger figure out it's like not just how something worked but why someone was doing that like it's important to me that David Greece and Earth came from a classical background he was always trying to make something that sounded like Boston Symphony Hall so if you looked at like oh wow there's these taps from the network here do those map to some reflections of Boston Symphony Hall in one seat or something like that john de toro was he published a couple of papers in the journal in the AAS and the first to jerk like he was the three part article the first two came out in 1997 and the third one came out ten years later because rumor has it that lexicon basically got really mad that he had published these papers because he had worked at lexicon and then left and then he well and then he also and then he worked at a place where they reverse engineered a lexicon reverb and then he left that place and moved from Penn State that was in Pennsylvania moved to California and they sued him and he was like is this whole thing about like intellectual property is again stuff for hire and it come turns out that in Pennsylvania if you do work for hire you can still talk about it like it's not like out here it's like if you decide for tie it higher I don't think you can speak about it all if they don't want you to but so he talks about this like this is he this is on some obscure corner of the Internet that's the 2-24 concert hall it's really weird and that's a really weird drawing of it it's this is what you see in DSP networks like so like if I either just a point like um it's at the microphone my still plugged in stay unplugged from the okay like it's hard to see but like that right there is a represents like a delay line of 239 samples and then it shows that you're adding them you're multiplying them and like the order in which these happen kind of shows you like the feedback network like this is a one pole filter this is a weird one pole shelving filter and then you see like these things like you know if you've learned enough of that you can say like oh these are nested all pass filters the crosses it's like well like there's improtant in the left everything for this Oh perfect thank you so much Wow like I think it's hard to see here and I'll post this so you can see it but it's like but I can also tell you where the link to it is like input comes in a certain place it's summed in like it goes through like it's filtered it goes through a couple of short all pass filters it goes through this longer all pass filter which is nested there's an all pass delay inside of an all pass lis I say longer because means this you can all pass delays these are all pass delays this right here is three all paths delays nested in there the shortest one that this thing indicates you modulate it then that this goes into this network over here and does a similar thing these are actually identical just different lengths but it's not drawn that way and then this here you can't read it all but this shows you it's like okay like the outputs for this network like you take a output out like from here somewhere in that delay and then somewhere in this delay and then somewhere in this delay and then same here like bam bam and bam and then there's also like that's like for like this is like left comes in here you take three left outputs and they also send something from here to the right and then vice versa here for the right but this is like this is something he has even like you know this is stuff where it's like you know when I saw this I recognized it because I'd kind of done some research on this stuff but it it's on some corner of the internet I don't know how many people know about it this is the one he published in 1997 he called it in the style of Greece enger' and he's just talking about this idea of like I call these all past loops which is that like you have like just like a bunch of all paths delays in one big loop and then you take the outputs out from some of them oh this one yeah it's a great sounding reverb it's not I don't know if it's actually based on any particular it might be based on one in the 224 it's about half the complexity of like most the 224 reverbs this one I hear like Keith Barr rest in peace he was the founder of alesis he's also founder of mxr he did a lot of stuff and he was like near later part of his life he was really open about how he did this stuff which is just like you can see it's like comes in goes through a couple of all passes like left and right goes out like total reverb time it's like it's really what it is is like two all passes one delay control the reverb time to all passes one two lake full reverb time repeat as necessary usually to have some sort of filter for those two or some sort of more complicated filter and those that think they sound different than the lexicon reverbs but I love the sound of the Lisa stuff I listen to so much music from the 90s that used alesis reverbs because they couldn't afford like any of the like the warp music bands they couldn't afford lexicon stuff I mean like you look at their gear now is like oh how can they afford that's like well those rolling boxes were 75 bucks or something exactly yeah now now these people exactly and and like remitted like the MIDI verb too and the quad reverb were huge in the 90s so much beautiful music done with those I loved it too and this is another drawing of the same thing so it can all pass and all passed away all passed away Tapout after each one although I really find like the the Keith bar one is much better because like all passes in the delay because if you don't have delay between the all passes we tap