Room EQ, do's and dont's

[Krunok]

I think @JohnPM the author of REW (awesome software and thank you for the latest fix!) maybe be able to answer the question about measuring both speakers at the same time...

On that topic, let's imagine a little experiment.. Let's assume you are measuring a system consisting of 2 subwoofers and one of them has inverted polarity. If you measure them separately each of them will produce a nice flat curve, but when you start to listen to them they will sound messy. If you measure response while both are playing it will be clearly shown that they combined resposne is not flat. So, which measurement better reflects the real situation?

Edit: PS. I use the default values in REW (500ms right window), sweep mode from 10 Hz to 24 kHz and 1/12 oct smoothing. I also use Acourate's microphone alignment tool for final mic positioning. I use a multi-client ASIO driver with the Lynx Hilo, so I can run REW and loopback the sweep signal into JRiver's ASIO line input and passes it through the convolution engine with the correction filters, digital XO, etc., in the circuit...

So your graph is showing a single sweep from left and right speaker?

 

[Krunok]

Here are the comaprison between RTA and average of 3 sweeps taken on the center and left and right boundaries of my sofa (app 1m of width), head height. RTA contained app 70 samples.

Left:



Right:




Both:




As you can see in all 3 cases measurements are pretty close. My guess is they would be even closer if I have made 9 instead of 3 sweep measurements and averaged them.

 

[mitchco]

On that topic, let's imagine a little experiment.. Let's assume you are measuring a system consisting of 2 subwoofers and one of them has inverted polarity. If you measure them separately each of them will produce a nice flat curve, but when you start to listen to them they will sound messy. If you measure response while both are playing it will be clearly shown that they combined resposne is not flat. So, which measurement better reflects the real situation?


So your graph is showing a single sweep from left and right speaker?

A "room eq do" is to not just look at the frequency response REW has many different views, one of them you should be checking is the step (timing) response which will show you if there is a polarity issue. Further, I use Audiolense, which sweeps each individual driver separately and one can easily see if there are any polarity issues by looking at each drivers step response (I use 3 way digital XO). Finally, Audiolense has a feature called "automatic polarity correction" which one can enable and never worry about inverted polarity ever.

Re: single sweep for each speaker - yes. The speakers I use have constant directivity from about 400 Hz on up, so the off axis response is as smooth as the on axis response. As mentioned in the other thread, I have tried several averaging techniques, including the moving mic measurement technique: https://www.diyaudio.com/forums/multi-way/262246-moving-mic-measurement.html and quasi anechoic measurements like: http://audio.claub.net/software/FRD...curate In-Room Frequency Response to 10Hz.pdf and beamforming technique: https://www.researchgate.net/public...ment_of_Loudspeakers_Using_Beamforming_Method

As mentioned in post 33, I have taken REW measurements at 14 different positions around a 6ft x 2ft grid area that represents sitting anywhere on my 3 seat couch and there is very little variability across the the area, based on a single analysis measurement. The constant directivity gives the smooth off axis response and Acourate or Audiolenses proprietary analysis algorithm gives me the smooth bass response across the seating area.

Your responses looks good below. To be consistent, REW best practices suggests the vertical axis scale should be 60 dB starting at 45 dB and ending at 105 db SPL. I expand the horizontal axis from 10 Hz to 24 kHz like this:

Anyway, as I say your response look good!

 

[Krunok]

A "room eq do" is to not just look at the frequency response REW has many different views, one of them you should be checking is the step (timing) response which will show you if there is a polarity issue. Further, I use Audiolense, which sweeps each individual driver separately and one can easily see if there are any polarity issues by looking at each drivers step response (I use 3 way digital XO). Finally, Audiolense has a feature called "automatic polarity correction" which one can enable and never worry about inverted polarity ever.

Re: single sweep for each speaker - yes. The speakers I use have constant directivity from about 400 Hz on up, so the off axis response is as smooth as the on axis response. As mentioned in the other thread, I have tried several averaging techniques, including the moving mic measurement technique: https://www.diyaudio.com/forums/multi-way/262246-moving-mic-measurement.html and quasi anechoic measurements like: http://audio.claub.net/software/FRD_Blender/White Paper - Accurate In-Room Frequency Response to 10Hz.pdf and beamforming technique: https://www.researchgate.net/public...ment_of_Loudspeakers_Using_Beamforming_Method

As mentioned in post 33, I have taken REW measurements at 14 different positions around a 6ft x 2ft grid area that represents sitting anywhere on my 3 seat couch and there is very little variability across the the area, based on a single analysis measurement. The constant directivity gives the smooth off axis response and Acourate or Audiolenses proprietary analysis algorithm gives me the smooth bass response across the seating area.

Your responses looks good below. To be consistent, REW best practices suggests the vertical axis scale should be 60 dB starting at 45 dB and ending at 105 db SPL. I expand the horizontal axis from 10 Hz to 24 kHz like this:

View attachment 22756

Anyway, as I say your response look good!

Well, my post was not about phishing for a compliment, but is definitely a pleasure to hear one, especially when it is coming from an expert, so thank you!

