Using REW to Find Parametric Equalizer (PEQ) settings

[Chris A]

I got a request to answer the following questions about setting the EQ target curve within REW:

11 hours ago, Droogne said:

I get the cutoff, but what about the "LF cutoff", "LF rise Start & End" and the "LF Rise Slope", for the moment, I just copied you but I'm not sure what they exactly, or to be more exact, how to use those to get what I want. I played with them, and I see a difference in the target, but still not sure on what to put it.  

"LF cutoff" sets the low frequency point at which the REW EQ optimizer stops trying to correct the EQ to the target curve settings.  For something like a La Scala or a Belle that have relatively high LF cutoff points do to their limited bass bin lengths, I usually set the LF cutoff at 60-80 Hz.  After the REW EQ optimizer runs and gives you automatic optimized PEQ settings, you can still add manual low frequency boosting PEQs manually and see their simulated effect on the resulting overall frequency response.

 

"LF rise" is the frequency at which any low frequency (ramping) boost or attenuation in the target EQ curve will begin.  "LF end" is the lower frequency below "LF rise" at which the low frequency boost or attenuation stops in the target curve, and the curve remains flat. The "LF slope" sets the target curve LF boost or attenuation slope between the "LF Rise Start" and "LF Rise End" settings.  See the following screen shot for an example:

 

 

The idea is that you have some control over the resulting shape of the low frequency target curve.  I usually set the low frequency target curve to have a 1.5 to 2 dB/octave rise in response (as you decrease in frequency) below 100 Hz, and I stop the LF target curve rise at the frequency where the bass bin no longer has the capability to produce flat EQ output beyond perhaps a 3-7 dB boosting EQ below the bass bin cutoff frequency. 

 

I set the "LF cutoff" at some very low frequency to keep the target curve relatively flat for as low a frequency as is possible.  The bass bin will roll off on its own at its low frequency limit, so the usefulness of the "LF cutoff" point is usually only for perhaps ported (bass reflex) bass bins that need rapid LF cutoff below the port frequency--in order to avoid unloading the woofers at low frequencies due to a lack of damping or air springiness to restore the woofer's cone back to its neutral position. 

 

I set the "LF rise" point to 100 Hz due to the acoustical size of my room (about 100 Hz for the so-called "Schroeder frequency", which is based on the room volume and its reverberation time--RT60).  If your room has a higher Schroeder frequency, you may wish to set the "LF Rise Start" to a higher frequency.  I set the "LF Rise End" to the approximate bass bin cutoff frequency. 

 

You will find that the REW EQ optimizer doesn't really add boosting or attenuating PEQs on the ends of the frequency response (i.e., extreme highs or extreme lows), and that you'll have to add boosting or attenuating PEQs manually and see their actual effects after running another REW measurement upsweep.  I find that it's actually nice to have some explicit control over the ends of the spectrum due to the more delicate nature of the problem--trying to get more out of the loudspeaker's tweeter or bass bin than you would otherwise get with a passive balancing network.

 

Chris

[Chris A]

As a separate topic from the above REW EQ target curve settings, boosting bass frequencies in home-sized rooms (i.e., "small rooms" acoustically) is one of those more mysterious discussions that usually touch on psychoacoustics of human hearing in small spaces.  Note that some of the extreme LF correction is due to the nature of stereo (but not multichannel) recordings that have been largely attenuated below 60-100 Hz. 

 

After demastering my entire CD music collection that restored the extreme low bass frequencies on my stereo music tracks, I can tell you that having a bass boost below 100 Hz also has a component of its purpose associated with the expectations of the human hearing system in acoustically small rooms.  I add about 1.5 dB/octave boost below 100 Hz...down to my subwoofer cut off frequency of 17  Hz.  This is something that you will need to experiment with.

