What is your favorite house curve

[dlaloum]

Can you tell Dirac and Audyssey not to perform corrections above a certain frequency range?

With Acourate, Focus Fidelity, and Audiolense you can perform different windowing for lower and upper frequencies. With very narrow windowing, you are effectively performing "broad tone controls" as Dr. Toole has recommended elsewhere. Or if desired, upper frequencies can be left out of the correction altogether. I would imagine that Dirac / Audyssey should have a similar function, it is a pretty basic feature.

Yes you can - Dirac allows you to specify "curtains" and it does not adjust outside of the "curtains"

But if you do not adjust the midrange and high frequencies, and you are not using identical speakers - then you are likely to have timbral matching issues... - hence there are (potential) benefits to be gained from a light touch of EQ, to match the timbral voicing of the differing speakers.

And that is my point - most of us do not have identical speakers all around... so that means optimal performance is going to take some degree of EQ to provide timbral matching. Setting the curtains and locking out the frequencies above schroeder, would not achieve this...

One alternative I have been considering (and experimented with a little some months back) - is to use the measured response of the front mains as the "target curve" - this would ensure minimal changes by the RoomEQ software on the mains, and would EQ the other speakers to the same room power response....

But the real objective is to EQ the direct anechoic response rather than the room power response, and I see no obvious way to do this with the tools I have available - so the next best, is to determine what might achieve the best approximation thereof!

 

[joentell]

Yes, some good information, mixed In with incomplete knowledge of the topic. Better than some, but still not the real story. He refers to my book - he needs to read it again.

It would be helpful to all of us to understand what you're referring to when you say "incomplete knowledge." I've re-read relevant chapters from your 3rd Edition as well as your AES papers regarding this topic more than most. I've done experiments to better understand the concepts. I frequently reference your work as well as other AES papers on the topic. I have the Kindle version of your book and many sections are highlighted. I don't think I need to read it yet again.

I've tried reaching out to you via phone call multiple times since December and via text several more times to clarify several concepts, so if if it's your opinion that my information is limited, it's not because I'm not actively seeking it. It's the opposite.

Your AES paper titled "The Measurement and Calibration of Sound Reproducing Systems" ends with several unanswered questions, in section 7.2 Research Questions, which I've set out to find answers to. Your last paragraph states, "We don’t yet have all the pieces of the puzzle, but we have enough to imagine that some carefully constructed research can provide the missing links and validations." I agree that we don't have all the answers, but I think I've discovered some extra pieces to that puzzle.

I think the reason people mistakenly use the "Harman curve" as a target is because previous JBL/Harman products have used a similar curve as a fixed target in the past. I haven't used the JBL Synthesis or Arcos products, but I know my JBL MS8 DSP/amplifier uses that as a target.

Paul Barton who did research with you at the NRC has his own variation of that Harman curve. It is available as a Dirac target curve for NAD products.

So, it's understandable that people mistakenly think it's a good idea to use these one-size-fits-all target curves on their systems.

 

[OCA]

what you're referring to when you say "incomplete knowledge

While I fully agree on your points about the difficulty of obtaining accurate information, I believe the main confusion stems from the suggestion of using high frequency equalization under certain criteria in the video.
By definition, you cannot measure a speaker's steady state "directivity" with an "omnidirectional" calibration microphone, let alone match it to speaker's anechoic directivity. Therefore, there is no valid way to match the high-frequency steady-state response to any predetermined curve, regardless of criteria. This practice of high frequency EQ to a target is essentially a marketing gimmick that has unfortunately become ingrained in consumers' minds, as Dr. Toole points out.

 

[joentell]

By definition, you cannot measure a speaker's steady state "directivity" with an "omnidirectional" calibration microphone, let alone match it to speaker's anechoic directivity. Therefore, there is no valid way to match the high-frequency steady-state response to any predetermined curve, regardless of criteria. This practice of high frequency EQ to a target is essentially a marketing gimmick that has unfortunately become ingrained in consumers' minds, as Dr. Toole points out.

