So I've been playing with rephase and room correction with my DEQX for a long time and I wanted to share some of my recent results in the hopes it helps someone else. I have had mixed results in the past using room correction
attempting to take measurements from a single listening position or even from 3 ft from the speaker and then trying to EQ the room.
I've always felt like the results were marginal or even made the sound worse. Well what Ive learned is there can be a huge learning curve (Im still learning) after years of playing with this stuff. SO dont give up on it ! Anyhow the best rephase tutuorial Ive found was by someone who goes by 'bear' (ive attached the tutorial).
Anyways, I added something of my own that I just thought of that made a HUGE difference to the results.
Previously Ive always just taken a measurement from a single point for each speaker and then tried correcting the phase and FR from that point. But I think the flaw is of course when you move your head 1 inch or more from that spot the FR changes and you're back at square one.
So I finally tried taking measurements from 48" and 78" at 9 degree increments (0 to 45 degrees) for each speaker and then vector averaging the measurements in REW and then correcting the resulting phase using THAT measurement. Also I started the sweeps at the lowest frequency wavelength from the distance measured. ie the 48" measurement is 285hz, the 78" measurement was 175hz. What a difference. I highly suggest trying this and creating fir filters this way if you havent already.
I think it does a way better job of averaging out the FR and phase in the room and the reflection come back more in phase which opens up the sound stage. Being that every room is different I dont think it would be possible for speakers to be perfect for your room out of the factory and Im not aware of any ARC that corrects using this method automatically.
I had almost given up on room correction until I tried this and its made enough of an improvement that Im going to continue attempting to fine tune it. Let me know if you have any questions and Ill try and answer.
attempting to take measurements from a single listening position or even from 3 ft from the speaker and then trying to EQ the room.
I've always felt like the results were marginal or even made the sound worse. Well what Ive learned is there can be a huge learning curve (Im still learning) after years of playing with this stuff. SO dont give up on it ! Anyhow the best rephase tutuorial Ive found was by someone who goes by 'bear' (ive attached the tutorial).
Anyways, I added something of my own that I just thought of that made a HUGE difference to the results.
Previously Ive always just taken a measurement from a single point for each speaker and then tried correcting the phase and FR from that point. But I think the flaw is of course when you move your head 1 inch or more from that spot the FR changes and you're back at square one.
So I finally tried taking measurements from 48" and 78" at 9 degree increments (0 to 45 degrees) for each speaker and then vector averaging the measurements in REW and then correcting the resulting phase using THAT measurement. Also I started the sweeps at the lowest frequency wavelength from the distance measured. ie the 48" measurement is 285hz, the 78" measurement was 175hz. What a difference. I highly suggest trying this and creating fir filters this way if you havent already.
I think it does a way better job of averaging out the FR and phase in the room and the reflection come back more in phase which opens up the sound stage. Being that every room is different I dont think it would be possible for speakers to be perfect for your room out of the factory and Im not aware of any ARC that corrects using this method automatically.
I had almost given up on room correction until I tried this and its made enough of an improvement that Im going to continue attempting to fine tune it. Let me know if you have any questions and Ill try and answer.
tutorial is too large to attach hopefully you an dl it from this link: Bear REW RePhase Tuto | PDF | Equalization (Audio) | Amplitude
About measurements have you seen MMM technique?
Like your approach it is based on averaging.
MMM homepage and blog of Jean-Luc Ohl
http://www.ohl.to/audio/downloads/MMM-moving-mic-measurement.pdf
An other averaging method is the 'beamforming measurement technique' described by Mitchba in his ebooks 'acurate sound reproduction'.
I advice to read on everything you could from Mitchba as he gives a method by step of optimisation which is clear and focused on important points.
Eg:
Audiolense Digital Loudspeaker and Room Correction Software Walkthrough - CA Academy - Audiophile Style
Have you considered using 'DRC' too?
DRC: Digital Room Correction
Wesayso had great results with it.
Like your approach it is based on averaging.
MMM homepage and blog of Jean-Luc Ohl
http://www.ohl.to/audio/downloads/MMM-moving-mic-measurement.pdf
An other averaging method is the 'beamforming measurement technique' described by Mitchba in his ebooks 'acurate sound reproduction'.
I advice to read on everything you could from Mitchba as he gives a method by step of optimisation which is clear and focused on important points.
Eg:
Audiolense Digital Loudspeaker and Room Correction Software Walkthrough - CA Academy - Audiophile Style
Have you considered using 'DRC' too?
DRC: Digital Room Correction
Wesayso had great results with it.
Last edited:
https://rephase.org/tutorials/rePhase Tutorial-en.pdf
scribd freely take other peoples work then charge for access.
