• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Philharmonic BMR Speaker Review

Snoochers

Active Member
Joined
Mar 28, 2020
Messages
187
Likes
70
See posts 37 and 38.

The score was 5.1 which is not terrible, but it seems generally acknowledged that the rating formula unfairly penalizes wide dispersion speakers like this one, so take it with a grain of salt!

Interesting!
 

Dennis Murphy

Major Contributor
Technical Expert
Joined
Mar 17, 2020
Messages
1,071
Likes
4,535
See posts 37 and 38.

The score was 5.1 which is not terrible, but it seems generally acknowledged that the rating formula unfairly penalizes wide dispersion speakers like this one, so take it with a grain of salt!

What feature of the scoring system penalizes wide dispersion? The predicted room response?
 

napilopez

Major Contributor
Forum Donor
Joined
Oct 17, 2018
Messages
2,109
Likes
8,420
Location
NYC
The speaker measures with an unusually constant directivity response. That is to say it does not narrow in pattern control at the top end much at all (wide dispersion across the whole audio spectrum rather than just in the bass like most speakers)

But it's also a ribbon tweeter. The ribbon is very wide pattern, but very narrow vertically resulting in at or above average total sound power radiated into the room, but in a unusual manner.

We are certain that the ear interprets patterns from the horizontal plane differently to the vertical (were MUCH more sensitive to horrizontal pattern changes)

This means that although the total radiated energy may not be that high in the trebble, it is likely perceived the same way a flat or tilted up power response is (similar issue to the behringer 2031a)

Basically, there is nothing wrong with the speaker, It will need a 'house curve'. -tilting the EQ down from 1khz to 20khz. By how much? To taste, but start with 2db slant, but as much as 6-8db is not crazy.

this may be more true with lots of hard surfaces in the room.

I disagree with this interpretation a fair bit, on two in particular points. My understanding is:
  • You're right that vertical and horizontal directivity are interpreted differently, but I believe it is the vertical that tends to have a proportionally larger impact on timbre. Horizontal data largely gets interpreted as spaciousness. a flattish power curve cause by wide horizontal directivity doesn't all go to timbre, so to speak. Anecdotally, I have heard narrower directivity speakers that have a more tilted PIR/power response that sound brighter than some wider, less tilted speakers.
  • EQing to personal taste is of course fine, but I strongly disagree that it needs a 'house curve' as a general rule, or that it would require a 6-8 dB tilt further than the existing one. In a home with a lot of house surfaces, a bit of an additional tilt may be required, but the same could be said for any variety of designs.
The ER/PIR/Power curves are ultimately contingent upon directivity, but you can have speakers with very similar tilts but quite different timbre.

Sean Olive, who establishes the common target curve in his preference score paper, acknowledges as much himself:

"The degree of tilt varies among curves for Test One and the larger sample. Test One includes mostly 2-way designs whereas the larger sample includes several 3-way and 4-way designs that tend to have wider dispersion (hence smaller negative target slopes) at mid and high frequencies. This suggests that the ideal target slope may depend on the loudspeaker’s directivity."

Furthermore:

"A speaker with constant, flat directivity could theoretically satisfy the flat sound power criterion and still achieve high preference ratings, so long as it had a smooth on-axis response well-maintained off-axis. However, such speakers are not widely available."

Note in this context he's talking about actual blind test preference ratings, not the predicted preference score.

What feature of the scoring system penalizes wide dispersion? The predicted room response?

Yep, my response above for more context. The preference score aims for a target tilt based on the highest preferred speakers, but that target curve is just an average. There's nothing to indicate a speaker has to follow the target tilt to be preferred. It's just more likely to given typical speaker directivity.
 
Last edited:

Dennis Murphy

Major Contributor
Technical Expert
Joined
Mar 17, 2020
Messages
1,071
Likes
4,535
See posts 37 and 38.

The score was 5.1 which is not terrible, but it seems generally acknowledged that the rating formula unfairly penalizes wide dispersion speakers like this one, so take it with a grain of salt!

