Ongoing Beta Beta Testing: Independent L/R Room Correction with Subwoofer

[dominikz]

I’ll review these Max Gain control improvements with our engineering team and aim to implement them promptly.

Thanks, that would be really amazing!

 

[slartibartfast]

There seems to be an issue where all the L/R PEQ filters sound and measure the same though. Nothing to do with my preference. It seems like the selected filter isn't being applied.

I half expected @WiiM Team to reply to this since they had replied to another post in this thread. I'll report it via feedback in the app when I get a chance as it seems quite important.

 

[dominikz]

Isn't the target curve meant to be the target for the listening position with both speakers playing though?

While I can understand why that might be expected, to my understanding it is not the case.

E.g. the often mentioned "Harman curve" was never meant to be a target curve for EQ - it is simply the idealized in-room response of a typical well-measuring, front-firing single loudspeaker (meaning flat on-axis response and smooth directivity).

(Image reference 1, from dr. Toole's book "Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms")

(Image reference 2, from dr. Toole's book "Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms")

Note especially the gray area in the 'Harman curve', meant to allow for variation of the amount of reflectivity in different rooms.

I.e. if you place different well-measuring loudspeakers in various rooms, measure their individual in-room responses at the listening position, then average all of those measured responses - you'd get something very much resembling the "Harman curve".

So when we use the "Harman curve" as a target for EQ, we're basically trying to make our in-room steady-state loudspeaker response match the in-room response of some 'ideal' of a well-measuring loudspeaker: which may or may not work well in the specific case, depending mainly on how even the directivity of our loudspeaker is, and how closely its directivity index (DI) matches the average DI of a typical well-measuring loudspeaker from the Harman study.

The "Harman curve" definitely represents a single loudspeaker response; no assumption seems to have been made when creating it on whether the loudspeaker is used in a stereo, 5.1 or some Atmos configuration with a large number of speakers. Actually, Harman famously (notoriously?) uses single loudspeakers in mono for a lot of their research and testing. So if we decide to use the "Harman curve" as a target, we should match each of our individual channel responses to it.

Let's take a different example - let's assume you want to EQ a pair of ideally-placed full-range loudspeakers as nearfield monitors in a highly acoustically damped studio space (or an anechoic chamber).
To do this you'd probably use the "flat" EQ target for each loudspeaker - since you are sitting in nearfield and reflections are absorbed, you expect to hear and measure mainly the loudspeaker direct sound at the listening position, which should be as flat as possible from 20Hz to 20kHz.
When you measure such a system, its left and right channels would individually measure as flat horizontal lines (which is what you wanted); but when played back together the combined L+R channels would measure +3dB in the low frequencies compared to high frequencies - for the same reasons described in my previous posts.

As hinted in the second image above, note that ideally we shouldn't really be using any of the 'standard' target curves for EQ, because an ideal target curve for EQ would be determined from the specific loudspeaker directivity index, how reflective the room is, the listening distance, and even personal preference on the amount of bass. No predefined target curve will perfectly match all loudspeakers/rooms/listening distances and preferences.
Though of course such precision may not really be required in practice to achieve very good results.

To view the same topic from a bit different perspective, let's look at anechoic measurements of some really well-measuring non-cardioid loudspeakers like e.g. KEF LS50 Meta, Revel M105, Kali Audio LP6v2, KEF Blade 2 Meta and Ascend Acoustics ELX Ribbon Tower.
From the anechoic spins we can see that most non-cardioid loudspeakers become basically omnidirectional around 100-200Hz - this is the point where their Sound Power DI becomes very close to 0 (note the right vertical axis is used for DI values in the spinorama graphs linked above).
An ideal individual loudspeaker target curve would therefore have a flat part in the low frequencies (usually below ~150Hz), and a downward-sloping part above it - as this would be a good match for the loudspeaker directivity.
Notice how similar that is to the B&K 1974 curve (reference to original B&K article).

For simplicity I didn't want to go into absorption of high frequencies in the air, but that is something that becomes a factor too, especially for listening distances well beyond 1m.

On the topic of preference - this is why it would IMHO be really beneficial to have a simple post-EQ bass/treble or "tonality tilt" control. This would allow users to easily dial in the preferred amount of bass after the system has already been calibrated with RC for a neutral response.
As we all know, not all recordings are equally well-made, so this would also allow the users to adjust the tonality to counteract deficiencies in the recording itself.

Sorry for a long-winded post (it is difficult to balance brevity and completeness in this case), but perhaps the information could be interesting to some!

 

[slartibartfast]

While I can understand why that might be expected, to my understanding it is not the case.

E.g. the often mentioned "Harman curve" was never meant to be a target curve for EQ - it is simply the idealized in-room response of a typical well-measuring, front-firing single loudspeaker (meaning flat on-axis response and smooth directivity).

