Equalizing loudspeakers based on anechoic measurements (community project)

[edechamps]

I'm aware of it, but the simple truth is that contradicts Olive's scoring formula.

Does it, though?

@Floyd Toole is often quoted as saying "if the direct sound is wrong, nothing else matters". You look at the Olive formula, see that it seems to prioritize PIR over ON, and you conclude that Olive contradicts Toole. I don't think that reasoning holds, for several reasons.

First, it's difficult for a statistical model based on PCA to encode something like "if variable X is bad, variable Y doesn't matter". Indeed that would imply a non-linear relationship between variables, something PCA can't deal with. So the modelling process couldn't come up with something like that even if it tried. Add that to the (already quite large) bucket of "possible reasons why the Olive model doesn't achieve a correlation closer to 1".

Second, the assessment that PIR is given a higher weight in the model compared to ON is based on the fact that the combined weights of NBD_PIR and SM_PIR are higher than NBD_ON. The problem is that the weird behaviour of SM throws a wrench into that reasoning. As has been discussed multiple times in the past, SM tends to be anti correlated with NBD because, if the slope increases, then SM improves, but NBD worsens, and vice-versa. This leads to the conclusion that a speaker cannot simultaneously achieve a perfect SM_PIR and NBD_PIR at the same time (unless you're gaming the model and doing something absurd like a perfectly flat PIR with an infinitesimal slope - which, by the way, is another reason why using the model as a target could give "surprising" results). Because NBD_PIR+SM_PIR can't be maxed out, a statement like "NBD_PIR+SM_PIR has higher weight in the model than NBD_ON" is questionable at best.

Third, another problem is that, in a typical non-EQ'd speaker, deviations in PIR and ON are somewhat correlated with each other. And not just because PIR includes ON in its spatial average (I agree that's negligible), but also because many defects in speaker frequency response will create similar deviations in both PIR and ON (resonances are the textbook example). And remember: we are dealing with a statistical model that only cares about correlation, not causation. This means that when model gives weight to PIR, it also gives weight to ON at the same time, because PIR is not independent from ON. This once again makes the statement "model gives more weight to PIR than ON, therefore PIR is more important than ON" highly suspect. Because PIR is influenced by ON, it could very well be the case than the total influence of ON in the model is higher than PIR when the model is applied on a typical speaker. By the way, this assumption that deviations in ON are correlated to deviations in PIR tends to break down when EQ is applied, because EQ can easily remove such "omnidirectional" deviations. This likely means the behaviour of an EQ'd speaker is not "typical" behaviour as far as the model is concerned, which brings me to my next point…

Four, again, the model is calibrated based on typical speakers. Most speakers (especially passive speakers) are not factory EQ'd, and that was even more true back when Olive generated his model. Non-EQ'd speaker design involves different tradeoffs than EQ'd speaker design. For example, if one doesn't use EQ, a speaker designer might be tempted to leave in a frequency response dip on the grounds that it would be hard to fix and is not very audible anyway. But Olive's NBD metric doesn't care about peaks vs. dips. Thus, the predictive power of NBD_ON is weakened, and PCA reduces its weight in the model. (The case of NBD_PIR is more complicated because of its interaction with SM_PIR.) In contrast, if we're talking about EQ'd speakers, then there's no reason not to fill in dips. One could easily come up with other examples of how EQ'd speaker design might differ from non-EQ'd. Another example off the top of my head is that frequency response deviations that affect all angles equally become less important because they can be EQ'd away, which means the speaker designer can decide to make tradeoffs such as EQ'ing a resonance instead of trying to physically remove it if that makes it possible to, say, achieve better directivity control. My point is: the model was likely not designed to assess speakers that are designed that way. They are not "typical speakers" as far as the model is concerned. Therefore, the predictive power of the Olive model is likely weakened on an EQ'd speaker, limiting its usefulness as a target for EQ.