 

Chris

[Droogne]

2 hours ago, Chris A said:

I got a request to answer the following questions about setting the EQ target curve within REW:

 

"LF cutoff" sets the low frequency point at which the REW EQ optimizer stops trying to correct the EQ to the target curve settings.  For something like a La Scala or a Belle that have relatively high LF cutoff points do to their limited bass bin lengths, I usually set the LF cutoff at 60-80 Hz.  After the REW EQ optimizer runs and gives you automatic optimized PEQ settings, you can still add manual low frequency boosting PEQs manually and see their simulated effect on the resulting overall frequency response.

 

"LF rise" is the frequency at which any low frequency (ramping) boost or attenuation in the target EQ curve will begin.  "LF end" is the lower frequency below "LF rise" at which the low frequency boost or attenuation stops in the target curve, and the curve remains flat. The "LF slope" sets the target curve LF boost or attenuation slope between the "LF Rise Start" and "LF Rise End" settings.  See the following screen shot for an example:

 

 

The idea is that you have some control over the resulting shape of the low frequency target curve.  I usually set the low frequency target curve to have a 1.5 to 2 dB/octave rise in response (as you decrease in frequency) below 100 Hz, and I stop the LF target curve rise at the frequency where the bass bin no longer has the capability to produce flat EQ output beyond perhaps a 3-7 dB boosting EQ below the bass bin cutoff frequency. 

 

I set the "LF cutoff" at some very low frequency to keep the target curve relatively flat for as low a frequency as is possible.  The bass bin will roll off on its own at its low frequency limit, so the usefulness of the "LF cutoff" point is usually only for perhaps ported (bass reflex) bass bins that need rapid LF cutoff below the port frequency--in order to avoid unloading the woofers at low frequencies due to a lack of damping or air springiness to restore the woofer's cone back to its neutral position. 

 

I set the "LF rise" point to 100 Hz due to the acoustical size of my room (about 100 Hz for the so-called "Schroeder frequency", which is based on the room volume and its reverberation time--RT60).  If your room has a higher Schroeder frequency, you may wish to set the "LF Rise Start" to a higher frequency.  I set the "LF Rise End" to the approximate bass bin cutoff frequency. 

 

You will find that the REW EQ optimizer doesn't really add boosting or attenuating PEQs on the ends of the frequency response (i.e., extreme highs or extreme lows), and that you'll have to add boosting or attenuating PEQs manually and see their actual effects after running another REW measurement upsweep.  I find that it's actually nice to have some explicit control over the ends of the spectrum due to the more delicate nature of the problem--trying to get more out of the loudspeaker's tweeter or bass bin than you would otherwise get with a passive balancing network.

 

Chris

 

2 hours ago, Chris A said:

As a separate topic from the above REW EQ target curve settings, boosting bass frequencies in home-sized rooms (i.e., "small rooms" acoustically) is one of those more mysterious discussions that usually touch on psychoacoustics of human hearing in small spaces.  Note that some of the extreme LF correction is due to the nature of stereo (but not multichannel) recordings that have been largely attenuated below 60-100 Hz. 

 

After demastering my entire CD music collection that restored the extreme low bass frequencies on my stereo music tracks, I can tell you that having a bass boost below 100 Hz also has a component of its purpose associated with the expectations of the human hearing system in acoustically small rooms.  I add about 1.5 dB/octave boost below 100 Hz...down to my subwoofer cut off frequency of 17  Hz.  This is something that you will need to experiment with.

 

Chris

Thanks! Will have to read though it a few times again, but I think I'm getting it  Will try it out later this evening. The optimisation of my sub is another thing I'm tring to figure out  thats for later. 

[Chris A]

Bumping for those new to using REW to equalize their systems using DSP crossovers.

[pbphoto]

For ROON users, you can also export the REW EQ filters as wav files and import them into ROON's DSP engine as a convolution filter to apply room correction or house curves.  REW plus a UMIK-1 is a very powerful combo.  Red=filter off, blue = filter on.

[Wirrunna]

Chris,

This thread inspired me to crank up Inguz Equate in my Squeezebox server and have a go at using REW to find PEQ settings.

Works well, ( http://forums.slimdevices.com/showthread.php?77084-RC-(Inguz-etc-)&p=904176&viewfull=1#post904176  )

 

Where do you advise the microphone to be placed to make the REW measurement with a K-Horn?