(Image reference 1, from dr. Toole's book "Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms")

(Image reference 2, from dr. Toole's book "Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms")

Note especially the gray area in the 'Harman curve', meant to allow for variation of the amount of reflectivity in different rooms.

I.e. if you place different well-measuring loudspeakers in various rooms, measure their individual in-room responses at the listening position, then average all of those measured responses - you'd get something very much resembling the "Harman curve".

So when we use the "Harman curve" as a target for EQ, we're basically trying to make our in-room steady-state loudspeaker response match the in-room response of some 'ideal' of a well-measuring loudspeaker: which may or may not work well in the specific case, depending mainly on how even the directivity of our loudspeaker is, and how closely its directivity index (DI) matches the average DI of a typical well-measuring loudspeaker from the Harman study.

The "Harman curve" definitely represents a single loudspeaker response; no assumption seems to have been made when creating it on whether the loudspeaker is used in a stereo, 5.1 or some Atmos configuration with a large number of speakers. Actually, Harman famously (notoriously?) uses single loudspeakers in mono for a lot of their research and testing. So if we decide to use the "Harman curve" as a target, we should match each of our individual channel responses to it.

Let's take a different example - let's assume you want to EQ a pair of ideally-placed full-range loudspeakers as nearfield monitors in a highly acoustically damped studio space (or an anechoic chamber).
To do this you'd probably use the "flat" EQ target for each loudspeaker - since you are sitting in nearfield and reflections are absorbed, you expect to hear and measure mainly the loudspeaker direct sound at the listening position, which should be as flat as possible from 20Hz to 20kHz.
When you measure such a system, its left and right channels would individually measure as flat horizontal lines (which is what you wanted); but when played back together the combined L+R channels would measure +3dB in the low frequencies compared to high frequencies - for the same reasons described in my previous posts.

As hinted in the second image above, note that ideally we shouldn't really be using any of the 'standard' target curves for EQ, because an ideal target curve for EQ would be determined from the specific loudspeaker directivity index, how reflective the room is, the listening distance, and even personal preference on the amount of bass. No predefined target curve will perfectly match all loudspeakers/rooms/listening distances and preferences.
Though of course such precision may not really be required in practice to achieve very good results.

To view the same topic from a bit different perspective, let's look at anechoic measurements of some really well-measuring non-cardioid loudspeakers like e.g. KEF LS50 Meta, Revel M105, Kali Audio LP6v2, KEF Blade 2 Meta and Ascend Acoustics ELX Ribbon Tower.
From the anechoic spins we can see that most non-cardioid loudspeakers become basically omnidirectional around 100-200Hz - this is the point where their Sound Power DI becomes very close to 0 (note the right vertical axis is used for DI values in the spinorama graphs linked above).
An ideal individual loudspeaker target curve would therefore have a flat part in the low frequencies (usually below ~150Hz), and a downward-sloping part above it - as this would be a good match for the loudspeaker directivity.
Notice how similar that is to the B&K 1974 curve (reference to original B&K article).

For simplicity I didn't want to go into absorption of high frequencies in the air, but that is something that becomes a factor too, especially for listening distances well beyond 1m.

On the topic of preference - this is why it would IMHO be really beneficial to have a simple post-EQ bass/treble or "tonality tilt" control. This would allow users to easily dial in the preferred amount of bass after the system has already been calibrated with RC for a neutral response.
As we all know, not all recordings are equally well-made, so this would also allow the users to adjust the tonality to counteract deficiencies in the recording itself.

Sorry for a long-winded post (it is difficult to balance brevity and completeness in this case), but perhaps the information could be interesting to some!

I am not sure the low frequency gets boosted 3dB compared to higher frequencies when using two separate speakers. The signals aren't correlated as closely as they are in a subwoofer. If that was the case then there would be no need to make a special case of L/R room correction with a subwoofer. It would act exactly the same as L/R room correction without the subwoofer.

Sean Olive says this
"One of the goals at the beginning of this target curve is we took a pair of good loudspeakers that measured flat and they were accurate.

We used a pair of Rebel consumer loudspeakers and we took a JBL M2, which is a professional monitor, and we put them both in this reference listening room. We measured at the person's eardrum – in this case, we're using an artificial mannequin and that became the basis of this reference curve.

The goal from the beginning was to make a headphone sound like a pair of good loudspeakers in a reference listening room."

Not just one speaker.

 

[dominikz]

Sean Olive says this
"One of the goals at the beginning of this target curve is we took a pair of good loudspeakers that measured flat and they were accurate.

We used a pair of Rebel consumer loudspeakers and we took a JBL M2, which is a professional monitor, and we put them both in this reference listening room. We measured at the person's eardrum – in this case, we're using an artificial mannequin and that became the basis of this reference curve.

The goal from the beginning was to make a headphone sound like a pair of good loudspeakers in a reference listening room."

Note that this appears to be a quote relating to how the Harman headphone target curve was created; not the "Harman target" for loudspeakers.

