MPB's LS3/5a Project
This project started with the intent to build a pair of the famous BBC designed LS3/5a speakers using all new components from Stirling Broadcast (cabinets) and Falcon Acoustics (most everything else, including new production drivers and crossovers). I decided to document the process, and examine the affect of some of the design elements of the speaker, such as impact of the perforated metal tweeter grille/protector. I've included numerous measurements to examine the affect of various crossover components. Note that this webpage discusses the earlier "15 ohm" version of the LS3/5a.
For background and reference material on the LS3/5a, the first and best place to go is Paul Whatton's site LS35a.com, which has a wealth of information on the design and history of the speaker. Paul's site includes numerous links to information on the speaker, but a good place to start is this publication by the BBC from 1976 which describes design and construction of the speaker.
There have been numerous reviews of this LS3/5a, but I'll refer you to the most recent review of the Falcon Acoustics LS3/5a, in Stereophile magazine here, as it contains many of the components I used for this project.
For related projects, I've built Jeff Bagby's very well regarded Continuum speakers (a modern interpretation of the LS3/5a), here, as well as refurbished a pair of Fried B2 speakers, here, and The Fried Model H, here.
As indicated above, I purchased new cabinets from Stirling Audio, website here. Doug Stirling is very helpful and highly recommended. The original batch of LS3/5a's were built by the BBC, and they re-used cabinets from the predecessor LS3/5 (no a suffix), which were constructed of 3/8" baltic birch, versus the 1/2" used for most LS3/5a production, and those first LS3/5a's had screwed on back panels, whereas most production units have glued-in rear panels. Those early BBC built prototypes are revered among LS3/5a lovers, and the Stirling cabinets seek to replicate them with thin walls and screwed on back panels. I chose a rosewood finish, although Stirling has a number of finishes available, including cherry, walnut and zebrawood and ebony.
Note that since these cabinets are the same outside dimensions as all other LS3/5a's, the opening for the baffle and the grille are slightly larger that cabinets made with 1/2" baltic birch. So the baffle and grille frame need to be a bit larger than generally available or spacers added (thanks to Doug Stirling for the heads-up on this). For the grille, I simply bought pre-assembled grilles from Stirling with the BBC spec Tygan material. For the baffle, In simply added some wood spacers around a standard baffle, as you'll see below.
Front/top view of the cabinets, showing the battens to mount the front baffle, and the foam damping material.
Rear view of the cabinets, showing the screwed on backs. Note I had already instated the terminals at this point.
Side view of the Stirling cabinet, showing the match veneer from top to side. As a bit of an amateur woodworker myself, this is very nice professional veneer work. These cabinets are a little pricey, but are very much worth it.
Drivers, Crossover everything else
Falcon Acoustics, website here, (also higly recommended) has provided parts and help for LS3/5a owners for many years. For the past few years, they've brought the original KEF B110 woofer and T27 tweeters back into production and have been selling their own complete LS3/5a's as well as those drivers. They also sell virtually all of the parts needed to build your own, including pre-cut cabinet kits.
The Falcon T27, which looks identical to some original vintage KEF units I have on hand. Note that in the LS3/5a, there's a perforated metal dome added atop the tweeter.
Rear view of the T27 tweeter. The blue lead is positive.
The Falcon B110 woofer, made to be identical to the KEF SP1003 version of this famous mid-woofer.
Rear view of the woofer. The only difference I noted was the connectors, which are a typical modern version, as the original vintage KEF connectors were quite fragile.
The LS3/5a crossover from Falcon. It is designed to match the BBC FL6/23 spec as originally designed. Note that if you are building speakers with vintage KEF drivers, Falcon can customize the values of some components to match the spec for the date your drivers were made, as the BBC changed values based on changes in production of the B110 and T27. I'll note here that in the tolerance of the Falcon capacitors was very close to the marked values.
Rear view of the crossover. No components can be mounted on the foil side as the BBC spec calls for a felt pad between the crossover board, which is mounted behind the tweeter, and the tweeter itself, to dampen vibrations.
The photo above shows all the parts to build a single LS3/5a (with the bonus of second crossover to show the back side). In addition to the cabinet, tweeter, woofer and crossover described above, there are several other key elements to the BBC spec. These include the front baffle and Tygan material covered grille (lower right), a perforated metal dome/grille to go over the tweeter (lower left), special U-channel rubber seal for the B110, a 1/2" thick felt diffraction square to around the tweeter (to prevent refections from the cabinet sides) and the afore-mentioned square felt pad to go between the tweeter and crossover PCB. In addition there are a couple dozen screws, nuts and bolts, conveniently sold as a package by Falcon.
One note: the speaker terminals shown in the photo above aren't the one's I eventually decided to use, as you saw above on the cabinets.
The front baffle, as sold by Falcon. Nicely finished with the velcro (to attach the grille) already applied.
One note: as received, the velcro on the baffle covers the mounting holes to screw the baffle to the cabinet. Before starting assembly, I laid the baffle face down and drilled thru the velcro into scrap wood to open those holes.
