Dr. Toole's - A Rational Approach To Calibrations

[Richx200]

While reading Dr. Toole's book "Sound Reproduction" I came across the attached chapter 13.2.3 "Room Correction" And "Room Equalization" Are Misnomers. If I'm correct, Dr. Toole is saying that very good loudspeakers should not be equalized above transition 300-400 Hz and doing so could degrade a very good loudspeaker. All that is needed are adequate specifications on the loudspeakers. Anechoic data on the loudspeakers would indicate that the loudspeaker was not responsible for the small acoustical interference irregularities seen in the curve.

So I went about in search of the adequate specifications for my Focal Aria 936. I couldn't get them from Focal, and only one place had them;

Focal Aria 936 loudspeaker Measurements

Sidebar 4: Measurements

www.stereophile.com

So I copied the data (attached) and it looks like I may not need to use any automatic calibration at all. I think I can correct the low frequency hump with the subwoofer's PEQ if I need a little more; I may have to touch it up with equalization; nothing above 400 Hz.

I hope I'm right about what Dr. Toole is saying and this information maybe valuable to others. Please excuse the sloppy attachments, I'm not very good at copying pages from books.

 

[dualazmak]

Thank you for the info!

I essentially agree with Dr. Toole (and you) even I DIY revised existing rather vintage but still-really-excellent 3-way SP (YAMAHA NS-1000) into fully active one (no change at all on the heavy-rigid cabinet), I mean each of the woofer, midrange and tweeter is driven directly (with no LCR network) by dedicated three amplifiers with DSP (XO/Gain/Group-Delay) controls in upstream digital domain.

What I did/do in my multichannel "active" setup (the latest system setup can be found here in detail) is to carefully "simulate" the YAMAHA's original passive LCR network configuration, i.e. mainly XO and polarities, with my upstream DSP where I configure the XOs (cross-over Fq, the slopes of filters at both side of XO Fq) as exactly similar to the original passive network as possible. No further EQ was implemented in my DSP settings. And then, I added subwoofers and super-tweeters to the SP system both also controlled (XO/Gain/Group-Delay) by upstream DSP.

As my next step-up, I intensively measured and tuned relative time-domain behaviors of all the SP drivers for 0.1 msec precision "time alignment" over all the SP drivers (summary ref. here and here).

In this way, the total sound quality was/is much improved thanks to the full elimination of passive LCR network and direct/dedicated drive by each of the amplifiers as well as thanks to the perfect "time alignment" at my listening position; I still highly respectful for whole of YAMAHA's original design on SP drivers, cabinet, as well as their passive LCR network (now completely removed though).

As you may agree with me that our individual listening room acoustic environment is always another major "critical factor" in addition to our audio ring including the SP system.

 

[Keith_W]

Rather than accept Dr. Toole's view on correction as gospel, it helps to think a little about why he said that. His view can be summarized as - buy good speakers, and leave the upper frequencies alone. The lower frequencies below transition are heavily modified by the room, so those should be corrected. So - what are good speakers, and why should the upper frequencies be left alone?

A good speaker is one that was tuned to be flat at 1m under anechoic conditions. It has smooth directivity, meaning the off-axis response is spectrally correct. This in turn means that reflections are also spectrally correct.

If we take such a speaker and put it in a room and measure from 1m away, we observe a flat response above transition. We then measure 2m away, and we observe that the frequency response starts to fall. Measure 3m further away, and the response falls even further, and so on. This is where the Harman curve comes from - it is the natural response of a speaker which is anechoically flat in the nearfield, listened to in the farfield. He backs this up by citing preference studies (including some of his own) showing that listeners strongly prefer these speakers.

He does not always think that upper frequencies should be left alone though. Some recordings may have a treble tilt. He feels that if necessary, "broad tone controls" can be applied to knock down the treble or boost it as required.

We thus see that achieving a Harman-like curve is dependent on two things - (1) that you own good speakers in the first place, and (2) listening distance (the further away you sit, the more treble tilt you get).

But what if you don't own good speakers? Then go buy a good speaker. However - if you DIY'ed your own speakers, they would absolutely require some upper frequency correction. Be aware that correcting the axial frequency response to flat will also affect the off-axis response, the two are inextricably linked and the only solution is a physical redesign. To minimize Toole's Circle of Confusion, these speakers should be corrected to be anechoically flat at 1m.

