Measure stereo channels separately or together to create filters

[Joffieb]

Hi All,
I looked into the question whether to do a combined measurement and create filters or do separate measurements for right and left and create separate filters for each channel.
The opinions online were varied with most people on the side of separate measurements. I felt that the opposite was better in that I felt that because the waves from each speaker interact and change the waves from the other, you are not measuring each channel properly when doing them alone because your sweep is not simulating the waves as affected by the other speaker’s waves.
I therefore tried an experiment. I measured the speakers combined and separately and created a universal filter for the combined measurement and separate filters for the single channel and re-measured. Interestingly enough, the measurements were very similar.
I am attaching my mdat file for you to peruse, it also has a few extra data windows. One is a re-measurement of the combined filters using the PEQ in a Wiim Mini as the other filtering had been done in a MiniDsp NanaoDigi 2x8. I also generated three target curves 5db apart to see if my sweeps fit in the acceptable region.
Please let me know what you think and if my filtered sweeps look good.
Thanks

 

[sm52]

For me, only measurements 2 and 3 are valuable for analysis. There are very strong reflections in the room. But after 3.77 ms they are in the measurement of both channels. And after 0.42 ms there is reflection only in the left channel. These are reflections at a level of 60% of the main signal. This greatly distorts the sound. There are others, but they are about 20%. Strange Step Response graph for both channels. Its appearance is typical for subs, for whom the main energy begins after 6-12 ms. But the presence of a subwoofer is not visible on the impulse graph. Based on what has been said, I would not pay attention to the correction for now while there are such oddities, but would find out and eliminate them. What can be done? In the window for selecting input and output devices, select devices whose names begin with Excl. And check in the operating system’s sound settings to see if any enhancers are enabled that would interfere with receiving correct data.

 

[Wayne A. Pflughaupt]

Hi All,
I looked into the question whether to do a combined measurement and create filters or do separate measurements for right and left and create separate filters for each channel.
The opinions online were varied with most people on the side

This is a tricky topic.

It’s fine (and useful) to take separate measurements of the left and right speakers. However, the two speakers should not have independent EQ filters above the Schroeder Frequency. My experience has been that individual filters above Schroeder whacks the stereo image, especially in the frequency range where the filters are applied.

The Schroeder Frequency will typically be between 200-300 Hz. It’s a bit different from room to room, I expect depending on how “live” the room is. To determine the exact frequency in your room, play filtered pink noise test signals (i.e. frequency-specific as opposed to broadband) through the speakers. The point where you can no longer localize the signal (i.e. sounds omnidirectional) is the transition frequency. Lacking filtered pink noise, assuming 250 Hz or so is a safe bet.

Thus, for the purposes of equalizing I run REW through both speakers, and above 200-300 Hz or so duplicate any EQ filters for both channels. But, go ahead and first take separate L/R measurements for a baseline reference. Compared to individual measurements, you might see a slight droop in the extreme high end when getting a measurement for both speakers played together, compared to the separate baseline measurements. This would be a result of the mic not being positioned perfectly between the speakers. If this is the case, ignore it – do not attempt to compensate by boosting the extreme highs.

Below Schroeder, it’s fine to apply individual (non-matching) filters to the left and right speakers.

Regards,
Wayne

 

[Joffieb]

Thanks everyone!!

 

[Joffieb]

For me, only measurements 2 and 3 are valuable for analysis. There are very strong reflections in the room. But after 3.77 ms they are in the measurement of both channels. And after 0.42 ms there is reflection only in the left channel. These are reflections at a level of 60% of the main signal. This greatly distorts the sound. There are others, but they are about 20%. Strange Step Response graph for both channels. Its appearance is typical for subs, for whom the main energy begins after 6-12 ms. But the presence of a subwoofer is not visible on the impulse graph. Based on what has been said, I would not pay attention to the correction for now while there are such oddities, but would find out and eliminate them. What can be done? In the window for selecting input and output devices, select devices whose names begin with Excl. And check in the operating system’s sound settings to see if any enhancers are enabled that would interfere with receiving correct data.

