- Improving the Performance of
a Classic Tannoy Driver -
This article describes my experiences with a pair of 15" Tannoy Monitor
Gold dual concentric speaker drivers. These were purchased at a garage sale
a few years ago in the cheapest and most basic original Tannoy box that
I have heard of. Now I have them in a much better and bigger box and have
made some other changes. The speakers started out sounding promising. Now
I think they sound extremely good. This article should have enough in it
that could be applied to other speakers too.
Among their advantages are that they are efficient enough to allow the use of low power class A amplification, in my case a home-built single-ended triode amplifier. High efficiency is a great virtue in speakers. Among other things it allows amplifier design to be freed of the compromises and/or expense that are required for high power and high gain. Also high efficiency speakers tend to sound rather more "effortless"; perhaps due to less heat being dissipated in their voice coils.
Unfortunately there is no free lunch. If you want high efficiency, you must either give up low bass or small, neat boxes; the physics says you can't have both. In my case I have bass down to 20Hz, high efficiency and boxes the size of domestic refrigerators, but don't think this means the speakers give only big sound. They can be detailed and delicate too.
The Original Boxes
The original Tannoy box was a Lancaster enclosure (54cm w x 31cm d x 84cm h) with a simple slot cut in the front panel. A friend bought a pair of 15" Monitor Golds in a nearly identical enclosure traded in at a local HiFi shop. In this case the port was sealed up, apparently at the Tannoy factory. These sound OK but with even less bass than my ported version. Both Lancaster versions sounded rather congested compared to the drivers in my new big boxes and with the various improvement described below.
One of the problems I had with the Lancaster box was that it is too short.
When I put it on some bricks to get the driver up to seated ear height the
sound opened up considerably. Otherwise it sounded a bit "sat on".
Not so easy to fix was the problem that the bass sounded rather tubby and
not very deep. The tubby sound was I think due to fairly large thin panels
without any bracing. The lack of deep bass was a problem due to the box
being simply too small, even though by modern domestic standards it was
big. Some of the congestion might have been due to the very heavy grill
cloth.
A Horn Enclosure?
I wrote to Tannoy asking them about horn enclosures for these drivers. They kindly sent me three enclosure plans. One of the horn plans (Professional or Rectangular GRF) didn't appeal because the driver was placed even lower than the Lancaster and I already knew that it was nice to get the drivers on axis at ear height. The other two were horn enclosures intended for corner placement (GRF and Autograph). The Autograph design in particular looked really nice. It has a short front horn and a folded rear horn. The currently produced Westminster enclosure seems to be a rectangular version of this box. It has approximately 600 litres total volume. The front horn loads the main cone from around 200Hz to 1KHz and the folded rear horn loads below 200Hz. The crossover allows the input of the tweeter (from 1KHz) to be stepped up to match the increased output below 1KHz. The rear horn splits part way through the box to exit next to the two side walls of the corner. Thus the corner extends the horn mouth into the room. Horn loading the bass cone should be a good thing in theory. It should reduce driver displacement for any particular output level and thereby reduce distortion. I expect the Autograph sounds very nice but have not heard it. Unfortunately placement hard into the corners wasn't going to suit my room and the wood-work looked very complex.
Back to a Simple Box
The horn designs were drawn in the 1950s so were intended for earlier versions of the 15" dual concentric but had been modified (only the driver mounting) for the 3838 and 3808 drivers from around 1980. This suggests to me that the parameters for the 15" dual concentrics have changed very little from model to model and published Thiele and Small parameters for later drivers could give an indication of what to expect from the earlier ones. [This seems supported by the measured HPD parameters] A friend with box modelling software plugged in the 3838 parameters to see what was suggested for a flat alignment. It came up with an impossibly small box, tuned impractically with a -3db point somewhere well above 100Hz. I suspect the program expects higher Q drivers when it tries to find an optimum. Be carefull that software doesn't tell you nonsense.
