GSonic Reference – Free Stereo Room Correction Tool (Measurement + FIR Export)

[john61ct]

Acourate Convolver is out because it uses proprietary .CPV files and not .WAV

Does this mean filters created by Acourate are non standard, or is this just an issue with Acourate Convolver ?

Any ideas why they went this route?

Are there any important filter creation tools whose output files that cannot be processed by HLC?

 

[flexy123]

Experimenting with Hang Loose Convolver currently, it's great.

A convolver that can't use FIR from REW, Gsonic, Camilla etc....seems absurd to me...

 

[john61ct]

A convolver that can't use FIR from REW, Gsonic, Camilla etc....seems absurd to me...

CamillaDSP is just a convolver, an alternative to HLC, right?

I see CamillaFIR creates filters, but is a separate tool

rePhase and MSO are of particular interest to me as well as GSonic, I assume HLC also handles their output well.

 

[flexy123]

CamillaDSP is just a convolver, an alternative to HLC, right?

I see CamillaFIR creates filters, but is a separate tool

Yeah, and to be honest CamillaDSP is super complex and difficult to setup as it seems to me. I didn't look too much into it. CamillaFIR is also interesting, and of course GSonic for filter creation.

 

[_giacomo_]

That dip looks like a 50Hz AC ground loop random phase shift issue. Are you in EU? Check if it persists or keeps shifting between measurements. If so, there's not much you can about it and it's not nearly as audible as it seems.

Filters are 0dBFS. REW needs a reference to convert them to dB SPL and usually uses 120dB - 3dB (division safety) = 117dB but that will change based on your PC's sound card and audio driver type, etc.

No need to add any offset to the filters in any convolution engine. They are normalized and clip protected.

I don’t think this is a ground loop issue. The measurement chain is fully isolated, since the laptop is running on battery power.

In my view:

1. The attempt to reduce the energy in the 60–160 Hz range also causes some unintended attenuation below 60 Hz.
2. If you look at the phase of the left channel before and after filtering (L vs. L+filter), the correction shifts the phase closer to zero. This increases the cancellation around 40 Hz, where there is roughly a 180° phase difference between the left and right channels.

Please find the measurement attached, if you’re interested.
Because of limits in the attachment dimension, I wasn’t able to include the final convolution of L×Filter and R×Filter, nor the (most important) vector average of the two, where the dip becomes apparent.

 

[jeffaegrim]

Well, after 2 weeks with GSonic filters - I continue to be amazed!

Dito, and I am convinced that it was the last piece of the puzzle towards my ultimate satisfaction with home audio. Nothing ever sounds bad anymore. I am also happy to report that you can, of course, route your L/R channel into the center speaker, treating it like a stereo pair with your other L/R speaker, to produce a correction for the center. I have been using a fully GSonic corrected 5.1 setup for several weeks now and it's everything one would ever need/want in terms of DSP, provided one has halfway capable/neutral speakers. Eagerly awaiting the release of the "premium" version

 

[greyscale]

After using Gsonic on a Windows 11 laptop connected via HDMI to a Denon X6800H AVR, the Windows 11 Audio Device Service must be reset via Windows Troublshooter. Am guessing, Gsonic takes control of the Windows Audio Service and turns this Service OFF. PERHAPS forgeting to turn it back on when Gonoc quits???

 

[OCA]

After using Gsonic on a Windows 11 laptop connected via HDMI to a Denon X6800H AVR, the Windows 11 Audio Device Service must be reset via Windows Troublshooter. Am guessing, Gsonic takes control of the Windows Audio Service and turns this Service OFF. PERHAPS forgeting to turn it back on when Gonoc quits???

It's usually the "Realtek" Audio service running in the background that messes up those things. GSonic uses rtAudio libraries which are proven and there's no way to keep the audio device engaged after use.

 

[OCA]

I don’t think this is a ground loop issue. The measurement chain is fully isolated, since the laptop is running on battery power.

In my view:

1. The attempt to reduce the energy in the 60–160 Hz range also causes some unintended attenuation below 60 Hz.
2. If you look at the phase of the left channel before and after filtering (L vs. L+filter), the correction shifts the phase closer to zero. This increases the cancellation around 40 Hz, where there is roughly a 180° phase difference between the left and right channels.