What if you have to sit in the nearfield or sit too far away? Can you correct the upper frequencies then? I would also argue that the answer should be yes. I see no difference between a tone control to correct individual recordings which are too bright, and speakers which are permanently too bright. It is after all, the same tone control. As we have seen, even a "good" speaker designed by Toole's criteria will be permanently too bright if it is used for nearfield listening. He says as much in this post on ASR - "Finally, pay attention to the overall shape of the room curve. Usually, at least for conventional forward-firing loudspeakers, the room curve will tilt gently downward. If the shape deviates substantially from the early-reflections spinorama curve then one can suspect something is amiss in the acoustical treatment of the room. If listening confirms a problem, then one is free to try modifying the shape of the spectrum with broadband, low-Q, tone-control kinds of equalization."

What about taking a measurement in the MLP and imposing a Harman-like curve on top of it? Dr. Toole is unequivocally opposed to this idea. He uses the word "absurd" and "farce" - in Ch 12.2.23 he says "this need to search for the right-sounding room curve is a farce. The installer and customer are involved in a trial-and-error search for a curve that seems to improve a situation that was most likely caused by an unfortunate choice of loudspeakers."

Finally, Dr. Toole notes that an omnidirectional microphone is not a substitute for human hearing because "we have two ears and a brain". More than that, we are not statues that sit still, we move around in our listening chairs. All you need to do is take two measurements at your MLP at 50cm apart and you will observe that the measurements are not the same. Correcting for every little notch with a preconditioning filter will produce a notch elsewhere.

So, I have my own take-away from Dr. Toole's advice. Buy good speakers. Or if your wife won't let you, do your best with your existing speakers. Apply a broadband low-Q equalization to the upper frequencies if necessary. Pay attention to the off-axis response. And finally, do not correct for a single point in space.

 

[dualazmak]

And...

Compensation (or not?!) the age-dependent hearing decline in treble (high Fq ca. 7 kHz - 20 kHz) would be one of the other aspects in equalization to the upper frequencies.

At least I myself do implement slightly upward SPL curve (can be safely and flexibly controlled on-the-fly in analog domain) in 7 kHz to 18 kHz to compensate my age-dependent slight hearing decline. I allow younger listeners invited to my audio system to flexibly adjust such SPL curve in high Fq using the integrated amplifier dedicatedly and directly driving super-tweeters, as shared and discussed here.

 

[Tom C]

It’s part of the unique greatness of this site that you can use the search function to focus on just Dr. Toole’s posts. I have found the exercise to be personally rewarding and quite worthwhile. I’ve read his book too, but reading his comments in response to specific questions has made it easier to grasp what he’s saying. If I understand correctly, when you measure with a mic at home in your room, you detect the direct sound and the reflected sound all rolled up in a ball that you can’t tease apart, and also that it doesn’t represent what you actually hear, because in the moment of listening, your brain is able, to some extent, to separate the direct sound from the reflected sound. There’s a little more to it, because if the reflections arrive really soon after the direct sound, your brain doesn’t separate the two, and the reflections muddy the sound. Except for that, what you perceive is “real time,” and different from the measured, “steady state” conditions we are able to perform at home. When you travel at the speed of sound, a second is a very long time, and you are dealing with things that are measured in thousandths of a second.
The upshot is that for frequencies above bass, the only reliable data for guiding EQ is the anechoic spinorama like what Amir publishes when he does a speaker review. And also that a well designed speaker simply doesn’t need any EQ above Schroeder frequency. For those cases where you are using a speaker but don’t have anechoic data available, just leave it alone because the data you need simply isn’t available. That is, you are stuck with what whatever flaws the speaker has, and if you try to EQ, you will be as likely to make things worse as make them better, because you are stabbing in the dark without a reliable guide.
For bass frequencies, you have to measure, because the room you’re in dominates the frequency response. Your room will boost some of what the speaker is putting out, and suppress/neuter/attenuate other parts of what your speaker is putting out. The bass peaks can be cut with EQ. The nulls can’t be boosted effectively using EQ, so that’s where positioning your speakers or subwoofers differently can make an improvement, and adding more than one sub can also help.

