Mostly copy/pasted from my review on my website. I took some out so if you want to see the full review go here:
https://www.erinsaudiocorner.com/loudspeakers/philharmonic_bmr/
Philharmonic BMR Bookshelf Speaker Review
Intro
Dennis Murphy designed the Philharmonic BMR some years back and it lit up the DIY circuit. This is a 3-way “bookshelf” (a big bookshelf) incorporating the following drive units: RAAL, 64-10 Ribbon Tweeter, BMR 2.5″ Midrange, and Scan-Speak 18W8545-1 Paper-carbon cone woofer. There are countless threads about this speaker and the design available via a simple Google search so I won’t spend my time going on about it. It is important to note, though, this speaker tested is in a different cabinet than current offerings. This shouldn’t effect the results of the test considerably (I assume the same volume and port tuning is used with the various enclosure shapes) but it is worth noting as some response differences can arise with a rounded enclosure with a flat baffle vs other designs some may come up with for their own enclosure.
The speaker can be purchased complete through
Salk Sound for $2395 - 2595, depending on what finish you purchase. Alternatively, this speaker can be built as a kit via
Meniscus Audio and you can also even purchase a “flat-pack” for the cabinet from
Speaker Hardware. Fully assembled (but unfinished), the DIY route would cost you in the neighborhood of $1150 to 1450 (depending on the flat pack order) and if you decided to upgrade components from Meniscus you could do so as well.
Foreword: Subjective Analysis vs Objective Data (click for more)
Product Specs and Photos
Grille On:
Objective Data
Unless otherwise noted, all the data below was captured using
Klippel Distortion Analyzer 2 and Klippel modules (TRF, DIS, LPM, ISC to name a few). Most of the data was exported to a text file and then graphed using my own MATLAB scripts in order to present the data in a specific way I prefer. However, some is given using Klippel’s graphing.
Impedance Phase and Magnitude:
Impedance measurements are provided both at 0.10 volts RMS and 2.83 volts RMS. The low-level voltage version is standard because it ensures the speaker/driver is in linear operating range. The higher voltage is to see what happens when the output voltage is increased to the 2.83vRMS speaker sensitivity test.
From the above data we can see the following:
- The tuning frequency of the enclosure is approximately 32Hz.
- The minimum impedance dips to about 4.5 Ohms around 1.4kHz with a nominal impedance at 6 Ohms. You might be able to power this with a standard AVR (as long as it is 6 Ohm rated). However, the low sensitivity of this speaker and a long listening distance means it will need ample power to get to higher output levels so choose how you power this speaker accordingly.
- There are no signs of resonance (typically visible as bumps/wiggles in the impedance that deviate from the curve).
Frequency Response:
The measurement below provides the frequency response at the reference measurement axis - also known as the 0-degree axis or “on axis” plane - in this measurement condition was situated at the ribbon.
The mean SPL, approximately 83.2dB at 2.83v/1m, is calculated over the frequency range of 300Hz to 3,000Hz.
The blue shaded area represents the ±3dB response window from my calculated mean SPL value. As you can see in the blue window above, the Philharmonic BMR has a ±3dB response from 62Hz - 20kHz. Even better, a tighter window of linearity is provided in gray as ±1.5dB from the mean SPL and this speaker stays within this window from about 80Hz all the way to 20kHz (ignoring the 0.1dB around 4kHz). That is impressive!
The speaker’s F3 point (the frequency at which the response has fallen 3dB relative to the mean SPL) is 62Hz and the F10 (the frequency at which the response has fallen by 10dB relative to the mean SPL) is 35Hz. This speaker has nice low frequency extension.
The one thing I find a bit odd is the 1dB shelf below 300Hz. I can’t imagine this being baffle step because, from what I can tell, the crossover point between the woofer and the midrange is about 550Hz.
Below are both the horizontal and vertical response over a limited window (90° horizontal, ±40° vertical). I have provided a “normalized” set of data as well. The normalization simply means that I took the difference of the on-axis response and compared the other axes’ measurements to the on-axis response which gives the viewer a good idea of the speaker performance, relative to the on-axis response, as you move off-axis.
