levels in a multifilter

[Dan McDonald]

When you run a multifilter, you have the option of several filters in a row, and I believe we reduce quantization errors by doing the math once, rather than once for each filter. I was wondering - let's say you have a Virtual Valve and using the audiophile setting, which increases the bass significantly. If you are near 0 dB (that is, normalized to 0dB) and run the VVA, you will get overload distortion. If you have that VVA in a multifilter and decrease the signal either before or after the VVA, there doesn't seem to be any distortion resulting. Is that right? I guess I'm just asking if the decrease in loudness after the fact (but in a multifilter, so essentially simultaneously) actually takes care of the overload distortion.

[Craig Maier]

I will let Rick weigh in on this one so that a more precise answer is provided. Nonetheless, the math is still performed for each filter just as it would be if they were run one at a time. The difference is that the data signal is transferred with a high level of precison, not at the native file format resolution. Between filters, the data stream is not kept as 16 or 24 bits in resolution, but more like dual 32 bit floating point. So, as a result, there are (for practical sake) no real quantization errors between each filter operation by using the Multifilter. Each filter has plenty of headroom using the multifilter. Of course, you can still distort a signal, especially using the VVA as that is its basic job, but it should not clip the signal when surrounded by other filters in the multifilter. But, some filters can produce substantial gain, like the various EQ's and can result in iternal filter clipping. Those filters have virtual LED overload indicators and should be monitored in a manner to preclude overloading of said filter.

I asked Rick to respond to this question too, so let's see how he approaches the answer.

Craig

[Dan McDonald]

OK - great. Thanks Craig.

[Marc Hildebrant]

Interesting question.
I have adjusted the gain so that the various "effects" don't overload in a multifilter but if the math allows some headroom, then maybe more gain could be useful.

Marc

[Craig Maier]

Hi Marc,

I still would not allow a LED clip/overload indicator to light. That indicates an overload within a filter or effect and not within the communications link between them in a multifilter. But, let's wait for Rick's take on it.

Craig

[rickc]

I think that I do clip the signal between stages of the multifilter. I don't necessarily have to, but I see code that I specifically added for that case. So I must have thought that was important at some point. Doug is right in that the signal is kept as a 32bit fixed point values all throughout the signal chain when in a multifilter so you don't get the constant re-quantization that you would get if you ran each filter individually.

[Craig Maier]

Rick,

I think that you mean to say that you can clip the signal between stages, not that you do. Put another way and if I understand this correctly, the signal pathway between filters or effects within the multifilter have a 192 dB dynamic range (each pathway). So, although it would be hard to cause clipping, I guess it still could occur if many cascaded filters or effects are placed in a row with each having significant gain (other than unity).

Craig

ps - as an aside, gains represented as linear multiplication factors will multiply through in cascaded systems (a 2X gain cascaded with a 10X gain produces a total system gain of 20X). Gains represented in terms of db (log factors) add through in cascaded systems (a 6 dB gain cascaded with a 20 dB gain produces a total system gain of 26 dB.

[Dan McDonald]

Thanks everyone for the answers. I think I've got it.