P.S. I did setup my vertical scale to be 45-105dB but for some reason 45 and 105 dB labels are missing

A "room eq do" is to not just look at the frequency response REW has many different views, one of them you should be checking is the step (timing) response which will show you if there is a polarity issue.

Can you please elaborate on this? I believe that all of us who are entering the world of room EQ would agree that it would be good to have here an example of good and not so good response and what actions can we take to make it better..

 

[mitchco]

In REW when you use the overlays tab, click on "set graph limits" on the top toolbar, it is the icon right beside the gears icon top right.

An old but good article on step response and time coherency. Shows why we use a step response instead of impulse response for measuring the timing response of a loudspeaker system. Shows what a good step response looks like. https://www.stereophile.com/features/100/index.html
Unfortunately, getting into what actions can make it better can take pages of writing, which I have done, but can't replicate it here easily... The easy answer is that most DSP software provides an overall timing correction towards the ideal step response target for a passive XO speaker.

Even better is to use linear phase digital XO, which convolves the individual drivers acoustic response into the digital XO so it sums perfectly in both the frequency and time domains. Another old but good paper on that: http://www.acourate.com/XOWhitePaper.pdf Therefore, very little excess phase correction is required overall, in addition to the time alignment... Anyway, a big subject for what can be written on a forum.

 

[Krunok]

In REW when you use the overlays tab, click on "set graph limits" on the top toolbar, it is the icon right beside the gears icon top right.

That's exactly how I did it and graph scaled correctly, only 45 and 105 labels were missing.

An old but good article on step response and time coherency. Shows why we use a step response instead of impulse response for measuring the timing response of a loudspeaker system. Shows what a good step response looks like. https://www.stereophile.com/features/100/index.html
Unfortunately, getting into what actions can make it better can take pages of writing, which I have done, but can't replicate it here easily... The easy answer is that most DSP software provides an overall timing correction towards the ideal step response target for a passive XO speaker.

Even better is to use linear phase digital XO, which convolves the individual drivers acoustic response into the digital XO so it sums perfectly in both the frequency and time domains. Another old but good paper on that: http://www.acourate.com/XOWhitePaper.pdf Therefore, very little excess phase correction is required overall, in addition to the time alignment... Anyway, a big subject for what can be written on a forum.

How do I measure step response with REW? Is it integral of IR? Is it what I get when I tick "Step response" on IR graph in REW?

Do you know how DSP software provides timing correction? Is it something I can do with rePhase?

Generally speaking, if I have speakers with passive XO and my listening room is my living room, so I can't take any actions with the speakers nor with the room, what exactly can I do to make step response better using rePhase and BruteFIR?

Or should i just leave it as it is?

 

[RayDunzl]

Generally speaking, if I have speakers with passive XO and my listening room is my living room, so I can't take any actions with the speakers nor the room, what exactly can I do to make step response better using rePhase and BruteFIR?

You can do this:

My panels are passive. And I've added subs overlapping part the range of the mains. And use an automated correction tool. Rephase or Brute FIR can surely be used.

Step response

If the electronics have a DC path/ability, the electrical response to a step is a step.

If not, then the step response jumps up and slowly tapers down according to the lowest frequency it will maintain.

A 1/2 second of preamp out step response:

Speakers don't have the low frequency bandwidth of the electronics.

A half second of speaker step response (uncorrected, and a little boomy on the low end):

With correction applied (AcourateDRC and minDSP OpenDRC-DI)

But when you look at a few milliseconds, taking into account the limits of what real speakers can actually do in a real room, you can get what looks something close to a "textbook" response.

Uncorrected at the same resolution:

What does the "filter" look like?

With no "correction" the SPL and Phase of a frequency sweep at the preamp out would be a flat line:

With FIR and IIR filters applied, the amplitude and phase output of the preamp looks something like this in my case (SPL at frequency top, Phase below):

I went to Automation because creating such a filter by hand is a bit time consuming.

Result:

Sweep (both speakers playing) without correction. Not too awful...

And with:

I picked flat, can throw a slope on it elsewhere, usually don't, as there aren't much in the way of reflections to pump up the treble in the room. Nobody complains.


48Hz is a problem with my non-rectangular room, the bass goes out of phase at that frequency at the listening position. 210Hz is probably the dipole bounce/cancellation from the wall behind the speakers. You don't notice it, or it contributes to the dipole sound, which, here, can be pretty good.

Not seen conventional speakers throw such a measurably flat phase to the listener (if they really do). Conventional speakers phase I'v seen is a steeply sloped line. Don't know if that's real, or a measurement artifact.

Preamp output step response with correction (for the speakers and room) applied:

A half second

Zoomed in:

 

[Krunok]

Thank you Ray! Although I don't think I follow..

Here is how it looks with my speakers:

Left



Right



So, what do I do next?

 

[RayDunzl]

So, what do I do next?

Show the first few milliseconds.

Review the graphs above.

Buy/use automated correction tool if you don't have one.

