I’ve spent hours searching for a commercially-made router bit that has the right dimensions to cut the dovetail slots in a traditional reed pan. It needs to be an unusually small diameter, but if you want to be able to cut the top slots after installing the chamber side walls as it was done originally (some of them undercut the walls), it needs to have a disproportionately long ‘neck’ between the cutter and the shank. On the plus side, the slot is quite shallow so the neck doesn’t need to be ridiculously skinny. In the end I decided to make my own.
I started with a piece of 1/4″ silver steel. After putting it in a collet and facing the end, I used the side of a threading tool to turn the tapered section, being careful to produce a sharp corner without significantly reducing the diameter of the base of the cone. I made it just long enough to be able to cut a 2mm deep slot, to avoid weakening the neck section unnecessarily. I set the tool holder over to produce the desired 60° taper:
Next I extended some more stock from the collet and turned the ‘neck’. On my first attempt, swarf obscured my view of the work and I accidentally retracted the carriage too far to the right and put a groove in the cone area. There was no option but to start again! The second time, I used the tailstock as a right-hand carriage stop to protect the cone.
The trickiest part of making your own router bit is producing the flutes without a special tool cutter/grinder machine. I cut three helical flutes by hand with a very small triangular saw file, then hardened and tempered it and sharpened the edges with diamond needle files:
Unfortunately it didn’t work well at all. It splintered the surface badly, then overheated:
Back to the drawing board. I studied a lot of photos of commercial dovetail router bits on Google Images and came up with a very different two-flute shape. Here’s a quick clip of me filing the relief angles on the second router bit with my saw file (click to stop it after you’ve seen it once, because the Instagram player auto-repeats):
A video posted by Alex Holden (@alexholdenmaker) on
And this is the finished bit, after heat treatment and sharpening. The thing it’s inserted into is one of my milling machine’s quick change tool holders:
This photo shows the reason why it needs a long neck (I made it a bit longer than would have been necessary for this Treble English, in case I want to make an instrument with deeper bass reed chambers at some point):
The second router bit works pretty well. Here’s a clip of it cutting a reed slot in a piece of scrap pine:
Programming the CNC mill to machine the slots is surprisingly complicated. The CAM software I’m using doesn’t understand how to cut a pocket with a tool that can’t plunge straight the workpiece and needs to enter and leave the edge of the material. I found a way to trick it into doing what I need, but the entire process filled nearly two pages of my logbook, and I need to do it all again for every size of frame I need to cut!
Another problem I ran into is that the outer dimensions of the antique Lachenal reeds I’ve been copying are a bit variable. Not by much, but a tenth of a mm change in width makes the difference between a snug fit and a loose one. This one fits very well – I can throw the block of wood in the air and catch it and the reed is still nicely seated – but the reed taken from the slot next to it (nominally the same frame size) is loose enough that it would fall out. I think when the instrument was built, somebody must have spent a while individually fitting each reed to its slot. Luckily my CNC mill (which I’m using for both the frames and pans) is repeatable to tight enough tolerances that I shouldn’t have this problem.
Quite a while ago I wrote a post about making nut plates for the doorbell project. Since I need twelve of them for a real instrument, and they will need to be accurately inset into the bellows frames, it made sense to program my CNC mill to produce them.
They are made from 2.5mm thick brass sheet. They have an M3 tapped hole in the centre, and two countersunk holes for fixing them to the frames using No 4 x 3/4″ brass screws. This batch of twelve took about 45 minutes to machine:
Of course, after breaking them out of the plate, they still required a bit of manual cleanup and deburring:
I tapped them all by hand using my ultra-high-tech tapping jig.
Here is the finished batch of nut plates, plus one of them set into a piece of scrap plywood the same thickness as the wood I’m going to use for the bellows frames. They have rounded ends so that I can cut the inset using a router bit in the CNC mill without having to manually square up the corners with a chisel afterwards.
