Sure!masterhiggins wrote:I hear a lot of big talk regarding the superior sound of hardware. You know what would kill this thread pretty quickly? A blind test.
-Sam
http://xhip.net/temp/match_this.mp3
Plain bandpass filter applied to PWM wave. Nothing else special going on here.
Now to do a blind test you'll probably need to match it at least close enough to satisfy a deaf test first. Not that this is impossible, but it is extremely difficult.
I'm able to reproduce it in software, but only approximating it of course. Given enough time and effort (and cpu power) it is easy to run the exact model. However actually producing an accurate model to run in the first place or coming up with an approximation that comes close enough in general to pass a blind test is not an easy task.
Urs said they're still working toward modeling a relative of this filter (KORG35) and described that it is extremely difficult to do so and may never be done. I'm not sure whether I know more about it myself or how much Urs may be aware, but I agree entirely with his assessment. I think however that it isn't a question of difficulty in producing a model alone but rather it's a question of the excessive amount of processing power required to accurately model this. Given one core per filter it is entirely possible with current high-end CPUs.
This is just an example of one case where it is obviously of great benefit to be running this particular filter as hardware rather than software. In software, regardless of where it is implemented it is likely to be unreasonably expensive. The electronics of the filter itself are worth only $5 and hundreds (thousands?) of these can be packed into a PC-sized box running on about 100-watts. In software you'd be burning in excess of that, worth $100s for one filter.