To Equalize or Not to Equalize: That Is the Question

[marty]

Any comments as to what I have been told about giving up some midrange and upper range tube magic if I went with TacT

Well, you knew I'd find this post, didn't you? The issue is really a very simple one, Steve, buddy and audiophile to the nth degree, what the hell are you thinking? You have one of the most perfect room responses I've ever seen. You are pretty much flat from 200-5KHz with a gentle roll-off thereafter and a gentle roll up below 200Hz to a "Jason Bloom/Agogee" bump in the 40-80 Hz range which tails off to sub 20Hz extension. In other words, a pretty much textbook "B&K microphone" response. So exactly what are you trying to correct? There is no free lunch. Whatever ultra minor anomalies remain may certainly be reduced, but you have to expect there is a price to be paid. Perhaps it's a small one or even a negligible one, but there is always a price. Leave well enough alone. Besides, if you fed your Lamm ML3s with a DSP'd signal, both Vladimir and Paul Stubblebine would probably have a heart attacks, and we wouldn't want that now, would we? But let's get real, your system sounds musical, and is utterly superb. It is arguably among the very best that exists on the planet earth. What exactly would you be trying to accomplish with digital signal processing that you believe is missing? If you can give me a cogent answer to that question, we can have further discussions. You know I'm a TacT advocate, only because the timber of my Pipedreams is unacceptable (at least to me) without DSP correction, particularly as it applies to integrating the Gotham subs. But my needs are very different than yours. In fact, I aspire to have a system one day that doesn't require DSP to sound its absolute best. Those systems are awfully tough to come by, but you already own one!! So again I ask, exactly what are you looking for? And I ask this with the gentle reminder that perfect can be the enemy of good, or in your case, damn good. So, be careful what you ask for!!
PS Tom's article was a masterpiece. More information, more accurately detailed, will be hard to come by.

 

[microstrip]

. Besides, if you fed your Lamm ML3s with a DSP'd signal, both Vladimir and Paul Stubblebine would probably have a heart attacks, and we wouldn't want that now, would we? .

I think the ML3s would go on strike. Even my modest VTL750s would be horrified with this DSP menace!
OK, I am now listening to a CD, but I put a tube preamplifier between the CD player and the VTLs ...

 

[amirm]

I have been to Steve's room and it does sound great. However, he has a slanted ceiling which goes in the wrong direction (right to left as opposed to front to back). So even though Steve says the two channels sound identical, I don't believe it . I suspect they sound different and room correction may be able to equalize them better.

Besides, Steve is a doctor so if his amp designer does have a heart attack, he will at least know the basics to revive him.

 

[audioguy]

Besides, Steve is a doctor so if his amp designer does have a heart attack, he will at least know the basics to revive him.

And that would be to call 911?

 

[amirm]

And that would be to call 911?

I hope so. I can't imagine him trying to use his Obgyn specific knowledge .

 

[Steve Williams]

Well from Marty's post you can see why I have reservations about using TacT or something similar.

 

[FrantzM]

I think the ML3s would go on strike. Even my modest VTL750s would be horrified with this DSP menace!
OK, I am now listening to a CD, but I put a tube preamplifier between the CD player and the VTLs ...

Microstrip ..

I would not qualify the VTL 750 of "modest" .. The ML3 may do certain thing better than the 750 but the 750 does surpass it in other areas so .. it is a give an take ..

Frantz

 

[FrantzM]

Well from Marty's post you can see why I have reservations about using TacT or something similar.

It would be interesting to audition at least ... ... You like .. You keep .. You don't like ...No harm done
Frantz

 

[audioguy]

It would be interesting to audition at least ... ... You like .. You keep .. You don't like ...No harm done
Frantz

+1

 

[tonmeister2008]

Nice article Tom! You really know your stuff. I only have 5 comments I'd like to make (which is a small number for me):

1. I agree with you that the audibility of resonance peaks is greater than dips of equivalent amplitude (e.g. 3 dB versus - 3dB) with the following qualification: as the bandwidth of the dip widens (lower Q), the audibility will become similar to the peak; the detection threshold will be the same. We discovered this in a research study on the measurement and detection of resonances/antiresonance at low frequencies.

2. I agree that you 100% that consumers need to start out with the best designed speakers before trying to equalize them in the room. In-room equalization above the frequency where they become directional (e.g. ~500 Hz) can be problematic if the directivity is not constant or smooth: otherwise, you may improve the sound power response of the speaker at the expense of the direct sound. That said, we have had some success demonstrating that full-band EQ can make a poor loudspeaker sound better. But without comprehensive anechoic measurement data on the speaker, it's difficult to know what the directivity of the speaker is, and you risk making it sound worse.