Now, are all these problems real obstacles to using the Olive formula as an EQ target? I have no idea. It's impossible to know without doing more experiments (or having access to Olive's raw data). All I'm saying is, I don't think naive, simplistic, jump-to-conclusions statements such as "Olive model was designed to correlate with preference on a typical sample of loudspeakers, and as such it presents a well defined design target" are useful. I think we should reserve judgment and not overconfidently extrapolate from studies that were never designed to be interpreted in that way. And again, I'm not fundamentally opposed to the idea of using the score as an EQ target; as I said before, I think that's better than doing nothing, and it might actually work. But summarily dismissing other approaches (such as targeting a flat ON/LW) based on a perception that the Olive model is always superior to everything else in every possible use case (such as EQ targetting) is unwarranted and overconfident IMHO.

 

[QMuse]

Does it, though?

@Floyd Toole is often quoted as saying "if the direct sound is wrong, nothing else matters". You look at the Olive formula, see that it seems to prioritize PIR over ON, and you conclude that Olive contradicts Toole. I don't think that reasoning holds, for several reasons.

First, it's difficult for a statistical model based on PCA to encode something like "if variable X is bad, variable Y doesn't matter". Indeed that would imply a non-linear relationship between variables, something PCA can't deal with. So the modelling process couldn't come up with something like that even if it tried. Add that to the (already quite large) bucket of "possible reasons why the Olive model doesn't achieve a correlation closer to 1".

Second, the assessment that PIR is given a higher weight in the model compared to ON is based on the fact that the combined weights of NBD_PIR and SM_PIR are higher than NBD_ON. The problem is that the weird behaviour of SM throws a wrench into that reasoning. As has been discussed multiple times in the past, SM tends to be anti correlated with NBD because, if the slope increases, then SM improves, but NBD worsens, and vice-versa. This leads to the conclusion that a speaker cannot simultaneously achieve a perfect SM_PIR and NBD_PIR at the same time (unless you're gaming the model and doing something absurd like a perfectly flat PIR with an infinitesimal slope - which, by the way, is another reason why using the model as a target could give "surprising" results). Because NBD_PIR+SM_PIR can't be maxed out, a statement like "NBD_PIR+SM_PIR has higher weight in the model than NBD_ON" is questionable at best.

Third, another problem is that, in a typical non-EQ'd speaker, deviations in PIR and ON are somewhat correlated with each other. And not just because PIR includes ON in its spatial average (I agree that's negligible), but also because many defects in speaker frequency response will create similar deviations in both PIR and ON (resonances are the textbook example). And remember: we are dealing with a statistical model that only cares about correlation, not causation. This means that when model gives weight to PIR, it also gives weight to ON at the same time, because PIR is not independent from ON. This once again makes the statement "model gives more weight to PIR than ON, therefore PIR is more important than ON" highly suspect. Because PIR is influenced by ON, it could very well be the case than the total influence of ON in the model is higher than PIR when the model is applied on a typical speaker. By the way, this assumption that deviations in ON are correlated to deviations in PIR tends to break down when EQ is applied, because EQ can easily remove such "omnidirectional" deviations. This likely means the behaviour of an EQ'd speaker is not "typical" behaviour as far as the model is concerned, which brings me to my next point…

Four, again, the model is calibrated based on typical speakers. Most speakers (especially passive speakers) are not factory EQ'd, and that was even more true back when Olive generated his model. Non-EQ'd speaker design involves different tradeoffs than EQ'd speaker design. For example, if one doesn't use EQ, a speaker designer might be tempted to leave in a frequency response dip on the grounds that it would be hard to fix and is not very audible anyway. But Olive's NBD metric doesn't care about peaks vs. dips. Thus, the predictive power of NBD_ON is weakened, and PCA reduces its weight in the model. (The case of NBD_PIR is more complicated because of its interaction with SM_PIR.) In contrast, if we're talking about EQ'd speakers, then there's no reason not to fill in dips. One could easily come up with other examples of how EQ'd speaker design might differ from non-EQ'd. Another example off the top of my head is that frequency response deviations that affect all angles equally become less important because they can be EQ'd away, which means the speaker designer can decide to make tradeoffs such as EQ'ing a resonance instead of trying to physically remove it if that makes it possible to, say, achieve better directivity control. My point is: the model was likely not designed to assess speakers that are designed that way. They are not "typical speakers" as far as the model is concerned. Therefore, the predictive power of the Olive model is likely weakened on an EQ'd speaker, limiting its usefulness as a target for EQ.

