Redefine your budget room EQ 'flat' target curve to Harman's pro curve

[Wayne A. Pflughaupt]

If a system (that doesn’t have reasonably linear response to start with) sounds worse after equalizing, then the equalizing wasn’t done correctly. That goes for manual as well as auto-EQ platforms. "Before" and "after" measurements, I'm confident, would show the problems with the poorly-done "after."

Regards,
Wayne A. Pflughaupt

 

[Wayne A. Pflughaupt]

Personally, I would always prefer the pointed up orientation in order to best capture the direct sound of the speakers and most of the relevant reflected sound from the ceiling and walls, including the wall behind you. They all influence the sound you hear.

I’ve heard this a lot, but remain unconvinced as to the “importance” of the reflected sound that the mic would capture at 90°, as far as meaningful measurements are concerned. Beyond that it seems like a bad idea.

For starters, a 90° measurement isn’t going to capture reflections from walls any better than 0° orientation will. That leaves the ceiling.

As discussed in my previous post #120, 90° orientation of the measurement mic requires an appropriate calibration file to compensate for the mic’s droop in off-axis response. The calibration file boosts the high end to overcome the off-axis high frequency loss. Thus, the highs of reflected sound arriving from overhead, virtually on-axis to an upright mic, will be unnaturally boosted and therefore more pronounced than they actually are if the appropriate 90° calibration file is in place.

So why would anyone think this is a good way to take measurements??

Above the transition frequency the ears give preference to the direct sound, not the reflected, as shown in Amir’s excellent Perceptual Effects of Room Reflections article. As such I can’t see any good reason to enhance the measured effects of reflections via 90° mic orientation.

Regards,
Wayne A. Pflughaupt

 

[emcdade]

If a system (that doesn’t have reasonably linear response to start with) sounds worse after equalizing, then the equalizing wasn’t done correctly. That goes for manual as well as auto-EQ platforms. "Before" and "after" measurements, I'm confident, would show the problems with the poorly-done "after."

Regards,
Wayne A. Pflughaupt

Then almost all broadband equalization attempts are done incorrectly because:

A. We don't hear with a single microphone, we have a head with two ears.

B. EQ systems don't account for the directivity of the speaker.

Sounds to me like an EQ'd frequency response, particularly in the higher frequency range, is merely fools gold. As long as it provides a smooth graph on paper afterward, then you guys can rest a bit easier at night, while the direct sound of your speaker moves further away from its anechoic response.

 

[Fitzcaraldo215]

I’ve heard this a lot, but remain unconvinced as to the “importance” of the reflected sound that the mic would capture at 90°, as far as meaningful measurements are concerned. Beyond that it seems like a bad idea.

For starters, a 90° measurement isn’t going to capture reflections from walls any better than 0° orientation will. That leaves the ceiling.

As discussed in my previous post #120, 90° orientation of the measurement mic requires an appropriate calibration file to compensate for the mic’s droop in off-axis response. The calibration file boosts the high end to overcome the off-axis high frequency loss. Thus, the highs of reflected sound arriving from overhead, virtually on-axis to an upright mic, will be unnaturally boosted and therefore more pronounced than they actually are if the appropriate 90° calibration file is in place.

So why would anyone think this is a good way to take measurements??

Above the transition frequency the ears give preference to the direct sound, not the reflected, as shown in Amir’s excellent Perceptual Effects of Room Reflections article. As such I can’t see any good reason to enhance the measured effects of reflections via 90° mic orientation.

Regards,
Wayne A. Pflughaupt

Wayne - I think we are splitting hairs. I have already conceded several posts ago that there is likely not much difference between 0 and 90 degrees with stereo. OTOH, there is multichannel, where I think you would concede 90 degrees would be better, so that the mike's less sensitive, less consistent bottom hemisphere is not oriented toward any speaker. Also, the omni calibration curves or calibration files I have seen do not show a big difference between 0 and 90.