3. Something you didn't mention about room correction is the need to do spatial averaging around the targeted listening seat(s) to avoid over-fitting the equalization to a too narrow localized point in space (note: your head with it's attached ears occupies a much larger space than a single microphone, so you need to move the microphone around and take the average amplitude response). The spatial averaging will also reduce steep notches from acoustic interference (which you don't want to equalize, as you mentioned).

4. I don't recommend using 1/3-octave equalizers with fixed center frequencies to equalize the speaker in the room. The reason you probably think these narrow band filters often sound worse compared to broadband EQ is that they aren't adequate for the job. For example, if the room resonance/mode doesn't have the exact same Q and center frequency as the 1/3-octave equalizer band, then the correction is not optimal, and you could end up making it sound worse or not as good as it could be. Parametric filters with adjustable Q, center frequency and gain are the better solution since they can dialed in to match the inverse of the peak/dip you are trying to correct.

5. We've done some research on evaluating different commercial room correction systems summarized here. Equalization can really improve the quality of reproduced sound if done right -- but it can also make it worse, if done incorrectly.

 

[Nyal Mellor]

Some clarifications from an acoustics perspective:
2) When we listen in a room we are typically listening to around 40% direct sound and 60% reflected sound. When a measurement mic is placed at the listening position it is measuring the 'power response' of our system above the room's transition frequency. The problem with EQing is that if you generate correction filters so that the measured response at the listening position is flat, typically the on axis direct sound will now not be flat. Solved one problem, created another.
3) Most dips are created by destructive cancellation of the direct and reflected wave from a boundary. If you try and EQ them out, most of the time it will not work because any increase in the direct sound is met by the same increase in the reflected sound. If you are not right in the cancellation null (e.g. a little to one side) then some impact on the depth of the null can be realized.
8) Like Sean said, not sure about your findings on this point. I find EQ to be best when it is applied surgically to remove resonances from room modes. In this instance you need a high Q filter (e.g. 20+) since the modal bandwidth is low (~3Hz). And obviously in order to locate the peak you need high resolution measurements, 1/24th octave preferably. Furthermore, whilst our ears do generally average over 1/3rd octave, this is not true at bass frequencies and can be easily demonstrated by playing say a note at 60Hz and one at 70Hz. Most people can detect the difference.

 

[tmallin]

Some clarifications from an acoustics perspective:
8) Like Sean said, not sure about your findings on this point. I find EQ to be best when it is applied surgically to remove resonances from room modes. In this instance you need a high Q filter (e.g. 20+) since the modal bandwidth is low (~3Hz). And obviously in order to locate the peak you need high resolution measurements, 1/24th octave preferably. Furthermore, whilst our ears do generally average over 1/3rd octave, this is not true at bass frequencies and can be easily demonstrated by playing say a note at 60Hz and one at 70Hz. Most people can detect the difference.

As to point (8): Of course one can hear the difference between 60 Hz and 70 Hz. An octave is a doubling in frequency. Going up an octave from 60 Hz is 120 Hz. There are thus only 60 Hz in that octave. From 60 Hz to 70 Hz is thus 1/6 octave, much more than a whole note on a scale. If one couldn't hear the difference between two consecutive whole notes on a bass scale, it would be difficult to appreciate the bass on any audio system.

My room with its concrete walls has exceptionally well-defined peaks and dips. The lowest audible peak is at about 27 Hz and, as you say, is about 3 Hz wide. But the octave from 20 Hz up to 40 Hz is only 20 Hz wide so the peak is about 1/7-octave wide, again more than a whole note. A 1/6-octave resolution parametric would be perfect. And even down this low, 1/3-octave EQ bands tend to work just fine, even if the equalizer is graphic with fixed bands rather than parametric. The typical considerable overlap between adjacent 1/3-octave adjustments on such devices actually works in your favor. All you need to get it right fairly quickly is a real-time display of the bass frequency response. You just adjust adjacent 1/3-octave bands until you flatten the peak without introducing a large dip. Even if you introduce a small 3-Hz-wide dip, you probably won't hear it because the ear is relatively insensitive to dips compared to peaks.

 

[tmallin]

I'm cross-posting something I wrote on another thread since it seems at least as applicable to the discussion of "to equalize or not to equalize: here.

Three other points occur to me:

(1) Some reviewers are beginning to argue that on-axis frequency balance of speakers really shouldn't be talked about so much in reviews since we are now in the era when the on-axis response can be corrected to a fare-thee-well by DSP or even analog EQ with so little fidelity loss due to the insertion of such a device that the gains should much outweigh any losses. The reviewer I'm thinking of now believes that the off-axis behavior of a speaker is much more important since this behavior cannot be well corrected by DSP.