Now, are all these problems real obstacles to using the Olive formula as an EQ target? I have no idea. It's impossible to know without doing more experiments (or having access to Olive's raw data). All I'm saying is, I don't think naive, simplistic, jump-to-conclusions statements such as "Olive model was designed to correlate with preference on a typical sample of loudspeakers, and as such it presents a well defined design target" are useful. I think we should reserve judgment and not overconfidently extrapolate from studies that were never designed to be interpreted in that way. And again, I'm not fundamentally opposed to the idea of using the score as an EQ target; as I said before, I think that's better than doing nothing, and it might actually work. But summarily dismissing other approaches (such as targeting a flat ON/LW) based on a perception that the Olive model is always superior to everything else in every possible use case (such as EQ targetting) is unwarranted and overconfident IMHO.

Where we dissagree is that, when doing speaker EQ, you would target flat ON/LW without looking at ER, SP or PIR.

I'm, on the other hand, vouching for keeping PIR smooth while taking care of LW as well as long as it doesn't influence the score too much.

The way PIR is calculated doesn't actually have anything to do with Olive's scoring system. PIR simply takes into account a fact that what happens in a typical room is that reflections are added to the direct sound in such a way that ER and SP actually dominate the PIR as a summed response of direct and reflected sound. Although you are trying to relativese it, Olive's scoring system, verified through listening tests, is without any doubt rewarding such approach, as @flipflop and @Maiky76 correctly pointed out.

What listening tests, performed in such detail and volume as Olive did, support your EQ strategy to flatten the ON/LW without looking at ER/SP/PIR at all?
And which one would you choose to flatten when those two differ significantly, ON or LW?

 

[TimVG]

Out of curiosity I imported the Salon2 spinorama in REW and overlayed my 'corrected' F206 file. They look very similar for two very different speakers. Could be a coincidence. ON, LW and ER overlayed.

 

[flipflop]

But summarily dismissing other approaches (such as targeting a flat ON/LW) based on a perception that the Olive model is always superior to everything else in every possible use case (such as EQ targetting) is unwarranted and overconfident IMHO.

What I consider unwarranted and overconfident is disregarding a PIR based approach, which has been shown to improve NBD_ON, NBD_PIR, and SM_PIR for all speakers, and instead favoring a direct sound approach, which improves NBD_ON and NBD_PIR, but decreases SM_PIR for a category of speakers. The former works while accepting the weights of the model or disregarding them entirely. The latter only works while hypothesizing that NBD_ON should have a greater weight than it actually does.

 

[edechamps]

The way PIR is calculated doesn't actually have anything to do with Olive's scoring system. PIR simply takes into account a fact that what happens in a typical room is that reflections are added to the direct sound in such a way that ER and SP actually dominate the PIR as a summed response of direct and reflected sound. Although you are trying to relativese it, Olive's scoring system, verified through listening tests, is without any doubt rewarding such approach, as @flipflop and @Maiky76 correctly pointed out.

The Olive model is rewarding which "approach" exactly? My understanding is that @Maiky76 is following an optimization approach that tries to come up with an EQ that maximizes the Olive score. Obviously such an approach will lead to an higher Olive score - that's tautologically true, and a circular argument. I'm not sure what @flipflop is doing.

What I was trying to communicate is, there are reasons to be sceptical that the Olive score is an accurate predictor of preference for speakers that are EQ'd in that way, because the model was likely not trained on EQ'd speakers, and EQ'd speakers behave somewhat differently than non-EQ'd speakers (see my previous post), thus they are not "typical" as far as the model is concerned. In other words, it's extrapolation. And like all extrapolation attempts, we can't be sure that they actually make sense until more experiments are run.

What listening tests, performed in such detail and volume as Olive did, support your EQ strategy to flatten the ON/LW without looking at ER/SP/PIR at all?