I hope members will appreciate how frustrating the preference scores are for speaker designers. I just checked the score for my mod of the little Pioneer BS-22 2-way. It clocked in at 5.3 without a sub. That's a fairly high score, and I think the mod is a fairly nice speaker. But it's not in the same league as the BMR. The BMR goes waaaaaaaaaaaay lower, throws a much wider and deeper sound stage, and reveals much more inner detail. If I really thought the Pioneer mod was a better speaker, I would be selling them and not the BMR. The only area where the mod surpasses the BMR is in vertical dispersion, and from what I can tell from the discussion on this thread, that's the explanation for the otherwise inscrutable difference in the preference scores. Until someone can show that wider (taller?) vertical dispersion is a critical factor in accurate sound reproduction, I think the best place for the preference scores would be in a footnote along with a disclaimer.
 

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
4,240
Likes
11,462
Location
Land O’ Lakes, FL
What feature of the scoring system penalizes wide dispersion? The predicted room response?
Yes, the formula was derived from your typical 2/3 way designs, and a BMR and a RAAL are non-typical, and the formula sees the shallow slope as sounding overly bright and the score drops, but even in the same paper he acknowledges that the target slope can change with dispersion differences.
 

TimVG

Major Contributor
Joined
Sep 16, 2019
Messages
1,181
Likes
2,573
If you could export the FR data for the Spinorama I could do that.

Thanks. Will take a look tonight at home. I asked because with the EQ it fits the "ideal" in-room target quite well, while also improving the direct sound, so I'd be surprised if the score didn't follow.
 

tuga

Major Contributor
Joined
Feb 5, 2020
Messages
3,984
Likes
4,281
Location
Oxford, England
I hope members will appreciate how frustrating the preference scores are for speaker designers.

Preference scores are frustrating for designers and misleading for consumers. Not much more than a science-driven, complex version of a star-rating system...
Consumers like (or have been trained to like) this sort of thing, think it'll make life easier for them. Might be alright for the average Joe but most audiophiles will be disappointed.
 

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
4,240
Likes
11,462
Location
Land O’ Lakes, FL
Preference scores are frustrating for designers and misleading for consumers. Not much more than a science-driven, complex version of a star-rating system...
Consumers like (or have been trained to like) this sort of thing, think it'll make life easier for them. Might be alright for the average Joe but most audiophiles will be disappointed.
The slope is an issue, but the main thing is showing designers that we at least roughly know what makes a good speaker (in regards to FR), and that designs like this are known not to sound as good. I am looking forward to seeing measurements of the new B&W 600 series, as they redid the crossover network and the last gen was indeed very poor in the crossover region, and that maybe due to all their internal issues this past year or two, they have gone back to making measurably good speakers.

But, I’d take knowing the Spinorama over knowing the score any day.
 

ctrl

Major Contributor
Forum Donor
Joined
Jan 24, 2020
Messages
1,616
Likes
6,086
Location
.de, DE, DEU
Until someone can show that wider (taller?) vertical dispersion is a critical factor in accurate sound reproduction, I think the best place for the preference scores would be in a footnote along with a disclaimer.
Evaluating the vertical dispersion characteristics of a loudspeaker is extremely difficult and far too little researched.

In my experience, I would even go a step further and claim that the heard vertical behavior of a loudspeaker is sometimes even counter-intuitive to its vertical measurements.

There are loudspeakers that are very constricted in vertical dispersion, in a certain frequency range, for example some Pseudo-D'Appolito speakers, but the timbre is very insensitive to vertical changes of the head.

Here is an example of a loudspeaker which, for my listening perception, has almost no tonal change when changing the listening position from sitting to standing.
Although the vertical dispersion pattern upwards is far from perfect. It even shows a comb filter effect at large angles.

Normalised vertikal FR 0 to +90deg:
1600253762192.png


In the simulation with an LR4 crossover, the spectrogram looked like shown below. This would surely get negative ratings from @amirm in a review. In the real loudspeaker I hear at the moment even primarily with an LR2 crossover, which would "worsen" the spectrogram even further, purely measurement-wise - although the heard vertical behaviour of the speaker is damn good.

1600254145600.png
 

richard12511

Major Contributor
Forum Donor
Joined
Jan 23, 2020
Messages
4,335
Likes
6,700
What feature of the scoring system penalizes wide dispersion? The predicted room response?
Yeah the PIR is what does it imo. Because of the super wide dispersion, it seems to end with a slope that’s less than what the formula wants.