Please find the measurement attached, if you’re interested.
Because of limits in the attachment dimension, I wasn’t able to include the final convolution of L×Filter and R×Filter, nor the (most important) vector average of the two, where the dip becomes apparent.

Can only do later, not a good time.

 

[flexy123]

Dito, and I am convinced that it was the last piece of the puzzle towards my ultimate satisfaction with home audio. Nothing ever sounds bad anymore. I am also happy to report that you can, of course, route your L/R channel into the center speaker, treating it like a stereo pair with your other L/R speaker, to produce a correction for the center. I have been using a fully GSonic corrected 5.1 setup for several weeks now and it's everything one would ever need/want in terms of DSP, provided one has halfway capable/neutral speakers. Eagerly awaiting the release of the "premium" version

I know it would be possible, but getting behind my AVR and swapping speakers is a total PITA....

Alternatively, until Gsonic for 5.1 comes out, one could also measure manually with REW? Which I didn't do yet for Gsonic since not 100% certain about the exact parameters like windowing etc. when exporting the IRs... and it's only a matter of minutes/hours/days anyway until that measuring tool is out

 

[OCA]

I don’t think this is a ground loop issue. The measurement chain is fully isolated, since the laptop is running on battery power.

In my view:

1. The attempt to reduce the energy in the 60–160 Hz range also causes some unintended attenuation below 60 Hz.
2. If you look at the phase of the left channel before and after filtering (L vs. L+filter), the correction shifts the phase closer to zero. This increases the cancellation around 40 Hz, where there is roughly a 180° phase difference between the left and right channels.

Please find the measurement attached, if you’re interested.
Because of limits in the attachment dimension, I wasn’t able to include the final convolution of L×Filter and R×Filter, nor the (most important) vector average of the two, where the dip becomes apparent.

You are right, this is not a 50Hz issue. Algo doesn't check for phase dips when cutting peaks with min phase filters and doesn't check for constructive cancellation of the left and right speaker summation and usually this is not necessary but due to your room conditions - possible asymmetry, you are getting a phase cancellation in the summation. Noted as a future addition to the tool (not an easy one!) but for the time being, try playing with speaker placement, small changes even can improve this significantly.

 

[Keith_W]

You just save it as a library item and play it no?

Hmm, I thought I had to "Open Live..." every time. And as you know from my posts in the JRiver forum, there is some weird incompatibility between RME and JRiver that has stopped me from doing that.

Does this mean filters created by Acourate are non standard, or is this just an issue with Acourate Convolver ?

Acourate has several filter output options, including standard .WAV files. It's only Acourate Convolver which requires proprietary .CPV files, and AFAIK only Acourate can generate those. I have no idea why they went this route, I have asked Uli many times to make Acourate Convolver accept standard .WAV so that more people can use it, but he's decided to stick with .CPV.

Are there any important filter creation tools whose output files that cannot be processed by HLC?

Dirac is the only one I can think of. It outputs in some kind of proprietary format, and it requires the Dirac VST as the final output in the signal chain. Otherwise, every single filter design tool outputs standard .WAV files.

rePhase and MSO are of particular interest to me as well as GSonic, I assume HLC also handles their output well.

You can not use MSO with FIR filters. MSO is an IIR filter design tool. Well, you can, but I don't see why you would want to. Linear phase FIR avoids many of the problems associated with minphase IIR which means the filters are much easier to manually design.

 

[mitchco]

Are there any important filter creation tools whose output files that cannot be processed by HLC?

Dirac is the only one I can think of. It outputs in some kind of proprietary format, and it requires the Dirac VST as the final output in the signal chain. Otherwise, every single filter design tool outputs standard .WAV files.

You can easily reverse engineer Dirac filters if you know what you are doing. But just because one can does not mean one should

Experimenting with Hang Loose Convolver currently, it's great.

Thanks - more features coming including a native ALSA engine and UPnP I/O.

 

[john61ct]

You can not use MSO with FIR filters. MSO is an IIR filter design tool. Well, you can, but I don't see why you would want to. Linear phase FIR avoids many of the problems associated with minphase IIR which means the filters are much easier to manually design.

Sorry, you say MSO "can" do Linear phase FIR ?

And these are better and much easier? than minphase IIR

I'm confused.