out it's just weird you get some artifacts because you've got like great this game very TWiki but there this right here is the feed forward path so if you have the input here and you have a feed forward path of 0.5 and if you for 5 to 0.5 right at that node you'd hear just the input signal with a gain of 0.25 if you run through a bunch of those you're just going to get some weird coloration things going on so a lot of ok this is less some kids love commercial reverbs have all past delays within a larger delay route loop I call them all pass loops they've also been called all pass rings they often have a lot of series all passes for the echo density some have less series all passes but higher numbers output taps were like they tap from the delays between them or they tap from the delays within them nested all passes that lexicon 224 used it and those things are sound cool but they are really get they're just sort of inherently rate ringy they're not super stable and I've just found that all past loops you can get them sounding quicker than FD ends with like they're not as twiki about like the delay links a lot of these things like you have to have the right tool a links and like an FD n is like the matrix you use has an effect on the delay links that you can use with it like some sparse matrices work best with random delay lace do that with four matrices and it sounds horrible so I just again I give this a class rest time about designing these things it's like I've written hundreds of these and that sounds like like sixteen years at this point it's 20 years I need to correct that and most of them are horrible like that's really how I learned it's just I get to the point where I code these things and just like find out what is bad and a couple of them are okay and then it's like but I learned it like I like to mix and match things it's really nice to like mix like you know you can mix feedback late networks in like the all past loops you can do all sorts of stuff okay what do you think oh I use um like C++ for the most part I've done a few things in the past and assembly but it's like my stuff is C++ I have my own low-level DSP library that's fairly simple it's not even fancy like C++ I basically use it as C with class like classes version of C so it's like the functions and like you know I don't do a lot of like I think as I'm coming from like I came from the embedded world where they just wouldn't even use C++ forever because they didn't trust the accuracy or like the speed of it so and then I have some customized like like Cindy intrinsic that I call you use anything like in gels I use apples vector library on the PC I just use I just use like I use like the Intel like SSE intrinsics and I don't use that much of them I mean I can show you that if you're interested in that like for example like my vector library is really simple oh wait that's I need to commit some files I thought I did that well I guess those libraries don't really have a lot of reverb they have convolution well like for example like this isn't like like I have a vector library where I just have an input output and I have the block size and some of these I'm like that one it's not even vectorized but like in some of them I'm calling like the accelerate functions I'm really happy I'm doing the accelerate in case Apple ever pulls the plug and just moves everything over to arm because they have the accelerate library on arm already yeah exactly oh there's like I kept like I used to do stuff for PowerPC as well so if they do that technically it's not gonna hurt it's just more the business model I worry about and then four or even have my PowerPC library there I could get rid of that for this plug it I'm not compiling for that anymore and then for like Intel you can see I'm actually calling like the like the intrinsics like SSE to intrinsics for stuff but like most that stuff I programmed back in 2010-2011 I've just been using it since there's like you know it's like once you figure out how multiply add happens and there's just a few things there's very few things like you know fractional there's very few things I need that's like actually requires any like reciprocal square root is neat and but you know like on you just call that on the on the Mac and that's just great like yeah how long would you try and fix it or just oh god sometimes I forget sometimes I forget why something sucked there's other times I'm like you know what there's something really good in here so like I'm air this is one like like this way it's like you know I've done like the two all passes one delay stuff in the past it's just that's not that good and it's just like you know for that particular ones like I figured out some good math for generating the all past links and then some other ways of tuning it so but all my stuff before that like I've have other ways of doing that and they just don't sound that good and you start like I hear much more I mean like you know I'm sure my senses are you know like my eyes are worse than I was when I was 20 years ago but my ability to hear reverb artifacts is much better I can hear like the patterns and stuff like that I can hear like weird resonances that I don't like so you know but but if there's but there's but the problem is when you come on things like sometimes you just keep coming back to ideas and it's like but and so it was like you know it's like my ability to kind of like know how to tune them is evolve but then my ability to hear the problems of them evolves as well like you're working on something listening listening this name and then just I can't I don't hear the different I don't know it's more