Airwindows

StudioTan

TL;DW: StudioTan is all the ‘non-dither’ dithers, up to date and convenient.

Let’s make things a little simpler.

If you were curious about the new work in quantization I’ve been doing… where I’ve put out a series of experimental plugins with names like Dither Me Timbers and RawGlitters and then said they weren’t dithers, or if you tried stuff like Dither Me Timbers and then found your limiter set to 0dB was now giving you overs, or if it just didn’t make sense at the time…

This is for you.

StudioTan is the sum total of all I’ve been doing with ‘dither’ that isn’t dither, for the last two years. It’s got three algorithms, StudioTan, Dither Me Timbers, and Not Just Another Dither. Each one is brought up to date, optimized, voiced to do what it does best. The first two begin with the use of quantization to apply EQ at ‘noise floor’ levels and the third applies Benford Realness statistics at noise floor levels, and they all use noise shaping to refine that behavior and get a specific tonal result. Not one of these have been quite available before, even from previous versions of Dither Me Timbers or Not Just Another Dither: it’s the latest finetunings of the algorithms. None of them can produce signal outputs beyond -1 to 1 so they can’t produce overs (if you’re ‘mastering’ so hot you get intersample overs, firstly that’s not a great idea and secondly StudioTan can’t help you there). All three come with 24 and 16 bit options right there in the plugin, as experimenter time is over and this is the final form, requiring no more fussing with plugins.

The effect of each is more pronounced at 16 bit, and that’ll give you a sense of what is being subliminally imparted at the far subtler 24 bit level.

StudioTan is like the posh, glossy, high class output. It sounds kind of like expensive studios and tape, brings up micro-detail but suppresses harshness. It’s slick and makes things sound more flawless and possibly more boring, but satisfying.

Dither Me Timbers is like the spatially enhanced, electrically charged output. It makes sonic events spark out of a darker, spacious background, and gives a little energy to transients and attacks. It makes things sound more human and attitude-laden, so it’s more dramatic and attention-getting. It’s got different noise shaping behavior and voicing but in basic character it’s the opposite of StudioTan, and can be approached from that direction.

Not Just Another Dither is like the holographic, hi-fi output. It’s a bit darker than previous Not Just Another Dithers and its purpose is to have detail go down to a digital noise floor that’s a balance of all types of artifact (since you’re always going to have a noise floor of something, no matter what you do) without changing character in any way. Since this perceived character is so balanced (this time, not even airy hiss as a ‘identifiable’ floor change) the new Not Just Another Dither is the best choice to seem like infinite resolution without even a noise floor being there. Instead of making ‘a sound’ like the previous two, it’s the one that sounds completely open and unaffected.

None of these give sensible results with test tones: you can’t take a sine and enhance high frequency details, so you get purely garbage data. Don’t use Airwindows non-dither dithers for scientific processing or your rocket ships will blow up on the launch pad ;) these are literally tone controls in two cases and an obscure data handling technique for the third, and they are re-voiced for 2019 to best deliver their sonic mojo. You can do test tone stuff to satisfy yourself they aren’t placebo effect: they really do the noise-floor-EQing or Benford Realness stuff, and the source code is open. But for evaluating how they really work, the proof is in the listening.

What they do cannot be done any other way, and they’re the final artistic polish on sonic creations done in or outside the DAW. Use them on mixes, or when mastering raw analog captures to lower word lengths for CD. They are ‘final dither’ in function if not strict definition, the way to crystallize high-res audio data into an output file that retains all the magic you intended, whatever flavor of magic that’s meant to be. The three categories of tone color ought to cover your bases there.

I know this is all I’ll be using from now on :)

I live off Patreon and look forward to growing in 2019: the more I earn through this hard, continuous work, the more I’m able to do. Increasingly I’m fascinated with audio hardware as well as software and if I can afford to (after all, I’m still living frugally on less than minimum wage) I’ll build physical gear for people to have, most likely using Eurorack format as a platform. That probably won’t be software-based but I know software folks such as Sean from Valhalla who’ve brought their software into the Eurorack format, so it’s not impossible. So if you like me doing this for a living, support the Patreon, not so much because it’s Patreon (if you paid me with PayPal, would you be supporting PayPal?) but because it’s me, and because Patreon is unusually good at letting me concentrate on my work rather than chasing individual ‘hit products’. That might end up with me having huge hit products (I think StudioTan may well be one of those!) but it needs to be idea-first, never sales-first. And you can’t fake that :)

Date January 14, 2019
Author Chris
Comments 9 Comments

TapeFat

TL;DW: TapeFat is the tone control from TapeDelay.

TapeFat

We’ll start off 2019 with something kind of fun :)

TapeFat is just the tone control from TapeDelay. It works like an averaging filter that you can use to either roll off highs (or eventually mids), or subtract the effect to create a highpass and take out the lows.