 

[GXAlan]

There is an argument coming from Magic Bean Audio’s approach.

If you have anechoic data, you can correct above the transition frequency. The SDP-75 does this.

Now imagine that I have three awesome speakers that have been eq’d anechoically with the SDP75.

Now, imagine my left and right speakers are free but my center is behind the perforated/woven screen.

Should you only correct the level? But that assumes the acoustically transparent screen is equally transparent and it’s not

Should the center channel receive a different EQ?

Skywalker Sound doesn’t mix with the center behind the screen….

Magic Beans is a fancy way of using nearfield measurements to minimize room reflections in your measurement and then measure at MLP to provide some correction.

The anechoic data won’t tell you how to fine tune above the transition frequency in room, and you might not want to simply ignore it…

 

[ZolaIII]

Let's put it this way; first fix room fundamental, then fix what's left to a knee and afterwards each and every time measuring again what's left above it and slope and adjust highs in many ways you can (acoustic treatment, FIR for less saturation and if need be slope them with wide PEQ or putting them off the horizontal angle). You might buy a rather good speakers but you won't have good big reference room or possibility of ideal placement in 99% of cases. There are of course many methods and approaches and you would be fool not to use them. You don't try to kill the room entirely, just get it under control making refractions similar to each other that none is standing out (waterfall plot to RT60 decay times, improving also overall clarity index), and that way used even the room fundamental can be useful as low bass reinforcement. Use psy equal loudness compensation to SPL and much more but I have to go to work now.

 

[Geert]

So I copied the data (attached) and it looks like I may not need to use any automatic calibration at all. I think I can correct the low frequency hump with the subwoofer's PEQ if I need a little more;

Looks like the frequency response of these speakers, on axis as well as of axis, has a bit of a downward slope (not only that 90Hz hump). If this is audible to you you could try to correct it. Don't know what automated correction would make of it, if possible it might be better to dial in the slope manually. If you like the sound above without correction above 400Hz, leave it as is.

Nothing wrong with a bit of experimenting with EQ supported by anechoic measurements, and a bit of tuning to preference of the highs and lows. In the end you want to enjoy the sound.

 

[Rick Sykora]

1. Yes, it is still best to measure the speakers in your specific room. That said…
2. No, you should not need to equalize your Focals based on Stereophile’s bass depiction. Note the text disclaimer from the review: “A large part of the upper-bass peak apparent in this graph will be due to the inevitable exaggeration of the nearfield measurement technique”. One can argue about the inevitability, but the bass hump is quite exaggerated. In my mind, given its very misleading appearance, would be best that the exaggerated response portion not be shown at all.

Btw, this exaggerated bass curve is a Stereophile staple. It shows up in every speaker review I have seen. I call it the Stereophile Bump. Klippel measurements would not have this bass bump and even speaker hobbyists know how to correct to more accurately depict bass response than these Stereophile misrepresentations do.

 

[SIY]

“A large part of the upper-bass peak apparent in this graph will be due to the inevitable exaggeration of the nearfield measurement technique”.

The frustrating thing is that this is easy to correct if you have the baffle size and geometry.

 

[youngho]

If we take such a speaker and put it in a room and measure from 1m away, we observe a flat response above transition. We then measure 2m away, and we observe that the frequency response starts to fall. Measure 3m further away, and the response falls even further, and so on. This is where the Harman curve comes from - it is the natural response of a speaker which is anechoically flat in the nearfield, listened to in the farfield. He backs this up by citing preference studies (including some of his own) showing that listeners strongly prefer these speakers.

The Harman curve (when it comes to loudspeakers in rooms) comes from an average of 11 listeners' adjustment of bass and treble when given the equivalent of two "knobs" (one for bass at hinge frequency of 105 Hz +15 to -5 dB and other for treble at 2.5 kHz +5/-10 dB) using a pair of Revel F208s that were equalized to flat. You can see some discussion at https://www.audiosciencereview.com/...er-preference-curve.50602/page-5#post-1848097, also the quoted text and subsequent posts by others.