I have measurement data that suggests otherwise. I'm working on Part 2 of my video which clearly shows this. If I take measurements of various speakers in-room using my NF method, and they all match closely with Klippel NFS measurements of the same speaker (above the transition region,) I don't see how it isn't valid from a room-correction standpoint. I can also extract directivity data using these methods. I've presented this to Charles Sprinkle, formerly at Harman, now at Kali Audio, and he seemed impressed with the results taken using the speakers that he designed and has anechoic data for.

Perhaps "incomplete knowledge" might just mean a disagreement, which is good to have once in a while as long as we're looking at things objectively.

I'm not designing speakers. If I were, I would get a Klippel NFS. We're doing room correction, so the margin of error of +/- 1dB across a small frequency range is acceptable in my opinion. Differences in mic calibrations have shown more variation.

BTW, good to finally talk to you. Good work on Audyssey One.

 

[OCA]

BTW, good to finally talk to you. Good work on Audyssey One.

Thank you and nice to talk to you, too!

I have measurement data that suggests otherwise.

Maybe with a mic array like this one (unless you are rotating a Klippel system at the LP around your head):

and using eigen vectors and matrix arithmetic to obtain a steady state response like this:

And even then there's evidence that our brains take spatial clues from these reflections and use them to increase intelligibility (at least for speech according to some papers) so, they might better be left untouched.

 

[OCA]

The "Dr Toole" curve - please stop calling it that - is explained in my book as the average steady-state room curve, above the transition frequency at least, generated by loudspeakers that have received high sound quality ratings in double-blind tests over several decades of hundreds of tests. This identifies loudspeakers with the minimum resonant coloration - the most neutral - which is what we want as a starting point. Achieving such a curve in your room indicates that you MIGHT have purchased excellent loudspeakers. Compensating errors could also generate a nice looking curve, but not sound as neutral. Equalizing a flawed loudspeaker to create this curve, using it as a target, guarantees nothing - don't bother. But people do and fill pages of forums discussing the results - a social media exercise, not science.

The so-called "Harman" curve was a tone-control exercise which allowed listeners to play with broadband spectral trends (not normal resonances) to satisfy a preference for the particular programs being used. Young, inexperienced, listeners boosted bass and treble - but was it because they liked that result or was it because they wanted it louder. Unfortunately, we don't know from that test and further tests were not done. The starting loudness was a good foreground listening level, not a rock concert level or a chamber music level.

Being able to adjust bass level is essential for fussy listeners, but is not always available without frustrating layers of menus and momentary silences while digital adjustments are made. Old fashioned analog tone controls were truly useful. Small changes in bass level matter a lot.

Thank you @Floyd Toole and it's been such a privilege to get this information directly from you. I will be looking forward to read the 4th edition of your book. I want to also thank to @amirm to provide the audio community with a platform with such exclusivity!

 

[joentell]

Thank you and nice to talk to you, too!

Maybe with a mic like this one (unless you are rotating a Klippel system at the LP around your head):

and using eigen vectors and matrix arithmetic to obtain a steady state response like this:

And even then there's evidence that our brains take spatial clues from these reflections and use them to increase intelligibility (at least for speech according to some papers) so, they are better left untouched.

My measurement method will either closely align with anechoic measurements above the transition region, or it won't. I've done enough comparisons that I'm confident that it does.

 

[Tim Link]

I have measurement data that suggests otherwise. I'm working on Part 2 of my video which clearly shows this. If I take measurements of various speakers in-room using my NF method, and they all match closely with Klippel NFS measurements of the same speaker (above the transition region,) I don't see how it isn't valid from a room-correction standpoint. I can also extract directivity data using these methods. I've presented this to Charles Sprinkle, formerly at Harman, now at Kali Audio, and he seemed impressed with the results taken using the speakers that he designed and has anechoic data for.

Perhaps "incomplete knowledge" might just mean a disagreement, which is good to have once in a while as long as we're looking at things objectively.