 

[amirm]

Then almost all broadband equalization attempts are done incorrectly because:

A. We don't hear with a single microphone, we have a head with two ears.

B. EQ systems don't account for the directivity of the speaker.

Sounds to me like an EQ'd frequency response, particularly in the higher frequency range, is merely fools gold. As long as it provides a smooth graph on paper afterward, then you guys can rest a bit easier at night, while the direct sound of your speaker moves further away from its anechoic response.

THere is no down side to broadband EQ as long as it is done in tandem with how it sounds to you. By definition if you like the correction then it is valid and anything but fool's gold.

 

[Robh3606]

You should not use EQ above a couple of hundred hertz. These auto EQ's do not take directivity into account. Your typical cone and dome 2 way is not normally designed to be a CD design and the off axis response does not mirror the on axis response. The sum of these 2 is what is being EQ'd and that messes up the on axis response. You don't want to do that. That has always been an issue with non CD speaker systems and why CD systems typically respond better to this type of EQ.

Rob

 

[Robh3606]

THere is no down side to broadband EQ as long as it is done in tandem with how it sounds to you. By definition if you like the correction then it is valid and anything but fool's gold.

Hello Amir

Do you want to EQ the on axis in this case?? I sure as heck wouldn't why corrupt the design??

Here is an example of a CD type speaker that if you look at the published on axis response you would be thinking you might need to EQ. These are from the JBL 1400 Array review in Stereophile. The first is on axis the second is an averaged measured in room response in the users listening window. It obvious from these plots that trying to correct the on axis response would destroy the in room averaged response.

Rob

 

[Fitzcaraldo215]

Then almost all broadband equalization attempts are done incorrectly because:

A. We don't hear with a single microphone, we have a head with two ears.

B. EQ systems don't account for the directivity of the speaker.

Sounds to me like an EQ'd frequency response, particularly in the higher frequency range, is merely fools gold. As long as it provides a smooth graph on paper afterward, then you guys can rest a bit easier at night, while the direct sound of your speaker moves further away from its anechoic response.

Most, but not all, EQ systems rely on a usually proprietary multipoint averaging scheme. Partly, that helps deal with the two ears thing, partly with other measurement issues of the complex sound field in a room.

Speaker directivity is only perceived indirectly at the listening position based on room reflections. We would only hear it directly only if we moved around in the room. EQ systems aim at measuring the result of that direct plus reflected sound. Speaker directivity is a useful comparative speaker performance measure, but in doing Room EQ, we are not attempting to measure speaker performance. We are measuring speaker plus room performance combined, which can alter anechoically measured directivity, sometimes dramatically.

Mikes do not have the same measurement problems in an anechoic chamber where there are no reflections by definition. The downward slope we see as desirable in the in-room target curve is not necessary anechoically, because there are no room reflections to affect response. But, of course, the degree of downward slope that is desirable in the room varies depending on the room, its cubic volume, reflective character, etc.

In a real room, unlike an anechoic chamber, the HF will be dominated by comb filtering because of the added reflections, provided you look at the output with sufficiently unsmoothed 1/X octave resolution. I agree with Wayne, Amir's excellent paper on this makes it clear.

Good EQ should not be attempting to follow all those micro, comb filtered peaks and valleys. It should rely on the psychoacoustics of the ear's ability to integrate and smooth those frequency variations, as Amir describes. However, like much of psychoacoustics, the measurement of the phenomenon is approximate. The smoothing of the comb filtering response done by EQ tools is also approximate. It never disappears in the room post-EQ, unlike in the anechoic chamber. But, well a designed EQ and target curve can provide an integrated sonic envelope for the micro variations of the comb filtering.

We do not really know exactly what a speaker will sound like in an anechoic chamber, nor do we wish to replicate that sound in our rooms. Anechoic frequency response is useful in designing and purchasing speakers, the key being consistently smoother frequency response is deemed better, almost universally. I do not know why anyone would meticulously override smoother response provided by EQ with peaks, valleys, frequency shelves, etc. in order to simulate an anechoic response which is not consistently smooth. But, to each his or her own.