Thanks again. My setup is near-field (about a2.75m triangle) Between me and the speakers is a glass coffee table with plants and other objects on the right side but the left side is/was completely uncovered. I ran new measurements and compared when I placed a few cushions on the bare glass and the 60% impulse on the left went away. Now I need to solve the 4ms ones. Am I right that if they are below the Schroeder Integral black line, that is okay? Please see my comparable measurements attached. Thanks

 

[sm52]

The measurements look better today. Almost good. You've sorted out the left channel. Almost everyone has a coffee table. It can be temporarily moved to the side of the sofa, perpendicular to the sofa, to the line of the back wall of the sofa. The problem with the right one remains. I don't think the distortion in the right channel is caused by the room. It looks like the driver is faulty, the wire is not secure inside or outside the right speaker, or in the amplifier terminal. This could be roughly the reason. The 4 ms impulse burst may be a reflection from the ceiling or floor. If the microphone is 2.75 m, and the additional distance is 1.25 m, then the entire path of the reflected signal is approximately 4 m.
What the room does can be adjusted a little by moving the objects in the room a little. Including speakers.

 

[Joffieb]

The measurements look better today. Almost good. You've sorted out the left channel. Almost everyone has a coffee table. It can be temporarily moved to the side of the sofa, perpendicular to the sofa, to the line of the back wall of the sofa. The problem with the right one remains. I don't think the distortion in the right channel is caused by the room. It looks like the driver is faulty, the wire is not secure inside or outside the right speaker, or in the amplifier terminal. This could be roughly the reason. The 4 ms impulse burst may be a reflection from the ceiling or floor. If the microphone is 2.75 m, and the additional distance is 1.25 m, then the entire path of the reflected signal is approximately 4 m.
What the room does can be adjusted a little by moving the objects in the room a little. Including speakers.

Thanks for this reply. I honestly don't know how to read for distortion. Please advise where you see the right channel distortion and what good values would be. In case there was a problem with my speakers I remeasured them with and without the subwoofer turned on and then swapped them and did the same tests again. The figures in the distortion window for each equivalent state (swapped or original position) seemed similar. The measurements are below. Please have a look and guide me.
Thanks so much!!!

 

[sm52]

Replacing the speakers shows the same distortion. The speakers are fine. Right and left play equally. The distortions that I wrote about occur when the right channel plays along with the sub. The left channel and subwoofer play fine. Therefore, I suggest measuring the sub separately in its original location without the speakers. Sub alone? Take the R_ref_L measurement by physically disconnecting the wires from the right speaker.

 

[Joffieb]

Thanks again. Which window are you reading the distortion in and what values tell you the info please? Thanks

 

[sm52]

This is the 'distortion' tab when the right channel works together with the sub. On January 22, the distortion level reached 27%. On January 25, the same conditions produced 17% of the main sweep level.

Without subwoofer, right channel distortion is 2.7%. 50 hertz is most likely interference from the power supply. The left channel with sub has 2.2% on January 25th.

 

[Joffieb]

Thanks again. WOW!

Learned so much!!!

Thanks

 

[peterb]

This is a tricky topic.

It’s fine (and useful) to take separate measurements of the left and right speakers. However, the two speakers should not have independent EQ filters above the Schroeder Frequency. My experience has been that individual filters above Schroeder whacks the stereo image, especially in the frequency range where the filters are applied.

The Schroeder Frequency will typically be between 200-300 Hz. It’s a bit different from room to room, I expect depending on how “live” the room is. To determine the exact frequency in your room, play filtered pink noise test signals (i.e. frequency-specific as opposed to broadband) through the speakers. The point where you can no longer localize the signal (i.e. sounds omnidirectional) is the transition frequency. Lacking filtered pink noise, assuming 250 Hz or so is a safe bet.

Thus, for the purposes of equalizing I run REW through both speakers, and above 200-300 Hz or so duplicate any EQ filters for both channels. But, go ahead and first take separate L/R measurements for a baseline reference. Compared to individual measurements, you might see a slight droop in the extreme high end when getting a measurement for both speakers played together, compared to the separate baseline measurements. This would be a result of the mic not being positioned perfectly between the speakers. If this is the case, ignore it – do not attempt to compensate by boosting the extreme highs.

Below Schroeder, it’s fine to apply individual (non-matching) filters to the left and right speakers.

Regards,
Wayne

This is exactly what I did, for the same reason you state: to keep the spectral balance between the L+R speakers, but with one twist: I had REW display its calculated EQ filter values, halve them, and then apply the halved filters to each channel. Before that, though, I ran a summed sweep to find common peaks below the Shroeder frequency and applied only max 6dB cuts to each channel in the minidsp input PEQs, freeing up more output PEQ channels.