- Thread Starter
- #381
Mic cal error seems to be caused by non-ASCII characters in your user name. Try saving it in an other folder and load from there.
-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
GSonic Reference – Free Stereo Room Correction Tool (Measurement + FIR Export)
shanghaifat
New Member
- Joined
- Oct 4, 2019
- Messages
- 3
- Likes
- 1
- Thread Starter
- #383
It can be avoided on the code side (with quite a bit of work in C++) and I added that in the to-do for the next update.
- Thread Starter
- #387
If you mean GSR FIR filters with XO (term normally used for crossover), pre-ringing is 60% quiter than IR peak and only present in the last 5ms each of which make it quite impossible to be audible:
Last edited:
Hi OCA,
I have a question regarding G-Sonic behavior during measurements and EQ sweeps.
When I open a G-Sonic project and run measurements/sweeps, I sometimes hear small crackling or popping sounds during the sweeps. It does not happen continuously, but randomly during some sweeps.
I also noticed that REW reports a relatively high clock drift during these measurements.
Is this behavior normal when using G-Sonic, or could it indicate a synchronization/buffer issue somewhere in the audio chain?
My setup uses:
Mic pointed toward speakers when measured
1 .File_20260518 ( measurements gsonic) Red fr is G-Sonic AND Green is no eq
2. Microfoon recht ( here is the rew file )
3. Schermopname (75) ( Screenshot clockdrift)
The measurements themselves still seem usable, but I wanted to verify whether occasional crackling during sweeps and higher clock drift values are expected or something I should investigate further.
Thanks in advance.
Danny
I have a question regarding G-Sonic behavior during measurements and EQ sweeps.
When I open a G-Sonic project and run measurements/sweeps, I sometimes hear small crackling or popping sounds during the sweeps. It does not happen continuously, but randomly during some sweeps.
I also noticed that REW reports a relatively high clock drift during these measurements.
Is this behavior normal when using G-Sonic, or could it indicate a synchronization/buffer issue somewhere in the audio chain?
My setup uses:
Mic pointed toward speakers when measured
- REW
- UMIK 2 microphone
- Convolution/EQ generated by G-Sonic ( sweep in Roon and measured with REW)
- Speakers Audio Physic compact 8
- Eversolo DMP-A8
- Amp Hypex ncorex 2*250rms 4ohm
1 .File_20260518 ( measurements gsonic) Red fr is G-Sonic AND Green is no eq
2. Microfoon recht ( here is the rew file )
3. Schermopname (75) ( Screenshot clockdrift)
The measurements themselves still seem usable, but I wanted to verify whether occasional crackling during sweeps and higher clock drift values are expected or something I should investigate further.
Thanks in advance.
Danny
Attachments
Last edited:
Keith_W
Master Contributor
For a start, you have two measurements with the same name that look very different. You have not indicated what has changed between the measurements.
And if you say your mic is pointing at the speakers, then you are using the wrong calibration file.
Hi Keith,
Thanks, I’ll update it. ( and thank you for your amazing write up how to work with Rew)
The red FR is with G-Sonic applied, and the green FR is the in-room response without EQ.
I also noticed that I performed the REW measurement using the 90-degree calibration file, while in G-Sonic I had selected the correct mic calibration file.
I’ll redo the REW measurements using the correct calibration file.
Last edited:
petersiemeel
New Member
- Joined
- Jan 11, 2026
- Messages
- 2
- Likes
- 1
Hi @OCA,
I'm new into the world of FIR DSP and your response triggered I question hope you can elaborate. You stated that the pre-ringing is -60dB quieter that IR peak. How do you see that from the Graph below? I see a graph in % and the purple graph (with FIR / pre-ringing) has a peak before t=0 of 40% (=-8dB) of the IR peak of 100%. Quite a difference from -60dB.
Am I missing something?
Also any comment from anyone on the auditability of this pre-ringing, I would appreciate from anyone with experience or maybe even some research evidence?