like frog vision to me like I like a frog it's like what you end up here like frogs like my understanding of frog vision is that they will look at something and just stare up at it and eventually everything fades away it's almost like it just like anything that's not moving fades away so if anything changes in that environment like a fly moving across there that's what you see so it's like when you're hearing things over and over it makes it's like you're not actually hearing like what you start hearing it's not you're not hearing the same thing over and over you just start hearing the differences over time but then at some point you have to stop your ears hurt but you know let me start again the next day and and I've actually taken pretty much three years off of designing reverbs so although I in it back in early 2017 at some point I'd like because I'm working on this new product I like I just took a breaks like I want it's like try to like make a really natural sounding reverb and I'm air I just took like a month and like just went through hundreds of algorithms and they sucked and then I came up with one that sounds good and so after I get the howl to lay out I want to try that you know like actually go back to that so it's like that's kind of on the back burner but it's like it's it you know so I spent years and years on it but I spent last three years doing mostly other stuff so you prefer headphones or a speaker speakers headphones are gonna like head like I have monitors headphones just end up hurting like hell over time that's right that's the thing is like I mean just even the pressure on the ears ends up it's kind of itching but like you can really get ringing if you listen to at least I do or as like with speakers not so much I have Mackey HR a two fours but I think that's my company I used to work for bought them for me back in 2001 I've just used them since when I really want to tune in something final I go to my to Don guns house he's a local engineer and he has a studio in his house in his place and he he always changes out his monitors but he used to use gen Aleksei sound good his space is really well tuned and that's something too it's like I just assumes like I have mathematical things I know how to do but I'm gonna when I do the final tuning I want to go to a space that is acoustically really nice because otherwise it's like you know if something otherwise you might just be like tuning to some reflection that's over here or something and your it really throws off stuff when people hear it where it's like you know just the frequency balance is not good so I could probably use better speakers at this point probably smaller ones by heaven I don't know I mean I guess like one thing yeah it's like so I would like I was really interested in computer music like the the concept of it like I remember like when I was an undergraduate I never could became the graduate but when I was like an undergraduate like I really wanted to take these computer music courses but I was just intimidated because it's like there's like there's like 16 people in this course when I first start out it was on like you know 1988 you everyone there's like a whole bunch of bt100 terminals and every terminal had two speakers on it all those speakers connected to the same DAC in the basement so everyone like there would be one person like that could be listening music at a time and so you either listen to that or you turn that off because there's like they had like a pdp-11 as well as a like refrigerator-sized computer as well as like a custom digital processor that was also refrigerator-sized so that's all they had and like is which also mean it's like it was a very collaborative environment it was like in this really nice building it used to be the former university presidents mansion it's up on this hill like always people doing that is really like you know just so people would listen to each other and other and all that but I didn't have the night I wasn't a double e I didn't like I was kind of intimidated by that so I only took some of the courses and then but technology changed like midway through they start replacing everything with next computers like the cubes and so mid 90s people were starting to use like the kid like they still have this like non real-time languages like it still would take time to do it but you could run it on a PC like for originally of Linux PC and then just a PC PC like running Windows 98 and so like I end up taking this year-long course at computer music at u-dub talked my way in there but one of the things like I really love the sound of granular synthesis like they would do things where they had like a hundred randomized grains processing the sound extending it over time I love that sound but it's like also that you know what I could take these for delay lines and feed them back on each other and modulate them and feed a sound into that and you get a sound that can last almost as long that has really similar characteristics so I think from the get-go I was always interested in kind of the compositional aspects of it and also just the sort of cheating of it because like you know Ford Ford like for delay lines even the even back then it's like okay press compile okay you're done the 100 tat like hundred grain granular stuff is like okay for a minute a couple minutes sounds like press compile exactly hacked away yeah I wanted like with the same stuff but quicker like don't be exactly and it's like I mean it's like and like these like I like that sort of like I mean I really like that like there's a definitely there's a kind of school of computer music there was much