The Olive study does not support any "EQ strategy". It makes no such claim. It only measures correlation of preference based on some summary frequency response metrics of existing, presumably non-EQ'd, typical, commercially available speakers. To turn that into an "EQ strategy", you have to assume that this correlation is also a causation. And again, that might be true, but there are also plenty of reasons why it might not be, see my previous post.

If you can come up with a study that shows that your "EQ strategy" (whatever that is, I guess smoothest possible PIR?) consistently results in higher listener preference than, say, an alternative strategy based on flat ON or flat LW, then I will be happy to concede the point. Until then, I'm not saying your approach is necessarily wrong, but I am saying that your overconfidence is unwarranted and it would be appropriate to exercise some caution.

 

[edechamps]

What I consider unwarranted and overconfident is disregarding a PIR based approach, which has been shown to improve NBD_ON, NBD_PIR, and SM_PIR for all speakers, and instead favoring a direct sound approach.

Point taken. I base my approach on the @Floyd Toole philosophy, which basically comes down to "we judge timbre/tonal balance based on direct sound, and we prefer reflected sound that is consistent with the direct sound". The latter part, "reflected sound consistent with direct sound", can't be changed with EQ. So I would focus on the former part, and simply EQ the direct sound.

Am I saying this is necessarily the correct/best approach? No, because I can't show you a study that showed this approach to be better than the alternatives discussed in this thread. What am I saying is that it's a reasonable approach given what we know, and it's not necessarily worse than EQ'ing based on the Olive score, which suffers from other problems (see post #101).

The latter only works while hypothesizing that NBD_ON should have a greater weight than it actually does.

As I shown in #101, simply reading the numerical weights from the model and directly concluding that PIR has higher weight than ON is overly simplistic, in my opinion. ON is very likely to leak into PIR due to spurious correlation, meaning that the weights are misleading because the PIR weights actually include ON influence as well.

 

[TimVG]

? My understanding is that @Maiky76 is following an optimization approach that tries to come up with an EQ that maximizes the Olive score.

He did that at first, but in listening tests I (someone has to be first) did not prefer the result. When we put the filters over the direct sound (ON/LW) in REW we saw it was compromised in favor of the PIR - but a well behaved constant directivity design will display a different PIR than a well behaved direct radiator design given both have a neutral direct sound - and a single metric for both will always favor either one or the other, in this case the direct radiator. The algorithm was then altered by him which fixed the issue (for me). Certain posters don't seem to accept that a sufficiently neutral direct sound is still a prerequisite even if it does not comply with the preference score in some loudspeakers.

 

[JIM_82]

PIR simply takes into account a fact that what happens in a typical room is that reflections are added to the direct sound in such a way that ER and SP actually dominate the PIR as a summed response of direct and reflected sound.

But a typical room doesn't exist.
Each room is specific and the only part of a non-room dependent spinorama are the ON and LW response.
It is considered by many that the critical distance is the ideal, in this case the direct = diffuse sound.
The power response of a spinorama is only the source of the diffuse sound because the reverberation specific to each room counts just as much (its linearity and its level).
Based a eq on a typical room that is not his own is a mistake.

 

[QMuse]

Difference between those 2 approaches (flattening the ON vs smoothing the PIR while also considering ON) cannot be well compared with speakers with good directivity, so let's have a look at the one that has a flaw in directivity, like Salk WOW1.

Here is how it looks when you flatten ON without looking at anything else (ON violet, PIR red):

And here is how it looks when you try to smooth PIR while looking at ON not to do too much damage (ON blue, PIR violet):

So, which one do you think would sound better in your room, when listening from say 3 meters?

Do you think that ON is all what we hear in room and that wide peak in PIR centered at 4100Hz caused by reflections adding to the direct sound will not be audible? Well good luck with that..

 

[TimVG]

Difference between those 2 approaches (flattening the ON vs smoothing the PIR while also considering ON) cannot be well compared with speakers with good directivity, so let's have a look at the one that has a flaw in directivity, like Salk WOW1.

