I'm not planning on implementing any kind of measurement and filter generation features. But I'm hoping that maybe some day some existing (or new) room correction app learns to control camilladsp. Then it could measure, create filters, apply them directly to camilladsp, measure again and if needed do more iterations.
I personally like the way DRC-FIR is structured, where one set of tools is only concerned only with producing a high quality impulse response and the other tool takes the impulse response to generate filters. Only a filter generating tool would have to communicate with CDSP, getting an impulse response could be decoupled entirely.
Maybe the tools included with DRC could serve as a starting point for something more user friendly. DRC's 'glsweep' generates log sweeps in raw PCM format, 'lsconv' can generate an impulse response from the log sweep and a corresponding recording. These two can be automated in a script, as long as all needed parameters are known: Input device, output device, gain settings...
Filter generation is a different matter as everyone has different tastes. Some want FIR filters, others want IIR. Some want to correct individual drivers, others want to correct an existing speaker. I doubt that there is a one-size-fits-all solution.
So far I've found the hard part to be getting a high quality impulse response in an automated and repeatable manner. I've yet to find a solution that I really like.
Automating filter generation has been comparatively smooth sailing as I can just take the impulse response and feed it into DRC to generate the FIR filters that I prefer. Easily done with a bash script. I don't know about generating IIR filters in an automated way, though
Last edited:
Yes I have. It's very cool, it's just not what I'm looking for. What I'm after is something that
- can run headless, on a computer without a screen
- is not resource intensive so it can run on something like a raspberry pi
- can be automated
OK!
A more interactive way via GUI perhaps? I envision a (one?) spline representing the correction curve where a constant pink noise is played and at the same time observing the system with a RTA FFT (both by e.g. REW) - and in CDSP being able to add points and handles to interactivly manipulate the "curve" - in the background Camilla creates the necessary IIR (FIR?) filters and adds them to the processing as soon as finished indicating with progressive feedback.
After a more scientific approach to fixing the room curve this "tuning tool" could be a very nice way of finishing off the job via a combo of "by ear" and measurement.
//
Last edited:
Right. I was thinking of what you wrote in your earlier post:
So what would a complete workflow look like? Something like this?Now, we all know what we want to do... place a mic in favourite position, enter room curve and watch the system slowly come into submission within +/- 1 dB(using pink noise!?) - is this something for CDSP in the future? 🙃
- Measure system
- Generate filter
- Apply filter - CamillaDSP goes here
- Optionally tweak by ear
- Profit?
I recently used Open Sound Meter on one laptop while I used a second laptop to make Rephase phase adjustments and apply them to the CamillaDSP config. Open Sound Meter realtime display eliminated the taking of measurements for each change. Open Sound Meter uses pink noise but thankfully it is one click to turn it off while going through the 'Rephase adjust / copy to CamillaDSP config / apply config to DSP' cycle.
A similar technique could be used to adjust the room curve, but good luck automating it.
https://opensoundmeter.com/ and how to set it up
A similar technique could be used to adjust the room curve, but good luck automating it.
https://opensoundmeter.com/ and how to set it up
How did you get the Equaliser at 96kHz ? On win 11 I can only get it at 48kHz.
Creating filters that fit a spline (or any other free-form curve) like that is quite a big project, not likely something I would ever find the time to implement. But in practice, wouldn't a graphic equaliser give you nearly the same functionality? That's something I have been thinking about adding, a normal graphic equaliser where you just tell how many bands it should have. In the background it gets built from Peaking filters with suitable q-values.
There's the AutoEQ project that does just that. Its purpose is mainly to equalise headphones, but there's no reason why it couldn't be used for speakers.
https://github.com/jaakkopasanen/AutoEq
It takes a frequency response measurement, a target frequency response and the type of EQ you want to use (FIR, IIR, static multiband EQ) and returns the EQ settings needed to reach the target response.
There is also a trimmed down python library available:
https://github.com/jaakkopasanen/autoeq-pkg/tree/master/autoeq
Hi all, i tried to search for answer, but did not find anything.
How long is Camila DSP latency if IIR filters only are used?
Does somebody used Camila for live production? Here you need latency less than 4ms at least - otherwise is almost impossible for musicians to play.
Standalone DSPs such as behringer DCX have latency ca. 1ms
I would very like use Camila but low latency is crucial for me.
Many thanks for answer.
Tomas
How long is Camila DSP latency if IIR filters only are used?
Does somebody used Camila for live production? Here you need latency less than 4ms at least - otherwise is almost impossible for musicians to play.
Standalone DSPs such as behringer DCX have latency ca. 1ms
I would very like use Camila but low latency is crucial for me.
Many thanks for answer.
Tomas