As I said above, the provided frequency response graphs were given with a limited set of data. I measured the response of the speaker’s vertical and horizontal axis in 10-degree steps over 360-degrees. Nearly 70 measurements in total are represented in my data. As you can imagine, providing all those data points in a single FR-type graphic below is a bit overwhelming and confusing for the viewer. A spectrogram is an alternate way to view this full set of data. This takes a 360-degree set of data and “collapses” it down to a rectangular representation of the various angles’ SPL. I have provided two sets of data: one set for horizontal and one for vertical. Each set consists of 2 graphics:
- Full response (20Hz - 20kHz with the angles from 0° to ±180°) with absolute SPL values
- Full, “normalized” response (20Hz - 20kHz with the angles from 0° to ±180°) with SPL values relative to the 0-degree axis
Normalized plots make it easier to compare how the speaker’s off-axis response behaves relative to the on-axis response curve.
The above spectrograms are the standard way of providing directivity graphics by most reviewers. Some prefer
not to normalize the data. Some prefer to normalize the data. Either way, it’s a useful visual to get an idea of the directivity characteristics of a speaker or driver.
However, these “collapsed” representations of the sound field are not very intuitively viewed. At least not to me. So, I came up with a different way to view the speaker’s horizontal and vertical sound field by providing it across a 360° range in a globe plot below. I have provided both an absolute SPL version as well as a normalized version of both the horizontal and vertical sound fields.
Note the legend provided in the top left of each image which helps you understand speaker orientation provided in my global plots below.
CEA-2034 (aka: Spinorama):
The following set of data is populated via 360-degree, 10° stepped, “spins” from vertical and horizontal planes resulting in 70 unique measurements. Thus, this is sometimes referred to as “Spinorama” data. Audioholics has a great writeup on what these data mean (
link here) and there is no sense in me trying to re-invent the wheel so I will reference you to them for further discussion. However, I will explain these curves lightly and provide my own spin on what they mean (pun totally intended). Sausalito Audio also has a good write-up on these curves
here. Furthermore, you can find discussion in Dr. Floyd Toole’s book “Sound Reproduction”.
Here’s my Amazon affiliate link if you want to purchase it and help me earn about 2% of the price. And, finally,
here is a great video of Dr. Toole discussing the use of measurements to quantify in-room performance.
In short, the CEA-2034 graphic below takes all the response measurements (horizontal and vertical) and applies weighting and averaging to sub-sets and can help provide an (accurate) prediction of the response in a typical room. If there is a single set of data to use in your purchase decision, this is probably it.
Alternatively, click this arrow, if you want my quick take on what these curves mean without going to another site.
The following are my takeaways from the data. More may be obvious to others or myself (with more time).
- There are various ways you can interpret the data, substituting the “dips” for “peaks” in analysis of the non-linearity. For example, I see the 400-500Hz and 700-800Hz regions as potential resonances (namely, the latter). However, if you view the 500Hz and 1kHz regions as dips, then one could argue the other bumps in response are not resonances. Looking at the nearfield data (in the Miscellaneous section), it appears the woofer has a notch at approximately 1kHz. I pulled up the driver data for the Scan-Speak 18W8545-1 and see there is a 2dB tall resonance spread from about 600-700Hz so I’m inclined to believe this really is a resonance, but not breakup, albeit small. Without disabling the other drive-units, however, it is hard to know for sure.
- The Listening Window lies mostly within the on-axis response but notably different at ~2.5kHz to 3.1kHz. This is within the region of the crossover between the midrange and the tweeter. You’ll notice in the above graphics that the off-axis response of the horizontal response yields a bump in this region compared to the on-axis response. Additionally, the RAAL is quite directional, vertically, due to its tall orientation. Therefore, it “beams” at approximately the frequency that equals the radiating height (70mm; 2.75 inches). Using the speed of sound and half-wavelength you can calculate the beaming point for this dimension as ~2.5kHz (SoS = 13500 under standard conditions: SoS/2/2.75 = 2449 Hz). So, the difference in the Listening Window vs the on-axis response is mostly caused by the horizontal response but the vertical directivity mismatch of the BMR midrange to the RAAL tweeter is a contributor as well.
- Outside of the two comments above, this speaker displays a very nice set of CEA-2034 curves with relatively wide directivity index curves. It has very wide horizontal directivity (which can be observed in the spectrogram and global plots presented earlier) with a horizontal response window of ±60° (non-normalized) up to about 10kHz. However, the vertical directivity is considerably narrower thanks to the RAAL tweeter’s height as mentioned above. The vertical window begins to narrow considerably above about 4kHz where the tweeter is the provider of response.
Below is a breakout of the typical room’s Early Reflections contributors (floor bounce, ceiling, rear wall, front wall and side wall reflections). From this you can determine how much absorption you need and where to place it to help remedy strong dips from the reflection(s). Again, as a pointer to the wide horizontal envelope, notice how the Rear Wall Bounces Curve is relatively high in amplitude (for a front-facing tweeter, at least) until about 10kHz.
