Menu Sidebar


Airwindows is one guy: me! I am an audio hacker and computer programmer from way back. I seek only to continue my life up here in Vermont, inventing things and putting them onto the internet, sometimes for free and sometimes for pay. My hope is that people richer than me (i.e. most people) don't rob me. My other hope is for another cup of Aeropress. One out of two ain't bad!


TL:DW; Like PaulDither, but more so.


This is for sort of a special purpose.

I’ve got an outrageous custom dither (well, wordlength reducer: there’s no noise and no randomness in it) called Not Just Another Dither. For final output, it’s that I recommend.

However, I’ve got a mastering engineer friend (Hi Bob!) who favors another dither of mine, PaulDither. The reason is, it’s traditional TPDF dither. It handles dithering duties perfectly (with a slight Airwindows-izing: the weight of the final noise floor is balanced to make one ‘bit’ function as zero, and to make the dither activate adjacent bits on either side for a better sound. This also gives a teeny least-significant-bit DC offset, if you assume ‘zero’ is between the two smallest bits, positive and negative)

I wanted to give Bob (and anyone else selecting traditional TPDF) something fancier that still counted as purely normal dither. So, here’s DoublePaul.

It’s TPDF highpassed dither, just like PaulDither. But, I’m also using some adjacent samples to skew the noise profile even further into the highs. It’s just a redistribution of energy, and it leaves the key pair of full-intensity noise samples to stay as TPDF dither. I just doctored the way you hear the noise, to push it a little farther into the highs. I’ve also backed off the teeny DC offset, and shaped every filter tap by ear. If you liked PaulDither, try DoublePaul and see if you like taking the PaulDither concept just a teeny bit farther, with no penalty to how the TPDF works. It’s just like a little sinc interpolation, a thing that’s trying to hide the TPDF energy by pushing it farther into the extreme highs. If you don’t trust NotJustAnotherDither, but you’d like something even quieter than PaulDither, try this.

All this is supported by my Patreon, so if you’d like to see me cheerfully continuing to think up this stuff indefinitely, help me get it to where it looks like a real job. I’m quite good at being a starving artist but money is a resource and it limits and distracts me when I’m starving too much.


TL:DW; Filter.


Sometimes half the fun is just inventing :)

People say all filter plugins are just biquads with different GUIs. It’s not true… or at least it’s not true at Airwindows. (actually I have to see what I can get out of biquads: I generally don’t even use ’em at all)

This is a new filter. When I say new, I mean that you don’t have it. I can demonstrate that, because due to the way it works it has a little characteristic response quirk at 1/3 the sampling rate: nothing that hurts the sound, just an odd little notch under some circumstances.

Because it’s a typically Airwindowsy, not-normal EQ algorithm, nobody has tried it, and because of the little quirk, if anybody did try it they knew they had to reject it, because it wasn’t perfect. It had a quirk! And even if they experimented, they probably weren’t that invested in concepts like the sequence of operations on stuff that, in theory, could be done in any order. Why care so much about the implementation details of an idea that had a quirk anyhow and was therefore useless?

More fools they :)

Because musically, this filter has tone for DAYS.

There’s almost an analog-synth-like effect to it: though it doesn’t accentuate the rolloff point, it behaves almost like a DJ ‘isolator’ in the obviousness of its filtering. It has a highpass, a lowpass, and dry/wet, and the high and lowpass are melded into each other as part of the algorithm to get the sound more fluid. It’s also got new experiments in rejecting zipper noise and clicks, because I’m figuring people will want to automate this one. But, even though it’ll work for crazy effects and ‘isolator’ madness, the fullness of the tone will also let you use it in mastering: if you need to tighten bass or roll off just a hint of brightness, you’ll pay no penalty in the body of the music. And the dry/wet is there to let you treat it as a shelf… or to create a presence boost by isolating an area of interest, and then leaving it mostly dry with a hint of the clarified zone.

I’m thrilled with this filter, and nobody else has it: nobody else would be allowed to do this one, because it has a quirk at 1/3 the sampling rate, and typical commercial developers are not allowed DSP quirks. But, since I’m supported by Patreon, I am. I’m developing a whole revised concept of what DSP in 2017 can mean, and I think you can hear what I’m up to in the plugins when you try them. Some might prefer the rather cumbersome methods and DSP of the commercial plugin industry, and if you like that, don’t let me stop you.

But if you like the stuff I started to get into when I brought out PurestDrive and that line of plugins, before Airwindows Patreon was even a thing, then it’s your lucky day. I’ve turned it into an analog-like filter, and it’s free. Have fun using it :)


TL:DW; Highpass on the side channel. Utility plug.


This one was by request. It’s not very fancy, but I hope it’ll come in handy. If it is, spread the word!