I made a batch of buttons for my first prototype instrument. For simplicity I decided to use solid black acetal (an engineering plastic, commonly called Delrin, though that is a trademark of DuPont) rather than metal. Acetal is used by most modern concertina makers and it has a number of useful properties; particularly ease of machining, low mass, low friction, and low thermal conductivity (i.e. they don’t feel cold to the touch). I believe the top quality instruments still tend to use hollow metal buttons though.
The acetal came through the post in 1m lengths protected by a plastic tube. Long lengths of it are quite bendy. I started with 6mm and turned it down to 4.8mm. Before putting it in the lathe I cut it into 250mm lengths, which was about as long as I dared (shorter would result in more wastage, any longer risks the unsupported left hand end whipping around dangerously). I got nine buttons from each length.
I did most of the work on my manual Taig micro-lathe. I did a few things differently than usual in order to increase efficiency. For instance I set up both a standard right hand tool in the front toolpost and a parting off tool in the back toolpost so I wouldn’t have to mess about changing tools twice per button.
I made a couple of simple length gauges to control how much of the stock was protruding from the chuck at each stage, then turned up to the Z axis stop (set up to allow the carriage to almost touch the chuck). The short gauge is for the peg on the bottom of the button, and the long gauge is for the main body of the button. I also made full use of the graduations on the cross slide handwheel to produce the two diameters without stopping to measure the part.
I made a special jig to hold the button while I drilled and countersunk the cross hole on both sides. It is built in such a way that you can turn it over 180 degrees and locate it using the two pins on the baseboard, which is clamped to the drill press table.
Although this photo shows a standard jobber drill bit, I found it worked better to first use a smaller, more precise drill press to spot the hole location with a small centre drill, otherwise the bit drifts to one side or the other and you end up with an off-centre hole.
Finishing the top of the button involved facing off the parting-off stub, hand-sanding to round it off slightly, then flame polishing with a pencil torch to get a smooth glossy finish.
(Close-up picture of the polished button didn’t come out well – it turns out that my camera’s autofocus struggles to lock onto glossy black objects!).
This video shows the whole process:
Here’s a finished button:
And the full batch (more than I need for the first instrument – I made extra because I wasn’t sure how many I would ruin in the process, and I can always use the extras for my second instrument):
After completing the buttons, I now had prototypes of all the parts of a concertina action, so I decided to put it all together in a little test piece:
As well as the crude box itself, I made the pad, samper, grommet, lever, post, spring, felt washers, button, and both bushes. It is currently sitting on my desk as an executive toy, and I find myself reaching out and pressing the button whenever I’m thinking about a problem!
Update: After a couple of days of pressing the button whenever I happen to be at my desk, it definitely operates smoother and easier than when I first assembled it. I think the pad may be sealing more tightly too.
I made a prototype action lever. It’s a Wheatstone-style riveted lever hand-cut from 1mm thick brass sheet (the post is 1.5mm; possibly a bit thicker than necessary, but I didn’t want it to distort when I hammered it in).
The hardest part was making a die tool to thread the pad end so that I could screw the leather grommet onto it. Because the lever is cut from thin flat sheet rather than round bar, an ordinary thread cutting die wouldn’t have worked, so I instead made a sprung die set to form the thread.
I started with a 15mm x 25mm x 100mm bar of O1 tool steel, drilled and filed a spring shape on one end, then slit it in half:
Next I clamped it tightly together in a vice, and drilled and tapped an M2 hole in the middle of the slit, near the opposite end to the spring:
I put a couple of M5 threaded holes in the bottom so I could bolt it to a chunk of angle iron, then hardened and tempered it to 200C, differentially tempering the spring end to a higher temperature with a blowtorch so it won’t break in use:
After a bit of experimentation, I found that I could get it to form an acceptable thread if I cut a section of the 1mm sheet to 2.5mm wide (this dimension is fairly critical: 2mm forms almost no threads, and 3mm distorts and creases badly). It works best to hammer the tool fairly hard four times: once with the lever vertical, once each at 30 degrees from vertical in both directions, then a final time with the lever vertical again.