more Vagan arian like you know you had like tons of complexity it's like you'd have thousands of grains and you'd have like hundreds of oscillators and they'd all make this stuff that's like you know what like like there's something about these recursive algorithms that what we call reverb that was used for reverb elements but it could just be abused to kind of create these sounds now it turns out it's like this wasn't like I was inventing anything by that it's like you know whoever could afford lexicon 224 1970s 9 was doing similar sounds like the dudes in the early 90s with a quad reverb we're doing similar sounds - it's like selected ambient works is basically you know a couple sips since through a quad reverb and you just crank up that reverb time so they're doing sort of similar compositional things but it's just like you know kind of like learning these sort of like things with it but it turns out that like you don't knit like you know one people don't necessarily like people like to buy the simpler things but - the simpler things can get kind of crazy like a quad reverb is kind of a commodity thing but you can get really that sort of complexity of it and even like when several years back people were into Paul stretch for massive time stretching it's like it really said it's like okay it's this massive sound that fades in over time and fades out and it's like you know like and that does with like giant FFTs and it takes like like it you know it takes a lot of like CPU to do that well it's like you know just take something like this turn up the diffusion turn up the size [Music] well it takes a little bit of time let that delay because it's tape-based on tape so this tray like [Music] let's even just clear that real quick takes a little bit of time to something where like actually fades in over time so it's like I mean it does a very similar thing to Paul stretch which is like the the artifacts you find from taking huge FFT windows and windowing them is very similar to what you get with just having a bunch of series all passes and modulating the delay lengths a little bit this thing you reached it in yeah this hack that you were like hey it is faster than this like kind of but then reverts just having the kid on that remained interesting up until like recently where you're like God I read you didn't well yeah yeah and I and there's so much Emily Emily and the plugin doesn't leave up to you it's like but it's like it's like it's but there's something about the mathematical nuances of it because there's like there's no one right solution and there's so many factors even if you have the same like we're doing a 16 delay fdn it's like what do you set the delay links that where you set the matrix at you have so many variable like what's the modulation there's so many variables and like just playing around that it's kind of like I don't know I've never played Sudoku but like some math matter like chess it's like some sort of metal mental patterning thing and you can just always work on it as I go for walks Ravenna Park and I'd be thinking about oh if you do this this and this and like it it ends up being of like there's just so much you can do with it and they sound and like and hopefully things sound different I mean I'm hoping that it may it's probably like you know a lot of things end up sounding the same you have different approaches to get similar sounds but it's I don't know it just ended up being like this a lot mentally to work on and I also like because it's kind of I like stuff that's born obsolete like I my dad like would in like his job was like in the 80's and 90's working like you know kind of recruiting people that knew technical skills and back then it was kind of scary because it's like as soon as like people hit their early 50s the jobs start drying up for them and is like you know because it's like there was massive evolutions like we're not using COBOL anymore we're like we're not using mainframes it's CAS 400 if you don't have a s 400 skills where it's like I'm working some my reverb algorithms are based on designs from 1975 if that's been obsolete forever and it still sounds good so there's something I really like working with a technology that is inexhaustible ends just and it's not as like much about like being in the moment and cutting edge it's like obsolete DSP still sounds great so is that a grainy quality to me doesn't work and I'll use like the waves plays you ever it sounds great but it wrecks my CPU and then you talk about these old harbour boxes that have so little processing power built into them I'm just wondering what your general thoughts are processing power find that toward reverb algorithms when it's necessary when it's helpful it's well I mean like one thing is like the like the amount of CPU is not necessarily proportional to the quality of the algorithm like for example like I'd say like in the Valhalla vintage verb like in a given algorithm about half of the CPU is being applied towards the cruft like you know it's like you have to do work to make it sound like a 16-bit algorithm at the 12 bit converter you have to emulate that stuff I mean I guess I could program in 16-bit fixed-point but that's I'm not gonna do that it's like I'm not gonna be programming an MMX or whatever that is so it's um you know like so there's a lot of work trying to emulate these artifacts and then something like the waves thing like I'm pretty convinced that the waves plate is actually convolution of the stuff at Abbey Road and most of the CPU is going towards modeling the analog inputs and outputs like that's happening real-time they've got saturation because if