Room correction, speaker correction anything above Schroeder a mistake?

[ernestcarl]

Correction of reflexes ( different than resonances ! ) higher than 80 Hz can sound better, but just in one listening spot.

I mostly agree. But I do have a relatively small but visually noticeable "peak" at 300Hz from my main right loudspeaker present in most of the 9 different listening positions I've measured in my room -- both from over 9 multi-point sweeps and 5 separate sets of MMM measurements for that single speaker alone. That peak is absolutely not present in any of my quasi-anechoic measurements. Audible or not, I reckon a small reduction would not be harmful or detrimental based on the data. *I forgot to mention I also have a 3dB boost at 100Hz for the left main and 126Hz for the right -- which, again, from my several measurements throughout several listening spots in the room proved slightly beneficial / at the very least largely benign rather than harmful.

2. Measure with a good measurement microphone and a program like REW on axis about 50 cm up till 1 metre distance from the loudspeaker . Use a pulse window of 5 ms. Measure at 0 degrees, 15 degrees, 30 degrees. Use the ”average” function in the loudspeakerprogram.

This is what I would generally do myself, but with the addition of some vertical curves. I also found it beneficial to use as an additional "reference curve" that is derived from multiple sets of MMM throughout the listening areas per individual speaker. One can load EQ corrections to both MMM and quasi-anechoic curves at the same time / side-by-side to see how changing one affects the other.

 

[j_j]

https://www.aes.org/e-lib/browse.cfm?elib=15846

This might help you. Note, there are some subtle results from using wideband spectra in some frequency ranges, so remember that.

 

[ernestcarl]

https://www.aes.org/e-lib/browse.cfm?elib=15846

This might help you. Note, there are some subtle results from using wideband spectra in some frequency ranges, so remember that.

Not an AES member or subscriber, sad to say, so only read the first front page of the paper.

Correct me if I'm wrong, but this strongly reminds me a bit of that weird looking Trinnov microphone technology in systems that use their Optimizer Room Correction:

...bespoke 3D microphone that can localise the speakers’ positions and to detect early reflections. It uses four microphone elements in a precise tetrahedral arrangement...

As each speaker is measured, the resulting sound waves pass through this array of microphones, hitting each one at slightly different times. Based on this timing, the Optimizer triangulates the location of the speaker from which the sound arrived. By testing each speaker in turn, a three-dimensional map is created of where the speakers are relative to the microphone at the main listening position.

The Optimizer captures a huge amount of acoustic information in each measurement and performs time domain and frequency analysis to better understand and characterise the behaviour of the speakers in the room. The time domain is required because the frequency response alone shows irregularities in response but doesn’t provide enough information to correct the errors properly.

... The strength of Trinnov's Optimizer is its ability to create a seamless 3D sound field, where the speakers are well-matched in terms of their amplitude and time domains. The Optimizer ensures that the effect of the room and any acoustical quirks in the speaker have been completely removed [Sic?] from the system. As a result, the bass should be perfectly integrated and the speakers tonally-balanced.

Besides the 3D mapping, the semi-automated algorithms (one still needs a high level of skill to operate the device) would be using some more complex types of filtering and analyses over the various impulses captured. Though, I highly doubt it fully capable of fixing just plain bad speakers to begin with...

But the immense price point for entry! -- not to mention the underlying science behind remains a little beyond me at this point in time. Fascinating stuff, nonetheless.

 

[j_j]

Not an AES member or subscriber, sad to say, so only read the first front page of the paper.

Correct me if I'm wrong, but this strongly reminds me a bit of that weird looking Trinnov microphone technology in systems that use their Optimizer Room Correction:




Besides the 3D mapping, the semi-automated algorithms (one still needs a high level of skill to operate the device) would be using some more complex types of filtering and analyses over the various impulses captured. Though, I highly doubt it fully capable of fixing just plain bad speakers to begin with...

But the immense price point for entry! -- not to mention the underlying science behind remains a little beyond me at this point in time. Fascinating stuff, nonetheless.

The system does not have to be expensive. Why it is is a discussion I am not privileged to have.

 

[ernestcarl]

The system does not have to be expensive.

I haven't been feeling too well the past few days (thought I had covid as I've been getting flu-like symptoms but tested negative) so didn't have much more thoughts about it to ask.

But you're are absolutely right that one can do similar stuff using relatively cheap/free solutions.

The more manual process can be rather tedious or as detailed and thorough as however one wants.

I'm pretty much just using a second hand multichannel DAC that cost me a little over $100 and a UMIK-1 along with REW and rePhase (while both software are free).

It did, however, took me quite a while to learn how to take raw measurements of each channel across a wide and/or narrow space (as well as quasi-anechoic IRs) and use this as a basis for any future corrections/adjustments applied to my simple 4.1 mch system. It seems to work well enough that the predictions often match the results almost perfectly.