I'm not designing speakers. If I were, I would get a Klippel NFS. We're doing room correction, so the margin of error of +/- 1dB across a small frequency range is acceptable in my opinion. Differences in mic calibrations have shown more variation.

BTW, good to finally talk to you. Good work on Audyssey One.

I'll be interested in watching that and learning what you are doing. As a DIY builder I'm interested in ways of getting a better picture of what my big and unorthodox speakers are doing. For now I indirectly target a flat and sloping room curve, no particular slope, by getting the drivers flat up close on axis, and then choosing horn flares and crossover setpoints that result in a flat but tilted response in the high frequencies at the listening position. It's making me pretty happy, and it's hard for me to imagine that the directivity could be too out of control since that's what's happening.

 

[007Shortz]

.. stereo recordings are made for loudspeaker reproduction .. what is an "absolutely neutral headphone".. Only statistical averages are possible .. It is a dilemma .. Now are you confused?

Of course I'm confused now!
But that's the reason why I'm in this forum. Thanks for your explanation!
I guess I need to unlearn some of my oversimplifications.
The audio chain appears like an under-determined system of equations.

 

[Sal1950]

So here is the challenge... knowing that the room response (AKA Target curve) is a consequence rather than a target

How does one leverage Dirac and/or Audyssey (or similar) to provide timbral matching of the speakers?

There's a ton of truth to what you wrote here.
But there's also the point of chasing perfection down a rabbit-hole to insanity.
Sometimes you just have to kick back, have a drink or smoke and know,
Man this system sounds great.

 

[dlaloum]

There's a ton of truth to what you wrote here.
But there's also the point of chasing perfection down a rabbit-hole to insanity.
Sometimes you just have to kick back, have a drink or smoke and know,
Man this system sounds great.

After adding the Penta's to the rear/surround, and doing the Dirac basic tuning - that was exactly my experience... kicking back, enjoying and thinking yep, it sounds great..... very satisfying.

But then the curiosity kicks in.... audiofoolery....

 

[Newman]

Thank you @Floyd Toole and it's been such a privilege to get this information directly from you….I want to also thank to @amirm to provide the audio community with a platform with such exclusivity!

Dr Toole is not exclusively present on ASR. I am pretty sure he has contributed on AVS for at least seven or eight years, where I have had past dialogue. Probably other forums too. Keep your eyes open, look around.

 

[Newman]

And that is my point - most of us do not have identical speakers all around... so that means optimal performance is going to take some degree of EQ to provide timbral matching. Setting the curtains and locking out the frequencies above schroeder, would not achieve this...

That is why I was pleased when, some year or three ago, I asked whether Audyssey MultEQ-X allows auto EQ below a specified LP frequency, plus subsequent manual edits above that LP point … and without losing the auto EQ.

…But the real objective is to EQ the direct anechoic response rather than the room power response, and I see no obvious way to do this with the tools I have available - so the next best, is to determine what might achieve the best approximation thereof!

Agreed, and hence my above statement. Does anyone know if Dirac allows this? I got the impression that the answer was negative.

 

[joentell]

While I fully agree on your points about the difficulty of obtaining accurate information, I believe the main confusion stems from the suggestion of using high frequency equalization under certain criteria in the video.
By definition, you cannot measure a speaker's steady state "directivity" with an "omnidirectional" calibration microphone, let alone match it to speaker's anechoic directivity. Therefore, there is no valid way to match the high-frequency steady-state response to any predetermined curve, regardless of criteria. This practice of high frequency EQ to a target is essentially a marketing gimmick that has unfortunately become ingrained in consumers' minds, as Dr. Toole points out.

I've just taken measurements for my video. This is a measurement of an Arendal 1723 Bookshelf S which we do have Spinorama data for.

Here is an in-room nearfield moving mic measurement 1/48 smoothing. Above the transition region, I'm comfortable using that as a starting point for room correction in-lieu of anechoic data which many people don't have for their speakers.