I’ve got a plugin called ToVinyl, which is rather fancy. It’s the Airwindows attempt at a vinyl mastering suite: elliptical EQ, highpass on the mid to help get levels, an acceleration limiter to tame sudden energy spikes without really cutting brightness otherwise, and even a groove wear modeler that was the precursor to Airwindows Aura. That’s one of the Kagi for-pay plugins, number 9 on the list when I reach $800 a month and begin doing those.

But, I got asked for something much simpler. A highpass filter on just the side channel, and one that would go up the full range from subs to highs, so at full crank it’d be purely mono.

Well, I’ve altered that a teeny bit: Sidepass is aware of sample rates, so if you’re at a 0.1 setting at 44.1K, it ought to be doing the same thing at 0.1 at 96K. But apart from that, here’s a little utility for AU, Mac and PC VST that just highpasses the side, from DC to 22K. It’s an Airwindows filter, same as what’s in Density and Drive as a highpass, and it ought to do nicely.

My work’s supported by Patreon, steadily growing as I keep on working. (My Patreon is new-ish but I’m not—I’ve been at this ten years, no sense stopping now!) If you’d like to see the fancier, for-pay plugins start to come out as free AU/VST too, join up for a buck a month (more if you’re flush) and we will eventually get there, without inconveniencing any one musician too badly! Patreon is a lot less lucrative than trying to make that one killer plugin, but it’s steady and it’s growing and I’m committed to the open/free model :)

Xiaomi Yi Tuning With autoexec.ash

Here’s another Airwindows Starving Artist posting! It’s kind of long, but this outlines what I’ve done to get high-performance results out of the Xiaomi Yi. I show the visual tools I made to focus and calibrate the Yi, explain how to set the saturation and vibrance controls, and make an impassioned plea for allowing the Yi to use its noise reduction when shooting compressed HD video. Even if you didn’t hack your Yi like I did, there’s stuff you can use here! :)

Not Just Another Dither/CD


The votes are in, and the Internet has spoken. We’ve got a name for the new noise shaping variant on Naturalize, after a January of wild suggestions. Seems most people simply wanted to emphasize that this was not your ordinary dither, not your daddy’s TPDF.

This is the plugin (two, in fact) that beats all the other ones, including the original Naturalize. In normal use, this ‘dither’ (in fact it’s not a dither at all: it’s the Benford realness stuff from Naturalize, run with a noise shaper, and all the added random noise taken OUT) will make any fixed-point output sound like infinite resolution. It’s AU, Mac and PC VST. The plugin comes in two forms: NotJustAnotherDither which is 24 bit, and NotJustAnotherCD, which is 16 bit.

So, if you’re sending an output to your DAC for monitoring, or to mix down a bunch of outputs using an external mixer, you’d use NotJustAnotherDither, for the 24 bit output.

If you’re making a CD or something else that requires 16 bit output, use NotJustAnotherCD. It truly will be not just another CD, because it will sound like 24-bit high-res audio, on any CD player or Red Book CD playing apparatus, with no fancy player or post-processing required.

The output levels are low, without a bunch of extra energy coming from the noise shaper: it’s pretty well behaved, though like any hyper-performance ‘dither’ this is really meant for output formats. I don’t think any harm can come from repeatedly using this in processing, but you can always use a normal TPDF if you want no change in tonal balance (plain TPDF dither only adds broadband noise: more opaque, but if it builds up it won’t accentuate anything: dithers with a lot of high frequency energy can end up exaggerating that.)

This is supported through my Patreon, so the more people hear about it the better: pretty much the whole industry can use Not Just Another Dither and Not Just Another CD, to enhance their monitoring and mixing, and to produce high-res output files and CDs. It should make your audio sound better, no matter what your audio is.

That’s my wish for you :)


TL:DW; Fluctuating saturation curves for a tubey tremolo.


So, you’ve probably got a tremolo built in to your DAW.

But, if you’ve heard tremolo effects off classic records, it’s a whole different animal. DAW tremolos are neat little volume animations, capable of many cool effects (try making ’em squarewaves synced to the tempo for a nifty sequencey effect). But they don’t have that organic pulsating thing that takes a sound and gives it a whole new character.

So I made this!

This Tremolo uses saturation and antisaturation curves like you’d find in Density, and does the tremolo with that. It’s the same trick I use on the compressor ‘Pyewacket’. The result is, the loud parts develop a density and thickness mere volume won’t give you, and the lean parts hang on to a skeletal form of the transient attacks so your music comes through. This is not just ‘analog color’ like a coat of paint, Tremolo works quite differently from your DAW tremolo. It doesn’t sync to tempo, but that’s partly because I don’t know what to read (in AU and VST) that’d give me that information: could be added in future if the secrets are forthcoming, but there’s no sense withholding Tremolo just because of that!

