The LS5/8 is the
flagship of the BBC designed monitoring loudspeaker systems and you might
occasionally find an odd pair still in use at the Beeb, some 40+ years
after their inception - albeit not in "frontline" applications any more.
Most are struck by their sound quality but few
would contemplate using them in a domestic environment. These
loudspeakers are BIG!
That didn't stop my good friend Paul, which will come as no surprise to anyone who knows him! Since first hearing a pair in TV Centre back in the 1980s, he has always wanted his own pair. He managed to find a single one some years back which made a rather effective center speaker, despite not actually being able to place it within 6 feet of his CRT TV set!
But when he finally did get the opportunity to buy a pair, they were offered without the matching amplifiers. Now, this is a problem. The LS5/8 is an active design - the bass and treble drive units are fed by the left and right channels of a modified Quad 405 which incorporates a BBC-designed assembly to perform the frequency splitting and equalisation at line level. This amplifier has the BBC designation AM8/16, and is an essential component of the LS5/8 system.
However, as we already had one working amplifier and a copy of the schematic, making AM8/16 clones seemed like a relatively easy matter, in theory at least. I foolishly offered to build a pair of crossover PCBs and fit them to second-hand Quad 405s. All Paul had to do was open his chequebook...
Contents
Background to the LS5/8
Before getting into the tale of cloning the BBC AM8/16 amplifier, it's worth a general discussion about the LS5/8. A good starting point is the comprehensive design report on the LS5/8 prototype, which can be downloaded from the BBC Research and Development website: BBC RD 1979/22. I'll summarise it here:
The BBC required a new loudspeaker that could be used for monitoring music at high levels. During the 1970s, the trend was for ever louder levels and the BBC's existing larger designs (such as the LS5/5) were unable to play at the required loudness.
One of the problems with existing designs was the use of Bextrene as a cone material. Bextrene diaphragms exhibit colourations that can be reduced by painting on a suitable compound that damps the material - this process must be done by hand after the cone has been vacuum-formed as there was no way to automate this process at the time. But although Bextrene was capable of good results when appropriately "doped", it was relatively heavy, which affected the sensitivity of the completed drive unit. To achieve the required sound pressure levels, more power is required, and this leads to problems with distortion and failures.
Dudley Harwood proposed using polypropylene as a diaphragm material. Polypropylene is lighter than Bextrene, and has better self-damping properties, meaning that no additional damping is required.
Swisstone (Rogers) introduced a 12 inch bass driver with a transparent polypropylene cone - over the mists of time, it's not clear exactly how much input from the BBC was received during the development of this, but it seems highly likely that Harwood's research was influential - perhaps essential - in the design. Whatever the exact background, this unit went on to be the heart of the LS5/8. According to the Design Report, this driver was 4dB more sensitive than their previous Bextrene-coned 12" unit, and this, in combination with high temperature materials in the voice coil, permitted much higher sound pressure levels to be reached with relative freedom from colouration and distortion.
The BBC was keen to make the loudspeaker active, and if possible, 2-way. This meant that a conventional stereo amplifier could be modified; the left and right channels would drive the woofer and tweeter separately. This woofer was reasonably well behaved up to 2 or 3kHz, and so a search for a suitable tweeter began. A well known 1" dome unit from Audax was considered, but it was found to be unreliable. The problem was solved when Audax introduced a 34mm dome tweeter which was able to handle a higher power level, and was more sensitive. This tweeter wasn't quite as good as the 25mm variant, but was considered to be acceptable.