I'm new into the world of FIR DSP and your response triggered I question hope you can elaborate. You stated that the pre-ringing is -60dB quieter that IR peak. How do you see that from the Graph below? I see a graph in % and the purple graph (with FIR / pre-ringing) has a peak before t=0 of 40% (=-8dB) of the IR peak of 100%. Quite a difference from -60dB.
Am I missing something?
Also any comment from anyone on the auditability of this pre-ringing, I would appreciate from anyone with experience or maybe even some research evidence?
Keith_W
Master Contributor
Please see this Siemens article on the audibility of pre-ringing. You want the section on "temporal masking".
- Thread Starter
- #393
No, you are right, 60% less amplitude but -8dB only in logarithmic scale, thanks for the attention but inaudibility is more about being within that 5ms.
I tried G Sonic Filter for 2 systems:
1. Monitor speakers: Just a pair of vintage monitor speakers from Kenwood — old, very cheap, used in my small working room. Sonos Connect Amp as the amplifier, connected to a MacBook Air M2 via ThinkPad Dock. ThinkPad Dock output is 3.5 mm, going to Sonos Connect Amp by a 3.5 mm to 2 RCA cable. I know this chain is very bad, but it's just a monitor speaker to play ambient sounds like jazz or lo-fi. Before applying the DSP, the sound was very bad. But with G Sonic Filter and CamillaDSP, it became very good compared with before.
2. Main system: KEF R3 + SVS PB-1000 sub. DAC: D90 III Discrete → Preamp: Pre90 → Power Amp: LA90D x2. I use XLR out from Pre90 to KEF R3 and RCA out to SVS PB-1000 (2 RCA L+R). I set LF on SVS sub at 85Hz, with volume and phase adjusted with REW.
Currently I measured this system with REW (UMIK-1) and applied PEQ (20–350 Hz, no boost, only cut peaks, Toole curve) for L+sub and R+sub separately on D90 III. For me, the sound is very good compared with MiniDSP Flex (with Dirac live/ both full range and limited range) or WiiM Pro Plus + SMSL D6s (Apply same PEQ on Wiim)
I tried measuring L+sub and R+sub with G Sonic and applying CamillaDSP to this system (limited range to 350 Hz, Toole curve, boost medium +6 dB), but I have to say the sound was lifeless. So I'm happy to go back to simple PEQ for my main system.
1. Monitor speakers: Just a pair of vintage monitor speakers from Kenwood — old, very cheap, used in my small working room. Sonos Connect Amp as the amplifier, connected to a MacBook Air M2 via ThinkPad Dock. ThinkPad Dock output is 3.5 mm, going to Sonos Connect Amp by a 3.5 mm to 2 RCA cable. I know this chain is very bad, but it's just a monitor speaker to play ambient sounds like jazz or lo-fi. Before applying the DSP, the sound was very bad. But with G Sonic Filter and CamillaDSP, it became very good compared with before.
2. Main system: KEF R3 + SVS PB-1000 sub. DAC: D90 III Discrete → Preamp: Pre90 → Power Amp: LA90D x2. I use XLR out from Pre90 to KEF R3 and RCA out to SVS PB-1000 (2 RCA L+R). I set LF on SVS sub at 85Hz, with volume and phase adjusted with REW.
Currently I measured this system with REW (UMIK-1) and applied PEQ (20–350 Hz, no boost, only cut peaks, Toole curve) for L+sub and R+sub separately on D90 III. For me, the sound is very good compared with MiniDSP Flex (with Dirac live/ both full range and limited range) or WiiM Pro Plus + SMSL D6s (Apply same PEQ on Wiim)
I tried measuring L+sub and R+sub with G Sonic and applying CamillaDSP to this system (limited range to 350 Hz, Toole curve, boost medium +6 dB), but I have to say the sound was lifeless. So I'm happy to go back to simple PEQ for my main system.
