Audyssey Manual Calibration “OCA’s REW + Audyssey Awesomeness”

[Mike Lima]

In an HT set up, if all speakers are fairly symmetrically positioned in a fairly rectangular shaped room with a central MLP at an acoustically valid position (ie not by the rear wall or at a room corner) then moving the mic around the MLP would always somewhat improve the calibration in a greater region. But when there are reflections higher than the direct sound itself in the measured responses (more common than you would think in calibration files I am sent!), cross correlation alignment of multiple mic positions (which is the only valid method to truly average out multiple responses) fails to correctly create the actual speaker response at the MLP resulting in inaccurate correction filters. In such cases, keeping the mic at a certain location (or closely around it) helps. On the other hand, even when everything is ideal, repeated measurements/measurement volume/length all improve SNR due to noise during measurements, hardware glitches, etc. I take 5-6 measurements at the dead centre mic position for each speaker in my stereo set up and average them after eliminating odd ones out (there usually will be 1 or 2 at the same mic position even with an Earthworks mic).

The new script (soon to be finished) eliminates Audyssey measurements that don't correlate properly with the rest of the measurements of that same speaker.

For best results with an Audyssey mic, I would strongly recommend using a mic stand (cheapest ones for about $20 would still work great) and keeping it at the central position and only rotating its arm for different mic positions keeping all at ear height and fairly close to each other. Spending time on the accuracy of the mic's upward position if you want a meaningful HF response, absolute silence during measurements and temporarily turning subs to always on during calibration rather than auto-on which might introduce extra delays as well as any extra processing on the sub itself like PEQ, LPF, etc. Speakers should all be directly facing the MLP. There's nothing that can ruin a speaker's sound more than boosting its high frequencies only because the low SPL at HF is in fact caused by wrong toe-in angle.

Also, you should not reasonably expect descent sound for all seats in a multiple row home theatre set up with just one pair of surround or height speakers.

Thank you soooo much for all you do. I plan to do 1 position for all tests. When using a high back, home theater recliner... is it ok to set the MLP where the center of my head would be? (2 inches from the headrest of the recliner)?

 

[EasyC]

Thank you soooo much for all you do. I plan to do 1 position for all tests. When using a high back, home theater recliner... is it ok to set the MLP where the center of my head would be? (2 inches from the headrest of the recliner)?

Beware of reflections from the leather of the backseat- you get better results with a blanket over it or reclining it out of the way.

 

[OCA]

Beware of reflections from the leather of the backseat- you get better results with a blanket over it or reclining it out of the way.

Given the size, I don't think anything below 600-700Hz would be reflected from a headrest but would rather diffract around it. Walls are your real enemies though.

 

[Mike Lima]

Beware of reflections from the leather of the backseat- you get better results with a blanket over it or reclining it out of the way.

Understood. But my head will be there... You think it still becomes a factor? I have put a towel there before. Thanks.

 

[Mike Lima]

Given the size, I don't think anything below 600-700Hz would be reflected from a headrest but would rather diffract around it. Walls are your real enemies though.

Thanks for your response. I have Auralex wall treatment. I know.. I know... it may not help

 

[OCA]

Understood. But my head will be there... You think it still becomes a factor? I have put a towel there before. Thanks.

It's called room correction for good reason. Wavelengths you can (and should) correct are between 1.7 and 17 meters. Below that, our brains are evolved to separate reflections from direct sound and even benefit from them for intelligibility.

 

[Peupeut]

In an HT set up, if all speakers are fairly symmetrically positioned in a fairly rectangular shaped room with a central MLP at an acoustically valid position (ie not by the rear wall or at a room corner) then moving the mic around the MLP would always somewhat improve the calibration in a greater region. But when there are reflections higher than the direct sound itself in the measured responses (more common than you would think in calibration files I am sent!), cross correlation alignment of multiple mic positions (which is the only valid method to truly average out multiple responses) fails to correctly create the actual speaker response at the MLP resulting in inaccurate correction filters. In such cases, keeping the mic at a certain location (or closely around it) helps. On the other hand, even when everything is ideal, repeated measurements/measurement volume/length all improve SNR due to noise during measurements, hardware glitches, etc. I take 5-6 measurements at the dead centre mic position for each speaker in my stereo set up and average them after eliminating odd ones out (there usually will be 1 or 2 at the same mic position even with an Earthworks mic).

The new script (soon to be finished) eliminates Audyssey measurements that don't correlate properly with the rest of the measurements of that same speaker.