The low-level crossover was, as already mentioned, fitted to a Quad 405. One advantage of this amplifier was the option to limit the output voltage - originally provided for use with Quad electrostatic loudspeaker, but here it protects the tweeter. Note that the voltage limit is lower than with the ESL, so the details Quad give in their documentation do not apply to the AM8/16. For the early versions, it's just a case of using a different resistor value for R11. But the arrangement in the 405-2 is different. Instead of the potential divider formed by the addition of R11, there is a series pair of zener diodes at the output of the op-amp (D8 and D9), and a link is inserted to bring those into circuit. But again, the voltage limit imposed by this is too high for the tweeter. To fix this, you might expect to find that the existing zeners have been replaced with lower voltage values, but I've never seen that in practice. Instead, when you look at one of these you'll spot a pair of zener diodes soldered onto the rear of the PCB, and you'd perhaps assume that these are in parallel with D8 and D9. But appearances can be deceptive! Instead, these additional zeners are in parallel with D1 and D2. These set the voltage supply to the op-amp, which is normally ±15V, but with the extra zeners it is reduced to something lower (I'll take a look at the values next time I take one apart).
The rest of the design followed well-established BBC practice; "thinwall" cabinets made from birch plywood and damped with bitumen pads. The baffle is removeable once the grille has been removed, but of course this is often easier said than done with BBC loudspeakers as the Velcro grips with amazing tenacity.
Beware: the grilles are further secured with screws in the bottom and dowels in the top. If you don't know this, you might assume that the resistance is coming from just the Velcro, and you can easily end up damaging the cabinet. Remove the screws from the bottom of the cabinet and carefully pull the grille out from the bottom, using a wide blade (such as a decorator's scraper) while taking great care to avoid marking the veneer.
There is an auto-transformer in the cabinet for matching the sensitivity of the tweeter to the agreed standard so that amplifiers and loudspeakers are interchangeable. Carrying handles are incorporated in the side panels, and special stands were produced with a shelf to hold the amplifier.
OB Variation
A special version for OB use was produced called the LS5/8AL (or R). These were fitted to the Radio OB fleet that, at the time of writing (2010), is just nearing the end of its life, and I have seen (and heard) them in action on a few occasions. Not only were the cabinets shaped appropriately (baffles face 30° in), but an additional equaliser unit was designed to tame the bass response in the confined environment of a small OB van. The examples I've heard are surprisingly good - no-doubt the different cabinet with fewer parallel surfaces and different bass tuning help, plus the vehicles themselves were fitted with comprehensive acoustic treatment.
This is a closer view of the right speaker, complete with years of dust!
Chord version
Once Quad discontinued the 405, the BBC was left with a problem. A search for a replacement amplifier resulted in a couple of options from Chord - much more here.
Cloning an AM8/16
For some reason R&D preferred to use transistors instead of op-amps. This is curious because the earlier LS3/7 used op-amps, although this was a Designs Department design. But it's important to remember that at this point in time, op-amps were not universal in the audio world. Much as I wanted to update the design with op-amps (and ditch the input transformer to move to electronic balancing), these weren't my loudspeakers and it wasn't my money. While there was a chance of a better end result, what if Paul didn't like the sound? The only option was to copy the existing amplifier, and then if there were any question marks over the final result, it wasn't my fault!
We had a problem sourcing the aforementioned input transformer - the original is a Beyerdynamic TR/BV 353 001 081, which is very small physically. Given that we couldn't find an equivalent, we had to raid the old stocks to find 3 identical line-level transformers, and adjust the input filters to get the best frequency response. This was the only modification to the design, but I'm sure it hasn't hurt the fundamental performance in any way - indeed the replacement transformers have a much better frequency response (especially at the high end) and saturation performance than the original.
Restorating the second-hand Quad 405
The Quad 405 is an interesting amplifier, not least for the current-dumping output stage. There is plenty of information about this amplifier out there already, so there's no point going into much detail here.
For now Paul decided to just buy one Quad 405 - the example he found came from Affordable Audio in Northampton. Unfortunately, the amplifier wasn't quite as advertised - someone had fitted a standard torroidal transformer as the original transformer had failed at some point in the past. This was the first thing we noticed as the unmodified Quad weighs twice as much - when we picked it up off the shelf we nearly launched it into the ceiling! The dealer assured us that he'd not realised it had been modified, but I can't see how you could fail to notice. Especially as there was an unmodified 405 next to it on the shelf (albeit at a much higher price). Hmm...