Yes the magic happened when I used vector averaging, but I think its also the way that I measured it. I dont think just using a parelelloid like Bear suggests or random measurement on the couch would have the same affect. I honestly dont know the deeper reasons for it (way above my paygrade) but my unscientific hunch is that the average phase becomes more of an ellipse than a straight line, so as you move right or left between the speakers you move along the other speakers "phase ellipse". rather than breaking a straight line of phase from a single measurement. Also I think starting the measurement at the lowest wavelength forces a kind of windowing that adds information to the measurement at each position. I also want to experiment with raising and lowering the microphone as this time I only did all measurements in the same horizontal plane. But it seems like so far that the measurement techinique is critical to the results. Also I prefer to use equalizer apo in windows rather than jriver. I also wonder what added results using a really high quality microphone could have.
Found the link to where I dl'ed the Bear tutuorial: https://rephase.org/tutorials/Bear_REW_rePhase_tuto.pdf
Hi Eliguy,
May i ask what kind of loudspeakers you use?
I wonder if it's not related to an improvement you feel: from Mitchba's reading and experience single position/ averaged gives different results. That said at the time of publishing he wasn't able to tell which one he prefered as both exhibits different but likible attributes.
If you was refering to 'random couch measurements' for Mitch's writing well, this is not what i get from this nor what he used as a technique to produce the FIR filter profile.
Those measurements are a validation that the FIR profile used is actually 'true' over a much wider spot than could be anticipated and not only for alignement of ways but even for low end anomaly (which are adressed through time domain processing).
The Fir profile were produced using either single point measurement either 'beam forming method'.
Wesayso iirc went averaged then gone back to single position measurements.
Both use fairly different loudspeakers a dual 15"+ constant directivity horn and a line array. They differs in a lot of aspects but offer one similar point: more directivity control than most loudspeakers.
The one i used for my drc experiments are 3 ways 80's design with 'wide' directivity ( with some design compromise/issues like 'waistbending' in the polar map) and i wasn't very pleased by result too.
So i wonder if it is not related?
It could be a room mode issue too which makes your approach working good to your room/loudspeaker/preference. In some rooms schroeder frequency can be high as 250hz.
Anyway thank you for sharing it.
I agree measurements are all what is about in this approach, many techniques needs experiments to find what suit you.
What do you call a 'good, measurement mic?
Something like an Earthworks, Dpa/ Bruel & Kjaer?
If yes then here is what i think of them: those are wonderful tools with improved spec ( over lesser models) on noise, spl tolerated and frequ range. Add to this the close tolerance between mics.
Some are even fine as mics for recording ( an A/B couple of Earthworks or Dpa are something to listen too!). That said for them to bring their full potential you'll need some very transparent mic preamp and adc.
This won't come cheap.
Within their own limitations ( they usually don't cope with high spl) entry level are ok if the calibration file you can trust.
Mid level isemicon models ( great quality/price ratio imho) are tempting me atm to replace the cheap one i use.
I already own clean preamp and good ADC.
May i ask what kind of loudspeakers you use?
I wonder if it's not related to an improvement you feel: from Mitchba's reading and experience single position/ averaged gives different results. That said at the time of publishing he wasn't able to tell which one he prefered as both exhibits different but likible attributes.
If you was refering to 'random couch measurements' for Mitch's writing well, this is not what i get from this nor what he used as a technique to produce the FIR filter profile.
Those measurements are a validation that the FIR profile used is actually 'true' over a much wider spot than could be anticipated and not only for alignement of ways but even for low end anomaly (which are adressed through time domain processing).
The Fir profile were produced using either single point measurement either 'beam forming method'.
Wesayso iirc went averaged then gone back to single position measurements.
Both use fairly different loudspeakers a dual 15"+ constant directivity horn and a line array. They differs in a lot of aspects but offer one similar point: more directivity control than most loudspeakers.
The one i used for my drc experiments are 3 ways 80's design with 'wide' directivity ( with some design compromise/issues like 'waistbending' in the polar map) and i wasn't very pleased by result too.
So i wonder if it is not related?
It could be a room mode issue too which makes your approach working good to your room/loudspeaker/preference. In some rooms schroeder frequency can be high as 250hz.
Anyway thank you for sharing it.
I agree measurements are all what is about in this approach, many techniques needs experiments to find what suit you.
What do you call a 'good, measurement mic?
Something like an Earthworks, Dpa/ Bruel & Kjaer?
If yes then here is what i think of them: those are wonderful tools with improved spec ( over lesser models) on noise, spl tolerated and frequ range. Add to this the close tolerance between mics.
Some are even fine as mics for recording ( an A/B couple of Earthworks or Dpa are something to listen too!). That said for them to bring their full potential you'll need some very transparent mic preamp and adc.
This won't come cheap.
Within their own limitations ( they usually don't cope with high spl) entry level are ok if the calibration file you can trust.
Mid level isemicon models ( great quality/price ratio imho) are tempting me atm to replace the cheap one i use.
I already own clean preamp and good ADC.
Last edited:
Hi Krivium,
Yes I do think that the type of speaker matters alot in this regard, especially its dispersion pattern. I'm using a pair of ATC SCM40's which utilize a dome midrange with a very wide dispersion upto 5khz. I have them crossed lower
however using my DEQX at around 3khz with a second order LR crossover.