I wrote again to Tannoy and asked for suggestions
for the best simple box for my drivers. They suggested a 300 litre box with
four ports, each of 100mm diameter and 250mm length and this is what I have
built. Tannoy claimed that this box should achieve a -3db point at about
35Hz with the 15" Monitor Golds. This is a BIG box and I think it is
about 50% larger than any of the commercial reflex boxes produced by Tannoy.
300L is about 10 cu.ft. Tannoy did not publish Thiel and Small parameters
for the Monitor Gold for the simple reason that the drivers predate Thiel
and Small's work. The free air resonance however was published as the same
as the 3838 (20Hz), obviously the drivers are about the same area, and being
used in some of the same boxes could not be grossly different. As a reality
check my friend modelled the 3838 parameters in the 300 litre box. The response
was reminiscent of a sealed box alignment with Q of 0.5, a transient perfect
but not perfectly flat response. Specifically, the response was very gradually
and smoothly rolling off from well above 100Hz and was 6db down by 35Hz.
Perhaps you could call that the -3db point if you allow the mid-range to
be plus 3db or perhaps the 3838 is not such a close model for the Monitor
Gold. Whatever the theoretical response should be the recommended box gives
strong, deep, well-balanced bass in practice using corner loading but the
bass is a little weak if the speakers are out in the middle of the room.
Minimising Box Colourations
Although a rectangular box is simple compared with horns there are still some important considerations. A major issue is minimising box colouration. I made the panels from some scavenged office partition material consisting of a chipboard core faced with plywood on both sides. Overall the material is 50mm thick, thicker than most speaker box material. It had 25mm holes running vertically through the core, presumably to reduce weight, which I filled with sand. The holes were plugged with bits of old mattress foam and then sealed with automotive "bog". In addition to thick panels and sand-damping, I placed two shelves inside the boxes for bracing. These have cut-out areas to allow air movement but the remaining wood ties opposite panels to each other, braces along the face of each panel and triangulates across the corners. Consider that even slight panel vibration is distributed over a much larger area than the speaker cone and could contribute significant sound output. I doubt it would be possible to go to far trying to kill panel vibrations.
When making these plans I received the interesting comment that the old Tannoy drivers can produce a very modern sound with such construction but that some people have been disappointed if part of what they liked about the older Tannoys was the warmth of box colourations. Personally I doubt that anyone would prefer the tubby sound of the Lancasters over the improved clarity, dynamics and extension of my big boxes.
The dimensions of the box were made as narrow as possible side to side without boxing in the driver very closely, mainly for visual aesthetics. I then tried to avoid making the height or depth of the box simple multiples of this internal width. This was to avoid multiple internal standing waves that might reinforce each other. The box ended up being 57cm (w) x 63cm (d) x 134cm (h) external dimensions. Inside the box the bracing shelves are just a little above and below the driver. The intention was to put most bracing where the driver might try to move the panels most but also avoid any very large unbraced sections. The bracing was also placed so that the sections of panel between the bracing were of different lengths to spread the frequencies of any residual resonances.
Damping material for the space inside the box seems to be a bit of a black art. In a vented box its function is to absorb higher frequency radiation from the rear of the cone that might otherwise bounce around and excape through the cone or vents. Also it should damp lower frequencies that might set up standing waves within the box. I applied a little reading and a little thought to this area. I am happy enough with the initial results that I havn't tried any experiments in this area. Ideas I used were 1) Wool is supposedly a better damping material than polyester, glass fibre etc. 2) Air particle motion is always at a minimum at the box wall, therefore damping material placed on the wall has least effect. 3) A clear though not straight path must be available between the driver and the vents and the vents should be well clear of damping material and the box walls.
