In all my close to 50 years' involvement with speakers, open back systems were not seriously considered due to the cancellation of low frequencies whose wavelengths were longer than the baffle.  Despite my stressing the effects of the room on any speaker system, for some unknown reason that fact was never considered regarding open backs.  Needless to say, I was quite surprised.  Had it not been for a New York friend's mentioning the Wharfedale SFB3, I'd have missed quite a sonic experience.  Having heard Magnepans in the 90's, I attributed their better bass output to the extremely large diaphragm area, about 13 ft^2 as compared to the 15 inch woofer's having a little less than 1 ft^2 or an 18 incher with 1.7 ft^2.

The first three photos show the SFB3 design but rotated 90 degrees for aesthetic reasons.  The panels measure 34" high by 31" wide.  The first two use the Beston ribbon tweeter; the third shows the Tymphany horn.

The same pair with the Tymphany horn tweeter which eliminated the 8dB drop.

Beston (green) vs Tymphany (red), both measured at 1 watt and 11 feet.

The peak between 1000 and 2200 was addressed later with a passive notch filter.  Being in a sensitive area of human hearing, it was found to be somewhat objectionable with female vocalists and violins.  However, brasses were quite realistic, especially if one likes early rock 'n roll or doowop where the saxophone was used quite frequently.

A bypass switch is being considered with this filter although a 31 band graphic equaliser works quite well, assuming one is available.

Six impedance curves of all the drivers used.  The irregularities may be caused by cone resonances. The curves of the 15s and 10s were run without any break in of the units which wasn't considered important given the lack of a cabinet.

The 0.4W difference between the tweeters was  ignored also due to its insignificance in this design.

Figure 3

Notch EQ3  0.33mh, 30uf, 16W

Impedance curves of the system with the horn tweeter and the notch filter (red) and without the notch filter (green).  The irregularities around 40hz may be due to panel resonances and/or sympathetic vibrations between the 15" and 10" units.  They are wired in parallel without filters; they cover the entire frequency range as best they can.

The hump at 1500 is due to the notch filter whose center frequency is 1500hz.  The wider hump is that of the 2uf filter in the tweeter circuit.  It actually comes into play around 9khz.  This was to tame an irritating rise in the octave between 4khz and 8khz.  This rise was caused by both the 10" and the tweeter but raising the tweeter high pass point proved more effective than inserting a low pass filter into the ten inch circuit. Raising the high pass point of the tweeter had no effect on the hump between 4khz and 8khz, hence the notch filter. (See Fig. 7)

This set of curves predates the raising of the tweeter high pass point and the insertion of the notch filter.  Note the rise between 1khz and 2.2 khz as well as the rise between 5khz and 9khz.

The thin (black & orange) traces are of the left and right units by themselves.  Neither shows the peaks as stated above but the two sum most objectionably and is quite irritating with certain music, especially female vocals.

A pair of passive notch filters was inserted later. (see fig.7)

ref: photo 3 and Fig. 4

These curves are those of Fig 4 smoothed 1/3rd octave for clarification.

The pair of aforementioned peaks is exacerbated by the dip between 2khz & 5khz.  Reversing system polarity or tweeter polarity on one or both systems had no effect on that.

This is most likely due to room reflections canceling as even an equaliser has little to no effect.

Tweeter reversed polarity

The result of reversing the tweeter polarity which didn't sound right probably due to the higher output product between 2khz and 5khz.  In short, it sounded very harsh.

These curves also predate the notch and the 9khz high pass filter which, at that time was a 4uf capacitor.

EQ-3 Notch   0.33mh, 30uf, 16W

The olive green curve is the same as the black curve of fig 5 and the purple curve of  fig 4.(Inot smoothed)  The blue  curve shows the effect of the insertion of the notch filter and the higher pass filter and also the smaller baffle.  The black curve is the blue curve with 1/3rd octave smoothing applied for clarification.