 

[emcdade]

THere is no down side to broadband EQ as long as it is done in tandem with how it sounds to you. By definition if you like the correction then it is valid and anything but fool's gold.

That's not what people do though. They show off a measured response in REW to validate their efforts/money spent.

 

[Robh3606]

Speaker directivity is a useful comparative speaker performance measure, but in doing Room EQ, we are not attempting to measure speaker performance. We are measuring speaker plus room performance combined, which can alter anechoically measured directivity, sometimes dramatically.

The room cannot alter a speakers directivity. That is fixed and by design. What the room does is combine the on axis and off axis response which is dependent on the speakers directivity. That is what you end up setting the EQ for the sum of the two. You cannot apply any form of equalization without altering the on axis response. You really need to take a close look at what these auto EQ programs are doing. They are not a cure all and should be used only as a last resort not as the status quo.

Rob

 

[Fitzcaraldo215]

That's not what people do though. They show off a measured response in REW to validate their efforts/money spent.

Yes, I have seen some who do exactly as you say, some being the key word. I do not think most do. Some also attempt to tweak the target curve, which is usually based on multipoint averaged response, in order to get a "nicer" single point REW response. Apparently, they believe that single-point REW is right and their multi-point EQ tool is wrong. I don't think that makes sense, either.

A few others I have seen are paranoid that the EQ tool of choice is boosting, not just cutting, output levels in some region of frequency response irregularity. So, they myopically rebuild a micromanaged target curve full of frequency specific valleys up an down the frequency spectrum to avoid any apparent boost whatsoever, not just to avoid an infinite null. The tools I have used personally, like Audyssey and Dirac, have built-in, automatic limits to the amount of boost they will apply. They are unlikely to blow up your system by attempting to follow an infinite response null in the bass. Dirac's boost limit is +8 dB. I have never had a problem with these automatic limits on boost. Others, like Anthem ARC, do not limit boost. So, sometimes care is necessary, although I have not yet encountered a true infinite null in any room I am familiar with.

People just do all kinds of crazy things.

 

[Fitzcaraldo215]

The room cannot alter a speakers directivity. That is fixed and by design. What the room does is combine the on axis and off axis response which is dependent on the speakers directivity. That is what you end up setting the EQ for the sum of the two. You cannot apply any form of equalization without altering the on axis response. You really need to take a close look at what these auto EQ programs are doing. They are not a cure all and should be used only as a last resort not as the status quo.

Rob

Yup, my choice of words was inexact. The speaker's anechoically measured directivity never changes. But, the effective directivity in the room may be altered measurably by room reflections, and it usually is.

Yes, EQ affects the direct response based on a measurement of the sum of direct plus reflected response. But, it is the latter, the sum, that we actually hear inseparably, not the former, the direct on-axis response alone. The reflections, of course, emanate from the direct sound of the speaker, but just going in different axial directions rather than straight at you. So, there is nothing to manipulate other than the direct sound of the speaker, the source of both direct and reflected sound.

Yes, you should also reposition the speaker and/or the listener and affect the direct and reflected sound before you EQ. You can also treat the room, but that can be a long, uncertain and arduous task, sometimes with unpredictable or disappointing results. Amir has said he had to rip out expensive treatments done for him by a seemingly professional outfit.

I have been using EQ for over 8 years. I have researched and educated myself fairly well about it in that time. I do not think it solves all problems, but it is still huge in terms of what it has done for me and most of my closest friends. None of us would ever be without it. The cool things about it are it is usually easy to install in a PC or it comes in a digital preamp/processor, etc., and you can easily switch it on and off on the fly to audition the before/after results with your own ears. Absolutely no one who has heard it in my room preferred the sound with it off.

