This is my new current system
After many years of happy listening to my Meridian DSP5000s with Meridian D1500 subs, I was looking for a new project and was thinking about building a new pair of loudspeakers, or perhaps taking some vintage Tannoys and making them active. However, things took a slightly different turn when I came across a pair of Bowers & Wilkins 801F 'speakers for sale. Back in my working days, the best sound I ever heard was from a pair of 801s at Danmarks Radio's IEC listening rooms. I also heard the 801s at Abbey Road and at various other studios, but back in the '80s, they were well outside my means. So, to cut a long story short, I bought them with the plan to make them active.
I first listened to them for about a week, (without any measurements) and they did some things better than the Meridians, like vocal clarity, possibly due to less distortion from the dedicated mid-range driver, and bass is different, but I couldn't decide if any better. This has been using an old Yamaha CR1000 receiver as the amplifier. I've worked out the best way to remove the crossover and Environmental Controls, and to bring out the three drive unit connections. I also thought I would repolish the woodwork, whilst I had all the drive units and inner stuffing removed.
The 801 original, with the factory frequency response plot. This shows that the variations from 50Hz upwards are contained in a 4dB range or close to. Pair-matching is very close, around 0.25dB
However, 30 years have taken their toll, and the frequency response I measured on S/No 6143, pseudo-anechoically above 200Hz look like this:-
It uses a 3-pin Bulgin connector as it commons up the ground connections for the tweeter and mid-range. I've replaced it with a 4 pin XLR connector as I don't want to common up the grounds. If anyone should use a bridged amplifier in the future, it would effectively ground half the amp, so not worth risking.
Three bags full! The box is stuffed with wool, and I removed three plastic bags full to get to the inside and remove the internal crossover.
Here's the passive crossover. I'll keep it in case I should ever need to reinstall it, or I could put it in an external box if ever I want to play passive against active.
Here's the bass unit. The unit is mounted with a rubber gasket and bolted through rubber shock-mounts so it's more or less decoupled from the box whilst still making a "perfect" seal. Any air leaks will affect the bass, as it relies on the internal air spring as part of the suspension.
A view of the internal bracing of the box. B&W obviously thought it was inadequate, as they then went on to Matrix construction, although they also abandoned the sealed box. The 801 Matrix had much better bass extension and greater efficiency, but I've had bass-reflex for many years and fancied trying sealed box bass. You can see at the back the 6 holes for the new loudspeaker terminals.
Finally, a view of the top plate. The five white spots are where the old crossover mounted, and I've had to block up the holes with some dowels to maintain air tightness. The holes at the front are where the head mounts and those are sealed with a rubber gasket which I'll have to make. The gasket both seals the joins and decouples the head from the bass box.
With the reassembled modified loudspeaker ready to be connected to the crossover and amplifiers, the first thing I had to do was to get the gain structure right so that my 201 pre-amp would drive the A-D converter fully whilst not “wasting” bits. I was also concerned about not having the volume control so far down that all my volume control was in the first small section, as the output noise of the 201 buffer is after the volume control.
I connected the output of the 201 into the analogue inputs of the DEQ2496, which would be used as the A-D converter and fine EQ adjustment. Its digital output would then drive the DCX2496 digital crossover, which in turn drives the power amps.
Plugging my CD player and Squeezebox Touch into the 201, I played a 0dBFS 1kHz tone and wound the volume up. I hit 0dBFS on the DEQ at a volume level of 59 out of 64. So far so good. The 201 would drive the DEQ fully, and my usual volume level of around -20dB would put it somewhere around the middle of the scale. At the other end, I arranged that with full-scale into the DCX, and with the A500 power amps’ volume at maximum, I wanted 30 volts (around 110w) out. To achieve this, I needed 16dB attenuation between the DCX’s outputs and the power amp inputs. It was easy to get a balanced attenuator into the XLR plug that went into the power amp. Using my normal stock of 5% resistors, all the cables were within a needle’s width on my meter of each other, so I didn’t need to use more precise resistors. My experience is that if all the resistors come from the same batch, then although they may be within 5% of nominal, they are a lot closer than that to each other. I bought a full set of 100 of each of the 5% values almost 40 years ago, and I’m still using them.
So, as already mentioned, just setting the crossover to the nominal values derived from the passive crossover, the results were pretty horrible. Time to get measuring.
I have a Behringer ECM8000 measuring microphone, which I calibrated against a calibrated Earthworks microphone and the calibration information is automatically used by the measuring software to correct for the microphone errors. The ECM8000 isn’t particularly good, uncalibrated, but is fine once a correction curve is derived.