Here is how it looks when you flatten ON without looking at anything else (ON violet, PIR red):

View attachment 75680

And here is how it looks when you try to smooth PIR while looking at ON not to do too much damage (ON blue, PIR violet):

View attachment 75681

So, which one do you think would sound better in your room, when listening from say 3 meters?

This is a valid argument. I believe a similar 'trick' was performed on the original Revel Salon and Studio if you look carefully at the spinorama.

The question becomes where the compromise lies - and if can we accept that the PIR curve for a constant directivity curve does not have to conform to the model that we are applying for what works with a direct radiator (smoothly increasing directivity) system, never mind for a system with severe directivity issues. Can we also accept that flaws of omission (vertical cancelation between two drive units) are potentially benign (such as my Revel F206/Salon2 example above) and not to be corrected while peaks (such as your example) are more offensive?

 

[edechamps]

So, which one do you think would sound better in your room, when listening from say 3 meters?

I suspect the former (i.e. flat ON). But again, I don't think these are settled questions. If anyone does a blind test (across a reasonable sample of listeners), and concludes that I'm wrong, then I will gladly stand corrected and thank them for the effort.

Do you think that ON is all what we hear in room and that wide peak in PIR centered at 4100Hz caused by reflections adding to the direct sound will not be audible? Well good luck with that..

This is where we disagree. I never, ever said that "ON is all we hear in room" - that's a ludicrous statement that contradicts the past decades of speaker science. You're saying that PIR smoothness matters; what I'm saying is, what matters is not the absolute smoothness of PIR, but the smoothness of the reflections relative to the direct sound (in other words, the directivity index - or PIR normalized to ON, if you prefer).

Is there a problem at 4.1 kHz with your example speaker? Yes, we both agree on that. Where we differ is, you see that as an issue with PIR, which can be fixed (or at least mitigated) with EQ; whereas I see that as an issue with DI, which fundamentally can't be improved with EQ.

In your second proposed EQ, the one where you take PIR into account, you're basically making ON worse without making reflections more consistent with the direct sound (you can't, since EQ can never have an effect on DI). You're saying that will sound better because the absolute PIR response looks better. I'm saying that will sound worse because you're making ON worse without improving the consistency of reflected sound vs. direct sound. Again, I doubt we will be able to settle these differences with the data currently at our disposal, unless someone is willing to organize large numbers of blind tests.

 

[QMuse]

This is a valid argument. I believe a similar 'trick' was performed on the original Revel Salon and Studio if you look carefully at the spinorama.

The question becomes where the compromise lies - and if can we accept that the PIR curve for a constant directivity curve does not have to conform to the model that we are applying for what works with a direct radiator (smoothly increasing directivity) system, never mind for a system with severe directivity issues. Can we also accept that flaws of omission (vertical cancelation between two drive units) are potentially benign (such as my Revel F206/Salon2 example above) and not to be corrected while peaks (such as your example) are more offensive?

Looking at those Salon2 responses it seems obvious that they optimised it for smooth LW. That dip around 2kHz is probably directivity error at XO region.

Looking at F206 responses it seems that smoothening the LW could make ON and ER even smoother than those of Salon2. Are they measured by you or digitalised from Harman spinorama measurements? In any case those small dips are looking very benign to me and I doubt they are audible at all, contrary to that wide peak at ER/PIR centered at 4100Hz at the example I posted.

 

[QMuse]

I suspect the former (i.e. flat ON). But again, I don't think these are settled questions. If anyone does a blind test (across a reasonable sample of listeners), and concludes that I'm wrong, then I will gladly stand corrected and thank them for the effort.

In your second proposed EQ, the one where you take PIR into account, you're basically making ON worse without making reflections more consistent with the direct sound (you can't, since EQ can never have an effect on DI). You're saying that will sound better because the absolute PIR response looks better. I'm saying that will sound worse because you're making ON worse without improving the consistency of reflected sound vs. direct sound. Again, I doubt we will be able to settle these differences with the data currently at our disposal, unless someone is willing to organize large numbers of blind tests.

Well, here we agree - the only way to settle our difference would be to organise large number of blind test to see which filter would be preferred by listeners.

