Airwindows

Righteous4

TL;DW: Final output stage for targeting dynamic range.

This one’s pretty eagerly awaited by some… all the more since I’ve incorporated all the latest tech, such as the changes in denormalization and noise shaping to the floating point buss (it uses the ‘more warmth, ease and depth’ version when set to 32 bit) and even the Spiral algorithm in its overdrive section.

But you might not end up using that noise shaping, because Righteous4 has NotJustAnotherDither (and the CD version) built right into it, along with an optimized ADClip and code from an old strange experiment I did called ShortBuss. Yep! Righteous4 is your all-purpose final output stage, which handles final clipping, saturating and even the dither to let you output 24 or 16 bit fixed point in the purest, most optimized way. It doesn’t even return to the DAW buss between these stages: everything is ‘long double’, so you put in your mix audio (from conservative levels to peaks that would clip a normal output file: it’ll soak up the peaks) and you get out literally the data you burn to the CD or upload to Bandcamp etc. as hi-res HD audio.

But it’s a little more complicated than that. Bear with me. Righteous4 will clog up and audibly distort under some circumstances, and it’s for a reason that might interest you.

Many music services these days (possibly all?) incorporate RMS loudness targets. They’ll turn up your stuff and limit it (maybe) which is awful but outside your control… but if you’re doing the loudness wars thing, they will turn you DOWN. And all your efforts to make loud masters will be wasted. Apple will cut stuff back to around -17 LUFS. YouTube goes for -13 LUFS. Spotify and Tidal do -14 LUFS. That’s loudness units short of full scale (similar to how many dB down from clipping your average (root-mean-square, or RMS) loudness is). Loudness war folks often push to -8 LUFS or even more, and that means YouTube will turn you down (and Apple will turn you WAY DOWN) to fit in with their playlists.

Here’s the thing: you can have all the peak energy you want (caveat: Apple doesn’t want your bright treble peaking over -0.4 LUFS in any case, so don’t go too bonkers with brightness when aiming for Apple music services). Peak is not RMS. In fact, music sounds great with lots of peak energy pervading it: this is one of the reasons old vinyl records sounded so great! I’ve measured RMS loudness behavior like -27 LUFS off old hit records. The energy pouring out of those old grooves is due to the way peaks and compression were handled: it was a different kind of limiting, handled differently, back then. Peak energy makes the music sound better. Righteous4 handles peak energy by clipping it with ADClip, so you can throw more at it in safety. This is not the audible clipping I’m talking about (at least, not if you’re talking percussive peaks etc. which are pretty cleanly clipped, especially by ADClip)

The audible clipping happens when you push your LUFS levels beyond the target. As you mix and set levels, you simply have to listen for whether Righteous4 sounds big and open, or whether it’s getting a little intense, or whether it’s obviously being pushed too far and breaking up (especially on bass, and mix fullness).

What’s happening is, the saturation from the Spiral algorithm is being fed into ShortBuss, which fills the energy back into the mix in the form of second harmonic. This is what gives Righteous4 its tone and extra warmth, and it’s a nice fullness and bloom UNLESS you’re slamming it (it’s calibrated using the slider, where you select your target). If you are too loud, you’ll hear it. Your peak energy will still be going up to the real clipping threshold, just as before, but your RMS energy will tell you how loud you should go.

It’s that simple. You don’t have to use it to hit LUFS targets, but since it’s got Spiral in there it also broadens the heck out of the ‘target range’ so you can pick whatever seems right to you, and it’ll guide you. If you find you need to push louder, you can set Righteous4 for a higher target, and that will clean it right up and let you push harder into the internal ADClip. If you’re looking for big dynamic range, you can set Righteous4’s target lower, or simply set it for Apple or YouTube loudnesses and then just don’t push into it very hard: this will give you dynamics, but the replay-gain services will be kinder to the result because it’s somewhat saturated and compressed, just in a very gentle way if you’re hitting it that gently. And of course the output’s optimized for whatever bit depth you’ve selected, so you can directly target CD or 24/96 or ’32 bit file to send to further mastering’ if you still need to mix into something more organic than a hard, brittle digital mix buss with all the charm of math :)