 

I tried 1m on axis with the squaker, then 1m from the bass horn out at half bass horn height, then at the sweet spot betwwen the speakers at the best listening position.

Vastly different readings at each.

Room details (https://www.stereo.net.au/forums/topic/118361-k-horn-crossovers-from-ak-3-to-universal-to-es/ )

[Chris A]

3 hours ago, Wirrunna said:

Where do you advise the microphone to be placed to make the REW measurement with a K-Horn?

 

If you're primarily EQing bass response below the Schroeder frequency of your room, you'll need to back off from each loudspeaker to  about one wavelength (i.e., 3.3 m at 100 Hz, etc.).  This is especially an issue with Khorns, which have a wide bass mouths spread angle (72 degrees) relative to a Jubilee bass bin (about 29 degrees). This bass bin mouth spread issue alone will create issues with Khorn microphone measurements between 120 Hz--the approximate frequency at which the bass bin loses its horizontal polar control--and 400 Hz--the center crossover frequency to the midrange horn/driver. 

 

If you're primarily EQing above ~200-300 Hz, then I'd recommend measuring at 1 metre on axis, with the microphone pointed upward at 45 degrees, and vertically located at the collar between the bass bin and the top hat.

 

For my room and the Jubs, I usually place the microphone at 1 metre on-axis and EQ flat, then I check at the listening positions below 100 Hz, making any obvious changes to the bass bin EQ that seem to be consistently presenting in the measured frequency and phase response plots.

 

Chris

[Droogne]

Hey Chris

 

I'm gonna be setting up everything over the coming few days and I was still wondering about something. How do I know what delays to use for the low vs high frequency drivers in the same speakers? And how much advantage is there to be gained by placing the drivers physically at the same distance, when having the ability to use digital delays? Should I place my K510/Faital HF200 right on top of the center of my LaScala, or even way more back due to the folded horn? I'm guessing I might be able to copy the delays used by other actively crossed LaScala users with different horns right (as long as the CD is at the same physical location in relation to the bass bin)? 

[Chris A]

On 4/2/2018 at 4:27 AM, Droogne said:

How do I know what delays to use for the low vs high frequency drivers in the same speakers?

You can see the phase or group delay response of each channel of the setup from each REW sweep.  The group delay plot, in particular, will tell you the amount of delay that you need on the higher frequencies relative to the bass bin. You can either physically move the HF drivers back by the calculated distance or you can dial in more delay on the Xilica for the HF channels. 

 

By way of example, here is the phase measurement of a Cornwall II (1979) mono-amped without time correction using a REW upsweep:

 

 

And the same thing for a K-402-MEH to the same scale:

 

 

If you believe that you can't hear the difference between these two loudspeakers, let me reassure you that once you hear the loudspeaker with very low phase growth, you'll know what the difference actually is after an A-B comparison.

 

By the way, a rule of thumb: you never delay the bass bin channels in a single loudspeaker setup (assuming horn loading)...you always add delay to the midrange and tweeter channels. 

 

One exception to this rule is the Klipsch Cornwall--which has a long midrange horn but a direct radiating woofer and a front-baffle mounted tweeter.  The correct thing to do in that particular case is to delay both the tweeter and the woofer channels, then you don't have to extend the midrange horn out into the room to achieve time alignment.  There is a particularly nasty time delay mismatch between the tweeter and midrange in the Cornwall--by several wavelengths at 5-6 kHz.

 

This special case is also true for other Klipsch models with a direct radiating woofer and horn loaded midrange and front baffle-mounted tweeters (like the Chorus and Forte series, etc.  I haven't checked the home theater loudspeakers for time delay since I've not owned these models.) 

