• dev

AT Behind the scenes

  • 17
  • 269

I thought it might be a good idea to write a blog-like thread where I record what I'm currently working on. I don't know if this is a great idea and if I find the time and motivation to keep this alive for long. Most of this will be quite techy or boring but maybe you find some of the information interesting.

Create an account or to write a comment.

  • Why has there been a lot of AT maintenance lately? Is something new coming here?

  • After the look updates on portal side, is the "studio" still in progress? The audio quality has deteriorated significantly. It seems that everything sounds with extra stereo filter and the resonance is very bad. Frequencies are absurdly out of control. Is it by any chance a desired downgrade? Please Administretor and Staff, please help me move my comment to the correct link and support location. Please, the aesthetic and user friendly aspect is ok, but the audio quality is the first requirement as well as the supreme priority of a music platform. Thanks.

    • Neither.

    • I can't confirm this behavior.

    • no changes have happend in the studio that should trigger such massive issues.

      "should" - in theory, at least.

      and please kindly post your issues in Bug Report - thanks!

  • Thanks for putting the joining dates of users in full, the "joined about xx years ago" was slightly frustrating ;)

    • On the bright side, now we can see the exact date a track was published

    • Yes, this is a " side product " of AT download indicator extension.

    • Nevermind, you're right this is Audiotool downloads indicator that apparently does also that

    5 more
  • I'll assist Martin a bit with the redesign so he can concentrate on the important stuff.

    I have the honor of finalizing the wave-forms. The most complicated thing will be preventing riots a wrong shade of blue may cause. Let's make it rainbow colored, so everyone is happy.

    34 more
  • I'm currently working on the timeline control (seek, loops, stop at end, show playback position). I think this is the last missing piece to make the prototype easier to test.

    That and allowing to edit things in the timeline without requiring a reload (should do this next)

  • Got the first numbers for multi-threaded audio: Up to 10× speed for certain scenarios but I think there's room for more (not for every draft or machine ofc)

    • Mine has 12 cores (24 threads) and that's just a normal value for a recent AMD machine.

      You can use as many cores as you like but at some point it doesn't scale anymore. There's some overhead for deciding which thread has to do which part. For me 20 threads performed worse than 10 threads for some loads.

      I personally don't think that 10 is the upper limit, but I'll have to wait for a real-world application to see how far I can get and if it's worth the time I have to invest.

    • Would this theoretically scale with CPUs that have higher core counts? Or would 10 be the upper limit of what it could utilize in its current state?

      Genuinely curious, since my current CPU has 6 cores, and I'm looking into getting a new one with either 12 or 16 cores sometime kinda soon, depending on what happens.

    • 1 to 20. The test case was bad for more than 10 cores. As soon as I have a better test case, this should get better.

    1 more
  • Pulverisateur inside out - this is a block chart of how I implemented the module.

    • yep, less lag. Not saying I am GOD level pro at MAX but this is very familiar, been tweaking around with it for 2 years now.

    • It would be cool for some users to be able to switch to this " simple layout ".. we could make much larger projects without all these visuals that need to be rendered all the time :)

    5 more
  • If y'all want to help: please try out the "Worklet only" mode (Chrome only rn) and let me know if there are any issues. This is going to be close to the operational mode of the new engine. Watch out for glitches and latency.

    details see:

  • Little big News: Pulv is complete!

    I also did some tests with the new engine running in the browser. Looks good so far: no glitches and very good latency (for a browser). Performance could be better. WebAssembly likely behaves different from what I expected. I have a working profiler. As soon as bigger tracks can be played I'll have a look at the bottlenecks.

    We're looking into a way to make this available to some testers, so we can get early feedback.

  • was anything new just implemented? Or did something go horribly wrong at AT HQ?

    • @Known As I oof, that sounds like a potential nightmare.. Happy to hear you guys keep good backups though!

    • @naut multiple disks died at the same time without S.M.A.R.T. giving us any notification in advance. Maybe a faulty hdd firmware or spikes in the power supply. That's what backups are for ;)

    • do you guys do your own hosting for certain services? no s.m.a.r.t. monitoring either?

    3 more
  • External Audio Input is also available now.

    The filters will be next.

  • The Pulv's noise-source is complete now (except for a combing issue which should be easy to fix).

  • Except for the waveform, osc1 is working now (panning, tune, octave, level, master level).

    • Sync and Kb-Track are working fine now.

    • No changes to the devices are planned for the first release. Some minor improvements for previously undefined behaviour or extreme cases are ok in my opinion. Non-audible changes are ok as well, if they improve internal stuff.

    • Thanks. I was wondering if this translation of the code would be used as an opportunity to implement improvements to the DSP. I know that the Heisenberg was rumored to need upgrades (mostly band limiting as well and improvements to the waveforms), or the issue with Slope's 10K cap in high pass mode. You're right that noise probably doesn't need extra band limiting. The generation mechanism probably implements it already.

    4 more
  • Midi integration is complete (it's easier to test a synth this way).

    I can finally play some boring sine waves on the pulv, but they don't stop voluntarily rn.

    I guess the amp-envelope would be the next logical thing, so the notes don't play forever.

    • Found the bug. Amp-envelope is now fully functional (including the decay-loop).

      Next up: make oscillator 1 sound non-sine-ish

    • The amp-envelope basically works - modulo a few minor bugs.

      Decay-looping is implemented but requires some testing.