*predicted summed responses pretty much matches actual sweeps at the MLP (wavelet spectrogram graph for instance)

 

[YSC]

sorry for resurrecting a ancient thread, read a lot about the EQ above Schroeder Freq. wonders that if my LP is so limited that I won't be able to move laterally for more than 20cm for real listening , and anyone else in the room to listen have to sat left to the left speaker, would it actually be more beneficial to do full FR EQ than <500hz to counter all reflection induced problems? or it's still recommended to just EQ the FR below?

 

[ernestcarl]

sorry for resurrecting a ancient thread, read a lot about the EQ above Schroeder Freq. wonders that if my LP is so limited that I won't be able to move laterally for more than 20cm for real listening , and anyone else in the room to listen have to sat left to the left speaker, would it actually be more beneficial to do full FR EQ than <500hz to counter all reflection induced problems? or it's still recommended to just EQ the FR below?

I'm always leery about "fine" corrections above 500Hz even though I do it myself all the time. What EQ filters are you thinking about? You mentioned before that you have the Genelec 8030c ("neutral" monitors) so we should be able to assess how much you are deviating from the native anechoic response using your corrections. And do you have any room treatment installed? Personally, I would try to at least acoustically treat the room even if minimally at first.

 

[YSC]

I'm always leery about "fine" corrections above 500Hz even though I do it myself all the time. What EQ filters are you thinking about? You mentioned before that you have the Genelec 8030c ("neutral" monitors) so we should be able to assess how much you are deviating from the native anechoic response using your corrections. And do you have any room treatment installed? Personally, I would try to at least acoustically treat the room even if minimally at first.

This is before correction. Basically above 500 hz I have done a mild boost at 400-500hz for 2-3db, same for 2-3khz, and shelve down the mild peak at 7-10khz, and also reduce the peak around 1.3khz by 2-3 db which I suppose is the rear wardrobe door reflection or the lcd screen slightly in front of my 8030c at play as moving mic don’t really change those peaks and nulls

Room treating is not possible as that will mean removing stuffs my wife puts there, and means I loose the minimal seating distance away from speaker as I am literally sitting with my back on the rear wall

 

[ernestcarl]

This is before correction. Basically above 500 hz I have done a mild boost at 400-500hz for 2-3db, same for 2-3khz, and shelve down the mild peak at 7-10khz, and also reduce the peak around 1.3khz by 2-3 db which I suppose is the rear wardrobe door reflection or the lcd screen slightly in front of my 8030c at play as moving mic don’t really change those peaks and nulls

Room treating is not possible as that will mean removing stuffs my wife puts there, and means I loose the minimal seating distance away from speaker as I am literally sitting with my back on the rear wall

What I had in mind was overlaying your own desired corrections over Amir's NFS on-axis trace data and your in-room measurements to have an idea what effect it would have over the pure "direct sound". The alternative is for you to apply frequency dependent windowing filtering unto your own measurements in lieu of the Klippel anechoic traces, and compare these as well to the raw unwindowed measurements.


I digitized the traces you provided and am merely guessing what you've done:

So now we've got a "visual idea" of the potential modifications applied to the native frequency response amplitude of your speakers.

We need more information to determine whether or not this is actually making things worse -- and for that we will also need more data: e.g. MMM and actual recorded time information found in your mdat file. The latter is not trivial to interpret.

A much easier way to assess the effect of your EQ correction would be via listening test comparisons, but you need a neutral reference point e.g. neutral IEMs or headphones. Playing band-limited pink noise and alternating between each set of transducers is one way of doing it. How close can you get them to match in sound via manual EQ manipulation? Even simpler would be to just compare how well the left and right sound match by switching quickly back and forth between left and right channels -- with/out EQ.

*range entered in noise generator above may be modified to be wider and filter type used even steeper.

Limit the range to better isolate the frequency band in question. Thereafter, when you are satisfied that the two channels are now matching as close as possible, do a full range PN comparison as well.

 

[YSC]

What I had in mind was overlaying your own desired corrections over Amir's NFS on-axis trace data and your in-room measurements to have an idea what effect it would have over the pure "direct sound". The alternative is for you to apply frequency dependent windowing filtering unto your own measurements in lieu of the Klippel anechoic traces, and compare these as well to the raw unwindowed measurements.


I digitized the traces you provided and am merely guessing what you've done:

View attachment 187121


View attachment 187122

So now we've got a "visual idea" of the potential modifications applied to the native frequency response amplitude of your speakers.

We need more information to determine whether or not this is actually making things worse -- and for that we will also need more data: e.g. MMM and actual recorded time information found in your mdat file. The latter is not trivial to interpret.