For best results with an Audyssey mic, I would strongly recommend using a mic stand (cheapest ones for about $20 would still work great) and keeping it at the central position and only rotating its arm for different mic positions keeping all at ear height and fairly close to each other. Spending time on the accuracy of the mic's upward position if you want a meaningful HF response, absolute silence during measurements and temporarily turning subs to always on during calibration rather than auto-on which might introduce extra delays as well as any extra processing on the sub itself like PEQ, LPF, etc. Speakers should all be directly facing the MLP. There's nothing that can ruin a speaker's sound more than boosting its high frequencies only because the low SPL at HF is in fact caused by wrong toe-in angle.

Also, you should not reasonably expect descent sound for all seats in a multiple row home theatre set up with just one pair of surround or height speakers.

@OCA , would you say the extra energy I see on my atmos in REW from 10 kHz to 20 kHz is due to the fact that for those channels, the microphone is not oriented at 90° but more closed to a 0° orientation? because I do not ear this extra energy...

And I have another question: if I want to make a try and extend the correction range to more than the limit of 200Hz you have set for Audyssey One, but keeping the wonderfull work it is doing for phase alignment, speaker delay etc..., can I:
- use the filter files I see in REW after One processing and extend the correction using the methodology you proposed last year for Atmos system and Mac users
- then use you HTML script "CorrectThem.html" to re-import the extended scope of the filters with the ady file "OneOptimized"?

Would it work or would I loose the phase/delay of One?

I know you do not recommend to extend the correction, it is just for testing and fun...

 

[OCA]

@OCA , would you say the extra energy I see on my atmos in REW from 10 kHz to 20 kHz is due to the fact that for those channels, the microphone is not oriented at 90° but more closed to a 0° orientation? because I do not ear this extra energy...

And I have another question: if I want to make a try and extend the correction range to more than the limit of 200Hz you have set for Audyssey One, but keeping the wonderfull work it is doing for phase alignment, speaker delay etc..., can I:
- use the filter files I see in REW after One processing and extend the correction using the methodology you proposed last year for Atmos system and Mac users
- then use you HTML script "CorrectThem.html" to re-import the extended scope of the filters with the ady file "OneOptimized"?

Would it work or would I loose the phase/delay of One?

I know you do not recommend to extend the correction, it is just for testing and fun...

I made it very simple in the next optimization tool - to be probably named Audyssey One Evo (to keep facebook group name intact) - to extend/revise/sharpen filters:

You can change any of these default parameters as you like. But I did that so people finally hear for themselves and understand why it will only sound throttled otherwise

It's almost finished now and I aim to post the new YT video in a few days at most.

Edit: Target curve is also replaceable at will.

 

[TheZebraKilledDarwin]

Audyssey Flat is hardcoded to bypass app customTargetCurvePoints so all filters will be deactivated. There's an anti high frequency roll off added to each filter in A1 that attempts to reverse the odd effects of reference curve.

I did a few tests and just wanted to let you know that it works and they were successful: It is indeed possible to use Audyssey FLAT calibration setting with a free personal custom target curve (and without any hf-rolloff applied by Audyssey), and additionally use the Audyssey Reference mode with "custom target curve" and the unavoidable hfr applied on top of the 'flat' + custom target curve-correction.

 

[OCA]

I did a few tests and just wanted to let you know that it works and they were successful: It is indeed possible to use Audyssey FLAT calibration setting with a free personal custom target curve (and without any hf-rolloff applied by Audyssey), and additionally use the Audyssey Reference mode with "custom target curve" and the unavoidable hfr applied on top of the 'flat' + custom target curve-correction.

I heard that from someone else, too the other day. It was definitely not possible (at least in my SR6015) until recently. This could be March 2024 app update or specific to some models because in the past some people claimed they can change "enTargetCurveType": to 0 instead of 1 or 2 and it worked. My Marantz completely got blocked if you send an ady with that setting at 0. I will try it out myself soon. I know it was possible with MultEQ-X since a long time maybe they decided to give that to Editor app, too. This is good news if true.

 

[kifeep]

I made it very simple in the next optimization tool - to be probably named Audyssey One Evo (to keep facebook group name intact) - to extend/revise/sharpen filters:

View attachment 370088

You can change any of these default parameters as you like. But I did that so people finally hear for themselves and understand why it will only sound throttled otherwise

It's almost finished now and I aim to post the new YT video in a few days at most.