I think I understand the point, but I'm not yet willing to go there. I have found equalization of on-axis mid and high frequency response to be quite problematic. I will therefore continue to audition speakers with the assumption that if I don't think the mid and high frequency response of a speaker sans EQ, I'm not going to take a chance with it on the assumption that I can "fix" it in my room with EQ.

(2) Take a look at the kinds of target curves almost everyone who uses a sophisticated DSP equalization device uses. I think you'll find that most everyone uses target curves which are a few dB up in the bass and which roll the top octave or two a few dB. Many keep the midrange flat, although some prefer a steady downward slope in response from bass to treble.

Now compare those target curves to the measured reported on-axis frequency responses you see of speakers in Stereophile, SoundStage or any other source which cares about reporting measurements. If you have never done this, I think you will be surprised at how few speakers start out with a raw overall balance that mimics what almost everyone favors in terms of tartget curve. The few that come close are the ones you want to most seriously audition if you haven't already.

(3) If you already own a sophisticated DSP unit, you can experiment with making the on-axis response of your speakers match the measured on-axis response of any measured speaker you find in the literature. Just adjust your target curve to match the reported measurement. You may find this sort of exercise enlightening and quite surprising. If you are like me, with some measured responses, you will be scratching your head wondering how ANYONE can like that sound.

Of course, you could just conclude that your EQ device is not correctly doing its job or that the off-axis response which you are not able to correct is messing up your judgment since your speakers' power response is probably quite a bit different from the power response of the measured speakers. But if so, you have just reinforced my conclusion in point (1).
​

 

[Nyal Mellor]

Some interesting points:

On 1) it seems some of the speaker designers really pushing the envelope e.g. Emerald Physics are designing not for frequency response primarily but for dispersion. This makes perfect sense when you can equalize the frequency response easily but it is impossible to correct a driver with poor dispersion characteristics.

On 2) Linkwitz has just published some interesting findings on his site which detail changes to the crossover in the Linkwitz Orion to move it AWAY from a target flat frequency response http://www.linkwitzlab.com/orion-rev3.htm with supposed gains to resolution, spaciousness and dynamic range

 

[tmallin]

Some interesting points:

On 1) it seems some of the speaker designers really pushing the envelope e.g. Emerald Physics are designing not for frequency response primarily but for dispersion. This makes perfect sense when you can equalize the frequency response easily but it is impossible to correct a driver with poor dispersion characteristics.

On 2) Linkwitz has just published some interesting findings on his site which detail changes to the crossover in the Linkwitz Orion to move it AWAY from a target flat frequency response http://www.linkwitzlab.com/orion-rev3.htm with supposed gains to resolution, spaciousness and dynamic range

Interesting about the Orion. I owned version 1.1 of the Orions for about a year. The eventual deal breakers were a hardness/brightness in the mid-to-tweeter crossover region from 1.5 to 4 kHz and a lack of midrange transparency.

The quad-amping arrangement makes for a particular type of clarity which drew me to the speakers in my pre-purchase audition. Eventually, however, I became disenchanted by what I heard as a lack of midrange detail in recordings, despite the tonal brightness. Rightly or wrongly, once I heard this, I attributed it to the magnesium midrange cone material. It was kind of the polar opposite of what I've called the "magic midrange" of the Harbeths with their Radial midrange cone material.

For all I know, however, the problem may have been related to the crossover and originally chosen frequency response as Linkwitz now says. Certainly in speakers which inevitably interact quite a bit with the room walls, as the Orion does (it is only dipole up to the mid-tweeter crossover; since the + version it has been bipolar in the highs, widely dispersing the treble to the back and sides with a back-firing tweeter) I have preferred a frequency response which "relaxes" considerably from 1 kHz on up. A dip in the 2 to 6 kHz region combined by a gentle roll off above that sounds nice with a lot of material on such speakers in my room.

 

[tmallin]

(2) Take a look at the kinds of target curves almost everyone who uses a sophisticated DSP equalization device uses. I think you'll find that most everyone uses target curves which are a few dB up in the bass and which roll the top octave or two a few dB. Many keep the midrange flat, although some prefer a steady downward slope in response from bass to treble.

Now compare those target curves to the measured reported on-axis frequency responses you see of speakers in Stereophile, SoundStage or any other source which cares about reporting measurements. If you have never done this, I think you will be surprised at how few speakers start out with a raw overall balance that mimics what almost everyone favors in terms of tartget curve. The few that come close are the ones you want to most seriously audition if you haven't already.