Several aspects of the "repair" were done badly - for a start the transformer was only rated at 220VA, but luckily that wasn't such an issue for us in this active setup. But the secondary windings gave 40V rather than 35V, which means the supply rails idle at ±58V instead of ±50V. This results in much increased heat dissipation, especially as the class A amplifier collector load is two 560Ω resistors in series (with bootstrapping) rather than a current source... Finally, and this was a big problem, the replacement transformer was bolted to the bottom panel. If you have ever worked on one of these amplifiers you'll know how dumb this is. This was the first thing to fix...
The easiest way to clean the 25 years of someone else's dirt from the heatsinks and prepare the way for restoration is a full teardown:
An aluminium bracket was found in my junk store. It came from a ¼" tape machine of a very similar vintage (for added "restoration sympathy"!). The bolts joining the bracket to the front extrusion use the holes that originally secured the transformer to avoid interference problems with the rear of the heatsinks (you'll see what I mean if you ever take one apart).
I've changed the original spring-clip speaker connections to 4mm sockets and made the tweeter output yellow - the original BBC amplifier uses a 5 pin XLR socket, but there isn't really space for it on the back panel of a standard 405, and they are relatively expensive items.
This view of one of the PCBs from the Quad shows how the hot-running R30 and R31 have discoloured the PCB. This early example uses a glass-fibre PCB - the existing amp (circa 1980) uses a cheaper consumer-grade material. After this was taken, I replaced R30/31 with higher-rated components, and spaced them further from the PCB using longer ceramic supports...
There are a few worrying design details - for example there's >100V between those two fuses that are very close to each other. I bent the fuse clips away from each other, just in case... Also, not obvious from the component side, the input and output nodes run parallel to each other in the bottom-right - luckily the amplifier is an inverting design, but converting it to non-inverting is a common tweak. The passive components are typical of the time and are generally reliable, but Ben Duncan rather disparagingly refers to them as "TV set resistors"; (see page 241 of his book - ISBN 0 7506 2629 1) - not an opinion I'd take terribly seriously in all honesty. Do watch out for the 3k3 resistors dropping the 50V rails down to 15V for the op-amp (R7 and R8). They dissipate about 0.37W, which is somewhat higher than the nominal ¼W rating of those Iskra carbon film resistors. This situation is worse on the 405-2 boards on the treble, bearing in mind, as discussed earlier, the voltage limiting is acheived by placing lower value zeners in parallel with D1 and D2. Because of this, I routinely replace these resistors with 1W metal-oxide types.
You'll notice the space for R11. As mentioned previously, this is the voltage-limiting resistor used by earlier versions. It works by attenuating the op-amp output so that it clips before the rest of the amplifier has reached full output. The feedback for the op-amp is taken from the junction of R6 and R11, so the overall gain isn't affected by this. With our original amplifier, I measured a limit of 10V RMS (12.5W into 8Ω) - it's perhaps surprising how little the tweeter needs in a nominally 100W system... And you might think that forcing square-waves into the tweeter during tape-spooling isn't such a good idea, but the peak power should be limited to less than 25 watts, which is certainly better than >>100W. Besides, clipping is not the only factor in tweeter damage (see here and here, for more background).
The cloned crossover modules
This is the rebuilt Quad next to the original (complete with a Perspex cover to satisy "Health and Safety", which is not a bad idea given the live terminals on the top of the transformer). Like the clone, the original AM8/16 has had its spring-clip speaker terminals replaced with 4mm jack sockets, meaning that both the amplifiers can use the same speaker leads. Note that the bass amp (left in the above rear view) has been turned around 180° to make space for the low-level crossover. I find it surprising that this is the bass amp - why not move that function (with all the extra heat it generates) to the other end of the amplifier where it has slightly more space to breath?
Here are the original and cloned crossovers. Things aren't quite finished yet; all the volume control potentiometers were changed for a Bourns conductive plastic model, and the input transformer was changed in the original unit to match the clones. The broken thing on the original (blue square, to the left of the pot) is, or rather was, the bass-boost switch. I managed to drop the assembly and the switch exploded - it's been hardwired to the flat position. The bracket is attached by a pair of M3 screws, which conveniently use the two holes that secured the input DIN socket on the original amplifier. I've simplified things slightly by removing the PO316 jack socket (which is simply wired in parallel with the XLR socket) from the original, having decided to not fit them to the clones. They take a lot of extra space and sit dangerously close to the bass amp PCB.