Edit: Target curve is also replaceable at will.

What is the relative benefit of capping the frequency of filter generation vs limiting the frequency to which corrections are applied via the built-in "MultiEQ Filter Frequency Range" setting? The advantage of the latter is the ability to make adjustments without rerunning the script.

 

[TheZebraKilledDarwin]

I heard that from someone else, too the other day. It was definitely not possible (at least in my SR6015) until recently. This could be March 2024 app update or specific to some models because in the past some people claimed they can change "enTargetCurveType": to 0 instead of 1 or 2 and it worked. My Marantz completely got blocked if you send an ady with that setting at 0. I will try it out myself soon. I know it was possible with MultEQ-X since a long time maybe they decided to give that to Editor app, too. This is good news if true.

I think there is a misunderstanding. My approach is to replace the SPEAKER MEASUREMENTS impulse responses, with the inverted wanted EQ filters impulse response. So instead of using a dirac impulse for a perfect speaker and applying the correction with the "custom target curve", the impulse response of the wanted EQ-correction is replacing the SPEAKER MEASUREMENT impulse response.

For example:
The speaker shows a +12dB mode @50 Hz in the room which is to be corrected.
REW calculates the correction with an EQ (-12dB @50Hz + a q-value).
That found EQ setting by REW is then inverted by hand (--> +12 dB @50Hz; therefore it is mimicking the measured 50Hz part of the speaker response).
Then this EQ in REW is exported as impulse response and used as SPEAKER RESPONSE in the ADY-file.
So it is de facto a speaker response, but with the difference to a Dirac or real speaker response, that the user can freely decide, in REW which EQ filters he wants and which room curve he wants applied.
Audyssey linearization sees the +12dB @50 Hz bump in the 'speaker response' and applies it's linearization/correction algorithm (in this example correcting the +12dB@50 Hz bump in the pseudo speaker measurmeent impulse response with the wanted -12dB @50 Hz correction).

That way Audyssey can be tricked in FLAT mode into correcting fully user defined corrections and target curves, camouflaged as (inverse) pseudo speaker measurements. With the result that the FLAT correction is resulting in the wanted user defined correction/target curve.
And the 'custom target curve' in REFERENCE mode is still available to be used additionally, because the wanted correction was injected as (pseudo) speaker measurement data.

 

[kifeep]

I think there is a misunderstanding. My approach is to replace the SPEAKER MEASUREMENTS impulse responses, with the inverted wanted EQ filters impulse response. So instead of using a dirac impulse for a perfect speaker and applying the correction with the "custom target curve", the impulse response of the wanted EQ-correction is replacing the SPEAKER MEASUREMENT impulse response.

For example:
The speaker shows a +12dB mode @50 Hz in the room which is to be corrected.
REW calculates the correction with an EQ (-12dB @50Hz + a q-value).
That found EQ setting by REW is then inverted by hand (--> +12 dB @50Hz; therefore it is mimicking the measured 50Hz part of the speaker response).
Then this EQ in REW is exported as impulse response and used as SPEAKER RESPONSE in the ADY-file.
So it is de facto a speaker response, but with the difference to a Dirac or real speaker response, that the user can freely decide, in REW which EQ filters he wants and which room curve he wants applied.
Audyssey linearization sees the +12dB @50 Hz bump in the 'speaker response' and applies it's linearization/correction algorithm (in this example correcting the +12dB@50 Hz bump in the pseudo speaker measurmeent impulse response with the wanted -12dB @50 Hz correction).

That way Audyssey can be tricked in FLAT mode into correcting fully user defined corrections and target curves, camouflaged as (inverse) pseudo speaker measurements. With the result that the FLAT correction is resulting in the wanted user defined correction/target curve.
And the 'custom target curve' in REFERENCE mode is still available to be used additionally, because the wanted correction was injected as (pseudo) speaker measurement data.

This is really interesting. I started playing with @OCA's methods (before A1 was released) because I had a notch at 70Hz that Audyssey was adding 9dB to and causing distortion. Using this method, one could limit the pseudo speaker response to only peaks above a smooth, minimum phase response (and only below transition) and avoid that problem.

How do you export REW measurements in a way that can be inserted into the ady file? The information that comes out of the ady file includes only the raw response data. Comparing a standard REW impulse export with the header data A1 adds before REW import, there are apparently some differences in peak value, peak index, and response length. The sample interval appears the same, however.