The jagged low frequency response is undoubtedly the effect of room reflections.  Despite how it looks, it sounds damned good.

These curves were obtained from the right channel speaker only.  It doesn't look too bad but it gets worse when two speakers are operating.  The change is due to room reflections.  The EQ3 removed the bump between 1khz and 2khz but introduced a dip around 1.5 khz, although that dip is not as bad as the wider bump.

Figure 8

5Vpk=3.54Vrms   Prms=2.9W (~94dB see fig 5) which is about 88dB at 11' in my living room.  This graph was captured with PicoScope Model 2205A in persistence mode.

These two curves show the effect of the sides (purple) and the sides  removed (green) from the original Briggs' SFB design..  The difference appears to be six of one and half a dozen of the other, so the sides were removed.  This led to the resizing the baffle (next photo) which not only reduced the weight by 30% but looks much better. (compare to photo 3)

The bass below about 400hz is much the same without the sides and the smaller baffle.  Compare the response below 400hz of this curve with the black curve of fig 7. 

These curves were obtained from the right channel speaker only.  It doesn't look too bad but it gets worse when two speakers are operating.  The change is due to room reflections.

Rear view showing the steel irons.  The wiring is temporary as is the baffle, the latter to be redesigned 3" wider and 3" taller.  It will be 1.5" thick  birch plywood which will probably negate the need for the angle iron but that may be added since that stuff cost me $28.

The two rear supports are temporary also and hold the baffle at a 15 degree angle to the floor, despite that it looks vertical, a consequence of my position when taking the photo.

The notch filter, E3 can be seen just under the magnet of the center speaker.

The two pink curves are those of each speaker run alone; the blue one is that of both speakers run simultaneously.  By run alone is meant that the only reactive component in the system is the tweeter high pass filter.  It shows the effect of the room reflections on the end result.  This would also be true even if the response of each speaker was identical, even if  quite flat.  Even if the room was perfectly symmetrical, the left side of the room center being a mirror image of the right side, the resultant response would be different that that of either speaker, which, if measured individually, would be identical in such a room.  However, there's no easy way to quantify the result of how the reflections from each side will interact, other than using a mic which will introduce other errors.  The mic will respond to pressure variations  at one point whereas ears respond to 2 points spaced several inches apart and with a head in between.

This is not to say that a flat response is the only one that will sound good.  Despite the irregularities in the blue curve, three friends loved the sound, as did I.  I suppose it's much the same as one's preferring their tea or coffee being stirred clockwise vs counter-clockwise.

Two curves showing the effect of a low pass filter.  The lighter curve is a copy of an original curve run earlier for the purpose of smoothing so as to retain the original #8.  The darker curve is that with the filter, a 0.28mh coil inserted into the circuit of the parallel 15" and 10".  The result of this can be seen just after 4khz (dark curve).  The intent was to reduce the peak between 4.5khz & 9khz.  It did, a little but raised the response around 10khz, the original dip being caused by destructive interference between the radiation from the 15 & 10 and that of the tweeter.  This could have been eliminated with the use of a two way second order crossover but I was hung up on the original SFB configuration of not using such.  As can be seen later, the result became more complicated as notch filters were used, which are more difficult to design than a 2-way crossover.  The good part is that of gaining more experience designing notch filters.

Such things are easier to incorporate into a speaker system when measuring the voltage levels at the speaker terminals but using a mic complicated things because of the introduction of the effect of the room. While there are available here 2 parametric and 1 graphic equaliser, it was preferred to avoid using them as that makes things too easy, tantamount to cheating.  Admittedly it's a play on words but in defense of my argument, equalisers are not part of the speaker system whereas notch filters in the speaker circuit are and I guess that can be debated also.  The end result is much the same, especially in this case because if the speakers are located in a different room,  the notch filters and the equalisers would have to be re-adjusted.

The top, photo 6L and center photo 7R are the result of the left and right speakers run alone.  The bottom photo 8L&R is that of both running simultaneously.