A much easier way to assess the effect of your EQ correction would be via listening test comparisons, but you need a neutral reference point e.g. neutral IEMs or headphones. Playing band-limited pink noise and alternating between each set of transducers is one way of doing it. How close can you get them to match in sound via manual EQ manipulation? Even simpler would be to just compare how well the left and right sound match by switching quickly back and forth between left and right channels -- with/out EQ.

View attachment 187123
*range entered in noise generator above may be modified to be wider and filter type used even steeper.

Limit the range to better isolate the frequency band in question. Thereafter, when you are satisfied that the two channels are now matching as close as possible, do a full range PN comparison as well.

Really thanks, maybe I will go pink noise L and R separately with EQ APO filters in the channels turning on and off above 300hz and see which works better

 

[ernestcarl]

BTW, just a note on my HF EQ corrections of post #125 in this thread: I use even fewer, simpler PEQs today; and no longer bother achieving a "flat" phase response in the sub-mains xo.

 

[j_j]

Really thanks, maybe I will go pink noise L and R separately with EQ APO filters in the channels turning on and off above 300hz and see which works better

Try a pseudo-allpass signal and FFT deconvolution.

AES PNW Meeting Report - Open Source Tools for FFT Analysis

 

[YSC]

BTW, just a note on my HF EQ corrections of post #125 in this thread: I use even fewer, simpler PEQs today; and no longer bother achieving a "flat" phase response in the sub-mains xo.

may I ask when you don't eq the highs, does that sounded better to you? or it's just more consistent across the LP

 

[ernestcarl]

may I ask when you don't eq the highs, does that sounded better to you? or it's just more consistent across the LP

I do EQ the HF, but using less PEQs. Unlike your 8030C, the S8 horn waveguide will have more edge diffraction peaks and dips -- the temptation to overdo/overcorrect at times has gotten the better of me. These small "micro-corrections" do not always clearly produce better results over more general shelving type of corrections.

I have saved both a "flat" and a "shelved" HF (-2.4 dB) preset, but mostly use the latter as my default.

 

[YSC]

I do EQ the HF, but using less PEQs. Unlike your 8030C, the S8 horn waveguide will have more edge diffraction peaks and dips -- the temptation to overdo/overcorrect at times has gotten the better of me. These small "micro-corrections" do not always clearly produce better results over more general shelving type of corrections.

I have saved both a "flat" and a "shelved" HF (-2.4 dB) preset, but mostly use the latter as my default.

I just tried removing anything EQ above 420hz above what looks to be the schroeder frequency (online calculator shows the theoretical schroeder frequency being ~450hz). likely will stay with that

 

[Abelard]

Breathing further life into this old thread -hopefully. I've read through the thread, which I found very interesting. Did not find a technical description for why room correction below Schroeder (and not above) is a thing. What is driving it - technically. Maybe someone can refer me to another thread? I've seen enough references to the concept to understand it to be legitimate. Would like to understand a bit more before I go off on a useless tangent.

My situation is a bit different, in that one avenue for me is to use Dirac only above Schroeder frequency. I am a bit new to Dirac and "automatic room / speaker correction" generally. I've been using REW / DSP for 15 years or so "manually".

 

[krabapple]

YES. More science please and contributors stop relying on just Toole, especially in a field that obviously has many variables related to listening environments and much uncertainty once the sound crosses the near-ear region and is subject to individual processing and perception.

Dr. Toole is not only an experienced primary researcher, he is a knowledgeable and comprehensive compiler, reviewer, and synthesizer of research by others, as anyone who's read his book and review papers knows. So I think I'm gonna keep 'relying on Toole'.

Can there be one method that covers all situations?? I doubt it.

Who says there is only one method? For example, Toole uses a 'room correction' (bass correction) setup that I doubt anyone here has access to. And he recommends against correcting above Schroeder *if* the speaker is a good spinorama performer.

 

[krabapple]

Breathing further life into this old thread -hopefully. I've read through the thread, which I found very interesting. Did not find a technical description for why room correction below Schroeder (and not above) is a thing. What is driving it - technically.

The inevitable existence of room modes -- modal peaks and dips -- affecting bass. 'Room' correction below Schroeder is mainly for lowering the peaks at the LP. Correcting big dips needs other methods.

 

[j_j]

Well, the human brain doesn't literally compute the trajectory of a cricket ball as it runs to catch it, but the end result in that application is equivalent. Are you saying that the end result in the human's brain is nothing like the separation of the direct sound from the ambience, or simply that the biological mechanism that achieves it is nothing like the mathematical process of deconvolution?

What part of highly nonlinear did you miss?

 

[luft262]

From what I'm reading people are saying it may be pointless to EQ speakers above a room's schroeder frequency? I currently cut of auto EQ at 500Hz in my room (EQ'ing only below 500Hz). Should I instead try to estimate the schroeder frequency for my room and place the cut off at that frequency instead? And why?

Thanks.