Further development of Tremolo and literally everything else is supported by Patreon, which as I expected is steadily growing month by month as I keep working. I’m not some dotcom that needs to be instantly rich or I go poof and vanish, far from it, but don’t forget to tell people about this project! I’m just about to release the Airwindows wordlength reducer (works like a dither) to dramatically eclipse everything else that’s been done dither-wise. And I even have an idea for slightly refining PaulDither, for those who really like that one (if you want TPDF it’s hard to beat it).

Thank you for listening, and I hope you enjoy Tremolo :)

Kick Drum Hacking

Much like my livestream on adding a Mudbucker to an electric bass, this video is about ways you can make custom drum mics using spare speaker parts, even if you’re not on a platinum seller budget! All about how a SubKick works, and a trick for getting DI beater click. Plus, these do not need a preamp to work! :)

In this video, I’m showing you how I made my ‘beater click’ and sub-kick ‘microphones’ and wired them. I’m assuming you either know how to wire things, or are willing to learn, and focussing much more on the specifics of how I’m implementing it. Because execution is everything with this stuff: placement of the tiny ‘subkick’ has a huge effect on its performance.

The beater click transducer is a piezo guitar pickup, probably a Fishman. In this context, any sort of piezo element ought to do. You mustn’t wire it parallel to the subkick or the low impedance will totally wipe out the piezo’s output (it’ll all go through the speaker, uselessly). An input with transformer balancing will also have this problem to a lesser extent. You can hear it all by itself in the video.

The subkick is a tiny acoustic suspension speaker where I cut away the basket with tin-snips: you can see in the video how I did it. I left some suspension to act as a sort of ‘hovercraft skirt’, helping to contain air near the drum-head. A larger driver might need support around the edges of the speaker, but this one is so tiny I could rely on the spider (suspension near the magnet) alone, without scraping. If it’s a driver this small, it must be acoustic suspension or the resonant frequency will be way too high! Removing the outer suspension also lowers the resonant frequency, which helps. When hacking apart a speaker be prepared for the possibility that you destroy the speaker: I’ve done this before, so I was pretty sure mine would be OK :) you can hear it wired up with the piezo in the video, it’s not all by itself.

I wired up the piezo, then did drum recordings with all mics, notably the normal kick mic (a modded 57 because it has to focus on mids in this scenario: it ends up as bass/mid/treble mics on the kick). I recorded with the piezo wired to the tip of the TRS jack, then flipped phase to see what was better. I liked the sound of inverted phase, so I rewired the piezo, connecting it to the ring of the TRS jack. That meant I could wire the speaker to the tip, giving me two independent inputs into the converter (one out-of-phase, being used for the piezo). I tried both polarities for the speaker and settled on my favorite: you can also wire up a phase switch on either of these, like you would on a guitar pickup.

Opening the video, and at the end, you can hear an example of all three ‘microphones’ mixed together, with the 57 padded down to not dominate the weirder transducers. This is done totally flat, which also means you get to have more immediacy from the sound and not run EQs on it: just balance the microphone types. If you are running piezo and speaker into a single input like I am, then you can only adjust their balance by moving the speaker nearer or farther away from the drum head, but that works quite well. It’s good for a ‘fixed drum setup’ where your studio drums are set up and left ready to go: finetune the positioning as you get familiar with how it works, and use the normal mic (doesn’t have to be a 57, but I like my SM57 in this role as midrange mic) to fill in the normal kick drum sound.

I’ve got an internal damping system in there: Evans kick dampers upside-down, with the corners of the pads screwed to lug bolts, and the part that’s normally along the floor of the drum, suspended in the air not really touching either head. This is to substantially damp low frequency ring while leaving the heads able to ring at higher frequencies: but they’re loose, so those higher frequencies are mostly ‘papery’ color, not basketball-bounce color from tight heads. This affects the real mic a whole lot, and you can hear it when that mic is soloed, but it doesn’t affect the piezo or speaker nearly as much. Exception: if your drum booms in the lows, the speaker/subkick ought to pick that up a whole bunch.

The great thing about the piezo and speaker ‘mics’ is this: these put out such a hot signal that they’re plugged directly into my MOTU 16A. If your mic pres are not exactly top of the line stuff, using this technique you can bring more thump AND more click to your drum recordings and not need more pres! Mic pres are maybe the hardest thing to do on a budget, and this lets you supplement your sound while running directly into cheaper line inputs, with good noise performance compared to mic pres. You could even get respectable recordings using just stereo overheads, and reinforcing those with the piezo and subkick mics run into line inputs, which might be a great way to get started on a budget interface that only has two built-in mic pres. I’ve tried this and it’s a neat sound with a lot of character and a very hyped, big kick drum sound (since all the direct sound is either subs or click, and all the ambient sound is the kick’s room sound, it’s huge in scale).

Have fun experimenting! :)


TL:DW; A tube recording console type tone coloring.


More classic Airwindows analog modeling with TubeDesk!

