B&W 804S quasi-anechoic measurements

[Ageve]

Equipment used:
REW v5.30.5 (mac)
Calibrated UMIK-2

Playback:
Asus Essence ST PCI soundcard (using ASIO in Foobar 2000, 24/48).
Marantz SR-7013 (analog 7.1ch in (front), pure direct to avoid the slow/poor DAC filter).

I started by measuring at tweeter axis (50cm):

To make sure the distance wasn't causing the 5 kHz dip, I then measured at 1m:

To make sure there was nothing wrong with the speaker, I measured the other 804S in the room. The bass response is not as even since it was closer to a wall, but other than that, the response is nearly identical (I had the other speaker in the middle of the room during all measurements (red)):

The only way to get rid of the 5 kHz dip, and still get a somewhat even response, was to measure at 9 degrees below tweeter axis (50cm):

Port/woofer measurements:

The frequency responses of the top and bottom woofers were nearly identical. I used the top one, and it looks very similar to the measurements on Audio.com.pl.

BOWERS & WILKINS 804S

Referencyjność, ekskluzywność, innowacyjność... i generalnie przewaga nad wszelaką konkurencją nowej serii 800 tworzona jest głównie ...

audio.com.pl

B&W could apparently make a speaker without port resonances 24 years ago, but the latest 804 D4 measures like this:

Bowers & Wilkins Diamond Series 804 D4 loudspeaker Measurements

Sidebar 3: Measurements I used DRA Labs' MLSSA system, a calibrated DPA 4006 microphone, and an Earthworks microphone preamplifier to measure the Bowers & Wilkins 804 D4's frequency response in the farfield. My primary axis for the measurements was level with the tweeter, but I also performed a...

www.stereophile.com

Horizontal directivity:

To again make sure the distance wasn't causing the weird results, I re-measured at 1m, but the results were the same, only less accurate below 1 kHz because of the room, so I didn't save them.

Vertical directivity:

Distortion:

The on-axis response is more even than 804 D4 overall, despite the big 5 kHz dip, and there's no port resonance. The distortion is also very low.

The directivity is... well... interesting.

The S/D-series was the first to use a 6dB/octave crossover for the tweeter. It starts to roll off at 5 kHz and continues to play all the way down to 800 Hz. I tried covering the midrange with a pillow and the response was still the same:

The midrange still uses a 12dB/octave crossover though:

 

[Ageve]

Here are near-field measurements of all drivers.

I also compared my directivity measurements of 804S to the 804 D4 (measured by Stereophile).

The angles are a bit different (6 vs 5 deg, and 12 vs 10), but still, the results are very similar (scaled to match, 275-20000Hz, 30dB range below 0).

Vertical:

Horizontal:

Bowers & Wilkins Diamond Series 804 D4 loudspeaker Measurements

Sidebar 3: Measurements I used DRA Labs' MLSSA system, a calibrated DPA 4006 microphone, and an Earthworks microphone preamplifier to measure the Bowers & Wilkins 804 D4's frequency response in the farfield. My primary axis for the measurements was level with the tweeter, but I also performed a...

www.stereophile.com

 

[kemmler3D]

Very cool, thanks for posting! What would you call that behavior around 5khz... reverse beaming or something? Very interesting. Looks distinct enough to be intentional...

 

[Ageve]

Very cool, thanks for posting! What would you call that behavior around 5khz... reverse beaming or something? Very interesting. Looks distinct enough to be intentional...

Maybe it's intentional, but I have no idea why.

702 S2 has the same on-axis dip, and off-axis behaviour:

Bowers & Wilkins 702 S2 loudspeaker Measurements

Sidebar 3: Measurements

www.stereophile.com

 

[kemmler3D]

Maybe it's intentional, but I have no idea why.

702 S2 has the same on-axis dip, and off-axis behaviour:

Bowers & Wilkins 702 S2 loudspeaker Measurements

Sidebar 3: Measurements

www.stereophile.com

So, if you are working in a DAW and want to make your sound more "wide" or "spatial", there are various techniques that are variations on taking high-frequency content and spreading it out. What this looks like is a huge dip on-axis with the missing energy showing up off-axis, which is not entirely unlike those studio techniques.

In-room this might contribute to a wide / enveloping soundstage. The pure on-axis sound would be a bit muffled but you'd get a lot of back via first reflections.

Anyway, just a casual theory on why they might have made the response like that.

 

[dfuller]

Very cool, thanks for posting! What would you call that behavior around 5khz... reverse beaming or something? Very interesting. Looks distinct enough to be intentional...

It's some kind of diffraction or resonance usually, that's what's going on here. Note it disappears with the nearfield measurement - that tells me it's diffraction related.

 

[kemmler3D]

It's some kind of diffraction or resonance usually, that's what's going on here. Note it disappears with the nearfield measurement - that tells me it's diffraction related.

I think you're right, but it's pretty intense compared to most (any?) measurements I've seen. You've got to imagine they knew it was there...

 

[YSC]

I know I should behave civilized, but...WT_ is that 5khz valley?

 

[bodhi]

I know I should behave civilized, but...WT_ is that 5khz valley?