 

[dallasjustice]

The room cannot alter a speakers directivity. That is fixed and by design. What the room does is combine the on axis and off axis response which is dependent on the speakers directivity. That is what you end up setting the EQ for the sum of the two. You cannot apply any form of equalization without altering the on axis response. You really need to take a close look at what these auto EQ programs are doing. They are not a cure all and should be used only as a last resort not as the status quo.

Rob

Your basic assumption about all EQ software is that there is absolutely no intelligent approach to deal with off axis in the higher frequencies. Of course, this isn't true. I believe JJ Johnston did alot of the the research on impulse windowing methods in relation to perception. I know that at least a few of the better EQ softwares I've used window the impulse dependent on the frequency (FDW). FDW, of course, can be changed according to the user's preference. Frequency dependent windowing does answer your theoretical objections to EQ, IME. If done right, EQ has no downside.


Michael.

 

[Robh3606]

Hello Michael

Let try to understand what exactly you are saying. If I get you right you can limit the bandwidth of the correction, so I could set-up the software to ignore the directivity dependent region so above say 500Hz or so as an example. It would let the speakers do their own thing above 500 hz and do corrections below that where the room dominates the response?? In that case I have no issues at all. It's a full bandwidth correction that concerns me.

Rob

 

[Rodney Gold]

I have been messing with a zillion DSP based equalisers , full time and freq correction suites and hardware and stuff .since 97 or earlier
My take is this

DSP to cure bass issues at listening position almost mandatory below 150 hz or so

Above 200 hz .. DSP is to taste
DSP above 200 Hz is best avoided if you can.. treat the room
DSP above 200hz can work to make a compromised system work well
Its all based on some target curve at the end of it all..which is a subjective preference based thing anyway

 

[Rodney Gold]

rob , Dirac , for example ,will let you define a start and end correction limit
Say you want to correct from 150hz down , you can do so , or from 50hz to 95hz , if you so wish
It doesnt force full freq corretion on you

 

[dallasjustice]

That's not quite it.

Let's agree that some off axis energy should be EQd and some should not. I agree with that proposition and I think you do as well. FDW is designed to estimate how our hearing system works so that we only EQ the sound which is most important and leave the rest alone. At low frequencies, our hearing system incorporates a longer impulse. At higher frequencies, we mostly hear sound directly from the speaker.

You are totally correct that it is a bad idea to EQ too much off axis energy at high frequencies. So we need a model to filter out this objectionable off-axis correction. IOW, the frequency response must be filtered to exclude off axis energy at higher frequencies before the inverse EQ is applied. If the FDW is well designed and tailored to the listener's preferences, one will be able to do full range correction is a very neutral way.

Of course, there are other ways to screw up EQ (eg. Trying to invert a phase reversal null). So I'm limiting what I'm saying to your objection concerning over-correction of high frequency off axis energy.

Michael.

Hello Michael

Let try to understand what exactly you are saying. If I get you right you can limit the bandwidth of the correction, so I could set-up the software to ignore the directivity dependent region so above say 500Hz or so as an example. It would let the speakers do their own thing above 500 hz and do corrections below that where the room dominates the response?? In that case I have no issues at all. It's a full bandwidth correction that concerns me.

Rob

 

[Wayne A. Pflughaupt]

Also, the omni calibration curves or calibration files I have seen do not show a big difference between 0 and 90.

I’d be interested in seeing those, if you could point me to them – thanks.

Regards,
Wayne A. Pflughaupt

 

[RayDunzl]

I’d be interested in seeing those, if you could point me to them – thanks.

Example:

http://www.cross-spectrum.com/cslmics/700-0589_mic_report.pdf

 

[Fitzcaraldo215]

Example:

http://www.cross-spectrum.com/cslmics/700-0589_mic_report.pdf

Ray - thank you very much. I believe the curves are quite accurate, and the variation we see in HF frequency response is to be expected. But, that is why we have mike calibration files to normalize these differences at 0 or 90. I happen to use a Cross Spectrum UMIK-1 with their cal file myself for my own calibrations.