This is how the crossover assembly fits into the amplifier. The bracket in the original is painted steel, but I fabricated my copies in 1mm aluminium, which is perfectly fine for this application. The original never fitted that well because some ribs in the casting of the aluminium end panel interferred, so I used a file to it to remove the worst of it. The clearance between this and the bass amplifier is still rather too close for comfort in my opinion (indeed, see later). The end panel of the (much earlier) second-hand amplifier was made from some sort of glass-fibre composite material, which made it much easier to work... Just as well, as I had to remove quite a lot of material to get a good fit as the casting pattern is rather different. This is clearly visible top-right below.
Hum...
Note also the improved earth wiring. Before stripping the second-hand Quad I noticed quite a lot of hum (and a 70mV DC offset) on the outputs. After replacing electrolytics and R30/31, and attending to some dry joints, this was much improved. Altering the earth wiring, as shown above, made an incremental difference. But some buzz was still apparent at very low level, and I spent two days chasing it, and the only cure was another mains transformer! This was most odd - indeed that was absolutely a last-resort test as I'd tried literally everything else. Unfortunately, my transformer was only 100VA, else it would have stayed in there. Luckily, Paul can't hear the buzz unless he puts his ear to the drive units...
Update
I've since come across the same effect when trouble-shooting an amplifier I was building. This time I wasn't under the same pressures (project deadlines, someone else's money, etc), so was able to think more clearly about it. Once I realised what was happening, I kicked myself!
Basically, torroidal transformers have a high leakage capacitance between primary and secondary windings, which means hum currents can be established in the ground or power supply system of the amplifier being powered. The effect isn't always noticed - the specific details of the amplifier concerned make a big difference, but with hindsight I'm sure it was the cause here. The original Quad transformer has less leakage current. To fix the problem in a torroidal transformer, an electrostatic screen normally does the trick!
Final mod...
Now, I wouldn't like to suggest that our customer is impressed by superficial details but, this was a real hit! I know that Quad purists will rightly be up in arms about it, but bear in mind this is no ordinary 405, and we felt it needed a visual cue to make that point...
LED technology has clearly moved on since 1976, so you'll definitely need to increase the resistor value. The original red LED is fed with 20mA, whereas this blue LED is being given only 8mA and it is still too bright.
Do bear in mind this project was done in 2002 when blue LEDs were relatively new and still something of a novelty - I doubt we'd do this today! If we did, we'd find that 8mA would probably be enough to light up the whole room - modern LEDs are astonishingly efficient.
Summary
The process took a surprising amount of time and effort as there were lots of details to attend to. We had to think carefully about component choice in the crossovers as every capacitor below 10nF is an (expensive) axial polystyrene. I get the feeling that these were used simply because the designers liked them - the 1% accuracy certainly isn't required in many cases. By sticking to 1% resistors and using polystyrene capacitors in the critical parts of the circuit we've emulated the response of the original rather well.
Once built, I was able to do a direct comparison between the new boards and the original by feeding the same signal into all three and comparing the outputs on my DM63 'scope, which has 4 input channels. As the input frequency is changed, all three traces stay perfectly superimposed - any errors were less than the thickness of the 'scope trace. This test was originally done on the bare boards minus the input transformers, but was repeated once the whole amplifier had been assembled, and I'm pleased to report that the results remained just as good.
Listening tests
The first listening tests were done before Paul had built the stands. I have to report that this was a real disappointment, and I had a hard job hiding this from him. Stood on the floor, the bass was massively overblown and very muddy, and the midrange was just a blur. My room is rather too small for these speakers, and I wondered if the results were at least partly because of that. Obviously, I realised that the stands would help, but as they are only a foot high, I wondered how much impact they would have.