you turn those off the CPU goes way down but I'm not sure what waves it what they're doing with their CPU because it's like a lot of stuff it sounds like but I call time-invariant which means it's not changing over time which means you could do it with convolution so but you know I think that that's and that's the whole thing it's like if you want to get real like I like there's for example like the per caste reverb uses it's far more powerful processing than this but then you know like and like then all of this then all of this apply towards one thing on all the cores but the question is do you need that I mean like if it sounds like it I mean like there's all sorts of ways to game the different sounds it's like if it's like you know like if you could do some technique that sounds like that at a fraction the CPU and it sounds like it then it is it because none of its real none of this is like actually modeling a real room that's still several orders of magnitude beyond what this can do like modeling everywhere in real-time would just be prohibitive and no one's doing that in real time so it's just you know I think it's it's more about like I think the CPU thing is like some people might you know there might just be programming things with stuff like that but I think so much of it comes down to the philosophy of what they're you know of what they want to do and I think a lot of it is not towards the Reaver of itself but modeling the kind of emulating the analog saturation stuff that's around it or like for example like Universal Audio they're exactly emulating the artifacts of the 224 at the original sampling rate which is like 20k it's really hard like you kid like our machines run it 48 4.1 48 96 they don't run it anything that evenly divides into 20k so they're doing some very complicated sample rate conversion so that's part of what's going on is like the lot of these people are trying to get things exactly where I'm just like yeah that's close fun I won't talk about anyone's I don't like it's like you know it's like everyone's got like different aesthetics and all that I mean like as far as one's eye I mean and achieve is like selling on the getting close I mean I have her per caste I haven't listened to it much because I've been working on this delay I really think exponential audio does really clear sounding stuff that's Michael Karns he used I mean he worked at lexicon since 1989 until until like I think 2011 or something they parted ways and so he formed his own company and so he's doing things that they don't they'd the lexicon ish but they sound clearer he isn't he doesn't have to make things sound exactly like what use doing there so I really like those and I think there's other really like really I mean I love the the fabfilter reverb is clear sounding and its GUI is beautiful this you know like I think but there's a lot of it has to do with like designs like I'm excited to hear Randy's three Namba thing because I love stompboxes I love things that are simple so that's part of it it's like I tend to be like lose my interest when something has like 50 controls or whatever oh no no I've got like I think I had like three lexicons at one point part of it is like I couldn't find these things it's like you know like I mean like you'd rent them it's like I'd go rent like a four ADL and at the time it's like you know I was not going to pay four thousand dollars to buy a for ADL you know like I Brent I'm in a studio that a 224 XL you know stuff stuff like that so you know well you go in like with you yeah I went into the studios kind of having a pretty good idea so just like I was like able to just like really do hardcore dong-gun with me we'd like spend eight hours or whatever we'd like at the end the day just like your brain is fried it's not like doing that sort of listening it's not fun at all it's just like because it's not like you're listening to music you're listening just like little signals and stuff and that and that's one of things like I mean like I remember doing that in 2013 with lexicon 224 and that's when I figured out like like there's this one sound like let me see if I can duplicate that that just freaked me out like mine is actually starting to hear the sound of the modulation as easiest just to do it and so let's do get the oscillators out just sine wave setup [Music] so let's get the attack let's turn it's hard to hear in here but there's like you started hearing these things like I was like it's let's see let's do in 70s mode yes it's well let me turn down this stuff to you I can't really hear it in here but if you're hearing it over good speakers it's like there's like it almost sounds like a modem so that's the sort of thing where it's like you know like start freaking out like oh it's a modem it's like oh that's because no matter what's going on they're using a z80 to generate the modulation and that's an 8-bit processor so it's like the modulation signals 8 bits which ends up creating like crazy quantization noise of its own sort so that's the sort of thing but it's like just having to go in and just like you know knowing enough of the theory and reading the service manuals and like so you're making those 8 hours at whatever the rate is count so that's more what he did like yeah I own stuff and figure stuff out but there's like a few things that's like you know we just rent and all that it was it wasn't even necessarily best of class I mean I knew I wanted to get that kind of like older lexicon sound but there were preferences like I remember I rent it to for ADL and like the one we rented was like had earlier roms like the one I could find in Seattle