I was able to buy batts of pure wool intended for ceiling insulation. Two of these went in the box supported by the shelf above the driver and so the top quarter of the box is full of loose wool. This should make it hard to sustain standing waves along the long (top to bottom) axis of the box. Another batt was stapled to hang loosely around three sides of the middle section of the box behind the driver. A fourth batt is in a U shape in the bottom section of the box. The four ports are on the front baffle in the lower section of the box. Immediately behind the port opening is a large acoustic foam tile, the sort with wedge patterns on the surface. Measurements of the port output indicated virtually no higher frequencies escaping the port. More subjectively, when I had my head and shoulders through the driver hole, hammering the staples holding the wool in place the metal on metal sounded like dull thuds.
Moving the drivers into these well damped and braced larger boxes gave extra octaves of bass and a big improvement in the tightness and clarity of the bass and mid-range.
Moving the Beasts
With thick, sand filled walls the boxes ended up
rather heavy! Two adult males struggle to carry one of these speakers. To
move them within the room I wanted some carpet sliders but none of the devices
sold for this purpose looked like they would take the weight. Then I noticed
the smooth, hard, rounded plastic buttons that are glued to the surface
of roads to mark lanes. Four of these glued to the bottom corners of easch
box makes them fairly easy to push about over carpet.
Speaker Placement
As mentioned already just putting the Lancaster
boxes on blocks so the driver was at ear height was a worthwhile improvement
for bigger, more open sound. In the big boxes the driver center is at my
seated ear height. This does introduce a potential problem.
Roy Allison has pointed out that a dip in the response of a speaker placed
at ear height can occur a bit over 100Hz as this is the distance to the
floor and back for half a wave length. Thus the reflection from the floor
comes back out of phase, unloads the driver and causes a 3db dip in efficiency.
Not thinking of any of this I had inadvertently placed the speakers in the
room such that the distance from the driver to the wall behind was about
the same as the distance to the floor. This would result in a predicted
6db upper bass dip. This led to a subjective effect that was initially surprising
for me till I worked out what I think was going on. With a bass instrument
only playing up and down through its range the sound was fine and no response
dip was apparent. If some mid-range instruments joined in the bass would
keep appearing and dissapearing with no obvious pattern. I think I learned
something about masking and the audibility of dips. 
So how could I avoid the dip with the height of the driver fixed? I used BestPlace software (Mac version) which was free from RDL Acoustics. It models just the loading/unloading effect of the three nearest surfaces; it is not a whole room acoustics model. By trial and error I found a position for the boxes toward corners that was predicted to give a gradually rising bass response without any serious dips, roughly equal and opposite to the gradually rolling off response predicted above (IE about +6db of boost at 35Hz). The position has the side wall of the corner closer to the driver than the distance to the floor and the distance to the wall behind is further away. Ultimately a corner gives +9db at the lowest frequency. Combining the modelled frequency response with the predicted boost gave an expected resultant -3db point around 30hz. The measured result was better than expected, flat to 20hz. I guess the wavelengths are long enough at 20Hz to get additional boosting from further surfaces, or perhaps the speakers are just better than expected.
Subjective impressions are consistent with all
this modelling; out of the corners the bass sounds a bit thin. In the right
position towards the corners the speakers sound well balanced with very
deep but not boomy bass. Bass instruments can be followed even with plenty
of other instruments playing but are not obtrusive. The speakers are along
the shorter wall of my lounge room so facing the speakers directly towards
the couch enables the drivers to be sufficiently close to the side walls.
The corner placement is also fortuitously a good way to not take up to much
floor-space. I point all this out because I started out with the received
wisdom that speakers need to be away from walls to "breath". Make
sure you don't put speakers equidistant to the surfaces of a corner. That
will cause a 9db hole followed by a 9db boost lower down.
Getting Rid of Dry Contacts
These speakers have quite a few non-soldered contacts
that might have corroded with time. Minimising the number of dry contacts
gave a subjective improvement in both treble quality and quantity, mostly
the former. The crossover has two switches, in series with the tweeter.
One switch ("energy") selects taps on a choke to step the tweeter
signal up or down. The middle position was initially selected for the best
tonal balance and hardwired. The other switch ("rolloff") gives
a flat position of various amounts of rolloff. I hardwired the flat position.
