A comparison of convolution engines

[HenrikEnquist]

I still think filter banks should be unlimited for camilladsp. It can easily implement an equalizer as well. And it has a loudness-compensated volume control that works the same way as the one in RME ADI-2 FS. Not exactly the iso226 curve but subjectively it gets quite close.
For special features one could mention:
- FIR filters can be chained, and mixed freely with IIR filters.
- It uses a yaml config that can be manipulated easily with for example python scrips.
- Can be remote controlled via a websocket connection.

Edit: also please remove "Convolver" from the name, it's just "CamillaDSP"

 

[mitchco]

@Keith_W output channel mask is not used... And you are missing my point, which I have made three times already. You as the user should not be writing a config file, period. The spec is designed for programmers to implement into their convolver and/or FIR filter designer software. so that you as the user don't have to do anything other than load the .cfg file which takes all of 2 seconds, including complex configurations like I noted in post 77. That's user usability.

Standards are about interoperability, repeatability, etc., and making it easier for the user to use software. If the convolver and/or FIR filter designer s/w can't read a .cfg file or output a .cfg file, then as far I am concerned, the s/w is incomplete.

Anyway, that's it for me. Good luck!

 

[HenrikEnquist]

You as the user should not be writing a config file, period. The spec is designed for programmers to implement into their convolver and/or FIR filter designer software.

In my opinion a good config file is both human- and machine-readable. That .cfg format fails miserably on the human-readable part.
I also dislike that it is a very fragile format where only the line number and position of a number determines what it is. Any structured format (yaml, json, xml, toml...) would have been superior.

 

[gnarly]

Looks like you covered all the options so just have to pick one even though none of them are ideal for you. As for acourate, you can certainly establish a repeatable/mechanical workflow which is easy to operate. UI can be fiddly but you get used to it.

Thanks. I've been afraid that I'd find I'm already reasonably aware of the current options...

Acourate has looked to be the most promising alternative so far.
How do you feel about it for multi-way, speaker tuning only, work? 4 & 5-ways.

 

[3ll3d00d]

How do you feel about it for multi-way, speaker tuning only, work? 4 & 5-ways.

V happy. I use it with some 3 ways (4 inc subs). I don't know what your budget is but I think acourate is pretty cheap in the scheme of things.

 

[mitchco]

Hi @HenrikEnquist nice to meet you virtually!

As mentioned above, the .cfg spec has been around since 2006 and successfully implemented in many free and commercial FIR filter Designer and convolver s/w products. It has worked reliably for 17 years, handles any possible convolution scenario, so there is no need to reinvent the wheel.

We can agree to disagree There is no reason to expose unnecessary complexity to the user. If the FIR filter designer s/w can output a .cfg file and the convolver can read the .cfg file, then there is no value in exposing that to the user. Just because one can doesn't mean it is a good idea.

It takes 2 seconds for the user to load a .cfg file and be up and running listening to music. If you ask me, I would rather be listening to music than fiddling with config files, YAML, XML, JSON or whatever.

 

[dc655321]

Hi @HenrikEnquist nice to meet you virtually!

As mentioned above, the .cfg spec has been around since 2006 and successfully implemented in many free and commercial FIR filter Designer and convolver s/w products. It has worked reliably for 17 years, handles any possible convolution scenario, so there is no need to reinvent the wheel.

We can agree to disagree There is no reason to expose unnecessary complexity to the user. If the FIR filter designer s/w can output a .cfg file and the convolver can read the .cfg file, then there is no value in exposing that to the user. Just because one can doesn't mean it is a good idea.

It takes 2 seconds for the user to load a .cfg file and be up and running listening to music. If you ask me, I would rather be listening to music than fiddling with config files, YAML, XML, JSON or whatever.

Just gonna leave this here, and back away slowly...

 

[AudioJester]

@mitchco - are there any alternatives to Dirac Active Room Treatment on the horizon?
Looks like the next big step in convolution?

 

[gnarly]

V happy. I use it with some 3 ways (4 inc subs). I don't know what your budget is but I think acourate is pretty cheap in the scheme of things.

Nice!
Acourate is within budget.
One quick question...can it run multiple instances?

 

[Keith_W]

Nice!
Acourate is within budget.
One quick question...can it run multiple instances?

I don't think so. I am not sure any convolver that needs to access ASIO drivers can, because they need exclusive access to ASIO devices.

Why do you want to run multiple instances? Also, don't forget that Acourate Convolver requires .CPV files for filters and it will not work unless you have Acourate software. I am not aware of anything else that can generate .CPV files apart from Acourate.

And BTW, I am a big fan of JRiver and if it did not lack one important feature (for me) it would have been perfect. I need an ABX comparator, and a means of quickly switching between filters, and JRiver doesn't have it. THE best ABX comparator I have tried is Hang Loose, filters switch instantaneously with no lag whatsoever and no audible pause. Acourate Convolver has a very slight pause, I am guesstimating about 1/4 of a second. It is enough to interrupt the music and my concentration when I am trying to focus on audible differences between filters.