If that is to ever happen I again strongly suggest to avoid speakers with good directivity as they will sound good in both EQ scenarios. It is the flawed speakers that need to be listened to see what more matters to listeners, flat ON or smooth PIR.

P.S. It goes without saying that best case is to have both, flat ON and smooth PIR. I would also like to say that IMHO smooth LW is a better design target than flat ON, exactly in the way it is shown on that Salon2 spinorama chart @TimVG posted above.

 

[edechamps]

If that is to ever happen I again strongly suggest to avoid speakers with good directivity as they will sound good in both EQ scenarios. It is the flawed speakers that need to be listened to see what more matters to listeners, flat ON or smooth PIR. It goes without saying that best case is to have both, flat ON and smooth PIR.

I wholeheartedly agree. If the speaker already has great controlled directivity, then there's no need to debate about smooth ON vs. smooth PIR - we can have both Hence why such speakers are described as "easily EQable" on the various review threads.

(I wonder if it would be possible to clearly demonstrate edge cases in the Olive model by finding (or simulating) a pathological spinorama, and a pathological EQ, such that applying the EQ makes both ON and PIR smoother, but makes the Olive score worse. I suspect it might be possible to find such an edge case if the EQ decreases PIR slope, as SM_PIR doesn't like small PIR slopes.)

I would also like to say that IMHO smooth LW is a better design target than flat ON

Sure. I'm not debating ON vs. LW here - my point was more about "direct sound" (in general) vs. PIR. In practice I agree it's probably preferable to use LW, to avoid overfitting the EQ to small features that are overly specific to the 0° angle.

 

[QMuse]

I wonder if it would be possible to clearly demonstrate edge cases in the Olive model by finding (or simulating) a pathological spinorama, and a pathological EQ, such that applying the EQ makes both ON and PIR smoother, but makes the Olive score worse. I suspect it might be possible to find such an edge case if the EQ decreases PIR slope, as SM_PIR doesn't like small PIR slopes.

I'd say it is possible to do that, but such demonstration would have little to no relation to real speakers and thus can be argued to be valid or at least to have no practical usage. For that reason I would rather see an experiment with the flawed speakers in terms of directivity, like the Salk example I showed.

 

[flipflop]

I base my approach on the @Floyd Toole philosophy, which basically comes down to "we judge timbre/tonal balance based on direct sound, and we prefer reflected sound that is consistent with the direct sound".

'Tonal balance' is not synonymous with 'sound quality'.

Revel M16 is a warm speaker:

Adam S2V is a bright speaker:

Dayton Audio B652-AIR is a neutral speaker:

The Revel and Adam are both superior to the Dayton.

The only thing he's advocating for in that quote is smooth dispersion.

As I shown in #101, simply reading the numerical weights from the model and directly concluding that PIR has higher weight than ON is overly simplistic, in my opinion.

Just to be clear: I don't disagree and I did not do this.
But let's for a moment ignore the weights and look at what the data tells us:

• A (reasonable) PIR based approach improves NBD_ON, NBD_PIR, and SM_PIR.
• A direct sound approach improves NBD_ON, NBD_PIR, and risks degrading SM_PIR.

It then can't be argued that the latter approach is superior to the former, unless you want to argue that SM_PIR is irrelevant, which is what you seem to be doing in #111.

 

[jlo]

When choosing to do correction based on ONaxis or PIR (or real In-Room ?), remember results of Room correction evaluated by Sean Olive :
"The room corrections that were able to fill in this sound power dip received higher preference and spectral balance ratings"
"In-room measurements spatially-averaged around the primary listening seat are good indicators of listeners’ preferences, perceived spectral balance,and comments"
In the evaluated case of B&W802N showing about 5dB dip in power response at 2/3kHz, the best corrections filled this 5dB dip and surely over-corrected the 2.5dB dip seen on axis.
By the way, I would be very interested (and I may not be the only one) if TimVG could show us some MMM measurements at listening position for his M105, KH80 and 8030C

 

[TimVG]