During the next day, Paul assembled them for me to try. They are made from 18mm MDF (double-layer for the bottom), 16mm threaded rod and highly polished scaffold tubes! As you can hopefully see from the pictures, they look much better than my description...
The threaded rods have a spike on the top that was precisely turned on a lathe - this connects to the bottom of the speaker. You can see the hardwood baton on the bottom of the cabinet - these were originally fitted to locate the speakers on the BBC-designed Speedframe stands, and have now been removed so that the speaker sits neatly on the top plate. We originally planned to add spikes to the bottom plate, near the corners, but found in practice that these weren't necessary as the ends of the threaded rod made good contact with the floor.
These simple stands completely transformed the sound - I was simply not prepared for the difference they made. At the time I was using the BBC LS5/9s in my normal system - the BBC design report for these stated that this smaller speaker was subjectively very similar to the LS5/8, but with less bass extension. This was now nearly the case - compared to before without the stands, the bass was much better balanced and relatively free from boom or overhang.
In my lounge, these enormous speakers worked very well indeed. They were able to disappear, allowing instruments and voices to hang in the air very naturally - there were no obvious colourations that would otherwise cause certain instruments to attract more attention than they should. I was genuinely surprised and delighted by the detail and transparency of this monolithic system.
The most interesting part for me was to compare the BBC family. All three of them have strengths and weaknesses - some of which are relatively obvious: the LS3/5a will clearly not have the bass extension or dynamic range of the LS5/8 or the LS5/9, but given the right conditions it's surprising what they can do. I played Paul a track from a Naim sampler CD - "Snow" by Acoustic Mania, taken from a highly recommended album called "Talking Hands". This is a slow track featuring two superbly-recorded acoustic guitars which is a brilliant test track as well as being a beautiful, moving piece of music. Half way through, I cross-faded between the LS5/8s and the LS3/5As (driven by my Musical Fidelity A1). Paul, understandably, refused to believe it!
With something more 'full-range', such as The Wall, the differences are rather more stark. The LS5/8s have the required weight and authority that the loud sections demand. "Hey You" is simply stunning, and the obvious test sequence from this album ("Another Brick In The Wall" - Parts 1 through to 2) has never sounded to powerful - I've heard it this loud before, but normally the school choir and guitar solos were painful at those levels - through the LS5/8s everything simply sounded right at any level...
Not my usual thing, but the album-version of Clint Eastwood (Gorillaz) is fun. Unbeknown to the majority of its intended audience, there is an amusing sub-50Hz bassline that simply isn't there with LS3/5As. Viewing it on an oscilloscope reveals that some of the bass notes are almost perfect sine waves! Suffice to say, it's well worth a listen, if only to watch the reaction of people who have not heard it properly before...
A recording of the Vivaldi Concerto for 2 Mandolins, Strings and Basso continuo in G major was interesting (Deutsche Grammophon 415 487-2). The Andante is covered with hiss and air-conditioning noise. While you can tell this low frequency noise is present on the LS5/9s, when played through the LS5/8s you can almost feel the air pressure in the room being modulated! It would have been very easy to high-pass filter this in the control room without affecting the sound of the two mandolins, but one can only assume that this recording wasn't monitored via full-range loudspeakers...
Another gem from the Naim sampler mentioned above is "Geronimo's Free", by Ted Sirota's Rebel Souls. This is a busy track featuring a full range of jazz instruments over a Ska rhythm, which borders on avant-garde. It is quite unlike anything that I have ever heard before, and I have to say that I am hooked! Check out Ted Sirota's website for details. Ted's percussion is fantastic - I defy anyone to listen to it and not tap their feet!
Whatever you might think of the music, on a good system this track sounds fantastic - all of the instruments are well separated and hang in the room in front of you in a quite breathtaking manner. It has now become one of my standard 'test-tracks', and the LS5/8s made a really good job of it.
However, in the final analysis, the LS5/9s driven by my A1 won the day, by a very small margin. Until I heard the 5/8s, the 5/9s were one of the smoothest loudspeakers I'd encountered but the 5/8s were even smoother. To a fault, I feel. The 5/9s showed more life, more detail and sharper better imaging, while still having the ability to 'disappear'. But these differences are relatively small, and I feel that BBC R&D were largely justified in their "LS5/8s with less bass" claim...