I measured the passive ‘speaker first, to get some idea of what my target was, and could I get it better. I struggled for a couple of days using different softwares, ARTA, REW, HolmInpulse and RMAA. Succeeded in thoroughly confusing myself , as all these packages do different things, in different ways, and trying to get consistent answers that make sense just wasn’t happening. In the end, the software that proved most useful was ARTA, once I got the hang of it. I found a tutorial, which helped enormously. I have the manual, but to understand the manual one needs to know how the software works. At first, I wanted to modify one loudspeaker only and keep a standard passive one as a reference, but decided that I might as well do both, and only think about going back to passive if I find I really don’t like the actives.
Before getting the hang of ARTA, I used it incorrectly, and using the EQ, I got some very nice flat graphs. Unfortunately, listening to the ‘speakers was horrible. There was a screaming top. Clearly something was wrong. This is what I mean by using listening as a reality check on measurements. If the measurements are good, then the sound will be right. If the sound is wrong, then clearly there’s something wrong with the measurements. It also showed the practical impossibility of setting up Active ‘speakers (certainly three way) by ear. There are just too many variables. Even sorting out the polarity of the drive units compared with what different crossovers need was hard work. The first frequency sweeps I did showed notches at the crossovers. This could be fixed by reversing the polarity, only Linkwitz-Riley filters don’t invert polarity, so it shouldn’t be necessary. This then got me going as to what the polarity of the drivers really was. Had B&W labelled the connections to make life simpler for production, or are they really labelled Red + and Black – as one would expect. I made a simple polarity tester from one diode and one resistor. This half-wave rectified a sine-wave from my generator, so I could tell on a ‘scope which way up it was. After a while fiddling with the equalisers on the microphone amplifier to sharpen up the display, I worked out that B&W had labelled the ‘speakers correctly for their polarity. So now, why was I getting nulls at the crossover points? Found that if I used pink noise rather than swept sines, I wasn’t getting the nulls, so I needed to redo the frequency response measurements to check.
I decided to use 8th order (48dB/octave) L-R filters as these have twice the cutoff that the original passive filters had, and when setting up the equalisers, there were smaller ripples to equalise with the 8th order as opposed to the 4th order. I didn’t try any other filter types as each one required different EQ settings, and life’s too short!
So the result as of now, is that I have modified both loudspeakers with the curves below. Note that I can’t measure below 200Hz due to the dimensions of my room. Compared with the original factory response, the Actives are flatter, the variations being contained in a 2dB range, whilst the originals are in a 4 dB range. The 30 year-old passives are in a 6-7dB range, 4dB if one excludes the dip at around 16k. I suspect that one of my tweeters may have lost a bit of upper HF output, but it equalised OK.
Left 'speaker (S/No 6144)
Right 'speaker (S/No 6143)
The final frequency response is now held within a 2dB window, so much more accurate than even the factory original, with pair-matching a little worse but still comfortably within 1dB, nearer 0.5dB. Pair matching affects imaging, and it's something I'm particularly sensitive to, or at least, I like pinpoint imaging. I could fiddle around with the EQ to try and get the pair matching closer, and still keep the overall response within +-1dB, but for the moment I'm happy as they are.
The last tasks were to check the effect of time-aligning the drivers and to measure distortion.
The first can be dealt with very quickly. The DCX2496 crossover includes an automatic time-align function. Using a measuring microphone, the machine sends out pulses to the drivers, measures the transit time and inserts delay to remove any time-alignment difference. I put the measuring mic at the listening position, pressed the OK button and it did it’s squeaking for a few seconds. Listening afterwards I couldn’t detect any obvious change, but it might have a subtle effect on imaging, see below in the listening tests. Nevertheless, it certainly did no harm and only takes a couple of seconds. Clearly, from an engineering point of view, time-alignment has to be a Good Thing, given that there’s no benefit whatever to them not being time aligned. If one takes it to extremes, if the tweeter were in Norfolk, then I would expect the time difference to be audible (even if the level difference was overcome).
Distortion was next. I measured at 31 1/3rd octave spot frequencies at 1m, and at a level of 90dB SPL. This level makes sense as firstly it’s about as loud as I’m likely to play and still be concerned about quality, and it’s the level often quoted in reviews or specs so I can make some comparisons.
I measured the 2nd and 3rd harmonic separately, but didn’t go to higher harmonics as they were all significantly lower with one exception as marked below.
Here’s the chart. I’ve also included the Meridian DSP5000 as a comparison.
Note 1: There was a spike in distortion at 8kHz to 2%, but this dropped to 0.3% immediately either side of 8kHz.
Note 2: There was a spike in distortion to 3% at 250Hz, but dropped either side to be 0.5% at 200Hz and 0.45% at 315Hz.
As these were both on the same loudspeaker cabinet and drive units, and I can’t find any reason for them, I can only put it down to a measurement error of some sort which I may recheck next time I have the test equipment set up. See * below.
I'm not sure what to make of the distortion figures I don't know how much better they should be when active compared to passive. The active 6143 is significantly better than 6143 passive, but unfortunately I didn't measure 6144 passive before modifying it. 6144 is still better than spec at LF, if no better elsewhere. They are, however significantly better than the Meridians (without subwoofer) at LF. This is probably a function of the much larger cones of the 801. How much difference the bass loading makes, I don’t know.