didn't even have like the most recent ROM so it's missing a fair amount the algorithms but I could hear the random halls kinda like it wasn't I mean it didn't really impress me that much whereas like I heard like the earlier - 24 - studio I just I've worked with one in college and just like that sound to me just spoke to me so it's like it's like my stuff it's like it's gonna be not necessarily the most realistic emulation it's like I want it to sound good to my ears and then a lot of things that's like just doing it for like look at like you like these old things have these like like multi slider Hardware interfaces and like you do stuff like oh you'd like with a couple of fingers you can adjust like the low frequency reverb and the high frequency like decay time like low and high decays and the crossover between them just with a few fingers and it's like you can't do that with a mouse so it's just like I changed how things were mapped so like the controls would work with that and then also just like what controls can I gang together and leave it's like you know a real 224 X 0 40 Ella has like so many more controls than this and you just have to like you know these step through them on menus and it's like I'm on everything on one panel and like you know not have to step through stuff so a lot of it is like choices made towards like one my own aesthetics and to like simplification of the interface and then yet another question I don't know thank you i I think that like part of it was working like I designed a reverb for a company audio damage like that like this licensed through them and their stuff was like 39 bucks 49 bucks and that's actually how I came to see them it's like you know it's like I you know it's like I didn't have a lot of money and it was something that I I think there's something of just like one like but is also something that's like you know they talk about a disruptive business model but I wasn't thinking about that I just basically it's like I remember I like I interviewed like I you know like we moved back here like I was working remotely and then got laid off my job did contracting but around like 2008 the economy started drying up so I was like okay is like I have a girl like interview for like a video game job at Microsoft and like I'm there like you know doing sound I remember they're like hey you go home early that's not good in a Microsoft interview I wasn't able to move the trains correctly or whatever it was that question like the little Martin Gardner question but but so with that it's like so I really it's like the $50 one is like well no one's there's no audio industry around here I can work for I could go get like a more junior job probably at Universal Audio but the money I'd make would be more than like look both my wife and I have to be working full-time someone have to watch the kids and like whatever money I'd make would be offset by the cost of living in the Bay Area which is just so dang expensive so it's like it's kind of like I had nothing to lose I had like literally nothing to lose with like doing my own company and it's like well you know what if I have nothing to lose and it's like I don't really have any particular motivation to like price things the same as other people it's like I feel like that doesn't help me I don't make any money if they sell it for 500 bucks so I mean it was just basically like you know and probably just like and and then also it's like a way of like look it's like I'm you know my GUI is they didn't look photo-realistic or anything like that it's like look it does this I don't want to promise too much so it's like it just seemed like a right pricing point plus I by but worked for like done some work for a company that a similar pricing point so like Chris Randall and Adam shoptalk definitely were named for something for that oh no I did yose yeah rad rad checklist rat Jack was christened Adam doing on their own so what is the difference between the 70s and 80s oh well it goes back to that stuff I was talking about like the seventies mode it's it runs internally it's down sampled so it's like running internally at like half like at if if this is running at forty four point one kilohertz this one's running the this will be running at twenty two point oh five and then which has a big effect just right there in the high frequencies there are nothing above 10k but then there's stuff like I I'm music you know in the 70s and 80s modes I'm running learning linear interpolation on the delay lines which basically gives you a sort of low-pass filtering that's dependent on the sampling rate so 80s mode that sounds kind of like in 80s mode is that makes a little bit darker in seventies mode that linear interpolation makes it a lot darker and then in now mode I'm running like a much you know just a all pass interpolation so it's very smooth like no high frequency loss this is the original GUI Bush we people like it hurts my eyes it's like well that's kind of a clue that's like that's what it sounds like it's like it's very you know I just want something just black and white this is like this really like there's no loss to it there's just it's colorless but then I realized it like okay black and white this way it looks colorless and people won't complain hmm there's gray I know I know but I really want like written aliy it was blinding white I've learned my lesson people don't like blinding white GUI people like we're living I'm not gonna use the now mode because I like well I wouldn't use it anyway I don't like the sound of it I put it in there just so you have it but but I've learned it's like yeah it's not like this is