These photos were taken with a tri-pod mounted camera set for an exposure time of a few seconds.  The analyser was set to fast response.  Each band was moving up and down in response to the noise generator's output, hence the blurred tops of each band.

The purpose here was to get a different and possibly more realistic look at what was happening.  From Wikipedia:  Pink noise or 1⁄f noise is a signal or process with a frequency spectrum such that the power spectral density (energy or power per frequency interval) is inversely proportional to the frequency of the signal. Pink noise is the most common signal in biological systems. In pink noise, each octave (halving/doubling in frequency) carries an equal amount of noise energy.

The entire article can be found here:   https://en.wikipedia.org/wiki/Pink_noise

PHOTO 6L

PHOTO 7R

PHOTO 8L&R

Just a piccy of the rear without the notch filter.  The horizontal tan board serves as a place to temporarily mount the barrier strip and input terminals as well as a place to wedge two boards at an angle to the floor.  A solid base has yet to be made.  This baffle, when I boost the bass and crank the volume to about 95dB will oscillate with the bass to  a visible +/-1mm as seen at the very top.  A solid base should alleviate that and if necessary, another support to the floor will be added from just under the tweeter.

Another pair of baffles is under consideration.  These will be about 3" wider and 3" taller as well as thicker, perhaps 2" MDF, made by gluing two sheets of 1" MDF.  

The crossover, actually just the 2.4uf capacitor (not marked) in the tweeter circuit.  EQ3 and EQ6 are the notch (band reject) filters.  They seem to work as the response in this room is flat within 6dB from about 300hz to 20khz. (fig. 13  green)

The following set of curves shows the responses with three different tweeters.

#27 green  Dayton 15"  Faital 10"   Tymphany (horn) This curve was run several days earlier)

#31 pink  W15CS  MCM 10"  Super 3

#32 brown  W15CS  MCM 10"  Beston (ribbon)

The W15 has a lower fs which might account for the increased bass but this says nothing of its transient response. However, transient response becomes less important as the frequency goes below about 45hz

All responses are measured with both speakers running in my living room.  The mic is 11 feet away and midway between the speakers.  It is also set at ear level, seated.  It appears that the venerable Super 3 gives the best response in the upper resister.  Also, all are driven at the same level of about 1 watt each, a combined total of 2 watts.  This would lower the curves by 3 dB.

The wavelengths of the frequencies at the peaks at 40hz and 72hz are 28ft and 16ft, resp.  This closely corresponds to the distance of the speakers to the opposite wall and the width of the room.  The dip at 115 is equal to a wavelength of 10ft, the distance of the mic from the speakers and the dip at 55hz is a wavelength of 21ft.  The mic is about 21ft from the wall behind it.  The room measures 16ft wide by 34ft long.

Based on the room dimensions, 34'L x 16'W x 8'H, the ratio of L:W is close to 2:1 and the ratio of W:H is 2:1.  To exacerbate the problem is the ratio L:H being 4:1.  The room dimensions correspond frequencies of 33hz, 71hz and 142hz which correspond nicely to the peaks and troughs below 300hz.

Figure 16

This figure shows two of the above response curves with the older (green) curve removed for clarification.  Surprisingly, the old Super 3 seems to have a more linear response than the ribbon, however, one must consider the conditions under which these curves were obtained.  A note worthy of mention.  The Beston ribbon in and by itself will outperform the Super 3, especially if tested under anechoic conditions.  Living rooms are anything but anechoic.  Also, there are two other speakers working in conjunction with the tweeters and two speaker systems operating simultaneously in a reverberant field.  Also, there isn't much of a crossover.  A second order three way crossover would flatten the response, under anechoic conditions, of course.  What the response would look, like with two speaker systems playing simultaneously in a reverberant field is anyone's guess.  These two systems were tweeked for this room.

Despite the irregularities in the two curves, both sound very good.  Any audible differences would be subjective.  These curves are here for purely academic reasons.