The maiden voyage
For those who don't know him, it's worth explaining that Paul is a rather excitable character. Like a child at Christmas, he couldn't wait to get these speakers home. And who can blame him? So, he arranged a big Saturday evening event at his house, complete with champagne... Of course, this was a big mistake - performance-anxiety set in and both Quads blew up!
We rigged both speakers up in the designated, less-than-ideal position, and got them playing. While they provided the soundtrack to "Shrek" for his young children, I assembled a phono-to-XLR lead for him, allowing us to dispense with the assortment of leads and adapters currently in use. The lead was carefully threaded behind his TV/AV stack, and before disconnecting signal connections, Paul pulled the power on the first Quad (the original BBC AM8/16). He was rewarded with a large bang! I noticed that the power LED went out instantly instead of fading out gently - that was a fairly strong indicator of major trouble inside...
Five minutes later the house was full of guests who were greatly amused by all this. We were relegated to mono for the evening - the "cloned" amp continued to work well while we ate and drank, and drank and drank. Later, Paul was demonstrating the difference between the LS5/8 and his existing Mission loudspeakers, but for reasons best known to himself he was muting the LS5/8 by pulling the mains out. After a few rounds of this, and with absolutely perfect comic timing, the 405 blew its mains fuse!
By now, the guests were well-oiled and highly amused by this - normally a mains fuse blowing should be a pretty mundane event, but that's the power of alcohol for you! Unfortunately for me, I was feeling the depressive effects, and spent the evening worrying about how I was going to source output transistors on a Sunday...
It was a happy ending, thankfully. The blown mains fuse was underrated, and the first amp only required two output devices and a few passive components (and PCB track repair!). Post-mortem findings revealed that +50V got onto the signal earth of the treble amp, which was most likely caused by the very tight clearances inside the amplifier. As mentioned above, I took the time to improve this by hacking away at the aluminium casting of the end panel, and during this repair I also redressed the signal connections because by default the screen of the coaxial signal wires aren't insulated, despite being very close to the cans of the output devices. This is the short version of the story - it took a whole day to do the repair because of the care needed when working with high power audio amplifiers - needless to say the bench power supplies and variac earned their keep on that day!
Update - January 2010
Not long after posting this back in 2002, I moved to a different area where there are many sets of LS5/8s still in use, amongst many others from manufacturers like ATC, Dynaudio, Genelec, PMC, etc. I felt that it might be appropriate to update this page with my current thinking about these speakers after having used them a lot more since this project.
I'm still amazed at how good these continue to sound some 30 years after their inception, but naturally with extended exposure you will begin to notice the flaws in anything. For example, the tonal balance: they are undeniably bass-heavy. The LS5/8 was initially praised for having well-controlled bass, but over the years the opinion of Group 2 studio managers and some television sound supervisors changed sufficiently to warrant a brief investigation to investigate whether there was a change in the behaviour of the LS5/8 itself, as opposed to staff expectations in light of experiences with other commercial monitors.
The results of the investigation were reported in "BBC Research Department Technical Memorandum No. S-1149, Low-Frequency Performance Of LS5/8 Monitoring Loudspeaker", by C. D. Mathers, published on 6 November 1992. Unfortunately, because it's a document that is internal to BBC R&D, I am unable to publish any part of it, not even after 18 years. I'm not sure I should even have seen it!
Two effects were identified: first, the port tuning, which could be altered slightly to achieve a flatter response and reduced group delay at the box/port tuning frequency of 50Hz, but it was felt at the time that the effects here were probably inaudible. But the thinking around these issues has changed over the years and I speculate that designers today perhaps wouldn't dismiss this effect quite so quickly.
