The first part lets look at ringing characteristics Above is a impulse response at 44.1 going through three instances of dualtech v2, this is made by passing a 1 sample pulse through the instances. Excuse the cursor in the middle. It's unbelievably clean! Like it never went through a convertor! Just a nice analog ring in the tail (transformer ring and eq ring) The lack of convertor ringing ensures incredibly good transient response.
Below are generic references to convertors (with linear phase filtering) and analog impulse response
Here, is a introduction to this part of another tech. here a very well isolated bell curve
The idea is we are only sampling the bell curve, so we don't stack up on the same signatures. The new technique insures to get the cleanest isolation possible. As you can see everything around the bell curve is flat.
So everything fits together will fit like a puzzle over all the instances. No repeated phase, frequency or ringing! We simply add a line instance over the top of the isolated bands giving a accurate emulation.
This is possibly the first time it's ever been seen in convolution!
Here, a phase plot of a high bell single band. This shows a very clean Total isolation of a single band. Notice the phase is zeroing from the low end.
myimo wrote:"The idea is we are only sampling the bell curve, so we don't stack up on the same signatures."
Genius idea looks like you have managed to do it.
The old way is to minimise the amount of times we go through the hardware, and or try to isolate as much as possible(i.e 1102 I isolated input transformer for eq sampling)...and trying to combine bands in sampling where possible, for example silk eq was great for this reason, because the whole eq is in two instances.
Now I can get a very good isolation, that will be very good for making a whole eq much truer. There is a whole list of improvements. In fact this next release will be like it never traveled through conversion, and thats even more important when stacking lots of instances.