Here is the spatially averaged natural response of the F208 at the main listening position in the Harman international reference listening room:

Interesting to compare this measurement with PIRR from measurements by @amirm : https://www.audiosciencereview.com/forum/index.php?threads/revel-f208-tower-speaker-review.13192/

Here is the preferred in-room response:

I believe that it's reasonable to say that there is a substantial boost in the preferred response below 100 Hz compared with the actual pre-EQ measurement. Interestingly, there was also a depression between 200-400 Hz (overlapping floor cancellations from the two woofers and midrange crossed over at 270 Hz?) in the pre-EQ measurement, also the pre-EQ measurement upper crossover-related dip (BBC dip, haha) seemed a little more prominent than in the PIRR.

We thus see that achieving a Harman-like curve is dependent on two things - (1) that you own good speakers in the first place, and (2) listening distance (the further away you sit, the more treble tilt you get).

And (3) room furnishings and treatments, as sidewall absorption may reduce effects of horizontal directivity issues, also addressing floor and/or ceiling for vertical.

 

[Basic Channel]

Rather than accept Dr. Toole's view on correction as gospel, it helps to think a little about why he said that. His view can be summarized as - buy good speakers, and leave the upper frequencies alone. The lower frequencies below transition are heavily modified by the room, so those should be corrected. So - what are good speakers, and why should the upper frequencies be left alone?

I am probably not adding anything to your great post, but can we not rationalise it with wavelengths of frequencies? If I sit 1m from a speaker and 1m from me to speaker to wall, the reflection will cancel out at a certain frequency. Anything below a certain frequency will not complete a cycle before the reflection from the wall hits my ear, so I will hear all that as one sound. By the time I have started to perceive the lower frequencies, I will already be hearing the reflection. Whereas at higher frequencies I will already have perceived the shorter wavelength before the reflection arrives. I presume no matter the room the distance between your ears means you can always separate the room from the source above 2k or so.

it's also quite rational if you think about the effect of speaker or listener placement. You can stand in certain seemingly random places and have the bass pumping. Whereas the sound of a hi hat seems quite connected to how far you are from the speaker.

 

[Curvature]

very good loudspeakers should not be equalized above transition 300-400 Hz

They should not be "corrected". They can be equalized as you like, usually by dialing in a tilt.

Correction implies fixing nulls or peaks in a steady state response according to some target.

Reasons are to do with how our hearing works and how rooms work. It's all in Toole's book, but takes some thinking.

 

[Richx200]

They should not be "corrected". They can be equalized as you like, usually by dialing in a tilt.

Correction implies fixing nulls or peaks in a steady state response according to some target.

Reasons are to do with how our hearing works and how rooms work. It's all in Toole's book, but takes some thinking.

I have read and thought about many of the subjects Dr. Toole presents in his book, but unfortunately this chapter is about how to avoid degrading very good speakers by using automatic room correction. Dr, Toole does cover the subject of personal preference based on individual hearing extensively else where in his book. However, he does compare a microphone to ears and a brain.

Dr. Toole does say "Consequently, the processors perform equalization corrections including non-minimum-phase acoustical interference irregularities, in order to hit the specified target curve." Could be semantics.

I do agree with Dr. Toole that individuals should tune their loudspeakers to their liking. The reason I post this information is to help those, interested, in not degrading their loudspeakers.

 

[Curvature]

I have read and thought about many of the subjects Dr. Toole presents in his book, but unfortunately this chapter is about how to avoid degrading very good speakers by using automatic room correction. Dr, Toole does cover the subject of personal preference based on individual hearing extensively else where in his book. However, he does compare a microphone to ears and a brain.

Dr. Toole does say "Consequently, the processors perform equalization corrections including non-minimum-phase acoustical interference irregularities, in order to hit the specified target curve." Could be semantics.

I do agree with Dr. Toole that individuals should tune their loudspeakers to their liking. The reason I post this information is to help those, interested, in not degrading their loudspeakers.

No, you are missing the point. I'll explain a little. This is not a semantic issue.

When you "measure" a room using something like REW you are making specific choices about:

• Bandwidth
• Level
• Window
• FFT length
• Averaging
• Other complex calculations

These affect the FR and other graphs. Based on what shows up in the graph, you can then make equalization adjustments using some specific kind of filter. Most manual PEQ uses IIR filters. There are consequences to all this stuff.

