Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[orangeart]

Can anyone point me in the direction of some instructions on the setup parameters of the config files for convolver? I've found some examples using search and have extrapolated some of the rules but I'd like to know everything if possible and I can't seem to find a definitive list.

Cheers

Stefan

[mwillems]

The config file format is described here: http://convolver.sourceforge.net/config.html

[mattkhan]

It might be easier if you state what you want to achieve and then people can comment on how you would do that.

IIRC jriver doesn't support one of the delay params, can't remember if input or output though. The answer will likely be somewhere in this thread - http://yabb.jriver.com/interact/index.php?topic=68828.50

EDIT: this post covers the features not supported, don't think it has changed since then -> http://yabb.jriver.com/interact/index.php?topic=68828.msg465919#msg465919

[orangeart]

Thanks Matt, that's brilliant stuff.

In answer to your question, over the last few days I've been gradually finding my way round JRiver demo, I've worked a few things out with regard to active crossovers and this is the next step really. I currently use a minisharc from minidsp and crossover filters which I generate as FIR coefficients in re-phase which are loaded onto the sharc. I want to be able to recreate this kind of setup in JRiver so need to be able to load filters per channel and ideally per sample rate into the convolver to achieve the same. I would need to use per channel delay though so hopefully that's supported.

Loving this forum, very helpful indeed

Stefan

[mwillems]

Thanks Matt, that's brilliant stuff.

In answer to your question, over the last few days I've been gradually finding my way round JRiver demo, I've worked a few things out with regard to active crossovers and this is the next step really. I currently use a minisharc from minidsp and crossover filters which I generate as FIR coefficients in re-phase which are loaded onto the sharc. I want to be able to recreate this kind of setup in JRiver so need to be able to load filters per channel and ideally per sample rate into the convolver to achieve the same. I would need to use per channel delay though so hopefully that's supported.

Loving this forum, very helpful indeed

Stefan

Per channel output delay is supported in the convolver, but JRiver can also delay channels in the parametric EQ module before or after convolution.  

Depending on your needs and plans you may be able to do many of the things you plan to do in the convolver through PEQ in JRiver which may have some benefits if you plan to use this setup for live playback. You mentioned trying to get latency low enough to use netflix in one of your other threads; that can be hard to do if you're doing all your EQ and routing with convolution filters.

At minimum you'll need to keep your tap count low for live playback, which depending on what you're trying to achieve, can be a non-issue or a deal-breaker.

[orangeart]

Hi mwillems,

it's just been my preference, I've found that if I perform my crossover / EQ etc some other way than FIR filters I have more phase to correct (obviously) and end up with just as many taps, that and I really like using rephase, it's an excellent bit of software. I've found that I can keep most of my tap counts below about 2048 which keeps the latency to a minimum for video and is how I've been running my MiniSharc and hence my other threads about keeping other latencies to a minimum, there is little point in doing this if other bits of the software or plugins etc introduce more latency than is desireable. I am hoping to experiment with steeper slopes which will probably require more taps but then maybe I can set up a different zone that uses less taps for video and another that uses uber taps for just audio, is that doable?

Stefan

[mwillems]

Hi mwillems,

it's just been my preference, I've found that if I perform my crossover / EQ etc some other way than FIR filters I have more phase to correct (obviously) and end up with just as many taps, that and I really like using rephase, it's an excellent bit of software. I've found that I can keep most of my tap counts below about 2048 which keeps the latency to a minimum for video and is how I've been running my MiniSharc and hence my other threads about keeping other latencies to a minimum, there is little point in doing this if other bits of the software or plugins etc introduce more latency than is desireable.

That's sort of my point; software will introduce some latency, and so will hardware, so you may find that your total latency budget for convolution may shrink a bit, meaning you might need to do "more with less."  2048 taps may be fine even when added to other unavoidable latencies once you get everything optimized, but it depends quite a bit on your soundcard. 2000 taps at 44k is about 20 milliseconds, which isn't a problem by itself, but even 10 additional milliseconds from hardware/software would start making lipsync potentially iffy; 22 milliseconds is the film standard for unacceptable delay, 45 milliseconds for TV broadcast.

And that's where PEQ can really come to the rescue.  I can confirm from my own measurements that doing minimum phase filtering using PEQ and then just using RePhase for phase correction produces less total latency than trying to do the exact same thing entirely in rephase by a wide margin.  It may not be an issue for you if you can do everything you want in 2000 taps or less, but there are other advantages to doing the "basics" to PEQ as well (see below).

If you haven't already, you might want to check out my how-to on speaker correction in JRiver over in the Sound Cards Sub forum for some ideas on how a blended IIR/FIR approach might work in practice with JRiver: http://yabb.jriver.com/interact/index.php?topic=87538.0

I am hoping to experiment with steeper slopes which will probably require more taps but then maybe I can set up a different zone that uses less taps for video and another that uses uber taps for just audio, is that doable?

Yes, zones are the recommended approach and work fine for that.  In addition to long and short filter zones, I actually have a zone where I just turn convolution all the way off for certain kinds of latency sensitive live content (video games, etc.).  Because I run an active system with JRiver as the crossover, if I didn't do most of my crossover work in PEQ that would be impossible: doing what you can do in PEQ makes turning off the convolution altogether (when necessary) a viable option  

[orangeart]

This is all starting to sound very much like a plan. Your line of thinking seems spot on. I'll give it a go. Presumably the channel config remains the same weather or not I am using the peq?

Stefan

[mwillems]

This is all starting to sound very much like a plan. Your line of thinking seems spot on. I'll give it a go. Presumably the channel config remains the same weather or not I am using the pew?

Stefan

More or less; it depends, because you can use PEQ for channel routing too (which is what I do); that's key to being able to turn off convolution.

If I had to make a conceptual diagram of my current DSP steps it would look like:
Convolution module (phase correction of crossovers and port wrap (in advance)) --> PEQ Module (Channel splitting/routing --> interchannel delay ---> Crossover filters --> Minimum Phase Frequency Response Correction)--> DAC-->Speakers

It might seem counterintuitive to put the convolution at the front, but I measured and it made no difference whether I did the phase shift "first" or "last"; the author of rephase (Pos) gave me the idea when he said you could use rephase to correct the phase of a passive crossover in a speaker by inverting it's phase pattern before sending the signal to the speaker.  It turns out that it works the same way in most contexts (at least the one's I measured).

Doing Convolution first provides the added bonus that I could use a simple stereo convolution filter (instead of a config file) by convolving everything before the channel split.