The reason this is interesting is, that tone control is completely bizarre. It’s an averaging filter, but on a pile of delay taps arranged according to prime numbers. Works more like an ambience control, but more densely packed. If you put an impulse through it you don’t get a smoothed-out lowpass so much as a bizarre micro-reverb. Since it’s using primes, it doesn’t reinforce any particular frequency. Since it’s an ambience, it doesn’t have any pre-echoes like linear phase EQs, and the artifacts it produces become a tone of their own (either in-phase, or inverted).

You can hear it on the video, which has a number of things updated, not least this: the new audio is directly captured analog sorta-house music out of my livestreams. This way you ought to be able to really hear the way my plugins retain analog qualities, because now the demo music is essentially AAD: not products of other DAW mixes or digital synthesis, but source material. You can also download those tracks at full 24/96 off SoundCloud. (I use an old Audacity to SRC them to 44.1K for plugin-video making purposes)

Patreon is how I do all this, and I appreciate people using it. I like it better than Kagi, which I used to use to get paid. If you can, please jump on the Patreon at a level equal to buying one this year (for $50 perpetual license, with no DRM, full support as long as I’m alive, you get the source code and the right to make derivative plugins: you know, NORMAL plugin licensing terms :lol: ) and after a year you’ll have paid $50 of which I get nearly all. Then, next year, we’ll look back and see if I made another plugin that you would have bought :) (and if there’s more than one, so much the better: I provide tiers up to five plugins a year and worked out the math up to seven)

See ya Monday at 11:00 AM EST for the Q&A livestream, and Tuesday 11:00 AM EST for the music jam!

Date January 6, 2019
Author Chris
Comments 7 Comments

Crystal

TL;DW: Crystal is a tone shaper and buss soft-clipper.

Crystal

We’ll step away from strange inventions this week, to plain old ‘phat buss mojo’ tone shapers.

Crystal’s the first of the Character reissues, by request: I know there’s the possibility for this to become people’s favorite plugin, because it already is one user’s favorite buss plugin and he begged me to rerelease it with updated code and VST compatibility. This is the result. Tonally it’s exactly the same as the classic ‘magic’ Audio Unit, but it’s got the denormalization and noise shaping to the floating point buss of 2018 and beyond.

The controls you’ll be interested in are Hardness and Personality. Hardness applies the same algorithm that was in ‘New Channel’: though it’s not a replacement for what made Channel special, it’s got its own uses. It lets you define the onset of clipping, whether soft-clip saturation or digital hard clipping. Though this dirties up the sound a little, it lets you dial the ‘fatness’ of the saturation effect and gives you a tonal parameter that no other Airwindows plugin gives you. Think of it as a slider for how much the roaring, overdriven midrange sticks out.

Personality is a precursor to what became BussColors (and there are other flavors to come) but in Crystal it’s a little different. The BussColors algorithms are taken from hardware convolution impulses, and there’s a time-constant making the interpolation between ‘loud’ and ‘soft’ impulses happen over several samples. In the Character plugins, this didn’t happen. It was sample-by-sample, so on the one hand there was no dynamic behavior, just each sample got a fixed convolution behavior.

On the other hand (and it took me a while to properly understand this) every convolution sample got its own, separate dynamic behavior. The curve was different for each one, so it became a more tightly controlled little kernel rather than a set of possible kernels. There are still people who swear these were the great ones, and I’ve learned to pay closer attention to such things.

And the thing is, Crystal’s not using a hardware sample. Unlike anything in BussColors, Crystal’s using a data set that comes from doing a brickwall filter: if I remember correctly, two different ones at different Q/steepness, and then generating the dynamic behavior out of that. So it’s doing a treble-restricting EQ behavior (a FIR filter), but then it manipulates that. The question is, do you like what it does? Some people really, really liked this one. Not everything about it is in line with how I usually do things. That’s why it’s different. Maybe it’s right up your alley? Let your ears guide you, and have fun checking it out.

All this is supported by my Patreon and the better that does, the more freedom I have to do interesting things. Right now, that means a couple of weekly livestreams on my YouTube channel on top of my plugin posting: Mondays at 11:00 AM EST I answer questions about my plugins and audio, a tech support stream that’s been running a couple hours long (I will go longer if I need to). Tuesdays also at 11:00 AM EST I’ll do a music-making stream for a couple hours, and then upload it to my Soundcloud at 24/96 FLAC, if I haven’t forgotten to hit record like last week :)

Date December 30, 2018
Author Chris
Comments 5 Comments

RawTimbers (and RawGlitters)

TL;DW: Just the quantization from Dither Me Timbers… and the opposite, as a brightener.

RawTimbers

So, Dither Me Timbers isn’t really a dither. I keep saying that even though it’s got noise shaping and gets a sound that stands up to comparison with Not Just Another Dither: it’s different, but good.

