KVR Audio

Ivan_C · Post by **Ivan_C** » Sat Aug 02, 2014 11:56 am

If someone is interested about the sound of the filter I have talked about before, you can hear it in my KVR DC contribution, Guitar Gadgets

I did a 0df version too, but the sound quality improvement wasn't that significant, so I prefered the "mystran's method" version.

Urs · Post by **Urs** » Sat Aug 02, 2014 10:36 pm

We're currently wondering what to call the method. One of the guys is finalising a set of papers that explain and compare various filter solving methods. At the moment he calls it "Cheap Method", but I'm leaning towards "Mystran's Method"…

What do you think?

aciddose · Post by **aciddose** » Sun Aug 03, 2014 12:44 am

The mysterious method. Include a graph shaped like a wizard's hat.

mystran · Post by **mystran** » Sun Aug 03, 2014 1:24 am

Urs wrote:We're currently wondering what to call the method. One of the guys is finalising a set of papers that explain and compare various filter solving methods. At the moment he calls it "Cheap Method", but I'm leaning towards "Mystran's Method"…

What do you think?

If I had to give it a "proper" name, I'd probably call it "fixed-pivot one-step linearization method" or something similar, since that's what it's about: draw a line through a chosen fixed pivot point (typically origin) and your initial guess (from the previous step), and just use that for solving the system.

I'd rather not have you call it "Mystran's method" since I have a bunch of other methods and I don't really think this is necessarily the best one.. although it's probably the cheapest one so if you insist on attributing it to me then calling it "Mystran's cheap method" would be fine I guess (although.. as I've pointed out previously, there is prior art so I'm not really the first one to come up with this approach).

Z1202 · Post by **Z1202** » Sun Aug 03, 2014 7:47 am

Urs wrote:We're currently wondering what to call the method. One of the guys is finalising a set of papers that explain and compare various filter solving methods. At the moment he calls it "Cheap Method", but I'm leaning towards "Mystran's Method"…

What do you think?

I'd also vote against "cheap method", if only to avoid confusion. In my works and posts I'm referring to applying nonlinearities on top of the linear solution as the "cheap method". Although, confusion-wise, I'd guess "mystran's method" is the most clear name

Urs · Post by **Urs** » Sun Aug 03, 2014 8:57 am

mystran wrote:If I had to give it a "proper" name, I'd probably call it "fixed-pivot one-step linearization method" or something similar, since that's what it's about: draw a line through a chosen fixed pivot point (typically origin) and your initial guess (from the previous step), and just use that for solving the system.

How about "Pivotal Linearization" method then?

The one that has its pivot at origin would be the cheapest, while other versions place the pivot at the previous result (which then makes it something like a single step Newton's method).

Hmmm… need more coffee...

Urs · Post by **Urs** » Sun Aug 03, 2014 9:02 am

aciddose wrote:The mysterious method. Include a graph shaped like a wizard's hat.

Nah, no magic - only divine stuff. We have lots of crosses and Ankh-shaped graphs instead.

andy-cytomic · Post by **andy-cytomic** » Sun Aug 03, 2014 9:29 am

mystran wrote:
Urs wrote:We're currently wondering what to call the method. One of the guys is finalising a set of papers that explain and compare various filter solving methods. At the moment he calls it "Cheap Method", but I'm leaning towards "Mystran's Method"…

What do you think?
If I had to give it a "proper" name, I'd probably call it "fixed-pivot one-step linearization method" or something similar, since that's what it's about: draw a line through a chosen fixed pivot point (typically origin) and your initial guess (from the previous step), and just use that for solving the system.

I'd rather not have you call it "Mystran's method" since I have a bunch of other methods and I don't really think this is necessarily the best one.. although it's probably the cheapest one so if you insist on attributing it to me then calling it "Mystran's cheap method" would be fine I guess (although.. as I've pointed out previously, there is prior art so I'm not really the first one to come up with this approach).

The correct name for such a method is "semi-implicit". In this case it is "semi-implicit trapezoidal" that uses f(x+c)/(x+c) instead of f'(x) as the g / derivative / conductance term, but the integration is still trapezoidal since the integration doesn't care about how you calculate the "derivative" of non-linear terms. So how about "semi-implicit fixed-pivot trapezoidal" ?

mystran · Post by **mystran** » Sun Aug 03, 2014 2:10 pm

andy-cytomic wrote: The correct name for such a method is "semi-implicit". In this case it is "semi-implicit trapezoidal" that uses f(x+c)/(x+c) instead of f'(x) as the g / derivative / conductance term, but the integration is still trapezoidal since the integration doesn't care about how you calculate the "derivative" of non-linear terms. So how about "semi-implicit fixed-pivot trapezoidal" ?

Sounds good to me.

antto · Post by **antto** » Sun Aug 03, 2014 2:14 pm

mystran's SIFPT method

Ivan_C · Post by **Ivan_C** » Sun Aug 03, 2014 5:09 pm

So, Guitar Gadgets use mystran's semi-implicit fixed-pivot trapezoidal method to simulate the EMS VCS3 filter, I said it

aciddose · Post by **aciddose** » Mon Aug 04, 2014 12:28 am

I like how we're nearly at the point where the bullshit marketing for plugins released in the late 90s is almost on par with what we can truthfully say about our plugins today.

xh3rv · Post by **xh3rv** » Mon Aug 04, 2014 1:03 am

I'd say call it Mystran's Pivot and be sure to properly define/categorize it somewhere.

Z1202 · Post by **Z1202** » Mon Aug 04, 2014 8:35 am

xh3rv wrote:I'd say call it Mystran's Pivot and be sure to properly define/categorize it somewhere.

I like that. This is very clear and doesn't cause any confusion (at least for those reading KVR's DSP forum regularly

) Besides that, even if Mystran wasn't the first one to introduce this method, he definitely was the first one to promote (and possibly even to apply) its usage in the field of VST plugin ZDF filter development, and this effort also deserves credit.

Urs · Post by **Urs** » Mon Aug 04, 2014 9:02 am

Well, the paper is supposed to be released in this forum, so I'd definitely give Mystran a credit.

Cheap non-linear zero-delay filters