It's musical. You wouldn't get it.

 

[Ageve]

It's some kind of diffraction or resonance usually, that's what's going on here. Note it disappears with the nearfield measurement - that tells me it's diffraction related.

I did some more testing today. Since it's very easy to remove the tweeter, I measured it with reversed polarity:

The 5 kHz dip is gone, but (of course) there's now a larger dip at the crossover frequency (3.5 kHz).

I did one more measurement holding an old iPad cover under the tweeter, just to see what happens when the sound from the tweeter doesn't interfere with the midrange driver.

The result: +/- 2.3 dB from 200Hz to 20 kHz.

Maybe that's what B&W did when they came up with the specifications (attached).

Frequency response: 38Hz – 20kHz ±2dB on reference axis

This is what the tweeter looks like btw.

It can be removed without using any tools. Just unscrew the long shiny bolt at the back and use it to push the tweeter assembly forward.

 

[richard12511]

Very cool, thanks for posting! What would you call that behavior around 5khz... reverse beaming or something? Very interesting. Looks distinct enough to be intentional...

Indeed. It looks to me like they've introduced that huge on axis dip to counteract that flaw in the off axis. Result may be a speaker that sounds more neutral than the on axis response would suggest.

 

[richard12511]

Here are near-field measurements of all drivers.

View attachment 370448


I also compared my directivity measurements of 804S to the 804 D4 (measured by Stereophile).

The angles are a bit different (6 vs 5 deg, and 12 vs 10), but still, the results are very similar (scaled to match, 275-20000Hz, 30dB range below 0).

Vertical:
View attachment 370463
Horizontal:
View attachment 370465

Bowers & Wilkins Diamond Series 804 D4 loudspeaker Measurements

Sidebar 3: Measurements I used DRA Labs' MLSSA system, a calibrated DPA 4006 microphone, and an Earthworks microphone preamplifier to measure the Bowers & Wilkins 804 D4's frequency response in the farfield. My primary axis for the measurements was level with the tweeter, but I also performed a...

www.stereophile.com

Thanks so much for all of these measurements. Comparing these to the 804D4, it looks like the newer version might have a slightly better off axis performance, but also looks worse on axis?

 

[richard12511]

This is what the tweeter looks like btw.

It can be removed without using any tools. Just unscrew the long shiny bolt at the back and use it to push the tweeter assembly forward.

View attachment 370642

The tapered tube . I've seen Lawrence Dickie demo that tech with quite impressive results.

 

[DMill]

That 5k measure isn‘t great. I’m being polite. Otherwise, it’s a lovely looking thing that checks most other boxes. I‘m A bit jaded by speakers that outperform this for a fraction of the cost. But they are by no means broken and maybe with EQ?

 

[richard12511]

maybe with EQ?

Unfortunately off axis mismatch flaws like that can't really be corrected with EQ.

 

[tmtomh]

I'm sorry, but despite being the owner of a pair of 705s from the same period that I was pretty happy with, these measurements are just brutal. I'm sure the speaker sounds more neutral than the measurements suggest as @richard12511 notes - because the measurements suggest a total train wreck.

But I don't care how much direct/reflected compensation and fiddling they did intentionally: IMHO it's just a bad design approach to create or tolerate such significant nonlinearities anywhere in the crucial 1-5kHz region, especially a nonlinearity that you can't EQ (as Richard also notes) because the off-axis response is so different from the on-axis. Yuk.

 

[kemmler3D]

IMHO it's just a bad design approach to create or tolerate such significant nonlinearities anywhere in the crucial 1-5kHz region, especially a nonlinearity that you can't EQ because the off-axis response is so different from the on-axis. Yuk.

I'm clinging to the idea that it's an intentional thing to create extra reflected treble while maintaining a decent power response, because I want to believe B&W really knew what they were doing at some point. But I haven't heard them so I can't really say if that idea would wash.

 

[MAB]

I'm sure the speaker sounds more neutral than the measurements suggest as @richard12511 notes - because the measurements suggest a total train wreck.

My experience is you need to have them pointed in just the right direction.
OP alludes to this:

The only way to get rid of the 5 kHz dip, and still get a somewhat even response, was to measure at 9 degrees below tweeter axis (50cm)

 

[tmtomh]

I'm clinging to the idea that it's an intentional thing to create extra reflected treble while maintaining a decent power response, because I want to believe B&W really knew what they were doing at some point. But I haven't heard them so I can't really say if that idea would wash.

I don't think "they knew what they were doing" is incompatible with this design being a trainwreck. If we make the assumption - I think a very safe assumption given their target clientele - that the customer isn't going to use EQ and doesn't want to use EQ, then it makes sense that they would be willing to sacrifice both on-axis linearity and accurate directivity for the sake of the spaciousness effect you describe. Because who cares if you can't EQ the speaker if your buyer is purchasing it precisely because it (allegedly) doesn't need to be EQ'd and the buyer doesn't believe in EQ?

 

[DMill]

Unfortunately off axis mismatch flaws like that can't really be corrected with EQ.

I was trying to be polite. Tough with these measured results though.