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

[kharan]

Not quite what I expected.

 

[OCA]

This morning I did the 3 measurements and generated the filters, what I can say is that the bass sounds better than with my filters generated with rew, but I don't like the mids and highs. I think GSonic corrects the speakers (Martin Logan esl 11a+2x Kef KC 62) too much and I don't know how to limit the correction to 300 HZ. OCA please look over the measurements and maybe you can help me. Thanks

There's very mild correction above 300Hz:

What mic are you using? I would suspect the correction above 15kHz could cause dullness with electrostatics since very few mics can capture that area correctly and it's a region I could never audibly test myself (due to age).

 

[cobby]

...what about GSonic + @VilhoValittu 's CamillaFIR ? (= full open source solution)

It has been a while since I last tried CamillaDSP on an x86 mini PC, so my experience there is a bit outdated. At the time, it was not nearly as smooth for me on x86 as I had hoped, although I am sure quite a lot has changed since then.

It looks like there are now also CamillaDSP options for PipeWire, which makes a lot of sense since PipeWire is basically the standard on modern Linux systems now.

For me, Easy Effects was, and honestly still seems to be, the easiest way to get DSP working on Linux. PipeWire is already standard on modern distros, and from there it is basically just a one-line Flatpak install.

Easy Effects is simple to use, and the filter support has been excellent in my experience. At the moment I just prefer the simplest possible solution.

That said, CamillaDSP is definitely very cool, and CamillaFIR also looks very promising. I could absolutely see myself trying that route again at some point when it fits.

 

[cobby]

This morning I did the 3 measurements and generated the filters, what I can say is that the bass sounds better than with my filters generated with rew, but I don't like the mids and highs. I think GSonic corrects the speakers (Martin Logan esl 11a+2x Kef KC 62) too much and I don't know how to limit the correction to 300 HZ. OCA please look over the measurements and maybe you can help me. Thanks

I can relate to that a bit. In my case as well, the bass improvement was obvious, while I am still less sure yet about the mids and highs. I need a few more measurements and more listening before I can judge that part properly.

I suspect there is still room for tuning through the target / house curve, but I also think there is a strong case for leaving the upper range more alone. If I understood OCA's philosophy correctly, that also seemed to be part of his idea with the minimum phase filters: correct what really benefits from correction, but do not overdo it across the whole spectrum.

Another thing I am wondering about is the measurement level. On the one hand, OCA recommends in the video to measure at the levels you normally listen at. On the other hand, by the time the GSonic app says the level is OK, it already feels quite loud to me, louder than my normal listening level.

So I am also wondering whether measuring at a higher playback level than the one you actually use later could lead to the filter sounding a bit overcorrected in practice?

 

[ciprian pop]

There's very mild correction above 300Hz:
View attachment 523623
What mic are you using? I would suspect the correction above 15kHz could cause dullness with electrostatics since very few mics can capture that area correctly and it's a region I could never audibly test myself (due to age).

I use Umik 1. I listened more and I think the filters are ok, I don't like how it sounds through the roon server, I never liked it. Too bad I can't load them directly into Eversolo because the bass is much better than what I managed to generate from rew. Thank you very much for the help

 

[john61ct]

My understanding is measurement level should be "your normal" OR the app's reco, whichever is higher.

So, is there a setting for "do not touch signal above 5kHz" for example?

 

[Plompudu]

My understanding is measurement level should be "your normal" OR the app's reco, whichever is higher.

So, is there a setting for "do not touch signal above 5kHz" for example?

I don't think there is, but you can use a custom target that is your speakers response above 5kHz, so that It won't get touched a lot.

I had to do something similar to get my Driver variance between my mains included in the left filter:
Purple : Dolby / Harman-ish Target
Red: New Target
Green: FL0 convolved with the new filter

To generate the Target use this:

Then create a left and a right channel target and run them in GSonic.
Then take the Mono Channels Left/Right generated with the right Target and use them.

 

[OCA]

I use Umik 1. I listened more and I think the filters are ok, I don't like how it sounds through the roon server, I never liked it. Too bad I can't load them directly into Eversolo because the bass is much better than what I managed to generate from rew. Thank you very much for the help

I think I figured why your electrostatics don't like the HF correction. They are acting like dipoles and front wall HF reflection is tilting the excess phase at the LP causing the algo to think of it as a point source driver in need of crossover phase linearization. I am adding a few new advanced options for cases like yours as there's no automatic and robust way to detect that kind of set up.

 

[OCA]

GSonic Reference v1.0.20 released!

Added advanced HF correction options (3 new parameters under the collapsible Advanced section on the Generate tab):

• Correction Strength (0.0-1.0) — Scale room correction intensity. Bass correction always stays at 100%; the strength only fades above the room's transition frequency. Default: 1.0 (full correction, same as before).
• HF Correction Limit (Hz) — Fade out all EQ above this frequency for a true bypass. Useful for electrostatic panels where dipole behavior should not be corrected. Default: Full Range.
• HF Phase Limit (Hz) — Bypass phase correction above this frequency, preserving the speaker's native transient response. Default: Full Range.

