Need help with left-biased soundstage (early reflection?)

[dasdoing]

you will never find a pro setup like this, because it will never sound balanced.
note that absorption coefficients are for random incidence (reverberant room)*, while you have a very specific incidence angle.
the best you could do is to simulate the left wall by placing another reflective surface to the right. you can then put the absorber to both sides and it would be much more balanced, at least in mid and high frequency range


*meaning it is the absorbtion of the sound running several times through the material at many angles

 

[dominikz]

Apologies for the somewhat off-topic question, but can someone explain how absorption coefficients can be > 1?

As far as I remember it has to do with measurement methodology for absorbers.

This is probably oversimplified but as far as I remember an absorber is placed in a diffuse field of a reverberant room (with known RT) and RT change is measured to estimate the absorption coefficient. Since many absorbers will also have some absorption from other sides (and not just the front) while typically only the main surface aread is used for calculation, you can end up with measured absorption coefficient > 1.

Also found this article with a quick search: http://www.sengpielaudio.com/AbsorptionsgradGroesserEins.pdf

 

[localhost128]

Apologies for the somewhat off-topic question, but can someone explain how absorption coefficients can be > 1?

additional losses from edge diffraction

 

[edechamps]

So… after doing some critical listening with the MondoTrap on the left wall and the foam wedges in front of it, and after my usual EQ calibration process, it sounds much improved. Sadly I still have to apply 2 dB of balance correction to center the vocals, but that's it - the imaging is much tighter, it doesn't feel like all the reverb is coming from the left anymore, and as a bonus I get the impression that some timbre problems I had identified before (mostly resonances in the mids, likely caused by interference from the reflection) are now gone. I'm satisfied with this outcome. Here's hoping it's not all placebo

 

[pos]

If you cannot add a physical wall on the right side, you could try adding a virtual one using convolution.
You could even simulate an absorption similar to the left wall (depending on the panel you are using) by altering the frequency response of the impulse.

 

[pos]

On the subject of assessing the imbalance by hear (which is a good methodology as these kind of phenomenon can be hard to isolate with measurements), maybe you could try doing your listening tests using filtered signals (mono music), restricting the frequency response to a given range each time.
That way you would be able track down the frequency range responsible for the imbalance you are hearing.
Given your configuration it might end up lower in frequency than one would expect.

 

[edechamps]

If you cannot add a physical wall on the right side, you could try adding a virtual one using convolution.
You could even simulate an absorption similar to the left wall (depending on the panel you are using) by altering the frequency response of the impulse.

How do you propose that would work, given that our auditory system is able to differentiate between a real side reflection and a "fake" one (since they will not have the same relative level and delays between the two ears)?

On the subject of assessing the imbalance by hear (which is of course necessary as these kind of phenomenon can be hard to isolate with measurements), maybe you could try doing your listening tests using filtered tests signals (or music), restricting the frequency response to a given range each time.
That way you could track the frequency range responsible for the imbalance you are hearing. Given the configuration it might end up lower in frequency than what you expect.

Yeah I tried that at the beginning, using a bandpass filter at various frequencies. It didn't seem to make much difference, leading me to suspect the problem is broadband in nature.

 

[pos]

How do you propose that would work, given that our auditory system is able to differentiate between a real side reflection and a "fake" one (since they will not have the same relative level and delays between the two ears)?

If the delayed impulse is only added on the right channel then it should be able to convey a somewhat convincing effect with such a short listening distance.
It should at least feel more "correct" than simple level or frequency adjustments on the "direct" (ie undelayed) signal itself.

 

[pos]

Yeah I tried that at the beginning, using a bandpass filter at various frequencies. It didn't seem to make much difference, leading me to suspect the problem is broadband in nature.

If it is broadband then the lower range should be addressed with EQ, as it will not be touched as much by the panel.
Maybe you should try doing measurements focused on the low frequency range (<300Hz or so), possibly using MMM.

 

[ernestcarl]

If the delayed impulse is only added on the right channel then it should be able to convey a somewhat convincing effect with such a short listening distance.
It should at least feel more "correct" than simple level or frequency adjustments on the "direct" (ie undelayed) signal itself.

This concept is quite interesting... many constrained setups (where one can’t put treatment) could probably benefit from it. Although I haven’t studied convolution enough to even have an idea how to go about something like it.

 

[pos]

This does not even require convolution as this is completely causal.
Convolution just seems more natural/easier/cleaner to me, but it can certainly be done with a simple set of plugins: delay, level and minimum-phase EQ, a bit of routing and tuning, and voilà...

 

[OCA]

you could try doing your listening tests using filtered signals (mono music), restricting the frequency response to a given range each time

Bringing this golden oldie back to life because the wisdom here is just too good to stay buried!

Are you suggesting listening tests with mono pink noise covering a variety of different frequency bands? And what exactly to do about it after finding out the problematic region and the phase anamoly causing it? Time-inverted 2nd order allpass filters and hope they will not pre-echo or do you have some better technique? I am quite certain my inherent left side bias will be a low bass problem and not solvable with wall treatment and probably not phase correctable either with anything less than Q25. I align L&R SPL of the post filter responses 4 octaves around 250Hz in REW (2 octaves each way) so at least the vocals are centered but it never felt right doing so although I often get the vocals centered fine.