Automatic room measurement and adjustment removes your ability to control some or all of these things.

I bring this up because, when you see line on an FR graph, unless you have some sense of what is being calculated, you will not know what physical events are being represented and how.

Next, hearing works, very roughly, in two ways: time sensitivity and level sensitivity, each being different per frequency. These separate into three (but really four) groups:

1. Below about 1.5kHz, frequencies are identified through the ear/brain phase locking to the physical waveform.
2. Above around 1.5kHz, this identification takes place through level.
3. Around 1.5kHz, both of the above mechanisms are active. Ability to resolve and localize sound is reduced.
4. Below around 100Hz, tactile perception is involved in bass perception. This involves skin pressure receptors as well as proprioception and other sensory pathways I do not understand very well.

For our purposes, only 1 (phase locking) and 2 (level) are important. We do not instantly become aware of particular frequency being played. Phase locking takes time, tending to be longer the lower in frequency the sound, and faster the higher (before it stops working once the frequency gets too high). This means for lower frequencies, what we hear is an average of sound energy over a short period of time. Once level sensitivity takes over at higher frequencies, the ear is mostly responding to the SPL of direct sound (so instantaneous SPL, with a small amount of time required for recovery before responding to the next wave). All of this is contained in what is called the precedence effect, the law of the first wavefront, or the Haas effect.

So, when looking at an FR chart, below the transition frequency, standard IIR filter EQ correction to a target curve will be effective because a standard steady state response will reflect what we hear (although the physical location of room modes complicates this). Around the transition frequency, IIR filter EQ will become ineffective because the peaks and dips shown in a steady state graph are nonminimum phase. Above the transition frequency, the steady state graph no longer reflects what we hear, and correction using IIR filter EQ to a specific target is wrong. Above, the transition frequency, different windows must be used to understand the FR of direct sound vs. the FR of reflected sound.

I apologize for not explaining everything, but like I said, it's all in the book.

 

[Richx200]

No, you are missing the point. I'll explain a little. This is not a semantic issue.

When you "measure" a room using something like REW you are making specific choices about:

• Bandwidth
• Level
• Window
• FFT length
• Averaging
• Other complex calculations

These affect the FR and other graphs. Based on what shows up in the graph, you can then make equalization adjustments using some specific kind of filter. Most manual PEQ uses IIR filters. There are consequences to all this stuff.

Automatic room measurement and adjustment removes your ability to control some or all of these things.

I bring this up because, when you see line on an FR graph, unless you have some sense of what is being calculated, you will not know what physical events are being represented and how.

Next, hearing works, very roughly, in two ways: time sensitivity and level sensitivity, each being different per frequency. These separate into three (but really four) groups:

1. Below about 1.5kHz, frequencies are identified through the ear/brain phase locking to the physical waveform.
2. Above around 1.5kHz, this identification takes place through level.
3. Around 1.5kHz, both of the above mechanisms are active. Ability to resolve and localize sound is reduced.
4. Below around 100Hz, tactile perception is involved in bass perception. This involves skin pressure receptors as well as proprioception and other sensory pathways I do not understand very well.

For our purposes, only 1 (phase locking) and 2 (level) are important. We do not instantly become aware of particular frequency being played. Phase locking takes time, tending to be longer the lower in frequency the sound, and faster the higher (before it stops working once the frequency gets too high). This means for lower frequencies, what we hear is an average of sound energy over a short period of time. Once level sensitivity takes over at higher frequencies, the ear is mostly responding to the SPL of direct sound (so instantaneous SPL, with a small amount of time required for recovery before responding to the next wave). All of this is contained in what is called the precedence effect, the law of the first wavefront, or the Haas effect.

So, when looking at an FR chart, below the transition frequency, standard IIR filter EQ correction to a target curve will be effective because a standard steady state response will reflect what we hear (although the physical location of room modes complicates this). Around the transition frequency, IIR filter EQ will become ineffective because the peaks and dips shown in a steady state graph are nonminimum phase. Above the transition frequency, the steady state graph no longer reflects what we hear, and correction using IIR filter EQ to a specific target is wrong. Above, the transition frequency, different windows must be used to understand the FR of direct sound vs. the FR of reflected sound.