To adjust the baffle dimensions to fit the thin-walled Stirling cabinets, I simply sanded down the edges of the baffle (so the wood glue would stick), and cut some thin strips of baltic birch ply to glue around the perimeter, shown here before gluing.
The baffles, front and rear, after gluing on the thin space strips. I then painted the edges to match the rest of the baffles.
Crossover Analysis and Tweaking
Since I planned to do some measurements before assembling the speakers, I built a simple jig to hold the baffle. This allowed me to mount the drivers and attach the crossover, sitting behind the baffle, with clip-on leads to experiment with changes.
Most of the following response curves were made with this jig, meaning no cabinet, and hence inaccurate low frequency measurements. So unless I indicate otherwise, ignore anything below a few hundred Hz.
The measurements below were made with OmniMic, available here, a system for time-gated FFT measurements with a calibrated microphone.
This photo gives an idea of my test setup using the baffle support jig. The baffle is held in the jig and the measurement microphone is lined up on the tweeter axis. You can see my laptop off to the right which runs the OmniMic software was used to capture the frequency response graphs below.
As you can see, the crossover is not mounted to the rear of the baffle where it normally resides, and is simply connected to the drivers with clip leads. As I changed parts on the board, I used more clip lead leads to swap components in and out.
The first test I made was comparing the two baffles/crossover assemblies for stereo matching. Since there is no cabinet, you should ignore any measurements below several hundred hertz. Also, we'll work on optimizing the frequency response shortly. But as you can see, the match between the two driver sets/crossovers is virtually identical, a tribute to the build quality of the Falcon drivers and crossover.
This is the BBC schematic for the LS3/5a. I'll refer to components shown here as we do some measurements and a bit of tweaking of the crossover.
There are 2 areas where components can be adjusted based on varying/drifting driver parameters.
1. We'll first concentrate on the few hundred Hz to 1Khz frequency area where the B110 has a peak that is corrected with a notch filter highlighted on the schematic. You can see the peak in the frequency response graph below in blue.
2. After working on the notch filter, we'll examine the tweeter level (and associated crossover point changes) shown on the schematic. Notice that the value of C2 was marked AOT for Adjust Under Test.
The Notch Filter:
As the parameters of the KEF B110 drifted over the years, the BBC increased the value of C5 from 6.2uf to 8.3uf and eventually to 10uf. The value of R2 was also changed, from 33 ohms to 22 ohms. We'll look at the impact of those changes on my speakers.
Components R2, L2 and C5 form the notch filter shown above to reduce that 600Hz-1K peak . The resistor R2 reduces the overall level to the B110, while the inductor L2 shunts the low frequencies around the resistor (so they pass unchanged), while the capacitor C5 shunts the high frequencies around the resistor (so they also pass unchanged), effectively creating a notch or dip in the signal fed to the B110 which, ideally, matches the peak inversely, resulting in a flat response. Thanks to Jon Risch's website, here, for that explanation.
This shows the extensive equalization that the LS3/5a crossover applies to the B110 woofer. The blue line is the B110's natural response with no crossover. You can see the several db hump in the response from 600Hz to 1KHz. The red line shows the response with the crossover. As you can see, the notch filter (C5, L2 & R2 described above) quite effectively reduces that big hump in the 600-1K range.
I didn't make this particular measurement, but is virtually identical to measurements I made some time ago on a pair of 1977 vintage Rogers LS3/5a's. It shows the signal fed to the driver terminals, including the effects of the crossover itself and the equalization (dip) in the 600-1KHz range to compensate for the peak in the B110 response. This heavy equalization also accounts for the inefficiency of the LS3/5a (83 db or so).
As mentioned above, the BBC increased C5 to deal with the drifting response peak around 1KHz. I experimented and found that in my case, increasing C5 flattened the response around 1K, both reducing the 900Hz peak and reducing the dip above 1K. C5 at 9.5uf is shown here (in blue), but I ultimately found 11uf gave the smoothest response in this area.
Next up was the value of R3, which was changed by the BBC from 33 ohms down to 22 ohms, again due to drifting parameters of the B110 over the years. This shows that, at least in my case, decreasing R2 to 22 ohms (in blue) actually increases the 1K peak. So my R2 stayed at 33 ohms.
Response with my final notch filter values (R3 unchanged at 33 ohms and C5 increased to 11uf) shown in blue compared to response with the default component values shown in red.
Next we'll look at tweeter level and impact of the metal tweeter grille.
Tweeter Level and Crossover Frequency:
One fairly unique aspect of the 15 ohm LS3/5a crossover is the tapped inductor, L3, which serves both to set crossover frequency (along with C2) and adjust tweeter level. I've read that each tap up or down on L3 is 1db of tweeter level. So I measured at taps 2, 4 and 6 (and made the required changes to C2 to keep the crossover frequency the same) and as you can see, my measurements made 2 steps apart are pretty close to 2 db of tweeter level each.
This graph show the effect of the perforated metal dome over the tweeter. The tweeter with no grille cover is in red, and you'll notice that the addition of the grille, shown in blue, increases the output from the tweeter. Kinda counterintuitive to me, but adding the grille actually flattens the response relative to the mid frequencies.