And below is the Predicted In-Room response compared to a general Target curve equaling -1dB/octave. The -1dB shelf below 300Hz effects the location of this target curve. Additionally, this, speaker didn't sound as bright as this overlay may indicate. Though, I did find it to be about 1dB too high in treble up until 10kHz where I wanted more "air". The combination of what I heard and this shelf would probably put the target and predicted curves more in line with what I heard.
You may ask just how useful the above prediction is. Well, I’d be remiss for not delving in to that a little bit here. Please see my Analysis section below for discussion on this.
Total Harmonic Distortion (THD) and Compression:
Distortion and Compression measurements were completed in the nearfield (approximately 0.3 meters). However, SPL provided is relative to 1 meter distance.
Extra Measurements:
Nearfield measurements.
Mic placed about 0.50 inches - relative to the baffle - from each drive unit and port. While I tried to make these as accurate in SPL as I could, I cannot guarantee the relative levels are absolutely correct so I caution you to use this data as a guide but not representative of actual levels (measuring in the nearfield makes this hard as a couple millimetres’ difference between measurements can alter the SPL level).
Step-Response.
Not Zoomed:
Zoomed and annotated for the arrival time of each drive-unit’s impulse. You can see the drivers are not time-aligned and arrive distinctly separately. The difference in time between the woofer (2.5ms) and the midrange (1.33ms) is approximately 1.2ms. To put this in perspective, 1.2ms is approximately 16 inches.
Subjective Evaluation:
Before I dive in to the subjective feedback let me first give you the layout of my room…
If the false wall part is odd to you, here's some background. I don't like seeing speakers when watching a movie. So, I built a false wall and used an acoustically transparent screen with speakers behind it. The wall is only 2x4's; no panels of wood or anything. Just a skeleton of a wall to give me something to attach the screen and acoustic treatment to. There is 2-inch wedge foam affixed to the 2x4 studs and between the false wall and back of the room are the front speakers (L/C/R & 18-inch subwoofers).
My demo music:
Subjective Analysis Setup:
- The speaker was aimed on-axis with the vertical listening axis between the mid and tweeter per Neumi’s manual.
- I used Room EQ Wizard (REW) and my calibrated MiniDSP UMIK-1 to get the volume on my AVR relative to what the actual measured SPL was in the MLP (~11 feet from the speakers). I varied it between 85-90dB, occasionally going up to the mid 90’s to see what the output capability was. In a poll I found most listen to music in this range.
- All speakers are provided a relatively high level of Pseudo Pink-Noise for a day or two - with breaks in between - in order to calm any “break-in” concerns.
- I demoed these speakers without a crossover and without EQ.
- Components: Oppo BDP-103 playing music off my thumb drive feeding signal via HDMI to a Denon AVR-X4000 which then feeds in to a refurbished Adcom GFA-545 for power.
I listened to these speakers and made my subjective notes before I started measuring objectively. I did not want my knowledge of the measurements to influence my subjective opinion. This is important because I want to try to correlate the objective data with what I hear in my listening space in order to determine the validity of the measurement process. I try to do a few listening sessions over a couple days so I can give my ears a break and come back “fresh”.
In the interest of time and due to no feedback discussing these matters, I am not providing photos of my notes (seems like no one cares and no sense in wasting my time if it’s not important to anyone). I did complete in-room measurements but sold the laptop without saving the results to my external drive so I am unable to provide those at this moment.
Here’s the rundown of my subjective notes (in quotes) and where it fits with objective:
- Overall, I found the max SPL I could drive the speakers to was around 98-100dB at my listening position, depending on the music. Which is quite loud at 11.5 feet! The tell-tale for me here was the woofer ran out of mechanical excursion. I didn’t notice any
- Depeche Mode’s “Enjoy The Silence”: really nice punch in the bass. When the lead singer says “… oh my little girl…” I heard modulation in his voice that I have never heard before. That’s a good thing. It means, at least to me, the resolution of the speaker is extremely high. This was a common trend throughout the rest of my demo.
- Steve Winwood’s “Higher Love”: The panning drums on this track were very nice. I wrote “WOW” in my notes. I was jamming to this track at about 95dB and it sounded incredible. Chaka Khan’s voice at the end literally gave me goosebumps. Goosebumps. I did note there was something in the 800Hz region that I hadn’t noticed before and I didn’t have a specific way to describe it. It stood out; I wasn’t sure if it was good or bad. Looking at the data, I believe I may have been hearing the resonance I discussed earlier.