Secondly, a region of bass boost was identified between 50Hz and 250Hz of about 4dB. There is a rapid slope down between 250Hz and 300Hz, followed by a relatively flat section between 400Hz and 900Hz (then things become somewhat ragged from 1KHz through the crossover region - also there is a 2dB peak at 50Hz). Apparently, this broad bass boost was present in earlier examples, but not at a level that was deemed significant enough to alter the crossover. At this stage, it wasn't possible to determine the cause of this, and I'm not yet aware of any further investigation into this.
Noticing that 250Hz is the baffle step frequency for an 18" baffle made me wonder if they'd "overcooked" the baffle step compensation in the crossover? Delving deeper, when you look at the raw driver responses posted in BBC RD 1979/22 (figure 6), you see that when mounted in the inclosure, the LS5/8 woofer has an rising axial response above about 250Hz that is caused by the expected "baffle step" loss. But instead of plateauing in in the usual way, it continues at roughly the same rate for reasons unknown. However, and rather conveniently, they were able to compensate for both these features with a single filter. However, if later production woofers had a different axial response that caused the response to plateau after the baffle-step effect, the filter could cause the characteristic bass-boost that we see today. Certainly, I have seen at least 3 different sets of component values used in this part of the low-level crossover, which suggests some experimentation was done in the early days.
Another factor concerns the path the design would have taken from R&D to finished product. The BBC had a department called Designs Department, who would take a prototype from R&D and turn it into a product that can be successfully manufactured in volume. It is entirely possible that modifications to the basic design would have been made, and although you'd think that R&D would have been involved in validating these changes, do remember that Dudley Harwood retired at from the BBC at this point, so perhaps no-one left there knew and understood the design well enough?
Whatever the cause, this
post on the Harbeth
User Group from Alan Shaw is especially interesting. [The thread has
long-since disappeared from the forum, but thanks to the Wayback Machine
we
can read it here.] The last sentence of the third paragraph sums
it all up succinctly - Alan Shaw independently has observed similar
behaviour to that identified by C. D. Mathers, but this large area of
bass boost came as a complete surprise to Dudley Harwood who had intended
the loudspeaker to have a flat frequency response.
Quite how this bass boost appeared and just how much it changed over the years is liable to remain a mystery. But the key message for people who might still be using them today is to remember that they aren't a tonally accurate loudspeaker, and if you fail to account for the bass excess when producing a mix, you'll produce bass-light recordings. Also bear in mind that they were meant to be flat in "free room", or anechoic conditions, so in a typical room they will tend to be bass-heavy before you begin to consider the extra 4dB discussed above!
Away from the bass end, I've noticed that the pair matching isn't very good. Given how carefully they were originally produced, this is almost certainly an aging issue - although I haven't been able to ascertain whether the problem lies with the amplifiers or the drive units. This manifests itself as an issue with the stereo imaging; most obvious when you hit the "mono" button and fail to get a well-defined narrow phantom image dead-centre between the speakers.
In terms of detail and transparency, as stated above these aren't quite as good as an LS5/9, and my ATC SCM20SL are much better again. While the LS5/8 makes a very pleasant sound, there are better options for critical work.
That said, I find that an amplifier service helps to clean things up. As even the latest amplifiers will be at least 20 years old, they will be suffering from worn out electrolytic capacitors. This can cause a variety of symptoms, the most common being an intrusive hum or an inability to maintain the stereo soundstage when being played at high levels. Luckily, it's not difficult or expensive for a qualified engineer to replace the capacitors. At the same time I would check the amplifier is performing correctly by measuring against the original acceptence criteria, but obviously not every engineer would have access to this. Following a soak test, a final listening test is a useful final check before putting them back in service. And aside from these obvious things, I would recommend carefully checking the wiring in the region of the crossover PCB to avoid a repeat of the fireworks we experienced above.
Update 2 - March 2021 - All about the bass - with measurements
About 5 years ago I did some basic testing of an LS5/8 - apologies for taking so long to write it up! I used my Omnimic measurement setup, and the lighting hoists in a TV studio - after (very!) carefully checking the load limits - to lift up the cabinet to get it away from the floor. Measuring indoors is always difficult, so I don't claim my results are perfect, but we're interested in the general trend here - the results are more than good enough for that. Indeed, given the amount of clutter, the results were a lot better than I expected!