 

Here is a group delay plot of a Jubilee showing the effect of the steep slope crossover used (48 dB/octave) without sufficient delay being added back to the HF channel to bring the bass bin into time alignment with the K-402/TAD 4002 channel.  Note the rise in "excess group delay" relative to the minimum phase group delay going from right to left at the crossover frequency of 475 Hz.  The rise in excess group delay is exactly equal to the time misalignment--about 4.7 ms--in this case due to the change in crossover filter steepness without paying attention to the time alignment issues that the steep slope filter introduced and was not corrected via added HF channel time delay:

 

 

Chris

[rjp]

Hi Chris,

Excellent write up. This tool looks really interesting. Hard to believe it is free!  I have been playing around with the application on my computer for a while and am amazed at the tools and visualization this software provides. I just ordered the UMIK-1 and would like to know if you could recommend a good source where I could learn to use it effectively. There is so much information online I scarcely know where to begin.

 

My primary interest at this point is for a 2 channel setup with no sub. I have a pair of Klipsch RP-260F in a room about 28x16. I plan to experiment with a miniDSP 2x4 HD to see what it can do to produce the best sound possible with these speakers in that room. The mic should arrive tomorrow.

 

Is there a special procedure for equalization of 2 Ch systems? Seems so much is geared towards home theater and sub integration.

 

Some basic questions:

 

Do I measure one speaker at a time? I would think trying to measure both with a monaural input signal would cause an interference pattern that would ruin the measurements.

 

Where should I place the mic? 

 

The freq response and PEQ correction seems pretty straight forward, but what is all this about time/phase correction?  Is the treatment of time of arrival errors separate from the PEQ generation shown in your examples above or is this already built into the PEQ filters generated? I assume the goal is to have the wavefront of signals at all frequencies arriving at the listener at the same time, so I would need some sort of frequency based time correction.

 

Thanks for any help

 

RIch

 

 

[Chris A]

3 hours ago, rjp said:

would like to know if you could recommend a good source where I could learn to use it effectively. There is so much information online I scarcely know where to begin.

I'd recommend going to this source to download a guide: https://www.dropbox.com/s/zdhq72a1puyyxpr/REW 101 HTS Current Version.pdf?dl=0

 

Also, there is a legacy REW support guide page: http://www.hometheatershack.com/forums/rew-forum/11707-room-eq-wizard-rew-information-index-links-guides-technical-articles-please-read.html

 

3 hours ago, rjp said:

Is there a special procedure for equalization of 2 Ch systems?

Not really.  It's actually easier than multichannel--all you have to do is match the performance of essentially identical loudspeakers in symmetrical positions in the room.  If you add a sub, it gets a little more complicated, if you add a center, it gets a lot more complicated, and surrounds are incrementally more complicated, etc.

 

3 hours ago, rjp said:

Do I measure one speaker at a time?

Yes. Always.  Sometimes you can add the subwoofer(s), but only if it is very close to the stereo pair of loudspeakers.

 

3 hours ago, rjp said:

Where should I place the mic?

For all measurements except subwoofers, always place the microphone 1 metre (yard) in front of the loudspeaker, centered on the midrange/tweeter interface, looking up at 30-45 degrees.

 

3 hours ago, rjp said:

what is all this about time/phase correction?

It's exactly what is appears to be.  You want the leading wavefronts from the bass, midrange and tweeter to arrive at the same time--an impulse. There are a few ways to see the time alignment of the various drivers/horns using REW, and there are a couple of ways to facilitate time alignment (physical alignment and time delay using a DSP crossover). 

 

3 hours ago, rjp said:

Is the treatment of time of arrival errors separate from the PEQ generation shown in your examples above...or is this already built into the PEQ filters generated?

Generally, the answer is the first one, but there is some effect of flattening the frequency response that also flattens phase.  The two subjects are generally treated separately, however.  I usually correct the frequency response first, then time alignment, then fine tune the frequency response again, especially in terms of controlling the timbre of the loudspeaker if you've got timbre matching issues.

 

3 hours ago, rjp said:

I assume the goal is to have the wavefront of signals at all frequencies arriving at the listener at the same time, so I would need some sort of frequency based time correction.