- The Cars “Magic”: Missing some “crackle (10k - 12kHz?)". Looking at the data shows a falloff in response above 10kHz.
- The Police “Wrapped Around Your Finger”: This speaker reproduced the bass line better than anything I’ve ever heard to date. I noted that it seemed a bit excessive in the 2-3kHz region, though. I see in the data that the off-axis response is a bit brighter in this region but I’m not sure if that’s the reason I heard what I heard or not.
- Peter Gabriel’s “Sledgehammer”: I could hear inflection in his voice that I’d never heard before which I felt was, again, attributed to the HF response.
- Portugal The Man’s “Feel It Still”: The chord progression on this track… I could feel it coming at me. Like it was coming out of the speakers. Crazy.
- John Mayer’s cover of Tom Petty’s “Free Fallin”: I see audiophile types say “you want to feel like you’re in the audience”. My response is typically “yea, if the recording is made in a way that will allow it”. The easy example is a singer with auto-tune: he/she shouldn’t sound “real” at all. But, sure, if the recording is of good quality and “real” was the objective then using that as a reference for a real sound is just fine. I feel this recording is one of those. But I haven’t heard a speaker yet that “puts me there” like these do. It was so cool, I even got my wife to listen in the sweet spot. Of course, she said “sounds cool” and went on her way. But the next day she had that song on one of her playlists. So… I take that as a win for this speaker.
- Paul Simon “Kodachrome”: Drums on the left side of the stage were clearly behind the speaker. Though, I did find something to be a bit resonant in the 250-300Hz region. The majority of the data doesn’t reveal any reason for me having heard this. However, when I look back at the normalized horizontal response I can see a small bump right in this region as the response trends off-axis. I am honestly very impressed that a) I heard this and b) there is correlating data. I keep saying… if the data doesn’t show it then either something is wrong or you’re not looking at the right data. This is a prime example of the latter.
- Tears For Fears “Everybody Wants To Rule The World”: perfect ~160hz vocal tone. “Behind You”… “Be-hind”… the “beee” portion of this word sounded a bit boomy. Might be the 250-300Hz resonance noted above. Might also be my imagination. Or might be the way it is supposed to sound. I believe it’s resonance, though. The breakdown portion starting around 2:40 has a three-chord progression; low notes. I have listened to this song thousands of times… thousands… it is my favorite song of all time… and I have never noticed the “pluck” of those notes until I heard it on these speakers. Of course, now that I’ve heard it, I notice it on other speakers… but I credit the BMRs for making that possible.
- Norah Jones “Tell Yer Mama”: Backup vocalist on the left is as wide as my room with the speakers placed about 3.5 feet from the side walls.
I chose not to run Dirac Live with these speakers due to time constraints. But with a little bit of EQ on the 4-8kHz region (where I felt it was about 1dB too high) and with room correction, I can only imagine how much better my listening impression would have been.
Bottom Line
These speakers are awesome. The tonality revealed things in songs I’d never heard before (in only one case did I find that to be a negative). The resolution reveals little details such as slight inflections in voices and even accentuates the breathing-in of a vocalist before they begin their next line. It was a new experience for me to hear these kind of details. The soundstage is incredibly large and well balanced and the reflections in the room help to extend the soundstage to a size I’ve yet to hear from another speaker to date.
For the kit price of about $1000, I can’t imagine anything coming close. I would have no qualms at all paying the kit price and ordering the flat pack and assembling this myself for about $1500/pair. And if I had more money and less time I wouldn’t mind ordering them complete from Salk. No matter what your experience level or desire to build, I think these speakers are a no brainer for their price.
However, I would like to hear what these speakers can do if one was to cover the tweeter so as to make it a small square rather than as tall as it is. That would theoretically bring the beaming point up to match the horizontal dispersion pattern and I would really be interested to see how that additional vertical window of dispersion would help (or potentially hurt) these speakers.
Kudos to Mr. Murphy on a very well designed speaker!
The End
If you like what you see here and want to help me keep it going, you can donate via
this PayPal Contribute link. Testing and reporting the data and analysis takes me 10+ hours. It’s definitely a labor of love. That said, there’s no fame or fortune in this and all my test speakers are typically purchased and paid for by myself with help from contributions or purchases made through my affiliate links (which is negligible). Your donations help me pay for new test items, shipping costs, hardware to build and test, etc. Even a few dollars is more helpful than nothing. If you don’t mind chipping in a few bucks now and again it would truly be appreciated.
Thanks!