- Thread Starter
- #395
Do you have impulse graphs of sub and one of the speakers handy. I wanna see the sub time alignment.
in this very forum there's a test provided by Dr. Ulrich Brüggemann (Acourate) in the topic "Pre-ringing with linear phase room EQ filters". in this forum in French you have a graph of audibility threshold and the reference it's deducted from : https://forum-hifi.fr/post-55888.html . It is using the energy time curve (ETC) in REW.
I made a bunch of variation with EP corrected and it's not black and white : although none of my filters failed the test provided by Dr. Ulrich Brüggemann and looked OK (though sometimes borderline), they sounded all different and at the moment I'm back to just a Low Shelf limited by a High Pass @ 23 Hz both channels + a single Pk for Left channel and 2 for the Right, No Phase correction. And unless the attached graphs (R channel) show obvious issues that could be corrected by altering phase, I'll say it's good enough for me.
It's more a philosophical stand point maybe : if the speakers are good use as little DSP/DRC as possible for satisfying results
Attachments
Apologies I forgot to mention I also have the brilliant VBA Ultra both channels of course.I also have a more bass controlled option, ie for solo cello, but the most tweaked channel just have VBA LS HP and 3 Pk : less is more
I also pay great attention to match channels, ie I have a Pk @ 382 for R but I don't correct over what is required to compensate for the LS there since I decided not to have a Pk in the same region for L
Attachments
Last edited:
Dave Loves Music
New Member
- Joined
- May 26, 2026
- Messages
- 4
- Likes
- 0
Hello,
I tried the tool in my system. After creating the filters, I imported them into REW and applied the calculation to see the effect (Left * FL Filters).
The result shows it did not impact frequencies under 100h. Rather it boosted where I was expecting cuts and HF corrected modestly where i expected it to cut more (some parts even boosted). I did the measurement twice just to be sure - one with speakers as is and second where I manually cut the speakers HF using the analog Eq. Both measurements show those weird result. What could have gone wrong?
P.S. Before running the measurements, I ran the Test and it keeps saying volume is low. I keep cranking the volume up much higher than normal normal listening level. At that volume the measurement chirps was ear-bleeding loud hence I aborted. I brought back the volume quite high but to listenable level before I took the measurements.
I tried the tool in my system. After creating the filters, I imported them into REW and applied the calculation to see the effect (Left * FL Filters).
The result shows it did not impact frequencies under 100h. Rather it boosted where I was expecting cuts and HF corrected modestly where i expected it to cut more (some parts even boosted). I did the measurement twice just to be sure - one with speakers as is and second where I manually cut the speakers HF using the analog Eq. Both measurements show those weird result. What could have gone wrong?
P.S. Before running the measurements, I ran the Test and it keeps saying volume is low. I keep cranking the volume up much higher than normal normal listening level. At that volume the measurement chirps was ear-bleeding loud hence I aborted. I brought back the volume quite high but to listenable level before I took the measurements.
Last edited:
- Thread Starter
- #399
Your measurements are corrupt. Something was wrong in your pcm audio path and there's no way GSR can generate any meaningful filters with that data:
Dave Loves Music
New Member
- Joined
- May 26, 2026
- Messages
- 4
- Likes
- 0
Thanks for the reply. I did a REW measurement beforehand which did not exhibit this issue.
Generally I use REW with a file measurement. For this measurement, I unfortunately had to link 3 USB cables into the DAC given my MLP is about 5 meters away (and my Umik cable is short).
Could it be that's the pcm audio path problem and is there any way I can use a sweep file for the measurement?
Generally I use REW with a file measurement. For this measurement, I unfortunately had to link 3 USB cables into the DAC given my MLP is about 5 meters away (and my Umik cable is short).
Could it be that's the pcm audio path problem and is there any way I can use a sweep file for the measurement?
Similar threads
- Locked
- Locked