When choosing to do correction based on ONaxis or PIR (or real In-Room ?), remember results of Room correction evaluated by Sean Olive :
"The room corrections that were able to fill in this sound power dip received higher preference and spectral balance ratings"
"In-room measurements spatially-averaged around the primary listening seat are good indicators of listeners’ preferences, perceived spectral balance,and comments"
In the evaluated case of B&W802N showing about 5dB dip in power response at 2/3kHz, the best corrections filled this 5dB dip and surely over-corrected the 2.5dB dip seen on axis.
By the way, I would be very interested (and I may not be the only one) if TimVG could show us some MMM measurements at listening position for his M105, KH80 and 8030C

Soon! Btw that preferred 802N curve which you refer to also had more and deeper bass response which will have definitely affected listener preference.

 

[edechamps]

'Tonal balance' is not synonymous with 'sound quality'.

Sure, I think I'm just being sloppy with the definitions. I think "timbre" would be more accurate. Basically, the perceived quality of the sound source, as opposed to any spatial attributes. I should probably read @Floyd Toole's book again to find the precise terms he used there.

The point I was trying to make was, we object to frequency response aberrations in the direct sound (not talking overall trend/slope here, that's another debate, I'm just referring to local deviations), and we object to the frequency response of reflected sound being different from the direct sound (which is measured by the directivity index, or, alternatively, PIR normalized to ON). I will of course agree that PIR is affected by both direct sound and reflected sound (that's literally the definition of PIR), but I do not agree that PIR is a better indicator of perceived sound quality than direct sound (ON, LW, whatever) and directivity index (or PIR normalized to ON, if you prefer) combined. To reiterate my position: if you EQ PIR at the expense of direct sound, then you're making direct sound worse - and since the directivity index is always a constant, the end result sounds worse overall.

But let's for a moment ignore the weights and look at what the data tells us:

• A (reasonable) PIR based approach improves NBD_ON, NBD_PIR, and SM_PIR.
• A direct sound approach improves NBD_ON, NBD_PIR, and risks degrading SM_PIR.

It then can't be argued that the latter approach is superior to the former, unless you want to argue that SM_PIR is irrelevant, which is what you seem to be doing in #111.

Obviously I will never object to an approach where NBD_PIR and SM_PIR are improved as a "bonus" of improving NBD_ON (which is what I would expect to happen at least some of the time, anyway). I don't think there's any debate on that front. But there will necessarily come a point where you can't improve one variable without compromising another, so at some point you need to make a choice. You could solve that problem using the Olive model's weights and maximize the overall score, but that comes with all the caveats I described in my previous posts (namely, that the model was never meant to be used that way was not trained for that purpose).

I'm confused as to why you'd say that improving NBD_ON will not make NBD_PIR worse, but that it might make SM_PIR worse. To be clear, I agree that it might make SM_PIR worse (again, past a certain point, no change you make can improve all variables simultaneously), I'm just confused as to why you'd treat NBD_PIR differently from SM_PIR in that regard. As far as I can tell, depending on the situation, optimizing NBD_ON could make NBD_PIR worse, or SM_PIR better, or neither, or both.

 

[Robbo99999]

The point I was trying to make was, we object to frequency response aberrations in the direct sound (not talking overall trend/slope here, that's another debate, I'm just referring to local deviations), and we object to the frequency response of reflected sound being different from the direct sound (which is measured by the directivity index, or, alternatively, PIR normalized to ON). I will of course agree that PIR is affected by both direct sound and reflected sound (that's literally the definition of PIR), but I do not agree that PIR is a better indicator of perceived sound quality than direct sound (ON, LW, whatever) and directivity index (or PIR normalized to ON, if you prefer) combined. To reiterate my position: if you EQ PIR at the expense of direct sound, then you're making direct sound worse - and since the directivity index is always a constant, the end result sounds worse overall.

Some good nuggets of information in that paragraph! One question, PIR, that's Predicted In room Response right? As an additional point, what's the point in EQ'ing a theoretical predicted in room response created from an average room when rooms are always different, wouldn't you just measure you own real in-room response using UMIK and then EQ that if a person thought PIR was important as the main EQ target. Your comments on EQ'ing direct sound make sense to me.