This is ‘essence of Dither Me Timbers’, no noise shaping… and the inverse, just because I can.

All RawTimbers does, is choose between truncating to fixed point upwards or downwards, from any given point. Like Dither Me Timbers, it runs one sample of latency in order to do this choosing. Unlike Dither Me Timbers, it has an opposite plugin: RawGlitters. I see this stuff as ‘level-dependent EQ’ and always thought it seemed useful to dull and soften the quietest sounds and let loud noises seem brighter: it didn’t make sense to me to brighten the quietest sounds and let loud noises be duller (might be worth a rethink of this, but I’ve been sick this week and can only do so much: this is after all vanguard stuff and I need to get it right, or try harder to get it right)

Anyhow, this is RawTimbers and RawGlitters. It does the exact same thing as Ditherbox ‘Truncation, 24 bit’ except it’s an EQ: RawTimbers softens the audio and rolls off highs (and actually generates more and different bright quantization noise, so it’s not like it makes the noise floor itself seem darker), and RawGlitters brightens the hell out of the audio and sort of merges it with the noise floor. Both drastically change the nature and quality of the truncation noise floor: they’re undithered, and each is super different from plain old truncation. I discovered making the video that if you hit RawGlitters first and run that into RawTimbers, you get a third sound that’s interesting and pretty good… and if you run RawTimbers into RawGlitters, it’s pretty terrible but I’m not your mom, go ahead if you want. :)

In listening to these, remember the intense EQ effect (as in, cut treble up to 90% relative to mids and bass) happens ONLY at a fixed loudness that is barely over the 24 bit noise floor. If you play around with BitShiftGain you can check it out more easily, as I did on the video. All the effects only touch the noise floor area. These don’t have noise shaping so they can’t change overs or add excess energy: they’re ‘safe’ in that they’re predictable behavior, though you might have an interesting time exploring what they do to the tone.

They can be used for 16 bit CD-making as seen in the video: BitShiftGain -8 bits, RawTimbers/Glitters, BitShiftGain +8 bits gives you exact CD dithering.

They can also be used for conclusive double blind testing: since this is strictly two opposite ways of quantizing audio to the same fixed point output, you can make matching files for use in ABX testers. Use BitShiftGain to dial in the output bit depth you want, save all the results at 24 bit if you like (the quantizations will be the same) and listen for the brightening and difference in atmosphere of RawGlitters. On an ABX tester you’ll be able to audition Timbers and Glitters all you like, and compare to X, which will be one of the two. You can use quiet sounds if you like, to help hear what’s happening: it’s totally legit to tailor audio that will reveal this stuff more easily, as it’s the threshold you’re interested in. I think it shows class to use volume levels where you could still stand the loudness of normal audio and not blow your speakers, even if the actual test audio is way quieter (cranking it up will help, and you don’t have to blow your eardrums up as part of the comparison process, just keep it sane enough that the loud bits wouldn’t be damaging)

If you do this and correctly predict X for whatever bit depth you’re attempting, you’ve conclusively shown that you can always hear stuff at that bit depth and nobody can contradict you. I’m pretty sure literally everybody will hear 12 bit even over YouTube on whatever casual listening stuff you’re using. I also think 14 bit is relatively trivial, and 16 bit is doable given the right example audio and some care and attention (this is a contentious claim, but I still think so). I don’t think anybody can actually do this with full-on 24 bit audio… but if you did, over whatever fantastic (and very treble-accurate, and very low-noise) mastering rig you used, it would be conclusive proof of it.

And you still ought to dither, even to 24 bit, ‘cos it’s the principle of the thing. But this is something different to do, and it’s a legitimate choice. I suspect there’ll be a lot of electronic musicians who take a liking to RawGlitters just because it airs up the digital noise floor in an interesting way…

All this is supported by my Patreon and the better that does, the more freedom I have to do more interesting things and expand my horizons. Right now, that means a couple of weekly livestreams on my YouTube channel on top of my plugin posting: Mondays at 11:00 AM EST I answer questions about my plugins and audio, a tech support stream that’s been running a couple hours long (I will go longer if I need to). Tuesdays also at 11:00 AM EST I’ll do a music-making stream for a couple hours, and then upload it to my Soundcloud at 24/96 FLAC. That way, if you liked what you heard, you can hear it in full-on serious mojo high resolution awesomeness. Those captures are NOT dithered, but they’re direct captures so it’s not like I have a higher resolution file to dither from: you get the jams hot off the MOTU 16A inputs, untouched. Sort of a ‘analog in digital bar’ to reach, in a situation where every little calculation takes a toll.

Date December 23, 2018
Author Chris
Comments 6 Comments

Chris

StudioTan

TapeFat

Crystal

RawTimbers (and RawGlitters)

handsewn bespoke digital audio

Kinds Of Things

The Last Year

Patreon Promo Club