The lever after sawing it out with a jeweller’s saw, forming the thread, and riveting it to the post:
The proportions were based on one of the shortest levers in a treble English; most of the levers will have longer straight sections. The straight section is 2mm wide; I had to make the threaded part a bit wider (the tool squishes it narrower and thicker):
After screwing the grommet on. It is necessary to enlarge the hole in the leather grommet to 1.65mm before it will screw on without using excessive force and damaging the grommet:
I made a simple machine for winding concertina springs, inspired by Bob Tedrow‘s video.
It has a drum with a mandrel sized for the desired coil diameter and a hook on the outside, driven by a crank handle. The small step at the base of the mandrel helps to get the first turn of the coil tight. The adjustable guide plate isn’t strictly essential, but it helps a bit with consistency.
The raw spring material; 22 S.W.G. (about 0.7mm) phosphor bronze spring wire. It bends easily, is fairly corrosion resistant, and I’m told it lasts a lot longer than brass. At some point I’m planning to experiment with stainless spring steel and other diameters, but I’m sure the phosphor bronze is going to work fine for my initial prototype instrument.
Step 1; use needle nose pliers to bend a right-angle that will form the ‘pin’ that you push into the action board:
Step 2; insert the wire into the machine as shown. It’s important that the hooked end is parallel to the face of the drum:
Steps 3 and 4; turn the crank handle clockwise about 2 1/4 times, then cut the wire off, using the guide plate to gauge where to cut.
Step 5; use small round nose pliers to form the hook:
Step 6; use needle nose pliers to bend the hook over at a right angle:
The finished spring:
Here’s a quick video of the process:
Sometimes it’s necessary to use an opposite-hand spring because of limited space on the action board. You make these in the same way but doing all the bends the other way and turning the crank handle anticlockwise:
Concertina pads are small discs that cover holes in the action board; when you press a button, it causes a pad to lift off its hole, which allows air to pass through a reed and produce a note. They are made from a sandwich of leather, felt and card. The leather forms an airtight seal against the hole, the card provides a rigid backbone and a surface for the action lever to attach to, and the felt acts as a buffer between the two that stops the pad making an audible slapping sound when it closes quickly.
It took quite a few experiments to find a combination of materials, glue, and procedure that produces satisfactory pads. Along the way I made quite a few pads that fell apart, were too hard or too spongy, and/or were too thick or too thin.
A pad ‘sandwich’ after gluing:
I eventually settled on hide glue with some urea added to extend the open time a bit. I soaked apart an antique Lachenal pad and I’m 99% sure it was glued with hide glue. PVA would probably work too, but when I tried it, it stuck well but it seemed to soak into the felt and make it harder. I know others have used sprayable contact adhesive successfully, but it barely stuck at all for me. There’s a bit of a knack to applying just the right amount of glue, and it’s important to brush it onto the card/leather, not the felt, otherwise it will soak up far too much glue and go hard when it eventually dries. Clamp the sandwich as lightly as possible and take it out of the clamp after an hour to avoid permanently compressing the felt. Leave it at least a few hours to dry before punching the pads out.
The leather is thin smooth sheepskin skiver, with the hair side out. The card is 1mm greyboard (I also tried millboard, but it turned out to be made of two layers that delaminated when I punched the pads out). I tried five different wool felts before settling on this one, which the supplier describes as 1.5mm 25 S.G., though it starts out significantly thicker than that and compresses down a bit when you glue it.
I’m punching the pads out using Priory wad punches (carefully resharpened), a lead mallet, and an anvil made from the smoothed end grain of a beech log soaked in boiled linseed oil.
It works best to punch with the leather side up, otherwise the card distorts and doesn’t cut cleanly.
It’s important to keep hammering until you’ve cut through the card all the way around.
A new pad next to a ‘retired’ antique Lachenal one; the new one is a bit thicker and softer, but I think it will quickly compress down to about the same thickness.