not my thing this is like I make it but it's like I'm trying to make something that people want to use and say well let's make something that still gets the idea of no color but it doesn't hurt your eyes when you look at it one of those one of those valuable lessons I've learned like in vocals but like you know but you have like stuff would save like sine waves and stuff like that get someone to sing or play you know you want the st. you want a sample you want the same thing played over and over and over again where it's like there's like that one little place where the singer's voice cracks you want to hear that the same every time so you can hit tell what part of it is like you know what like cuz me does something special there or maybe it doesn't and you know you you really like for deal I mean like I've listened to Tom's diner like ten thousand times I'm sure I don't know the words I'm like it's so weird like I'm like I'm saying there is like I am sitting in a corner at the huh feels like I should know this I'd like I've heard it like so many times but yeah it's like you just like you listen to the same things but it's like you know it's a combination of like kind of developing the stuff like sine waves same vocals over and over dry musical signals impulses and then once you start getting it live then you have to listen to more material and like hear how it works with like a bunch of stuff I'm sometimes I really like I tend to do the simpler version like like a fab filter which is it has of just a built-in spectral analysis mode well that sounds bad hey sorry so it's I find this very wow that's a lot of modulation hang on but like you can see like it's very useful for some things like you can see like you know like just running a sine wave through it you can see that Peaks are moving around and all that with a modulation or is like if you turn the modulation off [Music] it looks different you get a little bit of tiny motion but it's like I find like that's sort of spectral analysis I might do where as I climb and then it does really useful things like they put it after shimmer and you can hear an artifact and shimmer they've like how I did a noise reduction or you can you don't hear it but you'll see it hypothetically I'm am not seeing the right one oh there we go okay be that way I don't know why the spectral plot stops so let me just bring another one in oh because I'm not writing a signal through that like that's something to tell like oh well there's a lot of like low-frequency noise going on below the signal itself so like there's like spectral plots are used very useful for analyzing things but often for hearings like fruitsie seeing stuff that you wouldn't hear or like seeing like for example oh look at this it's like this sine wave here instead of being like a peak it has a much wider skirt and that tells you stuff about like the noisy processes and all that going on inside of it so yeah spectral can be very useful to like kind of suggest things but you still have to figure out what it is using your ears and theory and all that I don't let me just see if I have any more slides that are like important I think we're getting like like this is a design like I was talking about like this is an FTM that's got the all passes in it it sounds pretty good I just I listen to everything like you know like as far as getting stuff because like you know like this is what I do but then there's also like you know getting feedback and having several products is like you have more people listening to it I love surveys I hope I don't know if you notice I didn't mention forums in here but like I have a survey and it's like I've had thousands and thousands of people respond to it versus like in a forum it's like you'll have like a few people say like well general consensus is is that you mean you and those two other people they're just the posting that each have 20,000 posts and so that's like that's something I've like learn just like you know versus like surveys are neutral someone can sit there and rant all they want is like okay you've got one vote what I'm looking for is just the broader sort of thing and obviously you want to have people like you want to make people like you know really make people happy but the point is it's like you really can't make everyone on earth happy so trying to figure out like what's kind of like are there general trends and stuff like that that's not really like about designing reverb that's more just a business II think this is kind of something it's like I actually end up talking a few people like I'd love to know why people did this stuff like why like what was it like to make reverb back in nineteen seventy five or whatever and you know definitely had some people about it but it's like I don't know I still love to hear from some of these people because some of them are like starting to pass on or getting close to it so be like nice to me knows we're losing some of that sort of history now the history is there in the really the algorithms themself are in the roms and you can kind of figure it out by just I just wonder like why people made the choices they did like I can sit there and just sort of you know it's like you know I explained is like I kind of like I like computer music I like big complicated sounds as a way of cheating but I other people have their own reasons for doing things like why they got interested in it so I'd be nice to hear from other people how they did that

Channel: Seattle Music Machine Salon

Views: 9,687

Keywords: valhalla DSP, reverb, synth, SMMS

Id: aJLhqfHrwsw

Length: 79min 5sec (4745 seconds)

Published: Wed Aug 28 2019