I hope you like Righteous4, and it looks very much like we have hit another sort of threshold: since the Patreon looks to stay above $1000 as we enter June, you will now get two ‘list’ plugins a month from now on, beginning with Energy, then VariMu, then Ditherbox and so on: about two weeks apart, not just one a month! That should keep things exciting. I’ve also got Distance2 coming soon, with the ‘two different kinds of processing’: you remember, that’s the one that uses the new algorithm in Atmosphere but lets you crank it up to crazy distortion, and also lets you roll off the highs independently. Thank you! I’ve got plenty of stuff in the pipeline and I’ll keep new stuff arriving too, we’ve just doubled the speed of the ones listed on Patreon. Hopefully that will make this an even more exciting story to follow, because it helps me when new people hear about what I’m doing and join the Patreon! If you go there and look, you can now see a schedule that includes the two top-of-the-list plugins AND the bonus one: that’s three plugins a month not counting new work, and it’s now organized by month in order so you can see what’s dropping when, from now until Xmas!

Also it’s my birthday Tuesday. So I gave you a present <3 hope you like Righteous4, the first use of the Spiral algorithm in a plugin. :)

Date June 3, 2018
Author Chris
Comments 6 Comments

Golem

TL;DW: Blend a stereo track of two mics on an amp.

Golem

And back to earth we go, but some people will love this little utility plugin.

Golem takes in a stereo track, typically two mics on a guitar cab (a popular technique among metalheads), and mixes them to mono in the middle of the track.

But wait, there’s (very slightly) more!

Golem lets you balance the respective inputs so you hear more of L or R, and most importantly, Golem lets you apply a delay to only the side that you want to delay. The other side is always no-latency, straight through, but the one you’re delaying gets its little sample delay. Either one, it automatically switches.

Almost done…

The way this control is implemented, means that small adjustments give you a tiny, tiny linear-interpolated delay. I think linear interpolating is best for guitars because it’ll scrub off the tiniest amount of ‘hiss and rattiness’ while giving the least processed sound. That said, interpolating between samples in the way this does, allows you to dial in the subtlest imaginable phase relationships between your mics, and THAT has profound effects on the subtleties of your miked guitar tone. You are tuning where the cancellations between mics are, on a very fine level, in order to do specific things in the sound, and you need easily controllable subsample delays at your fingertips. Enter Golem, your simple tool for exactly that.

And THAT is the point, and also why this plugin has some cool marquee fans who swear by it. (Slippy, can I repost the ‘All Things Slippery’ mp3? I think there are lots of people who need to hear it, and it was a really soulful time in your life. I hoard it, since it fell off the Internet. But I digress)

There are some options for allowing a larger range of delay, or inverting one of the channels, but it’s basically just mixing two tracks and being able to micro-delay one of them. Some folks will be deeply unimpressed because you have to be a bit of a fanatic to get that worked up about delaying one of the guitar cab mics five-eighths of a sample in order to place a cancellation node JUST EXACTLY where you need it to be.

But you’re talking to someone who noise shapes a long double to the floating-point buss, so yay fanatics! You know who you are. This plugin is for you.

The Patreon is on the verge of another goal: if it is over $1000 when June begins, firstly that’s a cool birthday present (June 5!) and secondly it means you’ll get Righteous4 (with the new Spiral distortion algorithm) AND Energy in June, because >$1000 means I release two from the list each month from then on. There should still be time to include the new work I do, but the backlog goes twice as fast. Now, it probably won’t stay over $1000 because Patreon always has people drop off when billing starts, and that’s okay. I’m just saying, that goal is coming this month or next, and that means you start to get the ‘list’ released faster. And Righteous4 is next, regardless. (It should be possible when the list is all done, to go and start putting out specific legacy versions, including ones with legacy sound-code: sometimes people want things like the old bizarre antialiasing routines I was doing around 2007)

I think really the best birthday present is all of you. Thank you <3

Date May 27, 2018
Author Chris
Comments 5 Comments

Spiral

TL;DW: New best smoothest distortion algorithm.

Spiral

What if… the building-block sine-based saturation routine I build so many things on, which I thought was the smoothest distortion you could have, is NOT the ideal distortion?

