PC/MAC for Dirac Processor

[DWPress]

I've an old x86 box I test distros with and occasionally boot into Win, let me know when you get to multichannel noodling and I'll try to help out if I can.

 

[cobby]

I've an old x86 box I test distros with and occasionally boot into Win, let me know when you get to multichannel noodling and I'll try to help out if I can.

You're welcome.

My mind is focused on the current state right now, but I have some plans for sure.
My problem with multichannel is the sheer amount of possible configurations. If you like, let's clarify the specs you'd actually need a bit, so I can start making up my mind about it.

Don't want to get too much OT here, but I'd be willing to get something working.

 

[Music1969]

I do want to look at multichannel later though. I actually just bought a Scarlett 16i16 partly with that in mind (decided already before for a Motu M4), so it’s definitely on my radar. Just not something I want to promise yet.

+1 for multichannel. I use Focusrite 18i20 (10 channel output) that I use with DIY DSP active speakers, digital crossovers, speaker correction and room correction for each driver (I'm using all 10 channels).

I would use FX for FIR convolution on MacOS.

 

[cobby]

+1 for multichannel. I use Focusrite 18i20 (10 channel output) that I use with DIY DSP active speakers, digital crossovers, speaker correction and room correction for each driver (I'm using all 10 channels).

I would use FX for FIR convolution on MacOS.

Okay, cool — so that’s one vote for multichannel active crossover. I’ll definitely have a closer look at that use case and think about where it could fit into FXRoute.

macOS is another topic.

 

[cobby]

@DWPress @Music1969
I’ve thought about the multichannel crossover case a bit more, and it actually looks pretty straightforward.

FXRoute already has most of the building blocks in the current subwoofer crossover: LR24, gain, delay and polarity. So this would mainly mean expanding that to more bands and outputs, not building a whole new DSP path.

The existing stereo PEQ and convolution could stay as they are, with 2-way, 3-way, 4-way and maybe 5-way crossover modes after that. Then the bands just need to be assigned to the physical outputs.

The main thing I still need to think through is whether active speakers need more crossover options than LR24.

 

[DWPress]

The main thing I still need to think through is whether active speakers need more crossover options than LR24.

If you can offer all the flavors of XO options (Linkwitz, Bezel, Butterworth, etc.) and variable orders (6,12,18,24 - 72) that would make it the most useful.

 

[DWPress]

@cobby if you're in the US and are serious about wanting a Motu M4 I just happen to have one available - replaced it with the 6i6 to get digital I/O.

 

[cobby]

If you can offer all the flavors of XO options (Linkwitz, Bezel, Butterworth, etc.) and variable orders (6,12,18,24 - 72) that would make it the most useful.

Yeah, that’s roughly what I had in mind. The current crossover is LR24, but for active speakers I’d want filter type and slope to be selectable.

Something like Linkwitz-Riley, Butterworth and Bessel, with a sensible range of orders, should cover most use cases.

I’ve also got the Scarlett 16i16 now and it’s already running. I’m exploring it at the moment, and the channel profiles I’m working on are already leading naturally towards an output matrix, which I’ll bring into the regular version anyway.

I’d keep multichannel experimental at first. Once I’m a bit further along, I could make an experimental branch and ISO if you’d like to try it.

Sorry for the off-topic. Hope that’s okay.

 

[cobby]

@cobby if you're in the US and are serious about wanting a Motu M4 I just happen to have one available - replaced it with the 6i6 to get digital I/O.

That’s a really nice offer, thanks! I’m in Germany though, so we rather avoid dealing with customs.

The MOTU M4 shouldn’t be a problem anyway. It’s very well supported under Linux, widely used, and with four channels there’s nothing particularly unusual about the setup.

The larger Scarlett interfaces, like the 16i16 and up, have a few additional specifics to deal with, mainly around ADAT and samplerate. But for interfaces with up to eight channels, the standard profile generally works without any trouble.

 

[Music1969]

@DWPress @Music1969
I’ve thought about the multichannel crossover case a bit more, and it actually looks pretty straightforward.

FXRoute already has most of the building blocks in the current subwoofer crossover: LR24, gain, delay and polarity. So this would mainly mean expanding that to more bands and outputs, not building a whole new DSP path.

The existing stereo PEQ and convolution could stay as they are, with 2-way, 3-way, 4-way and maybe 5-way crossover modes after that. Then the bands just need to be assigned to the physical outputs.

The main thing I still need to think through is whether active speakers need more crossover options than LR24.

I persnally don't need crossover designs inside your software - it would be cool and I see others would like it.

My crossover is designed by room correction software - I just need to be able to add the FIR convolutions files to a convolver.

I just upload mono 64-bit .WAV files into your program (convolver).

 

[cobby]

I persnally don't need crossover designs inside your software - it would be cool and I see others would like it.

My crossover is designed by room correction software - I just need to be able to add the FIR convolutions files to a convolver.

I just upload mono 64-bit .WAV files into your program (convolver).