 

[Chromatischism]

Speakers should all be directly facing the MLP. There's nothing that can ruin a speaker's sound more than boosting its high frequencies only because the low SPL at HF is in fact caused by wrong toe-in angle.

Agree with everything you said except this . Toe angle is critical for many speakers, mine included, and this is another reason I don't correct above about 300-400 Hz.

 

[CrazyDwarf]

Are you using open baffle speakers by any chance? The error means for at least one of the speakers, there was no point below 250Hz where frequency response of the speaker could reach the target level. I have seen this only once with an open baffle speaker and had to customize the code for it. For normal speakers, my best guess is that a speaker's response wasn't measured correctly during the original calibration because of some glitch.

I have the same issue. My in-ceiling speakers (Polks RC60i) are installed in a suspended celeing (I made no back box for them unfortunately, their backbox is a plastic tube which is then screwed to a concrete ceiling - i had no other way to attach them to the ceiling - they are designed to be attached by plastic fittings to the plasterboard). Because of that, I suppose, they have literally no bass response, and the receiver puts them at 250Hz Crossover point. I cannot remove the ceiling already, and to do backboxes as of now. Now, your scripts returns an error: Error: Speaker has no bass response according to your measurement file. Please check! - program execution has stopped! What should I do now? It is just stopping doing it's magic and that's it.

 

[CrazyDwarf]

Thanks. Started again from scratch and still end up with the following
error: speaker has no bass response according to your measurement file. Please check! - program execution has stopped

I have the same issue, have you been able to resolve that? Thanks.

 

[mosco]

Tengo el mismo problema, ¿pudiste resolverlo? Gracias.

Hola. Soy el del openbaffles al que OCA le hizo la versión que menciona. Yo simplemente fui adaptando las líneas diferentes a las siguientes versiones. No soy usuario de programación pero funciona. Te dejo un enlace a la última que tengo https://drive.google.com/file/d/1tYP7Zq7o55106odhpLAiSsxpzewwNjK_/view?usp=drivesdk

Translation by Moderator:
“Hello. I'm the guy from openbaffles for whom OCA made the version you mention. I simply adapted the different lines to the following versions. I'm not a programming user but it works. I leave you a link to the last one I have”

Please use English going forward @mosco .


Sorry. The translator retranslated...

 

[CrazyDwarf]

Hola. Soy el del openbaffles al que OCA le hizo la versión que menciona. Yo simplemente fui adaptando las líneas diferentes a las siguientes versiones. No soy usuario de programación pero funciona. Te dejo un enlace a la última que tengo https://drive.google.com/file/d/1tYP7Zq7o55106odhpLAiSsxpzewwNjK_/view?usp=drivesdk

Hola, le envié una solicitud para obtener acceso. Gracias por tu ayuda.

Can you please also elaborate if possible? Did OCA made you a modified script version, so that it works despite the speaker having no bass? Or, did you change something in the ady file? Or in the HTML one?

Thanks.

P.S. I understood you. I changed the settings for maximum frequency per this error in the HTML file and it worked. Thank you.

 

[Hyacin]

Windows PCs run the height channel sweeps without problems. You should check Windows Sound/AVR settings.

What should they be? I can't seem to find them anywhere, and I've been searching high and low, watching videos, and trying every setting combination I can imagine for hours now :'(

I've set it in Windows to 'Stereo', '7.1' and 'Atmos', and not one of them will play my highs or surrounds (not surround backs) correctly. I'm almost in tears I'm so frustrated here and can't think of what else to try.

I remember seeing something in one video I think about keeping the receiver on "Multi Ch 7.1 in" or something, so I've done that, as well as making sure the AVR is the 'default device', which it has been the entire time I've been struggling with this.

Is there a page or portion of a video that goes over specifically and exactly these settings?

Thank you. <3

Update: If it helps at all, with the "speaker configuration" set to 7.1, when I play a known Atmos movie with Media Player, the receiver kicks into Atmos mode and seems to play as expected. When I play, say, "FWR.mp4" from the updated 256k sweeps in the second post of this thread, the receiver stays in "Multi Ch In 7.1" mode, and the sweep comes out of my FR speaker, not my FWR speaker. Same with the original sweeps linked from the video description.

 

[Chromatischism]

Hmm if my PC is set to Atmos, I get my heights when I play Atmos or DTS:X demo clips. On past Denon AVRs it would say Multichannel In but on my X4500H it displays Atmos. Have you downloaded the Atmos add-on for Windows?