​

I think it is instructive to look at the anechoic frequency response graphs done for SoundStage by Canada's NRC on the SoundStage Speaker Measurements page. These graphs show on axis, off axis at various angles, and room averaged responses. I think you will find that very few of the on-axis graphs come anywhere close to the type of target curve most users of DSP equalization units end up using to adjust the response of their systems for most recordings. Most speakers measure either flat on axis or actually a bit up in response in the top octaves. The bass response is rarely elevated at all and usually is flat or sloping off well above the ultimate low-frequency cut off.

Now, some may argue that anechoic measurements are not important since that's not how speakers sound in real rooms, especially in the bass. True, real listening rooms will add low frequency peaks and dips at resonant frequencies related to room dimensions. And, unless speaker placement is well tended to, the Allison effect tends to produce a dip in the 100 to 200 Hz region, robbing much music (especially classical orchestral) of bass power since this is the "power range" of lower brass. In addition, there is a lot of talk about listening rooms adding "room lift" to the low bass.

But, frankly, "room lift" only is significant in the lowest octave, below about 40 Hz, and most speakers are anechoically weak in that octave, so room lift helps except for corner-placed subwoofers. And placing speakers well out into the room, as most discriminating listeners are wont to do, further ameliorates any tendency of the listening room to boost bass response. Further, if, as I do, you listen in the near field and toe the speakers in toward your ears, from the listening position you will hear something quite close to the on-axis anechoic response from the midrange on up.

Of all the speakers measured by SoundStage, I think you will find that the Harbeth Monitor 30 comes closest to having an anechoic on-axis measurement which mimics the type of target curve most listeners prefer for home listening of a majority of recordings. That could be one reason I find the Harbeth sound to be so well balanced, frequency response-wise.

I've also made no secret about my fondness for the balance of old Acoustic Research speaker systems. Again, the AR-303 (the 1995 remake of the vintage AR-3a) is one of the few speakers at any price which Stereophile's measurements have shown to have a measured on-axis response mimicking common target curves.

 

[VernNeal]

Normally I would say no to equalization. But my JL subwoofers have adjustments for the room. And my room is tile with carpet out front of the speakers and some room tunes in the corners. Aren't we equalizing with each cable change, componet change etc... My Mac preamp has bass/trble. I just leave them flat. Do what makes you happy

 

[flez007]

This thread is a"Paradigm Killer" in the good sense... Educational and VERY interesting content, a second round reading it is my Sunday task!

Thanks!

 

[Phelonious Ponk]

Aren't we equalizing with each cable change, componet change etc... My Mac preamp has bass/trble. I just leave them flat. Do what makes you happy

Either that, or we are avoiding equalizing with every cable, component, etc.... It's one or the other.

P

 

[stellavox]

What do you equalize

Hello gentlemen,

I have been following this thread for a number of months and wish to add my own comments. I've been using equalization for over a year since I built a DIY version of the Cello Palette preampifier from parts found in a CT warehouse containing the "remains" of the Cello operation.

From reading the posts to this thread, it appears that the discussion so far has centered around the pros and cons of equalization to correct system and or room problems. When successful, the results are then applied in what I would call a "static" manner, to create what has been labeled as "digital (or analog) room correction".

Not much mention has been made of the manner in which I use the Palette - to perform "dynamic" corection to individual music selections or, more typically, whole albums. I "insert" the Palette when listening to a lot of what I will call popular digital (CD) music - especially that with female vocals. Now turning down the 2 and 5Khz bands by just a few dB or so removes enough "glare" on the voice that I can listen through a whole album instead of immediately switching to something different, as used to be the case. I call this the "subtractive" mode of operation; the "additive" mode is used when listening to older music, a lot of that produced in the 50's or early 60's, which apart from having little or no real bass, has little mid-bass. In this instance, turning up the equalization at 120 hz or so brings that spectrum and my overall enjoyment level "alive". Correction below that (at 20Hz) does little.

The big problem with applying this "dynamic" equalization relates to what I will call "ease of use". To make it really practical, you need to be able to control it quickly AND right from your listening position - and that's where the operational characteristics of every equalizer I've ever seen fall down. They all require you to get up; go to your equipment rack; fiddle with a control; then sit back down. In my case, I separated the equalizer portion of the Palette from the preamplifier function and mounted it in a "roll-around" right in front of my couch, with a 30ft umbilical that connects it back to the rest of the electronics. Now that I'm up the learning curve regarding how the controls change the sound, I can "equalize" a particular selection within the first few seconds, and then settle back to enjoy the music.

Thank you

Charles