Re: [webaudio] possible error in DRC processing graph

Yep, the code definitely reads such that the knee comes after the
threshold. Sees a bit strange as it gives that boost relative to the
threshold, but it makes sense from a graphing standpoint. Thanks for fixing
this.

Curtis


On Mon, Nov 16, 2020 at 8:49 AM Paul Adenot <padenot@mozilla.com> wrote:

> That was my expectation as well (and what you're saying sounds logical to
> me), but we've largely inherited code that was already running, and had to
> retro-specify it.
>
> Here is the code we had to work from, for reference:
>
> -
> https://searchfox.org/mozilla-central/rev/ff82c973f8ccb0475ec32439e9ec07014b3a681f/dom/media/webaudio/DynamicsCompressorNode.cpp#88
> -
> https://source.chromium.org/chromium/chromium/src/+/master:third_party/blink/renderer/platform/audio/dynamics_compressor_kernel.cc;l=189;drc=d0e87c5b5ac8f1d884f5899f4b63967fea784d07
>
> the WebKit code was forked by Mozilla long ago because there was no
> specification, and it was already shipping in Chrome and Safari, and we had
> to ship a compatible DynamicsCompressorNode. Style has diverged but this
> bit is the same.
>
> Hope this helps,
> Paul.
>
> On Mon, Nov 16, 2020 at 4:44 PM Curtis Malainey <cujomalainey@google.com>
> wrote:
>
>> Thanks Paul,
>>
>> I am not 100% sure about the second graph. Isn't the knee the width
>> around the threshold? The first graph in my mind looks correct if you just
>> move the threshold forward to the hard knee so the curves remain the same
>> after the knee. Otherwise it looks like you are factoring in a boost post
>> threshold which seems to be the opposite of a compressors purpose. See
>> slide 10 (ignoring overall boost)
>> http://c4dm.eecs.qmul.ac.uk/audioengineering/compressors/documents/Reiss-Tutorialondynamicrangecompression.pdf
>>
>> That being said, don't take my word at face value, I am trying to fix
>> some of our compression curves and making sure I understand this enough
>> before I do it.
>>
>> Curtis
>>
>> On Mon, Nov 16, 2020 at 3:53 AM Paul Adenot <padenot@mozilla.com> wrote:
>>
>>> Yes, you're right, I don't know why I did it like that in the first
>>> place.
>>>
>>> Thanks for the report, I've updated with a new proposed diagram:
>>> https://github.com/WebAudio/web-audio-api/pull/2277
>>>
>>> Paul.
>>>
>>> On Mon, Nov 16, 2020 at 6:11 AM Curtis Malainey <cujomalainey@google.com>
>>> wrote:
>>>
>>>> Hello w3
>>>>
>>>> I was browsing the DRC web audio spec and I believe there is an error
>>>> in the graph in section 1.19.4 figure 14. The graph denotes the difference
>>>> between a hard and soft knee. But shouldn't a hard knee be on the threshold
>>>> marker? The figure has the threshold marker at the beginning of the soft
>>>> knee.
>>>>
>>>> Curtis
>>>>
>>>