Another set of dry contacts is a plug and socket arrangement on the chassis
of the driver. The socket can be reversed to act as a convenient set of
soldering lugs, like a little tag strip, to solder the speaker connections.
Opening up the Crossover Box
The crossover was sealed up in a plastic box. I opened it by drilling out the four rivets in the corners. Two screws hold a board in the larger part of the box. These were undone to allow the electrical equivalents of the switch positions described to be hardwired directly on the board. Initially I retained the original wire except the lead to the switch box. Later I replaced the wire with van den Hul wire and can't claim to have heard any improvement. Hardwiring seemed to make far more difference. When I rewired I moved a few wires around so that I could later biamp or biwire if I so chose. That is not too hard to work out.
I replaced all the crossover caps with equivalent
value MIT or Solen polypropylenes. This was not too expensive because the
largest value required is 16µF. The 16µF was made by paralleling
a couple of smaller values. This also was a subjective improvement in treble
quality and quantity. The larger caps don't allow the lid of the crossover
box to be put back on. I put some caps on the inside and hung others off
the underside of the board, then screwed the board back into the plastic
box without the lid. This seems quite robust. The whole thing now lives
loose in the bottom of the speaker box on a short set of wire soldered to
a removable binding post plate on the back of the box. Note that the plastic
crossover box is important; it helps to hold the chokes together. Otherwise
they would rely on just a bit of tape.
Surgery on the Dustcap
Removing the dustcap in front of the tweeter gives
a worthwhile improvement in treble quality, and a little more quantity.
I think the earliest and latest versions of the dual concentric do not have
the dust cap and the 1980ish 3838 has a more transparent looking dustcap.
The idea was suggested to me by a guy I met in a TV repair shop in a little
town in Tasmania who was quite a Leak and Tannoy enthusiast. He insisted
that the glue holding on the dust cap was readily soluble in acetone. It
is soluble but not quite so easily as I was led to believe. The method I
was told was to paint acetone liberally along the glued edge of the dustcap
to soak the paper and glue, then slide a knife between the dustcap edge
and the cone to neatly lift the cap off with no damage to the cone or cap,
and therefore reversible. I tried this. It seemed to be working OK but then
I lost a little bit of cone paper when I hit a patch of less well dissolved
glue. In my experience the glue softened but didn't really let go completely.
Work very carefully at this! Acetone is very volatile so you have to keep
painting it on as you work. When it was done the improvement was considerable.
More open and airy top end. Also just plain more top end, perhaps a little
too hot.
Getting the Balance Back
Various changes had improved treble quality but
also seemed to give just a little more quantity too. Eventually the cumulative
effect was just a little too much top end. At that point measurements showed
the tweeter output to be broadly about 2db above the woofer output. To fix
this I went back to the crossover and moved the hardwired crossover to the
second lowest tap on the "energy" choke. The choke does two things.
Firstly it is part of a conventional LC filter. Also it acts as an autotransformer
with 5 taps. The second lowest is the driven tap connected to the input
via a capacitor. The nominally flat position that was originally prefered
is the middle tap. Moving down to the second lowest tap retored the correct
sounding balance and as a bonus the treble clarity seemed better still.
Perhaps this was from using the direct position on the choke, neither stepping
up or down.
Tidying Up
I worried that dust, insects, bits of metal filings
etc. might find their way into the magnet gap having removed the dustcap.
I put some double-sided sticky tape along the foam on the front edge of
the driver chassis and stretched some fine black speaker grill cloth over
and trimmed the edge. The cloth was further secured, and the ragged edge
hidden, by an open black-painted metal grid that is clamped to the front
of the chassis with the 8 bolts that retain the driver in the box. The metal
prevents excited kids from crashing into the driver. The cloth keeps out
dust etc and is more acoustically transparent than the original dust cap.
I think it probably helps being at the horn mouth rather than closer to
the horn throat. I noticed no loss of sound quality with the addition of
the cloth and it looks nicer too.