On another note ... my dream convolver would be able to host VST plugins like @pkane PKHarmonic, store different settings in different banks, and instantaneously switch between them. I run different sets of VST's with different settings for headphone listening and listening to speakers, and I have to reconfigure JRiver every time I switch.

 

[Keith_W]

Spreadsheet updated. The latest version can be viewed here.

After all that back and forth about config files I decided to stop giving them all a red mark and instead simply note whether it is required or not.

There is a new section at the bottom to remark on some features:

 

[mitchco]

On another note ... my dream convolver would be able to host VST plugins like @pkane PKHarmonic, store different settings in different banks, and instantaneously switch between them. I run different sets of VST's with different settings for headphone listening and listening to speakers, and I have to reconfigure JRiver every time I switch.

Fyi, HLConvolver comes with HLConvolverHost that can also host any/multiple VST3 plugins. You can wire up the plugins pretty much anyway you want...

Wrt VB-Virtual Audio Cable issues, you may want to evaluate this alternative: https://vac.muzychenko.net/en/

 

[mitchco]

@mitchco - are there any alternatives to Dirac Active Room Treatment on the horizon?
Looks like the next big step in convolution?

Yes, you may want to look into Thierry's Room Shaper plugin I reviewed a while back. I know he has an updated version. It can reduce the low frequency decay time in the room. It can both be measured and heard.

 

[AudioJester]

Yes, you may want to look into Thierry's Room Shaper plugin I reviewed a while back. I know he has an updated version. It can reduce the low frequency decay time in the room. It can both be measured and heard.

Thanks! Interestingly this seems to predate Dirac ART by years

 

[3ll3d00d]

Acourate is within budget.

Note there are two products here, acourate (the filter designer) and acourate convolver (the actual convolver). You need the former if you use the latter but you can feed acourate generated filters to any convolver (that doesn't use a proprietary format). My comments are about acourate as I use jriver as the convolver.

 

[HenrikEnquist]

The new "Uses Sourceforge config?" row has several errors. Should be no for at least camilladsp, brutefir and EqualizerAPO.
The row name is also weird, what does soureforge have to do with it? Call it "uses Convolver .cfg config format" or something like that.

 

[Keith_W]

Thank you Henrik, updated.

 

[gnarly]

Note there are two products here, acourate (the filter designer) and acourate convolver (the actual convolver). You need the former if you use the latter but you can feed acourate generated filters to any convolver (that doesn't use a proprietary format). My comments are about acourate as I use jriver as the convolver.

Yes. You clearly understand what I'm looking for, ........Acourate just for filter design,
and not to be used as a convolver.

I don't think so. I am not sure any convolver that needs to access ASIO drivers can, because they need exclusive access to ASIO devices.

Ideally, I'd like to use Acourate without even seeing convolver section. (see above reply to @3ll3d00d)

Why do you want to run multiple instances? Also, don't forget that Acourate Convolver requires .CPV files for filters and it will not work unless you have Acourate software. I am not aware of anything else that can generate .CPV files apart from Acourate.

I like to have a filter design instance running for each section of the multi-way I'm working on.
It helps to quickly generate sets of FIR files with varying xover types/points/orders for each driver section.

The Acourate FIR wav files are all i need. CSV would be better, but wav works.

THE best ABX comparator I have tried is Hang Loose, filters switch instantaneously with no lag whatsoever and no audible pause. Acourate Convolver has a very slight pause, I am guesstimating about 1/4 of a second. It is enough to interrupt the music and my concentration when I am trying to focus on audible differences between filters.

Qsys, with its nearly unlimited preset capability, allows instantaneous glitch free switching between filter sets, even for FIR files.
I find the preset design capability quite extraordinary.
I'm currently comparing a 5-way speaker who's sections have so much frequency range overlap,
it can be used as a 4-way, or at the extreme as a 3-way.
Qsys allows instant switching between those scenarios; changing all the necessary xovers, filters, level adjustments, amp mutings, etc.

The only thing I've found it can't change instantly, is the wiring between speaker sections and amps Lol.
So bottom line for me...not looking for a convolver or a comparator.

 

[3ll3d00d]

I like to have a filter design instance running for each section of the multi-way I'm working on.

We are heading off topic.... Acourate is workspace based so you switch between them as you like. You would typically look at multiple ways in a single workspace but a workspace is just a directory containing some files so you can do whatever you like. I don't think it lets you run multiple instances of the app concurrently though atm.

 

[Keith_W]

Sadly for you, Acourate (the filter design program) costs €340 (for EU countries) and €286 (for non EU countries) - link.

It is Acourate Convolver (the convolution program) that costs €126. So all up you are looking at between €412 to €466 for Acourate + Acourate Convolver. Acourate itself is cheaper than Dirac or Audiolense, and probably equally powerful. But it is less automated and requires more manual work.

I was thinking of doing a filter generator feature comparison, but that is a massive topic and it wouldn't fit easily on a spreadsheet. A quick search for the terms "Acourate", "Audiolense", "Dirac" on ASR will yield many discussion threads.