All defaults produce identical output to previous versions. Settings persist across sessions.

 

[boXem]

GSonic Reference v1.0.20 released!

Added advanced HF correction options (3 new parameters under the collapsible Advanced section on the Generate tab):

• Correction Strength (0.0-1.0) — Scale room correction intensity. Bass correction always stays at 100%; the strength only fades above the room's transition frequency. Default: 1.0 (full correction, same as before).
• HF Correction Limit (Hz) — Fade out all EQ above this frequency for a true bypass. Useful for electrostatic panels where dipole behavior should not be corrected. Default: Full Range.
• HF Phase Limit (Hz) — Bypass phase correction above this frequency, preserving the speaker's native transient response. Default: Full Range.

All defaults produce identical output to previous versions. Settings persist across sessions.

Do I understand correctly that correction strength 0.0 and HF Correction limit at room's transition frequency will produce the same result?

 

[OCA]

Do I understand correctly that correction strength 0.0 and HF Correction limit at room's transition frequency will produce the same result?

The difference comes down to target curve.
Strength = 0.0 will still apply target curve to shape the overall tone.
HF Limit = F_s will not touch audio at all above the bass region for true bypass

 

[Coltr@ne]

HF Limit = F_s will not touch audio at all above the bass region for true bypass

Forgive my ignorance, but in this case would the taps be reduced, allowing use in devices like the Eversolo DMP 8, with only 2000 taps?

 

[OCA]

Forgive my ignorance, but in this case would the taps be reduced, allowing use in devices like the Eversolo DMP 8, with only 2000 taps?

Tap count will not change unless you compromise frequency resolution at the same sampling rate.

 

[olieb]

Strength = 0.0 will still apply target curve to shape the overall tone.

The meaning of "apply target curve to shape the overall tone" is "magnitude (only) correction"?

HF Phase Limit (Hz) — Bypass phase correction above this frequency

And this means similarly that magnitude (only) will still be corrected? How does this happen? With linear phase, or in a "natural" way, like an IIR (implemented in FIR)?

 

[OCA]

The meaning of "apply target curve to shape the overall tone" is "magnitude (only) correction"?

And this means similarly that magnitude (only) will still be corrected? How does this happen? With linear phase, or in a "natural" way, like an IIR (implemented in FIR)?

GSonic filters are true mixed phase FIR and gradually shift to pure linear phase at HF. IIR is not really natural above room transient and introduces group delay smearing.

 

[ernestcarl]

IIR is not really natural above room transient and introduces group delay smearing.

Uh, if one were correcting minimum phase deviations then there is really no additional GD "smearing". The phase shift for most EQ done up there is also generally low Q e.g. shelving filters.

 

[olieb]

GSonic filters are true mixed phase FIR and gradually shift to pure linear phase at HF. IIR is not really natural above room transient and introduces group delay smearing.

Well, it depends.
But to put it differently, my question was, what happens when the FR has a "wiggle" from a driver resonance?
And what happens if the FR has a wiggle from interference with a (room) reflection.
For the former an IIR-style filter would be quite "natural". The latter I would prefer to be left alone in HF.
But your answer is helpful, thanks.

 

[ernestcarl]

The latter I would prefer to be left alone in HF.

I would agree, although occasionally if it's really audibly annoying (sounds harsh or peaky or too reverberant) then I see no reason why one could not apply a bit of a cut over the offending limited HF band.

 

[Gorgonzola]

There is little problem with Gsonic that I've encountered with one of my Win 11 PC that I've dedicated to music playing. The gain/sound volume I was able to get with output via USB to a DAC. Gsonic's 'Test Sound' required I set my DAC (used a preamp) to 0.0 dB; 'Measure FL+FR' sweeps sounded very quite, and the resulting .WAV filters didn't seem to work in Equalizer APO.

 

[ciprian pop]

I think I figured why your electrostatics don't like the HF correction. They are acting like dipoles and front wall HF reflection is tilting the excess phase at the LP causing the algo to think of it as a point source driver in need of crossover phase linearization. I am adding a few new advanced options for cases like yours as there's no automatic and robust way to detect that kind of set up.

I loaded the measurements into GSonic v1.20 and limited the frequency and phase correction to 300Hz. The bass remained unchanged (which is good) and the panels are now uncorrected. Subjectively, the sound is better, it seems more natural and that slight lack of transparency that I felt before when the correction was full range is now gone. Now the system sounds better than ever, despite the fact that I used several correction programs (Arc Genesis, Dirac, Audiolense), including my own filters in REW. I don't know how to thank you, OCA you are a great guy!