I was watching a youtube video by Brady Haran, on his ‘Numberphile’ channel. It was about the golden ratio, and it showed the little drawing you’ve maybe seen, where you take a golden rectangle, and then make a square on it and put a circle segment (like when I use sine curves for my distortions), and then make a smaller square next to it and a circle section on that which connects, and so on ad infinitum. A ‘spiral’ like a seashell, a golden ratio spiral.

And Brady says, ‘of course this isn’t actually a spiral’ and my mind: blown.

Because of course it isn’t. It’s circle pieces butted up against each other, and only looks like it’s connected because it’s pretty flat going through those points. But the rate of curvature changes really suddenly and drastically at those points… and does it the same way, at the zero crossing of ConsoleChannel, Density, PurestDrive, PurestWarm, Channel, and lots and lots of other plugins I’ve made. It’s part of the ‘fatter, smooth’ sound of some of my distortions. Seemingly really fat and analog-sounding, but there’s a discontinuity as you go through zero, which is why it makes the audio sound obviously different.

People do like making the audio sound different, but people HATE the sound of discontinuities. An old version of Channel where I hacked in a ‘flat’ undistorted section, got me a huge blowback of dismay and outrage until I put it back to the simple sine-based one. People are incredibly sensitive to second order discontinuities, where the output number will be 0 but the direction things are changing will suddenly be the opposite. That’s what made the old ‘New Channel’ be such a failure: the center of the wave was flat no-distortion, and then without making a visible discontinuity it would suddenly change to ‘tighter sine-based saturation curve’, both on the negative and positive sides of the wave. It would look perfectly normal but people just hated the sound. Now I know why. And now it turns out the sine based one HAS a discontinuity, at the zero crossing, right where you’d get class AB and B distortion, and it never occurred to me.

And I can fix it, and turn the code for that part into ONE line of reinvented original code, which will be open source because I’m Patreon-supported and don’t have to stop people using my good audio code (they only have to credit me when they use it: it will be very possible to tell when they do. They don’t have to pay anything since it’s an MIT license, and they don’t have to open their own source, just publically credit that they used my Spiral code).

And of course I did: here, have Spiral, free. You can just install this and listen to it, if the video and the post are too long. This is the proof of concept which can be used in several useful ways. There are no controls, at all. You can gain stage into it and do stuff with it and sit it on the top of every track like it was PurestDrive or Channel, or use it as a 2-buss clipping stage, where it will clip to around -1.4dB with about -0.4 intersample peak maximum (so it is Mastered For iTunes friendly, used as a final clip). It also has a ‘freak out mode’ if you massively overdrive it, and you can do that with things like uncompressed drums, and it makes noises you’ve never heard before, or when used more gently it just sounds like the ultimate analogification.

I really had no idea I could do a basic saturation algorithm (which still uses long double precision sines as part of it) that was that good. It’s a considerably bigger sonic improvement than the new noise shaping technique, because what it does is on a far higher level… though of course it also is using the new noise shaping, for good measure. All the latest everything, right here.

Bear in mind, the original sin()-based one in Console5, PurestConsole etc. is still optimal for Console5 encode/decode because it can be lossless and has significant effect at low levels. My tape emulation stuff uses the ‘fattening’ effect of that on purpose. I can’t just go through and replace everything because all the sounds will change. I have to re-voice everything that would take advantage of the new code, and I’ll do it, and it’ll take time and effort, and probably become new versions of things so you can still have the ‘sine fatness’ versions if you want them. I don’t like taking sounds away from people, and old tones shouldn’t be removed or made inaccessible.

But what if there’s a WAY better analog-sounding distortion effect based on the way that the ‘constructed’ golden ratio seashell/spiral made of circle sections in boxes that get smaller, is NOT correct because much like the simplest sine-based overdrive, it doesn’t start with zero distortion but with the same tiny distortion the whole time (which, in joining to the opposite pole, makes a discontinuity you can’t see much like the golden ratio ‘spiral’ has discontinuities you can’t see?) …and I fixed it?

Try Spiral, and I will get to work incorporating this into my library of audio plugins that I make available using my Patreon. I’ve put out more than 100 plugins for you to freely use, since I started doing it. If even one of them is so useful that you’d have paid $50 for it as a commercial plugin, please join my Patreon at the equivalent of $50 a year. Or, you know, if you think I do good work and I’m better off continuing to do this stuff full-time with awesome resources, and not taking a day job or trying to fix my own porch with time I should spend coding stuff for you. It’s as simple as that.