In addition to the tweeter grille, the LS3/5a has a square of thick felt around the tweeter, to avoid diffraction reflections from the sides of the cabinet. This shows that the addition of the felt (blue response) adds a bit of smoothing to the response with the tweeter grille only (shown in red).
Note that this measurement was made without the cabinet, whose sides extend beyond the baffle, so the response would look much worse with the cabinet and no felt.
After the above crossover changes, this is the response of the baffle/crossover assembly, along with the responses of the individual drivers. Again, this is the baffle only, so ignore the frequencies below a few hundred Hz),
Crossover Parts "Upgrades"
In addition to the tweaking of values to optimize the frequency response of the crossover discussed above, I also made some parts changes that some readers may find controversial (or just plain unnecessary). First, I am not a fan of the sand cast resistors used on the original, so I substituted 12 watt Mills 1% wirewound resistors for R1, R2 and R4. The latter was originally 5 watt, but the LS35a.com webpage indicates this R4 is prone to overheating, so I increased it to the 12 watt. The remaining resistor, R3, was replaced with a 5 watt Mills 1% wirewound. The Mills resistors may not sound better, but I believe they are more reliable. In addition, the original caps used on the crossovers appear to be mylar/polyester, and although they were VERY tight tolerances to their marked values, I replaced all caps in series with the signal (C5, C2, and C4) with Soniccap polypropylene caps, matched channel to channel within 1%. I've used these in many speaker projects and find them a very good value for the money, and I do believe they sound better than polyester caps. The new caps are larger than the originals, but they fit on the board with little problem. The board, by the way, is very well made and easy to work on, given a high wattage iron.
The stock FL6/23 crossover as received from Falcon.
The crossover after my modifications with the Mills resistors and Sonicap capacitors. You can see my markings of the exact values of the new caps, which along with the smaller bypass caps was used to get <1% left to right matching. I added a bit of silicone glue beneath the larger caps to prevent vibration.
Assembly is actually pretty simple. The drivers are first mounted to the baffle (which I had done earlier to test & tweak the crossover). Then the felt damping pad is added behind the tweeter, and the crossover mounted and wired up. Thats it!
The tweeter mounts very simply in the baffle atop a foam seal that's included with the T27. The woofer requires a special plastic channel strip that fits on the edge of the B110, and when the driver is bolted to the back side of the baffle, it forms an airtight seal. We are building an acoustic suspension speaker, so it needs to be airtight.
As you probably know, the woofer of the LS3/5a mounts from the back of the baffle, so the seal is at the front of the frame, not the back, as with the T27 tweeter or any front mounted driver (which is almost all).
This photo shows the plastic seal (from Falcon) in place on the B110. Note that during assembly I applied a bit of silicone glue at the joint (visible at the bottom center in this photo) to ensure no air leakage there.
Here's the baffle with both drivers mounted. Note that the felt square around the tweeter is applied over two of the bolts that hold the crossover, so that and the tweeter grille will be applied later, after the crossover is mounted.
At this point, the baffle was mounted in my stand, and here you can see the rear of the drivers. The T27 tweeter comes pre-wired, but the B110 needs 5 or 6 inches of wire attached to connect to the circuit board. Mounting the woofer in the baffle with the connectors facing the right (from the rear, as shown here) allows the shortest run to the crossover board.
You can also see the bolts and standoffs surrounding the tweeter that will be used to mount the crossover.
But before we mount the crossover PCB, the BBC spec calls for a felt square (again, from Falcon), about 1/2" thick, between the back of the tweeter and the PCB, to dampen any vibration in the crossover board.
This shows the baffle with the crossover board in place and wired up, except for the connections to the input terminals on the rear of the cabinet.
This is the front of the baffle with the felt diffraction square (self adhesive) attached and the metal tweeter grille attached. For the latter, I've seen recommendations for fairly permanent adhesives, but I choose to use silicone glue, which is removable if I ever have to replace the tweeter.
They're starting to look like every other LS3/5a ever made...
Back to the cabinets to add the input terminals. I decided to use Superior BP30-2BR gold bind posts, here. They mount on 3/4" centers, which is a bit smaller than the 30mm standard spacing that the Stirling cabinets are drilled for. I drilled and countersunk new holes, and sealed up the original holes with silicone glue to ensure a good seal. The base of the Superior posts just barely covers the old holes.
This shows the complete speaker, with the final wiring from the rear input terminals to the PCB. For that and for the B110 wiring, I used Kimber TCSS with teflon insulation.
The Finished Speakers:
The assembled speaker pair. Except for the slightly thinner cabinet walls, looks like every other pair of LS3/5a's.
With the grilles in place.
One final measurement, comparing my new build LS3/5a (blue) with a 1978 vintage Rogers LS3/5a (yellow-orange). These measurements are made in the near-field for the woofer and in the far-field (at 50 inches) for frequencies above 250Hz (so the low frequency data IS accurate here). The graphs are remarkably similar, especially the low frequencies. Also note the efficiency of the two speakers (made 40+ years apart) is almost identical.
I did another measurement of my 2 new build speakers, but the left and right are virtually identical to each other.
Another shot of the front and back of the completed pair.