Yes, you can get that with a DSP crossover and bi-amping or tri-amping.  If you're really serious, you can get special types of DSP crossovers that have something called "FIR filters" or "finite impulse response".  These can correct phase and amplitude at the same time.  But they're usually more expensive and they need a little more depth of understanding to use effectively.  Generally, you can get there without using FIR filtering, but if you've got a nasty problem, you can use it to get out of issues. 

 

Chris

[rjp]

Great. Thanks Chris.

 

Regarding the time alignment. Is this just for aligning the time of arrival from physically  different drivers, i.e., a bi-amped woofer and a tweeter fed by a time-adjustable crossover, or can it be used to correct for frequency dependent TOA in a speaker system fed by a single amp? Does that make sense? The latter would be much more complicated if I'm explaining it correctly.

[Chris A]

6 hours ago, rjp said:

Regarding the time alignment. Is this just for aligning the time of arrival from physically  different drivers, i.e., a bi-amped woofer and a tweeter fed by a time-adjustable crossover, or can it be used to correct for frequency dependent TOA in a speaker system fed by a single amp?

Mainly, it's the first issue-- time aligning drivers/horns to each other. 

 

The second issue is handled mostly by simply flattening the frequency response of the loudspeaker, which also flattens the phase simultaneously...they're coupled together.  The reason this happens is something called "minimum phase" behavior of the loudspeaker drivers/horns and the room.  This property holds (usually) until mid-bass and bass frequencies are reached due to room modal reflections which become predominant at lower frequencies.  This usually happens below the so-called Schroeder frequency of the room, at which point frequencies in the room are dominated by early reflections from room walls instead of optical-like wave behavior above the Schroeder frequency.  This is usually below 200 Hz for many home-sized listening rooms, but it it can be a lower frequency.  Below the room-loudspeaker Schroeder frequency, the sound wavelengths become significant, and usually the room's dimensions cannot hold more than one or two wavelengths before interacting with a room boundary, i.e., everything in the room is in the "near field". 

 

Note that minimum phase correction filters (the type found in most everything--such as EQ boxes and most DSP crossovers--also referred to as infinite impulse response, or IIR filters) work well down to the Schroeder frequency to correct phase.  This is a strong reason why loudspeakers with flat frequency response are also preferred by listeners--their phase in-room is also typically fairly flat. 

 

If the phase isn't flattened enough using minimum phase filters, then FIR filters can be used to flatten the phase more completely, but this can also bring about a phenomenon known as "pre-ringing" which is undesirable, and is the reason why FIR filters are so rarely used in EQ boxes and in record mastering EQ.  "Linear phase loudspeakers" are typically achieved using FIR filtering, but they usually don't sound any different or better than minimum phase-corrected loudspeakers that are well set up in-room.

 

Also, non-minimum-phase room behavior (e.g., room modal behavior below the Schroeder frequency) cannot be corrected using FIR or IIR filters.  You can move the phase around using all-pass filters, but you can't correct the phase for anything but a very tiny listening spot in the room.  All you can do is EQ down the peak behavior at specific listening points, but you cannot boost the nulls. This is what "room correction" is trying to do, such as Audyssey, YPAO,  etc.  These software/firmware-based EQ systems usually don't work very well in my experience, especially when something "automatic" is correcting for non-minimum phase room/loudspeaker response at the listening position.  The most effective approach is to use multiple subwoofers spaced around the room to "fill up the room modes" and get fewer frequency response nulls in the room. 

 

Chris

[Schu]

Chris... how are you supposed to read this?

 

 

 

Is the drop in the graph at 450Hz the shift in phase? What is the second graph line that extends outside the screen?

 

[Chris A]

The curve that you see is phase, the other that is off-screen is SPL (amplitude, in the form of sound pressure level or SPL).

