KVR Audio

soundmodel · Post by **soundmodel** » Mon Apr 25, 2016 8:07 pm

If you have WDL-OL, then I can share the code.

To me it seems like everything's ok. I didn't build with Python in order to run the examples.

Miles1981 · Post by **Miles1981** » Mon Apr 25, 2016 8:17 pm

I have WDL-OL (my plugins are based on the framework), but I don't have Windows at the moment (on vacation), so it may take some time though before I can test it properly.

soundmodel · Post by **soundmodel** » Mon Apr 25, 2016 8:20 pm

Well I can make a quick XCode project as well.

Miles1981 · Post by **Miles1981** » Mon Apr 25, 2016 9:02 pm

It's idl-ol, so I should be able to figure it out

Send me a PM with a link, and I will have a look in the next few days, it's in my best interests that problems with my library are solved!

soundmodel · Post by **soundmodel** » Mon Apr 25, 2016 9:03 pm

So do you want only the code or a working Xcode project?

Miles1981 · Post by **Miles1981** » Mon Apr 25, 2016 9:10 pm

The code will be enough, I can work out a idl-ol project out of it.

soundmodel · Post by **soundmodel** » Thu May 12, 2016 11:25 am

Can you use ATK's filters so that you can process 1 sample at a time and write your own interpolation for parameter changes?

Miles1981 · Post by **Miles1981** » Thu May 12, 2016 1:27 pm

Yes, you can of course, just say process(1) on your pipeline and then update the paraters the way you want. But I would suggest to use several samples each time like the small snippet I put on github.

soundmodel · Post by **soundmodel** » Thu May 12, 2016 4:40 pm

What's the reason for using several samples? Speed?

camsr · Post by **camsr** » Thu May 12, 2016 5:05 pm

Miles, could you explain what the atk-rel filter with hysteresis does? What does the hysteresis modify?

Miles1981 · Post by **Miles1981** » Thu May 12, 2016 5:26 pm

Fluky wrote:What's the reason for using several samples? Speed?

Yes. You don't have the virtual calls overhead and you can do vectorisation. And also less parameters recomputations (they can cost a lot).

Miles1981 · Post by **Miles1981** » Thu May 12, 2016 5:28 pm

camsr wrote:Miles, could you explain what the atk-rel filter with hysteresis does? What does the hysteresis modify?

There are 2 hysteresis, actually:
- attack, more or less just for the sake of being there, it only increases the output when the input is above the last output * the hysteresis factor (with a slow AR filter, the output never goes to the maximum value, so less interesting)
- release, mimicks a sustain, as the release is decreased only if the input is lower than the last output * the hysteresis factor.
Full description and graphs on my blog: http://blog.audio-tk.com/2015/09/29/aud ... ysteresis/

camsr · Post by **camsr** » Thu May 12, 2016 7:45 pm

Interesting. What do you mean by increased and decreased with attack and release? The amplitude of the output?

Miles1981 · Post by **Miles1981** » Thu May 12, 2016 9:32 pm

Yes, I meant apply the update function with the attack or release respectively. It's possible that no update is done if the attack hysteresis is 1 and the release is 0.5 for instance.

camsr · Post by **camsr** » Thu May 12, 2016 10:23 pm

What isn't coming to mind immediately is (refering to your graphic) why the attack stage holds quite a bit below the peak crest. Something about this doesn't seem right... I think it's related to a lack of a set threshold for the state change. This is how I know hysteresis, but it may be quite simplified,
True if greater than 1, False if less than 0. This should describe a state change that follows crest of the input. But now thinking about how there is a filter involved in your equation, the strange hold levels may be the result of phase lag, and I may have answered my own question

AudioTK library