Ah yes, I see what you mean now. For that use case FXRoute would need a separate convolution file for each physical output/channel, right now, mono for L and R.

That’s not supported at the moment, and it needs a bit more thought than the normal crossover case. Definitely an interesting option though.

 

[DWPress]

For that use case FXRoute would need a separate convolution file for each physical output/channel, right now, mono for L and R.

Yes, 8 channels of individual IIR and FIR with selectable XO orders would be ideal. Before better solutions became available around the Juce code (Hang Loose Convolver, Element) I used X-MCFX-Convolver plugin where I had to create 8ch .wav files in Audacity and implement it in a hand coded .cfg file. A pain in the ass but worked.

Good luck with the project, I'll be following along developments. Maybe start a new thread here at ASR when you Beta the multichannel so it gets its own traction since we've pretty much hijacked this thread.

 

[PaulF2020]

Yes, volume control on my Okto which also has a remote.

Analog/digital inputs can be done with the Scarlett - with the bi-directional USB connection thus giving you the benefits of system wide EQ. You technically wouldn't need the WiiM anymore and just use desktop apps on the mini controlled by the companion phone apps or just Airplay to the mini.

The Scarlett does have volume control but no remote. Focusrite also has a phone/tablet app that could be used for switching sources and outputs but it's not as eloquent as the WiiM app as it's meant for pro audio.

So I got Scarlett 6i6, It seem to work with Dirac fine, however I cannot figure out how to route SPDIF from Scarlett to Dirac virtual device. Do you have any suggestions?

 

[DWPress]

Do you have a digital input? If not then the USB audio should present itself as an option, you may need a virtual cable like Blackhole (free) for routing if you're using the Dirac standalone app or as a plugin. Which way are you using Dirac?

 

[PaulF2020]

Do you have a digital input? If not then the USB audio should present itself as an option, you may need a virtual cable like Blackhole (free) for routing if you're using the Dirac standalone app or as a plugin. Which way are you using Dirac?

Yes I have digital input. I connected Wiim to scarlet coax input.
So I want to use Wiim as a source =>scarlet SPDIF input=>Dirac virtual device=> output to Scarlett

"Dirac virtual device=> output to Scarlett" this part is already configured and seems to be working I can start calibration and do test tones there etc.

Ai suggesting I need something like Reaper or Audio Hijack to route Scarlett SPDIF to Dirac

 

[DWPress]

I use Hang Loose Convolver which hosts the Dirac plugin. You might try Element which is free, can be buggy last time I used it but will depend on your Mac OS version. Reaper would work (outside my experience though) but overkill.

Routing would be your "input source" > Dirac plugin > output device in a graph like this (HLC in this case). Here's an example of what I do, output is to my 8ch DAC:

Audio input > Dirac Processor (L/R 3 way XO +2 subs on the right side routed through MEqualizer for HP subsonic filter). The mvMeter plugins are just bling. You would need 4 channels for your L/R + 2 subs, 5ch for L/R + 3 subs etc. Not sure what settings you'll need to configure in Focusrite Control to get input from WiiM looped back via USB.

 

[PaulF2020]

Ok, I have measured everything, but Dirac seems to give me odd results, I think it is cancelling my bass response in MLP by setting those delays:

​

Not entirely sure what I'm doing wrong. Delay between subs and mains I can potentially explain, KEFs have their own DSP and signal probably arrives after subs, but why there is so huge delay between subs?

 

[DWPress]

Yes, all 3 of those delays figured by Dirac are way off. Is your mic input and speaker output both going through the 6i6 or are speakers routed through the WiiM? A lot of measurement apps don't like separate interfaces where the clock can potentially be reported wrong. Typically the main speakers are delayed before the sub, if you have internal DSP happening with the KEFs it will delay them more but not 253ms.

Dirac is not always the most intuitive app to work with. Try another session with just 3 measurements and skip to the design stage, if the delays are reported with a sane number you can always go back and take additional measurements.

 

[PaulF2020]

Yes, all 3 of those delays figured by Dirac are way off. Is your mic input and speaker output both going through the 6i6 or are speakers routed through the WiiM? A lot of measurement apps don't like separate interfaces where the clock can potentially be reported wrong. Typically the main speakers are delayed before the sub, if you have internal DSP happening with the KEFs it will delay them more but not 253ms.

Dirac is not always the most intuitive app to work with. Try another session with just 3 measurements and skip to the design stage, if the delays are reported with a sane number you can always go back and take additional measurements.

My mic is UMIK1 via USB connected directly to Mac mini. Speakers connected via Coax from 6i6 and subs connected directly to 6i6 TRS outputs.

 

[DWPress]

I switched from using my UMIK1 to my XLR phantom powered mic specifically because of timing issues for my 3 way XO. I'd gotten reliable results from REW with the UMIK but conflicting results sometimes from Dirac. Your set up is typical though and even Dirac recommends the UMIK so I'm at a loss why the big discrepancies. All I can say is keep trying. Can you share the results measured in REW so we can see what's going on?