 

The way that you read the phase curve is, first knowing where the crossover frequencies are located for a Cornwall (i.e., 800 Hz and about 5 kHz), you look on either side of those frequencies for big changes in phase.  You can also see it in the first derivative of phase--group delay, which shows you the changes in the slope of the phase curve:

 

 

In both cases around the crossover frequencies, there are disturbances that are otherwise called "phase growth" (in the sense of going from higher frequency to lower frequency).  The 800 Hz crossover looks symmetric about the center frequency, but you see big phase growth on either side, indicating that someone dialed in the crossover at that point to be in phase, but out of time alignment by one or more full wavelengths. (This is the case with the Cornwall in that the woofer is actually leading the midrange horn/driver by about a wavelength.)  The phase shown is REW's best guess as to what the relative phase actually is,

 

In the case of the midrange-tweeter crossover at a center frequency of ~4.5 kHz, it shows a very wide crossover interference band (due to low order crossover filters) from about 1-10 kHz, where the phase jumps by about 5 full wavelengths (the number of cycles corresponding to the difference in the midrange driver's acoustic center vs. the tweeter's acoustic center).  The jump in the phase curve at this point is in-phase, but wildly misaligned time-wise, with the tweeter getting in 5 full cycles at the crossover frequency before the midrange begins to add its output. 

 

Chris

[Chris A]

Contrast the above with a dialed-in JuBelle, which is tri-amped and time-aligned using a K-510 midrange horn, K-69-A compression driver, and a Beyma CP25 tweeter (notice the difference in vertical scales and curve smoothing...the Cornwall group delay curve is much noisier and much higher amplitude):

 

[Droogne]

Been rereading through it, and I tried to implement it on some of my readings. Tried this out in particular: some quesitions in bold.  

On 4/2/2018 at 6:02 PM, Chris A said:

 

Here is a group delay plot of a Jubilee showing the effect of the steep slope crossover used (48 dB/octave) without sufficient delay being added back to the HF channel to bring the bass bin into time alignment with the K-402/TAD 4002 channel.  Note the rise in "excess group delay" relative to the minimum phase group delay going from right to left at the crossover frequency of 475 Hz.  (is it possible that you didnt toggle the excess? Or did you mean the "1/12 curve"?) The rise in excess group delay is exactly equal to the time misalignment--about 4.7 ms--in this case due to the change in crossover filter steepness without paying attention to the time alignment issues that the steep slope filter introduced and was not corrected via added HF channel time delay:

 

 

 

Chris

 

So not sure which graphs to look at. But here are the results of my measurement done on my Center LaScala (PH4525/Faital HF200). EDIT: these are sweeps with delay or PEQs.. I'm not able to do readings right now, so I'm gonna leave them here for now. It's more about the theory behind it, and knowing I'm doing it right. I do however have one reading with PEQs and an improvised delay (same as the K510 LaScalas). At my crossover (550hz) I come at these results:

 

does this mean a 2,57ms delay on the HF channel? (I subtracted the Minimum GD from the Excess GD). Or am I wrong / is it not as simple as that? (EDIT: I come at a 7ms difference on the PEQd channel... I'll include the FR and GD graphs below here too.)

 

 

This is the excess delay as you also showed it: 

 

 

PEQd channel:

 

1/48 smoothing

 

Psychoacoustic smoothing

 

 

 

[rjp]

Chris, Mathematically it seems the ideal phase would be any linear function of frequency since this would give a single constant time shift to the entire signal, thus preserving TOA across the board. It shouldn't matter what the slope of this phase is as long as it's linear. Does this hold true for audio systems as well? You mentioned group delay above as a figure of merit. Is this simply the derivative of phase as a function of frequency?  If so then constant group delay is the goal I assume?

[Schu]

didnt Klipsch, at some point, reverse phase on mid's and uppers (that used a passive network) in order to more seamlessly align time?

[Chris A]

1 hour ago, Droogne said:

does this mean a 2,57ms delay on the HF channel? (I subtracted the Minimum GD from the Excess GD). Or am I wrong / is it not as simple as that? (EDIT: I come at a 7ms difference on the PEQd channel... I'll include the FR and GD graphs below here too.)

 

Looks like you need an additional 4 ms of delay, but you have to also look at what that's doing to the crossover region--most of this delay may be due to the crossover filters themselves.

 

Chris