Tubes aren’t necessarily ‘mellow’. They’re also known for clarity, realism. TubeDesk isn’t a mud-ifier, but it might bring you some effects reminiscent of vintage recordings.

Like the other Desk plugins, it’s got a kind of saturation going on, a distinct flavor to how it distorts. However, unlike TransDesk, its power supply is very different. TubeDesk is so old school it acts like there’s a vintage tube rectifier, imparting an obvious rectifier sag. This conditions the sound, affecting how dynamics work through the plugin.

You have a distinct ‘analog modeling’ tone then, which is no specific console, no arbitrary color: just sort of retro vibe, generalized. It’s not calibrated to work with Console (that’s the for-pay version of Desk that consolidated these and calibrated them all) and it doesn’t have special requirements for where it should go. Place it where you want that style of tone coloring: like Desk, you can use it on auxes and submixes (including ones ‘inside’ Console) to better emulate running through a lot of circuitry.

And of course, this is supported through Patreon. I’ll try to ramp up production at least somewhat (doing eight things a month plus the patron-oriented content is pretty demanding, and I’ve been dealing with extraneous stuff that’s come up, but I think I’m still doing at least a thing a week :) )

Why I’m Kind Of Stuck Running Youtube Advertising

TL;DW: They’re a monopoly and I expect them to quietly blacklist and suppress anybody NOT running Youtube Advertising.

As you might imagine, on YouTube this video is known as ‘Why To Run Youtube Ads’. Follow along with my reasoning in the video, and you might come to the same conclusion: you’ve got to put across the appearances of somebody who’s naive and available to be exploited by YouTube’s promises, so you can piggyback on it by trying to do work that’s realistic (and be paid in ways that are more practical, like Patreon).

So if you wonder why the change… for one, you’re apparently not ad-blocking, and I think that’s a mistake. You should be blocking, and I don’t want you to turn it off in order to ‘support’ me in some way, because that’s what my Patreon is for. Much like Kagi before them, Patreon has paid me and (mostly) not jerked me around, and that’s what I’d ask you to support, for those of you who are well-off enough that you have a credit card that can stand twelve bucks a year in charges. I don’t trust YouTube to pay me, much less pay me honorably, and I’m not trying to get people to shut off their blockers. On the contrary: do as I do.

But, also do as I do with this. I’m convinced the algorithms already weight things using this, and there’s no reason to believe it will get better and every reason to expect the pressure to get more intense. If you’re trying to reach an audience through YouTube (or any ‘free’ service), you must play along and put across the appearances that you want to make money the way they want you to make money. It’s a scam, because these things follow a power-law distribution and even PewDiePie is unhappy with YouTube (and he’s at the peak of it), but you’ve got to pretend for the algorithms that you’re a true believer.

Or they will find someone who is.

Welcome to 2017! I promised I’d do ‘starving artist’ posts, and this is the first one. Be sure and engage with these huge, algorithm-bound entities in a way they can understand, because you cannot piss off a non-human gatekeeper and expect to skate by on the strength of your personality… or even your content! Find ways to recognize the world you’re really in, while making it plain what your expectations are. Mine’s this: I hope everybody has sense enough to block, and hope people understand what I mean by all this. You can’t communicate without a network, and you don’t own the network, and virality is an illusion: something is deciding what catches on, same as it ever was. Don’t get needlessly filtered out. :)


TL:DW; More of a transistory, rock desk tone coloring.


Continue the look into classic Airwindows analog modeling with TransDesk!

There’s a famous analog console known for rock mixes. It might not have the preamplifiers to hang with the APIs and Neves of the world, but it’s been a watchword for mixdowns, both for the sound and for the extreme flexibility it offers (automation, effective EQ, compression). I’m not going to name it, but I was tuning TransDesk to get into a similar area: in particular, I wanted to get a comparable aggressiveness into the highs. This isn’t a plugin for making things soft and sweet, it’s for rocking out.

That said, there are many paths to this ocean of sonic mayhem, and what distinguishes TransDesk is that it gets its sound with very little processing. You don’t lose much mojo just to get that coat of sonic paint. In some ways the immediacy of this approach is closer to the analog truth. In other ways, it’s less a would-be clone of a classic big console, more a way to get some of that energy. (Technically, I’m doing it by trying to match the overload characteristics, including power supply idiosyncrasies that affect the way energy can be drawn for the highs.)

The result is another Desk-style plugin, with a completely different sound. It’s not calibrated to work with Console (that’s the for-pay version of Desk that consolidated these and calibrated them all) and it doesn’t have special requirements for where it should go. Place it where you want that style of tone coloring: like Desk, you can use it on auxes and submixes (including ones ‘inside’ Console) to better emulate running through a lot of circuitry.

And please support my Patreon. This is just scratching the surface.

Older Posts


handsewn bespoke digital audio

Kinds Of Things