I am not sure the low frequency gets boosted 3dB compared to higher frequencies when using two separate speakers. The signals aren't correlated as closely as they are in a subwoofer.

My guess is that it is due to the fact that at low frequencies the wavelength quickly becomes larger than the distance between individual loudspeakers, making them act as correlated sound sources and summing accordingly (i.e. +6dB). At high frequencies the wavelength is much smaller than the distance between the loudspeakers, making them sum like uncorrelated sources (i.e. +3dB).

If that was the case then there would be no need to make a special case of L/R room correction with a subwoofer. It would act exactly the same as L/R room correction without the subwoofer.

As I mentioned previously, this effect is completely unrelated to WiiM Room Correction, it happens with any loudspeakers I measured so far - with or without DSP applied.

 

[slartibartfast]

Note that this appears to be a quote relating to how the Harman headphone target curve was created; not the "Harman target" for loudspeakers.

My guess is that it is due to the fact that at low frequencies the wavelength quickly becomes larger than the distance between individual loudspeakers, making them act as correlated sound sources and summing accordingly (i.e. +6dB). At high frequencies the wavelength is much smaller than the distance between the loudspeakers, making them sum like uncorrelated sources (i.e. +3dB).

As I mentioned previously, this effect is completely unrelated to WiiM Room Correction, it happens with any loudspeakers I measured so far - with or without DSP applied.

I guess it is easy to check whether L/R room correction without sub does indeed add 3dB by measuring. Most of my early room correction experiments were hampered by an iMM6 microphone with roll off at low frequencies so all target curves gave me an unwanted bass boost

 

[slartibartfast]

@dominikz Did you see my post about getting the same measured response regardless of which saved L/R PEQ I selected? Is it just me or do you see the same? I'll do more measurements when I get a chance and report my findings through the app.

 

[dominikz]

@dominikz Did you see my post about getting the same measured response regardless of which saved L/R PEQ I selected? Is it just me or do you see the same? I'll do more measurements when I get a chance and report my findings through the app.

I did see it but haven't yet had a chance to test. I'll try it as soon as I have some spare time and report back!

 

[slartibartfast]

I did see it but haven't yet had a chance to test. I'll try it as soon as I have some spare time and report back!

I have now satisfied myself you do indeed get a 3dB increase at bass frequencies relative to higher frequencies with separate speakers so now I am totally confused about what the targets should be for stereo measurement and individual measurement. Maybe in real music signals are not well correlated but bass does tend to get mixed as if it was mono.

 

[slartibartfast]

I did see it but haven't yet had a chance to test. I'll try it as soon as I have some spare time and report back!

Here is my measurement. You can see that Red is B&K and Orange is Harman and they are sitting on top of each other.

 

[canard]

Here is my measurement. You can see that Red is B&K and Orange is Harman and they are sitting on top of each other.
View attachment 17999

but the choice of color on pseudo vumeter....

 

[slartibartfast]

but the choice of color on pseudo vumeter....

Eh?

 

[dominikz]

so now I am totally confused about what the targets should be for stereo measurement and individual measurement. Maybe in real music signals are not well correlated but bass does tend to get mixed as if it was mono.

These are all really good questions.
The research I'm aware of allows for a lot of freedom in setting the bass level, so I'm afraid there may be no one-size-fits-all solution to it.

As previously explained, the amount of bass that sounds "natural" in a room with a specific audio system will depend on the room reflectivity, loudspeaker directivity at low frequencies, and listener preference.

Which is one of the reason why I advocate for an easy to use post-RC "tone tilt" control that goes from 'warmer' over 'default' to 'brighter'. Note how I cunningly avoided to use the terms "flat" and "natural".

 

[canard]

Eh?

shouldn't match, right?

 

[slartibartfast]

shouldn't match, right?

Well yes, that was my point.

 

[dominikz]

Here is my measurement. You can see that Red is B&K and Orange is Harman and they are sitting on top of each other.
View attachment 17999

That is indeed very close. :/
Are these measurements with filters generated by the new RC algorithm, or some manual EQ configs you saved before the beta?
If it is the latter, do you have the equivalent measurement prior to the beta?

 

[canard]

Well yes, that was my point.

so explains my reaction...tired.... ;-)

 

[EddNog]

Here is my measurement. You can see that Red is B&K and Orange is Harman and they are sitting on top of each other.
View attachment 17999

Looks more like Harman than B&K.

-Ed

 

[canard]

with rew ..filter a 3ms 330hz 0r 2-25.ms.. for see
..

 

[slartibartfast]

That is indeed very close. :/
Are these measurements with filters generated by the new RC algorithm, or some manual EQ configs you saved before the beta?
If it is the latter, do you have the equivalent measurement prior to the beta?

I think the Harman filter was generated using the previous beta L/R algorithm and the B&K was with the new algorithm.