OTOH, Toole says that wide dispersion with flat on axis is preferred, and is why the Salon2 beat the M2.
 

napilopez

Major Contributor
Forum Donor
Joined
Oct 17, 2018
Messages
2,109
Likes
8,420
Location
NYC
Evaluating the vertical dispersion characteristics of a loudspeaker is extremely difficult and far too little researched.

In my experience, I would even go a step further and claim that the heard vertical behavior of a loudspeaker is sometimes even counter-intuitive to its vertical measurements.

There are loudspeakers that are very constricted in vertical dispersion, in a certain frequency range, for example some Pseudo-D'Appolito speakers, but the timbre is very insensitive to vertical changes of the head.

Here is an example of a loudspeaker which, for my listening perception, has almost no tonal change when changing the listening position from sitting to standing.
Although the vertical dispersion pattern upwards is far from perfect. It even shows a comb filter effect at large angles.

Normalised vertikal FR 0 to +90deg:
View attachment 83203

In the simulation with an LR4 crossover, the spectrogram looked like shown below. This would surely get negative ratings from @amirm in a review. In the real loudspeaker I hear at the moment even primarily with an LR2 crossover, which would "worsen" the spectrogram even further, purely measurement-wise - although the heard vertical behaviour of the speaker is damn good.

View attachment 83204

Yep, vertical dispersion seems to be the frequency domain aspect that most requires further research at this point. There seems to be little conclusive about its effects. I suspect in a case like the above you were probably listening from far enough that there was enough 'room sound' for the uneven vertical dispersion to not matter so much? For me it tends to be a bigger problem in the nearfield, though still perhaps not as much as one might expect.

I hope members will appreciate how frustrating the preference scores are for speaker designers. I just checked the score for my mod of the little Pioneer BS-22 2-way. It clocked in at 5.3 without a sub. That's a fairly high score, and I think the mod is a fairly nice speaker. But it's not in the same league as the BMR. The BMR goes waaaaaaaaaaaay lower, throws a much wider and deeper sound stage, and reveals much more inner detail. If I really thought the Pioneer mod was a better speaker, I would be selling them and not the BMR. The only area where the mod surpasses the BMR is in vertical dispersion, and from what I can tell from the discussion on this thread, that's the explanation for the otherwise inscrutable difference in the preference scores. Until someone can show that wider (taller?) vertical dispersion is a critical factor in accurate sound reproduction, I think the best place for the preference scores would be in a footnote along with a disclaimer.

I totally get this. I simply think that if a speaker has exceptionally wide or narrow directivity, the score needs to be taken with a grain of salt. It simply wasn't designed for those speakers as there aren't many of them on the market.

And to be clear, it is the exceptionally wide horizontal directivity that penalizes the BMRs the most. The PIR curve isn't tilted as much as the score likes.
 

Chromatischism

Major Contributor
Forum Donor
Joined
Jun 5, 2020
Messages
4,765
Likes
3,703
Yep, vertical dispersion seems to be the frequency domain aspect that most requires further research at this point. There seems to be little conclusive about its effects.
Does the Atmos research tell us anything? We know that sounds bouncing off the ceiling and coming down from above need modification to sound correct. That's why up firing Atmos speakers have an altered frequency response to deal with the transfer function.
 

Frank Dernie

Master Contributor
Forum Donor
Joined
Mar 24, 2016
Messages
6,445
Likes
15,779
Location
Oxfordshire
I hope members will appreciate how frustrating the preference scores are for speaker designers. I just checked the score for my mod of the little Pioneer BS-22 2-way. It clocked in at 5.3 without a sub. That's a fairly high score, and I think the mod is a fairly nice speaker. But it's not in the same league as the BMR. The BMR goes waaaaaaaaaaaay lower, throws a much wider and deeper sound stage, and reveals much more inner detail. If I really thought the Pioneer mod was a better speaker, I would be selling them and not the BMR. The only area where the mod surpasses the BMR is in vertical dispersion, and from what I can tell from the discussion on this thread, that's the explanation for the otherwise inscrutable difference in the preference scores. Until someone can show that wider (taller?) vertical dispersion is a critical factor in accurate sound reproduction, I think the best place for the preference scores would be in a footnote along with a disclaimer.
I am not a fan of the preference score so never look at it.
IMHO the measurements are very interesting and the preference score adds nothing to the review.
I know this opinion may not be in a majority though.
 