For context, I'm looking to establish multichannel and multisub* crossovers and timing / phase tuning first,

then Gsonic Reference for final "as if just stereo" DRC

*likely stereo bass but mono LFE content (for say 15-50Hz) blended in. If needed maybe adding mono sub(s) minimalistically to tackle room modes, all subject to REW feedback as well as ears+brain

 

[Keith_W]

Sorry, you say MSO "can" do Linear phase FIR ?

And these are better and much easier? than minphase IIR

MSO CAN NOT do linphase FIR. The output of MSO is in PEQ's, delays, and all-pass filters. Perhaps this thread will clear up any confusion between FIR and IIR.

GSonic, Acourate, Audiolense, Focus Fidelity, rePhase are all linear phase FIR design tools. REW can be made to do linear phase, but it's not easy and it needs rePhase to fill in missing functions, so you could consider it to be primarily a minphase IIR design tool. Because IIR is minimum-phase, it means that every correction (PEQ) by necessity rotates phase, which is OK if all you are doing is using a minphase DSP to correct a minphase response. Remember that each PEQ looks like a peak with left/right symmetry. If you want to correct a complex looking bass response, you will need to stack PEQ's. Since most IIR processors (e.g. MiniDSP) don't have many of them, you need to figure out the most efficient way to use your limited resources. This is where tools like MSO comes in, it uses brute force computation* to try to find the best solution using the resources you have available.

* the author does not like this term. He likens it to ants wandering around finding food. Some ants will go down a different path, and if it's a better path then more ants will follow. Either way, it's a "try and discard" type algorithm.

 

[john61ct]

MSO CAN NOT do linphase FIR.

Great thanks!

I am starting out doing the FR part crossovers via active analog, and only want to add DSP minimally if needed, particularly for timing / phase / delay tuning

If I use IIRs only, with a low-resource convolver like miniDSP then it might make sense to leave the analog xovers in place for the FR side, to lighten the processing load?

 

[Keith_W]

Great thanks!

I am starting out doing the FR part crossovers via active analog, and only want to add DSP minimally if needed, particularly for timing / phase / delay tuning

If I use IIRs only, with a low-resource convolver like miniDSP then it might make sense to leave the analog xovers in place for the FR side, to lighten the processing load?

I am not going to drag this thread off topic any further. I'll quickly answer your question, then I am out. I will only answer "general FIR related questions which are somewhat related to GSonic" out of respect for @OCA.

There is no point leaving the analog XO's in place to "lighten the processing load" for MiniDSP. If you look closely at how MiniDSP allocates biquads, you will find that nearly all of them have up to 18 biquads per channel, of which four are reserved for XO duties. You can't use them for PEQ's even if you want to. MiniDSP locks them out.

Each biquad is a two-pole filter, meaning 2nd order. So if you have two of them, you can have up to 4th order. And since you have 4 of them, you can have one 4th order LPF and a 4th order HPF. I don't think it is possible to combine them and make a single 8th order LPF for example. The only reason to leave analog XO's in place is if you want a higher order filter than what MiniDSP offers, or you want some kind of strange XO type that MiniDSP can't do, e.g. Chebychev. Or if you don't feel like modifying your speakers to bypass the analog XO.

 

[3ll3d00d]

Hmm, I thought I had to "Open Live..." every time. And as you know from my posts in the JRiver forum, there is some weird incompatibility between RME and JRiver that has stopped me from doing that.

Minor OT but just to close that one... Fwiw you can save a given open live settings as a Library item, name it (eg to reflect the source that connects via those channels) then play that via any remote as usual

 

[3ll3d00d]

of which four are reserved for XO duties. You can't use them for PEQ's even if you want to. MiniDSP locks them out.

It's 8 and you can use them

 

[_giacomo_]

You are right, this is not a 50Hz issue. Algo doesn't check for phase dips when cutting peaks with min phase filters and doesn't check for constructive cancellation of the left and right speaker summation and usually this is not necessary but due to your room conditions - possible asymmetry, you are getting a phase cancellation in the summation. Noted as a future addition to the tool (not an easy one!) but for the time being, try playing with speaker placement, small changes even can improve this significantly.

Yes exactly, there is a room asymmetry that is killing a whole wide band around 40Hz when the speakers are at 180cm from the front wall. I already played a lot with positioning. Now the speakers have ended up at 120cm (much closer). I lost some imaging, I gained some bass.
The addition would be great, and maybe trading (peak) correction close to a dip already may help.
I understand this is not going to be easy, yet I will definitely be waiting for this feature.