I also mentioned that we can extract directivity information above the transition region as well. This is our speaker response exported to REW (psychoacoustic smoothing.) As you can see, below the transition region, the room takes over. We are able to determine what that range is, and we transition from a correction of the nearfield pseudo-anechoic response, to a correction of the steady-state response. So, I don't think it's a marketing gimmick. We simply have more data that we can use to base our corrections off of than most auto-room correction software. (The high frequency rise above 16kHz in these measurements is due to a using the wrong mic calibration. Regardless, the directivity tracks pretty well since the calibration affected the nearfield and the farfield response equally.)

@Tim Link I think this was the information you were interested in.

 

[goat76]

Here is an in-room nearfield moving mic measurement 1/48 smoothing. Above the transition region, I'm comfortable using that as a starting point for room correction in-lieu of anechoic data which many people don't have for their speakers.

It seems like your method works. How near in front of the speaker do you wave your microphone, and how large area do you cover on the vertical plane?

 

[joentell]

It seems like your method works. How near in front of the speaker do you wave your microphone, and how large area do you cover on the vertical plane?

Nearfield MMM starting directly on-axis to the reference point (typically the tweeter or between the tweeter and midrange. If the manufacturer specified the optimal point, use that.) Start at a distance of 0.5 meters and move straight back to 1 meter. The starting difference might be further back depending on the spacing between the furthest driver and the tweeter. In that case, 1X that distance might be a good starting point to ensure proper driver integration.

 

[ernestcarl]

Nearfield MMM starting directly on-axis to the reference point (typically the tweeter or between the tweeter and midrange. If the manufacturer specified the optimal point, use that.) Start at a distance of 0.5 meters and move straight back to 1 meter. The starting difference might be further back depending on the spacing between the furthest driver and the tweeter. In that case, 1X that distance might be a good starting point to ensure proper driver integration.

I’ve been too preoccupied with other things (and just lazy to haul back) to test this with my FX50, but I do believe this method at least gets one reasonably close enough to the direct anechoic response average as well as power response and reflections included (with usually some more minor room+speaker stand effects) around and above 500Hz or so.

The EQ applied below that point or maybe 700Hz or so increasingly depends on how one wants to weigh the difference between the nearfield spatial average vs the more distant measurements.

 

[goat76]

Nearfield MMM starting directly on-axis to the reference point (typically the tweeter or between the tweeter and midrange. If the manufacturer specified the optimal point, use that.) Start at a distance of 0.5 meters and move straight back to 1 meter. The starting difference might be further back depending on the spacing between the furthest driver and the tweeter. In that case, 1X that distance might be a good starting point to ensure proper driver integration.

Thanks, I will try this out tomorrow and compare it to my gated measurements. Regarding my 3-way ATC speakers, the acoustic center is right at the mid-dome driver.

 

[goat76]

I’ve been too preoccupied with other things (and just lazy to haul back) to test this with my FX50, but I do believe this method at least gets one reasonably close enough to the direct anechoic response average as well as power response and reflections included (with usually some more minor room+speaker stand effects) around and above 500Hz or so.

I don't think the speaker stand will be a big problem, and you could easily place the speaker further away from the walls during the measurement to make sure the microphone is always at a closer distance to the speaker than the first reflection point.

 

[garyrc]

My current AVR has Bass and Treble adjustment buttons on the remote... allowing for "instant" adjustments to taste...

I would prefer a Quad style "tilt" button - but old world bass and treble does the job fine! - salt & pepper to taste

Some manufacturers ceasing to equip pre-amps and receivers with tone controls was the most harebrained (or hairbrained, as in hair penetrating the skull and growing into the brain) move in the history of audio. The idea that recordings somehow have flat frequency response was the second most harebrained notion. I just love it when people who have tone controls say something like, "I always leave the tone controls flat because I want to hear the music the way the engineers heard it," as though they are sitting in the engineers' control room, listening through the engineers' speakers.

My tone control settings vary from the extreme case of Bass +6 dB, Treble +3 dB, through "Flat," to Treble - 6 dB (for the over-bright Blu-ray of How the West Was Won, SmileBox version).