And yes, my porch is still broken, but this has been a good day. :)

Date May 20, 2018
Author Chris
Comments 11 Comments

Channel5

TL;DW: Channel for 2018, with new subtleties.

Channel5

Channel has long been a sort of test-bed of mine, and so when I started working with new methods for some of my stuff it was an obvious choice. I’ve been asked for an extra Output level control, which it has (for those of you who are really into level matching, can’t fault you for that though sometimes I think it’s not relevant) but if you set Output to 1.0 it bypasses itself and becomes exactly, exactly the same as Channel4.

Except for two things, which first appeared in Atmosphere: this lets you A/B them with the original versions, using Channel 4 and 5.

One, I introduced the concept of live air denormalizing: before this (and on plugins I think are suited to it), Airwindows plugins watched for total digital black silence, and if it was present, produced an impossibly faint hiss sound, at around 300 dB down. If there was any audio at all that wasn’t digital black, this got switched off. With Atmosphere, it’s different: the denormalizing signal is always there, and if the result WITH the signal is digital black, then the signal’s subtracted again. Same result, but there’s a permanent noise hundreds of times quieter than typical ‘noise adding plugins’. The idea is that there can be no ‘dead’ silence, only ‘alive’ silence: you can’t save the file in normal formats and retain it (24 bit audio saving will simply turn this noise to digital black again), you most likely can’t hear it ever no matter what you do, but I wanted the deepest darkest silence to be alive rather than dead black nothingness.

Two, it noise shapes to the floating point buss differently, and on the one hand I bet you can’t consistently hear it but on the other, I think it matters. Before, I used an interleaving technique much like I often use in IIR filters and plugs such as Capacitor. It related to the sample rate by forcing everything to be considered as the interaction between two or more samples, and the energy being brought back into the output audio provided corrections at high frequencies. What began with Atmosphere, and what you’ll (maybe) hear with Channel5, is a completely different and more radical approach. These plugins noise shape ALL the error energy back into the audio. It accumulates, sticks around, and so it can contribute to bass energy in tiny amounts because it persists over thousands and thousands of samples. There’s a ‘pull-back’ in the form of multiplying the amount by something like 0.999999, which will bleed built-up energy away, but this multiply doesn’t happen on a per-sample basis. It happens per BUFFER. So, in the world of noise shaping floating point error, this is the one that stores up every tiniest bit of error and applies it, not interleaved but very smoothly, to the audio.

What should happen as a result of this is, there’s a sense of ease from switching from interleaved to the smooth way. It might seem a little less lively, and I’ll have some plugins that retain the other way. But due to the greatly increased time during which the error is held, plugins that use the new noise shaping will be crazy good at presenting analog-like, seamless bass and mids. So, more soothing, possibly less exciting, but a huge jump in depth and naturalness.

…in RELATIVE terms to what was there before. That’s the thing. I feel it’s possible that when you get used to it the change can be dimly sensed as a good thing. I don’t think you, or I, or anyone would be able to pick this out double blind. I can think of a mastering guy or two that MIGHT, but they’d struggle: these are people who take pains to dither to 24 bit and say that they sense something wrong if they don’t. It’s on the order of that, or even more subtle.

But I do feel that one eventually senses the full range of a sound one works with, and more importantly if I can make something better I will do so, without hesitating.

Have fun playing with Channel5. And if anyone needs to sound the alarm and say HEY! You need to stop using the new stuff! That noise is driving me crazy, and your noise shaping sounds awful now!

…well, that’d be pretty funny :)

All this is supported by my Patreon. I’ve put out more than 100 plugins for you to freely use, since I started doing the Patreon. If even one of them is so useful that you’d have paid $50 for it as a commercial plugin, please join my Patreon at the equivalent of $50 a year. Then, next year, we’ll see if I’ve done just one additional plugin that you’d also buy as a commercial plugin. I might make as many as 52 more by then, so the odds are kind of in my favor :)

Date May 13, 2018
Author Chris
Comments 6 Comments

Chris

Righteous4

Golem

Spiral

Channel5

handsewn bespoke digital audio

Kinds Of Things

The Last Year

Patreon Promo Club