OP
hardisj

hardisj

Major Contributor
Reviewer
Joined
Jul 18, 2019
Messages
2,907
Likes
13,908
Location
North Alabama
I am not a fan of the preference score so never look at it.
IMHO the measurements are very interesting and the preference score adds nothing to the review.
I know this opinion may not be in a majority though.

Ditto. That is why I never provide a preference score. IMHO, it's just not ready for prime time and only serves to muddy the waters and detract from the discussion about the data itself; instead turning discussion toward the merit and use of a single value to rank performance. (and, no, I don't think the discussion about the merit of the score results in further understanding about the speaker itself)
 

HooStat

Addicted to Fun and Learning
Joined
May 11, 2020
Messages
856
Likes
933
Location
Calabasas, CA
That is why I never provide a preference score

At the most basic level, a preference score is just a weighted average score across the dimensions used in the model. The dimensions were chosen based on other research and experience, but there are many other scores that could be created.

A score no different from a subjective assessment of a speaker -- at the end of the day, a human makes an assessment based on their own implicit weights of different factors they evaluate.

I think both add some value, including highlighting areas where there is discordance.

The benefit of the preference score is that it is applied the same across every evaluation and it allows the comparisons to be internally consistent. Whether they are generalizable to other people is the issue. Again, there is some evidence that they are generalizable (which is not the same as saying they work for everybody).

Given that it is the only game in town, I don't see any reason not to provide it. But what I would really like to see is a correct measure of variability that goes with it. In my opinion, the main issue isn't the score -- it is that we can't see the uncertainty in the score, which leads to unrealistic comparisons and assessments. The higher the score, the greater the uncertainty. But we would need the original model's variance/covariance matrix to be able to estimate it properly.
 

ctrl

Major Contributor
Forum Donor
Joined
Jan 24, 2020
Messages
1,616
Likes
6,086
Location
.de, DE, DEU
I suspect in a case like the above you were probably listening from far enough that there was enough 'room sound' for the uneven vertical dispersion to not matter so much?
For the size of the speakers (HxW 1.15mx0.45m) not very far away (<3m), but certainly more in the diffuse field than in the free field.
I could easily imagine (and speculate) that we are even less sensitive to dips in the frequency response vertically than horizontally, since our shoulders are responsible for a permanent comb filter effect when sound is coming from above.
If the narrow-band frequency dips are ignored in the above example, the vertical frequency responses are not so bad anymore.


For me it tends to be a bigger problem in the nearfield, though still perhaps not as much as one might expect.
Yes, if the free field is dominant, loudspeakers certainly behave more as one would expect (IMHO).


I am not a fan of the preference score so never look at it.
IMHO the measurements are very interesting and the preference score adds nothing to the review.
I know this opinion may not be in a majority though.
Very unusual, today I have to argue in favor of the "olive score" o_O

The prediction of the average loudspeaker rating is demonstrably quite well made by the "Olive-Score", within the scope of the test execution - one should therefore not dismiss it as irrelevant.
But you should be aware that the score partly predicts the listener preference exactly, but the predictions are on average 20% off - that still leaves a lot of room for interpretation.
1600276725638.png

But one should also not forget that the "measured preference rating" itself is again an average of ratings by listeners (if I understood correctly).

For a rough sound classification of a loudspeaker, the Olive-Score can be used safely and the more the tested loudspeaker corresponds to a "fictitious average loudspeaker", the more reliable the score will be.
 

Dennis Murphy

Major Contributor
Technical Expert
Joined
Mar 17, 2020
Messages
1,071
Likes
4,535
Yep, vertical dispersion seems to be the frequency domain aspect that most requires further research at this point. There seems to be little conclusive about its effects. I suspect in a case like the above you were probably listening from far enough that there was enough 'room sound' for the uneven vertical dispersion to not matter so much? For me it tends to be a bigger problem in the nearfield, though still perhaps not as much as one might expect.



I totally get this. I simply think that if a speaker has exceptionally wide or narrow directivity, the score needs to be taken with a grain of salt. It simply wasn't designed for those speakers as there aren't many of them on the market.

And to be clear, it is the exceptionally wide horizontal directivity that penalizes the BMRs the most. The PIR curve isn't tilted as much as the score likes.
Point taken.
 
Top Bottom