I apologize for not explaining everything, but like I said, it's all in the book.

I'm sorry but, I don't see what this has to do with taking the loudspeaker's anechoic data on/off axis and comparing is to the REW readings taken from the room; Equalize up to 300-400 Hz and if necessary treat the HF irregularities with acoustic products to not degrade the speaker. I have read most of your reply in Dr. Tool's book and know it would apply to calibrating loudspeakers; It's just that Dr. Toole's chapter 13 is trying to save the performance of a "good loudspeaker." All of your post can/should be taken into account, just don't break the speaker.

 

[Keith_W]

Next, hearing works, very roughly, in two ways: time sensitivity and level sensitivity, each being different per frequency. These separate into three (but really four) groups:

1. Below about 1.5kHz, frequencies are identified through the ear/brain phase locking to the physical waveform.
2. Above around 1.5kHz, this identification takes place through level.
3. Around 1.5kHz, both of the above mechanisms are active. Ability to resolve and localize sound is reduced.
4. Below around 100Hz, tactile perception is involved in bass perception. This involves skin pressure receptors as well as proprioception and other sensory pathways I do not understand very well.

Excellent post. To express the above point in another way, hearing localisation works via ITD (interaural time delay) and ILD (interaural level delay). Both work together at the frequency thresholds described by @Curvature - the so-called "duplex theory of sound localisation". ITD has a lower limit due to the width of the head, which separates the ears. The average human head is between 15-20cm wide. It is traditionally thought that long wavelengths below 80Hz can not be localised. I linked to a fairly recent paper in this thread which showed otherwise.

BTW, I found some articles on bat hearing Sound localisation is even more important for bats because they use sound to navigate. But they have tiny heads and massive ears. Because of this, it is thought that bats do not use ITD for sound localisation, instead bats almost exclusively use ILD.

The most important point IMO is that the non-minimum phase response should not be corrected. @Curvature made a very important point - automated software removes your control of what should/should not be corrected. IMO it's a good crutch for beginners, and it will produce acceptable results for some people. But there is a strong likelihood that it will do something inappropriate and degrade the sound. Just look at that Dirac thread and all those people complaining about strange sounding results. You can correct it to a target curve and it still sounds harsh. What's going on? Well, read @Curvature's post again and again until all of it is digested.

Can't give it enough likes. Once again, loved your post.

 

[Justdafactsmaam]

Let's put it this way; first fix room fundamental, then fix what's left to a knee and afterwards each and every time measuring again what's left above it and slope and adjust highs in many ways you can (acoustic treatment, FIR for less saturation and if need be slope them with wide PEQ or putting them off the horizontal angle). You might buy a rather good speakers but you won't have good big reference room or possibility of ideal placement in 99% of cases. There are of course many methods and approaches and you would be fool not to use them. You don't try to kill the room entirely, just get it under control making refractions similar to each other that none is standing out (waterfall plot to RT60 decay times, improving also overall clarity index), and that way used even the room fundamental can be useful as low bass reinforcement. Use psy equal loudness compensation to SPL and much more but I have to go to work now.

That’s one approach. I take a different approach. I do kill the room and I use highly directional low distortion speakers. It a no compromise approach but it is the road less traveled. Killing a room is no small task.

 

[ZolaIII]

That’s one approach. I take a different approach. I do kill the room and I use highly directional low distortion speakers. It a no compromise approach but it is the road less traveled. Killing a room is no small task.

I don't want headphone effect, keeping it in order with good clarity and RT60 decay times is fine with me. I initially forgot to mantion ISO 3382-1 (back to front refractions ratio) importance which are also very hard to improve in normal environment (small to medium sized room).

 

[youngho]

Next, hearing works, very roughly, in two ways: time sensitivity and level sensitivity, each being different per frequency. These separate into three (but really four)

1. Below about 1.5kHz, frequencies are identified through the ear/brain phase locking to the physical waveform.
2. Above around 1.5kHz, this identification takes place through level.

You wrote about identifying frequency, but you seem to be describing localization (like how would one identify higher frequencies through level?). Pitch perception and processing are different. I put together a list of links on psychoacoustics with some selective quotes and summary here: https://www.audiosciencereview.com/...acoustics-self-education-links-sharing.45583/