Uncheck the first box to get only step response.

 

[Krunok]

Show the first few milliseconds.

Review the graphs above.

Buy/use automated correction tool if you don't have one.

Uncheck the first box to get only step response.

Here they are..

Left:



Right

 

[RayDunzl]

Not pretty (if it makes a difference).

Yours might be inverted (probably doesn't make an audible differerence). The controls can invert the display. Could be the amp/preamp/mic/DAC/speaker leads that cause it to be inverted.

First graph above shows "theoretical ideal" shape. Timing would be in the few millisecond range depending on the low frequency response of your speakers.

Something like this:

 

[Krunok]

Not pretty (if it makes a difference).

Yours might be inverted (probably doesn't make an audible differerence). The controls can invert the display. Could be the amp/preamp/mic/DAC/speaker leads that cause it to be inverted.

First graph above shows "theoretical ideal" shape. Timing would be in the few millisecond range depending on the low frequency response of your speakers.

Something like this:

View attachment 22776

On the bright side, both look equally "not pretty".

Ok, so what can I do to make them look more like that red line?

 

[RayDunzl]

Thank you Ray! Although I don't think I follow..

Read captions more slowly.

 

[Krunok]

Read captions more slowly.

I don't unerstand the concept. Am I supposed to correct phase to get better looking step response or ..?

 

[RayDunzl]

I don't unerstand the concept. Am I supposed to correct phase to get better looking step response or ..?

You correct amplitude at frequency, and phase, too.

This image (from above) shows how the frequency and phase correction filter adjusts the output going toward the speakers. If no correction, it would be just two flat lines.

The adjustments to the frequency response and phase will ultimately be applied to the music you play.

Running a constant amplitude sweep through the correction filter results in this jaggedy looking mess, which, if you think about it., reveals the "shape" of the filter.

Here, the filter does a couple of deep cuts in the low frequencies, wiggles the higher frequencies a little, and the sloping phase line brings the bass forward in time relative to the higher frequencies, where little or no phase correctionis needed, to adjust the timing of the output signal.

The automated tools do it quickly and precisely because they can. They use complex math instead of eyeballs and fingers.

Doing it manually, fix something, remeasure, see what you get, try again, etc. You can fiddle with it for hours.

Although I have the option to use rePhase, and used it for a while just to see, the automated solutions are much more convenient.

Click to measure. Wait a minute, Click to set a desired frequency response curve. Click to compute. Wait a moment. Click to send correction file to device. Measure the result to see if it "looks good". Play music, have lunch.

 

[pierre]

I don't unerstand the concept. Am I supposed to correct phase to get better looking step response or ..?

You can export the filter from REW and import it into Rephase. You can then play with the phase and generate a filter.
Acourate / Audiolense do the same but are easier to use.

 

[Krunok]

You correct amplitude at frequency, and phase, too.

This image (from above) shows how the frequency and phase correction filter adjusts the output going toward the speakers. If no correction, it would be just two flat lines.

View attachment 22778

The automated tools do it quickly and precisely because they can. They use complex math instead of eyeballs and fingers.

Doing it manually, fix something, remeasure, see what you get, try again, etc. You can fiddle with it for hours.

Although I have the option to use rePhase, and used it for a while just to see, the automated solutions are much more convenient.

Click to measure. Wait a minute, Click to set a desired frequency response curve. Click to compute. Wait a moment. Click to send correction file to device. Measure the result to see if it "looks good". Play music, have lunch.

Amplitude is ok as the response is linear but I can't correct phase because I can't get a good phase graph in REW. Here's what I get when I set Right window to 3,5ms (under IR Windows):



It would look better if I can shift that brown curve by -360 deg.

 

[Krunok]

Ok, there's a -360deg button in REW which made things better.



I still don't know if this is correct phase measurement so making any corrections based on this is not a good course of action.

 

[Krunok]

After time aligning both responses and setting IR window to 3.5ms, as that is when first noticeable reflection peak occurs:



Can somebody please comment if this is how phase response is supposed to look when measured correctly?

 

[mitchco]

That's exactly how I did it and graph scaled correctly, only 45 and 105 labels were missing.

How do I measure step response with REW? Is it integral of IR? Is it what I get when I tick "Step response" on IR graph in REW?

Do you know how DSP software provides timing correction? Is it something I can do with rePhase?

Generally speaking, if I have speakers with passive XO and my listening room is my living room, so I can't take any actions with the speakers nor with the room, what exactly can I do to make step response better using rePhase and BruteFIR?

Or should i just leave it as it is?

I think Ray has you sorted out. Re: scales - you are showing 10 dB per division on the frequency response chart, should be 5 dB per division. I would use the overalys chart... Also John's REW forum at Audionirvana has several tutorials on the various ways to use REW and view charts.

May I humbly suggest a bit of reading to some article links already posted? Back in post 24 are a couple of articles that show you how to manually do this in REW and rePhase. Have a look at the JRiver article and the dropbox article. As others have mentioned, it is time consuming as compared to the commercial DSP tools...