For while the boxes stayed rough looking but eventually I got around to
sanding back the plywood and glueing on some thin decorative strips of wood
over the exposed chipboard edges. The whole thing was given a coat of a
stain/varnish and I think they now look quite presentable in a solid, slightly
old-fashioned kind of way, which is only appropriate.
These speakers really rewarded this tweaking. Non-audio types are often
amazed that I can get the best sound they have heard from gear that is mostly
home-built valve electronics or 1960s vintage. A friend has Monitor Golds
still in Lancaster boxes and the sound is not nearly as good.
PS: What is Going on in that Cross-over?
A few people have suggested to me that the Tannoy
crossover is unduely complex and suggested simplifying it. I think the crossover
is actually just about right. Everything is there for a good reason. The
various level and roll-off controls provided on different Tannoy models
don't require any extra components other than the switches themselves; they
just take advantage of parts that have to be there anyway. So, how do they
work? Firstly there is an LC low pass filter for the bass section and a
corresponding CL filter for the high pass section. So far very conventional.
Altering that would require some carefull thought about the offset of the
tweeter to the rear and the phase-shifts associated with the filter. Five
tappings on the winding of the high pass inductor enables it to double as
a single winding transformer (autotransformer). The "energy" switch
selects the different taps allowing the signal to be stepped up or down
a few db. A feature for the price of a switch!
From here it gets a bit trickier. The rest of the crossover is there to
equalise the response of the tweeter which otherwise would be far from flat.
Some horns tweeters have a flat on axis response but that (like most speaker
drivers) is at the expense of more and more narrow dispersion with increasing
frequency. Instead the tweeter in the Tannoy is a constant dispersion horn.
In order to get constant dispersion a falling response with increasing frequency
is unavoidable. At the low end the efficiency is extremely high and still
not bad in the top octave. Thirty ohms in series with the tweeter give a
lot of attenuation for the low end and the high end is recovered with a
capacitor that progressively bypasses the resistance at higher frequencies.
If the cap were disconnected one would get a treble roll-off. If the 30
ohms were partly bypassed one would get a lesser roll-off and if the cap
were connected in parallel with the tweeter the 30 ohms and cap would form
and RC filter to give an even steeper roll-off. Thus an extra switch makes
a feature out of necessity and provides a selection of treble roll-off choices.
Finally, in parallel with the tweeter is a resistor, a capacitor and an inductor in series. These, I'm told, are there to pull down a response peak at about 2 Kohms. In the Monitor Gold crossover these are fixed but later versions of the crossover have turned these into a control feature also providing a switch that varies the value of the resistance. Thus the later crossover gives the possibility to reduce the effect to give a mid-range peak, or increase it to give a mid-range dip.
I can only think of one way to get rid of some
of the crossover bits (other than the switches) without causing more trouble
that it is worth. That would be to bi-amp the speakers and build the tweeter
equalisation into the amplifier. That might have some nice consequences.
The tweeter is so efficient without the equalisation that one could choose
a class A amplifier design with virtually no contraints. One could pick
the most linear triode capable of putting out a couple of watts and have
plenty of power.
Comparison with 3838 Sound
How do the older Monitor Gold drivers sound next to the more modern 3838 drivers? A friend brought some 3838s to my place in reflex boxes about 2/3 the size of mine. The first obvious difference was lower efficiency. Equal volume required an extra 2 clicks on my stepped volume controls which amounts to about 4db difference. Overall the Golds sounded more "effortless". "Effortless" is a common description of the sound of efficient speakers but another factor could have been my low power amp being 4db closer to clipping with the 3838s. There was a small region of the mid-range that seemed a little smoother on the 3838s, perhaps the effect of the struts on the rear on the cone preventing some cone breakup that the Monitor Golds might have. The older driver was prefered at the frequency extremes, deeper bass and more "airy" top end. Otherwise the family resemblance was clear and the sound was quite similar.
Acknowledgement
I would like to thank Peter
Campbell for his kind permission to reproduce this paper at this web
site.