At above 100Hz, the distortion doesn’t seem to benefit from active operation which makes me question whether at very low frequencies, the crossover is the limiting factor, whilst above 100Hz, the drive units themselves and their linearity are the factors that affect distortion most, These are 30 year old drivers, and looking at distortion plots of loudspeakers of that era, these are pretty good numbers, but bettered by modern designs. I wonder how the latest B&W 802D would fare on a similar test.
*One final comment about the distortion measurement:- This was done using 500mS tone-bursts, measured at 1m. This therefore includes the room’s response. If, for example, the room had a 6dB peak at 2fHz compared with f, then the 2nd harmonic would be boosted by 6dB (or cut, if the room had a dip at 2fHz. Ditto at 3f compared to f. Consequently, these distortion measurements can only be taken as a broad indication rather than as an accurate figure. However, they were done in more or less the same position for each loudspeaker, so comparative figures should still be usable. Nevertheless, a future measurement is needed, as the effect of the room may explain the two anomalous readings mentioned above. Anyone have a spare Anechoic Chamber I could borrow?
This is where I get out of my comfort-zone in trying to put into words what I hear, so apologies in advance if it’s not as articulate as I would like.
Firstly, the precision of image struck me as staggering. Mono sounded almost as if it were coming from a dedicated loudspeaker rather than being a phantom image. Stereo is of a precision I hadn’t previously heard. I can put this down to a combination of the narrow almost diffraction-free head and accurate pair-matching. Level-matching between the two ‘speakers is typically within 0.5dB, and everywhere within 1dB above 200Hz (I haven’t measured below 200Hz, so don’t know.) One of Beethoven’s Razumovsky quartets on Radio 3 had the four musicians arranged in an arc, with the cello clearly forward and to the right of one violin, with the Viola to the left of the cello. It was slightly irritating that one of the violinists was swaying about as they played. I had not appreciated image precision of that nature before.
Although it’s been written before, even to the point of cliché, this stuff about “removal of veils”. I never thought I’d be writing anything like this, but I can’t think of a better way of describing the improvement in solo singers. I suspect that the dedicated mid-range driver, carrying 380Hz to 3.5kHz, is contributing to this extra clarity. That and the superior imaging. I don’t think it’s anything to do with better distortion performance, as distortion is pretty unexceptional, although the very flat frequency response must help.
Finally, the bass end:- Here, bass isn’t as extended as with the Meridians and subs as it starts rolling off above 50Hz, but being a sealed box, it will roll off at 6dB/octave, i.e. pretty slowly, rather than the 12 or 24dB/octave of the Meridians. Although positioned more or less exactly where the Meridians were, the 801s also don’t seem to set off the room boom in the same way the Meridians did, and the bass does have a more “gentle” character than the seemingly more aggressive bass of the Meridians. Rock music isn’t as “up-front” in the bass with the 801s, but nor is it particularly lacking or weedy, just less prominent. It does seem easier to follow complex bass lines, but I’m still not sure which I prefer, but I can play with EQ once I’m happy about every other aspect.
In summary, I’m very happy indeed with what I now have. Musically it is more satisfying than before, and technically, given the age of the drivers, I don’t think I can do better. Whether changing the drivers for modern lower distortion units would be worthwhile, I don’t know. The chances of getting something modern that fits in the same mounting holes without butchery is slight, I think, so perhaps I won’t bother unless I have to. B&W can still supply new tweeters, which might have lower distortion levels than ones 30 years old but with my increasingly limited hearing at HF, I question whether I would derive any benefit from the change. B&W's Technical Support advised me to dismantle the tweeters and clean out the magnetic gap as dust, fluff etc might have accumulated. Also, if my tweeters could be one of the earliest to have Ferrofluid, and this may have dried in 30 years. I'm reluctant to dismantle the tweeters unless I have a pair of spares in case of damage, so I'll leave it for the time being.
Now to enjoy them!
I have now changed the tweeters for new ones from B&W. Firstly, congratulations to B&W for making parts available still for 30 year old loudspeakers! Exemplary!
The new tweeters were for the 801M S3, so although were physically the same size, they are some 2dB less sensitive, and are metal-domed rather than fabric domed. However, in my application with active crossovers, there was no issue matching the new sensitivity. I also had to re-equalise the loudspeakers for the now flatter response of the new tweeters. I achieved the 2dB envelope as above, but with much less EQ needed, and both tweeters were far closer in performance than the old ones.
I've also now had the 'speakers professionally re-veneered in a lovely Zebrano veneer by a local furniture restorer.
I think they now look as good as they sound, and the brown Zebrano sets off the brown grille cloths very well. Definitely keepers!
Current System >