I spent several hours yesterday doing a repeat of the process and refining it. I want to take it even further next time The second time I was unable to effectively knit together the measurement from different distances because they
were too far apart and I was getting phase anomalies, so Im going to try again soon.
But my process is basically this:
1) Find ideal listening axis in the horizontal AND vertical planes and align speakers to listening position. (In this case the best FR and Phase Response was on axis at 0 Degress horizontal and about 7 degress vertical). I used an equilateral triangle between the speakers and listening position. Its important to use a measuring device here such as a measuring tape or a lazer pointer because our eyes often decieve us as too what is aligned.
2) Find best relationship to rear and side walls with most balanced FR and PR
between both speakers (not necessarily each individual speaker) But essentially we're trying to get both speakers to have the most similar response in the room
3)Then take measurements at ear height from 8", 36", and listening distance (say 80-100") then take some 0hz-100hz measurements from around the room 100" inches away if you have the space. Measurements are 9 degree increments from 45 degrees to 45 degrees (10 measurements)
4) Vector average measurements for each speaker and create EQ filter file for Rephase as well as export Vector Average measurment as text for rephase.
5) Import Vector average measurement into rephase and apply REW EQ filters.
6) Export Impulse Response from rephase into REW
7) Create Excess Phase measurement from Corrected Vector file and export into rephase again
8) Use phase EQ to align phase in rephase and create correction file
9) Do measurements and make sure both speakers are time aligned to within 100th of a millisecond.
10) Take measurement from main listening position and apply minor peaking eq filter to any major bumps (no more than 3db or so)
My next step that I want to experiment would happen before Step 1, which is to take each speaker outside first and apply DEQX anechoic flattening and time alignment to each speaker and THEN do the room measurements for FIR filters. I wonder how much better we can get
I'm also curious about how much better it could get with a really good microphone (Im using a cheapy Umik from minidsp right now)
I wonder if I could rent one? Supposedly most microphones are only really accurate from around 600hz to 8khz or so.
The great thing is (without the DEQX of course and an expensive mic) experimentation this way is free. Cheers
Yes I do think that the type of speaker matters alot in this regard, especially its dispersion pattern. I'm using a pair of ATC SCM40's which utilize a dome midrange with a very wide dispersion upto 5khz. I have them crossed lower
however using my DEQX at around 3khz with a second order LR crossover.
I spent several hours yesterday doing a repeat of the process and refining it. I want to take it even further next time The second time I was unable to effectively knit together the measurement from different distances because they
were too far apart and I was getting phase anomalies, so Im going to try again soon.
But my process is basically this:
1) Find ideal listening axis in the horizontal AND vertical planes and align speakers to listening position. (In this case the best FR and Phase Response was on axis at 0 Degress horizontal and about 7 degress vertical). I used an equilateral triangle between the speakers and listening position. Its important to use a measuring device here such as a measuring tape or a lazer pointer because our eyes often decieve us as too what is aligned.
2) Find best relationship to rear and side walls with most balanced FR and PR
between both speakers (not necessarily each individual speaker) But essentially we're trying to get both speakers to have the most similar response in the room
3)Then take measurements at ear height from 8", 36", and listening distance (say 80-100") then take some 0hz-100hz measurements from around the room 100" inches away if you have the space. Measurements are 9 degree increments from 45 degrees to 45 degrees (10 measurements)
4) Vector average measurements for each speaker and create EQ filter file for Rephase as well as export Vector Average measurment as text for rephase.
5) Import Vector average measurement into rephase and apply REW EQ filters.
6) Export Impulse Response from rephase into REW
7) Create Excess Phase measurement from Corrected Vector file and export into rephase again
8) Use phase EQ to align phase in rephase and create correction file
9) Do measurements and make sure both speakers are time aligned to within 100th of a millisecond.
10) Take measurement from main listening position and apply minor peaking eq filter to any major bumps (no more than 3db or so)
My next step that I want to experiment would happen before Step 1, which is to take each speaker outside first and apply DEQX anechoic flattening and time alignment to each speaker and THEN do the room measurements for FIR filters. I wonder how much better we can get
I'm also curious about how much better it could get with a really good microphone (Im using a cheapy Umik from minidsp right now) I wonder if I could rent one? Supposedly most microphones are only really accurate from around 600hz to 8khz or so.
The great thing is (without the DEQX of course and an expensive mic) experimentation this way is free. Cheers
Last edited:
600hz to 8khz? You are kiding!:
M50 - 50kHz Measurement Microphone | Earthworks Audio
4006 Omnidirectional Microphone
Yes you could rent. It'll depend where you live: close to a big city there is chance there is distributor or dedicated pro shop ( tv/studio/live dedicated rental shop).
Try to contact an audio engineer ( free lance or not) they usually know where to ask.
M50 - 50kHz Measurement Microphone | Earthworks Audio
4006 Omnidirectional Microphone
Yes you could rent. It'll depend where you live: close to a big city there is chance there is distributor or dedicated pro shop ( tv/studio/live dedicated rental shop).
Try to contact an audio engineer ( free lance or not) they usually know where to ask.