And here is the result, taken on the tweeter axis at 1m. Click the image to see the full-resolution version. The vertical axis is 2dB per division.
)
So everything discussed above is visible here, though the 4dB lift happens a bit higher than 250Hz - it's more like 350Hz. It's quite close to the chart in that report I mentioned earlier, but my measurements are a bit smoother through the midrange. That could be just caused by the software applying more smoothing than whatever was used by R&D. I also note that the treble lift - peaking at 11kHz - is much more pronounced.
I took a measurement of just the woofer, fed from a standard audio amplifier, so with no equalisation or crossover - this is the red curve on this graph, along with the black curve shown immediately above for reference. I shifted the red curve vertically to normalise the curves at 60Hz to make it easier to see the effect of the equalisation in the crossover.
)
I think these results verify my theory about what's happening - the "raw" response of the production LS5/8 has the expected bass loss caused by baffle-step, but as you go up in frequncy, it plateaus above that region - as you would expect - and is pretty flat before falling away at ~1.5kHz.
But look at the raw response of the prototype - it carries on rising above the baffle-step region, before peaking at roughly 900Hz, at some 6 or 7dB above the 300Hz region where the production driver stops rising. This chart is 5dB per division, so the curves might not look quite as dramatic as mine, but the overall rise from 100Hz to 900Hz is some 10dB!
)
As RD 1979/22 shows, the crossover incorporates a lift that starts somewhere in the 800Hz region and carries on until about 70Hz. The total boost is about 12dB.
)
But at the higher end of this region, the production woofer is effictively less efficient than the prototype woofer, meaning the woofer is too quiet relative to the tweeter. But that is easy to fix by simply adjusting the pre-set resistor on the BBC crossover card to bring down the level of the tweeter to match the woofer in the crossover region. But doing that brings up the bass end, relative to the midrange, and because there is too much lift in the bass slope circuit, there is too much at the bass end. Looking at my measurements of the production woofer, you need only 6dB of bass boost to compensate for baffle-step loss, but the crossover is providing at least 10dB - in other words, 4dB too much. Which that explains the 4dB bass lift.
It would be quite easy to correct for this - assuming you have access to a large enough space to be able to make the measurements. There are just 2 resistors and 1 capacitor in the crossover circuit doing this (shown in Figure 8 of the design report), and it should be easy to arrive at the required values empirically in about 10-15 minutes once you've got everything set up. But if you did so, then it wouldn't be an LS5/8 any more. Or at least, a production LS5/8.
While all of this explains some of what happened, it still doesn't explain why. I don't think we'll ever get an answer to this, but at least we have some measurements to take us out of the realm of subjectivity.
I can't help thinking about the slot the R&D prototype had in front of the woofer - which production models clearly lacked. Without a slot, a loudspeaker that is reasonably pistionic will "beam" at higher frequencies, but a slot makes the driver behave like a smaller unit, broadening the dispersion at the higher frequencies. I did some experiments with an 8" PA driver a few years back, comparing the on-axis and off-axis behaviour with and without a 100mm slot, and the results neatly confirmed the basic theory. Of course, these effects have been known about for many decades, and several older BBC models included slots - the LS5/5 being a good example - so it's actually quite surprising that the slot originally included in Harwood's prototype was dropped. But I've not yet managed to convince myself that this is the answer because actually the removal of the slot should cause the on-axis response to rise because of beaming - that's the result I got in my experiments - and we're seeing the opposite. Of course, the beaming effect is usually less severe in polypropylene drivers as they naturally decouple at higher frequencies, so my results with a paper-coned 8" driver won't necessarily give the same results...
Ultimately, it does seem more likely that the drive unit changed in some way between prototype and production models. One thing that adds weight to this theory: according to the data gathered here, there are two woofers for the LS5/8 - the LS2/9 and the LS2/11. I have no data about the differences between these two, but I can't help wondering if the LS2/9 was the original unit used by Harwood during the development of his prototype, and the LS2/11 is what went into production... Anyone know more? Answers on a postcard, please!