I just finished fixing up a Bastari G/D Anglo for a friend who plays for Anonymous Morris. The aim was mainly to repair a few faults and tune it up, to make it more playable rather than carry out a complete restoration.
The first problem was that the action boxes had split apart in several of the corners, so I cleaned up the joints and glued them back together, adding reinforcing blocks to strengthen them. This was a bit of a delicate job because they were originally constructed with PVA glue and I couldn’t remove any significant amount of material from the joints when cleaning off the remains of the old glue or they would have got smaller and would no longer fit the rest of the instrument.
The chrome-plated brass end plates were a bit grubby so I gave them a quick polish:
This instrument’s Achilles heel is the aluminium pivots where the buttons are attached to the action levers. Note that these are different from the rubber tube type commonly found on Stagi instruments. Most of them were rather wonky, and the most-used buttons were sloppy due to wear; a few were almost worn through:
After discussing the problem with the client, I agreed to make replacements for the most-worn levers. I cut the new levers from 1mm brass sheet with riveted pivot points, so they are unlikely to wear out again. After designing the two sizes of lever and the button insert in CAD, I printed a template on sticky paper and cut them out by hand with a jeweller’s saw.
I re-used the top parts of the buttons, which were made from chrome-plated brass with the aluminium pivot glued into a hole in the bottom. The first step was to break off most of the pivot with a pair of needle nose pliers:
Then I put the button in a drill chuck in the lathe and drilled into it with a centre drill:
Then a 3mm drill (this was a tiny bit bigger than the original hole, so it left a nice clean surface on the inside of the hole):
At a certain point, the drill stopped cutting; this meant that the tip of the old pivot had come loose and was stuck on the end of the drill bit. After withdrawing the drill and removing the loose piece I was able to finish cleaning up the rest of the hole:
I sawed the new pivot pieces slightly wide, then carefully filed them down until they were a snug fit in the hole:
I glued the new pivot pieces in with Araldite Rapid Steel (an epoxy resin that is specifically formulated for gluing metal):
I found it easier to work on the action if I removed all the levers apart from the one I was working on at the time. I glued the new levers to the original pads using hot melt glue (this seemed to be how it was done originally), fitted the spring, put the end plate on, marked the position of the hole on the lever, took it apart, drilled the rivet hole in the lever, cut it shorter, riveted the button on, put it back together, and bent the lever until the button was directly below the hole.
Here’s a quick video clip of me riveting a button to a lever:
And here’s the resulting pivot. I actually made a ‘snap’ tool from hardened silver-steel to form domed rivet heads, however I found that it inevitably made the joint stiff if I hammered it enough to take out all the play. By using lots of light taps with a small ball peen hammer instead, I was able to make pivots that work and feel just right. Note that the mushroomed end of the rivet doesn’t turn; it expands enough so that it is a tight fit in the lever, but there is just enough play in the joint for the pivot to turn smoothly without any noticeable wobble.
This video shows the difference between a sloppy worn-out pivot and one of my improved brass riveted ones:
As well as replacing ten of the levers, I adjusted the remaining 21 as best I could, straightening and tightening them up as much as possible. Rather a fiddly, painstaking task, and it’s impossible to get them perfect without replacing them all.
The bellows had quite a few worn corners, some of which were leaking air, so I glued thin patches on them. I tried to dye the new leather to match the old, but it didn’t work very well: I managed to get the leather slightly darker, but it seemed to quickly reach a point where it didn’t want to absorb any more of the dye.
Tuning the instrument proved much more difficult and time-consuming than I expected. Some of the reeds were no trouble, but many of them behaved illogically, randomly going flat and muffled, then suddenly going sharper again when I fiddled with them or just after playing them for a while. In hindsight, stiff/sticky valves were mostly to blame for this. Some of the reeds were rather dirty; this one went five cents sharper when I wiped the sticky black dust off it:
There was one reed that nearly had me pulling my hair out: it kept going flat by six cents whenever I tightened the instrument’s end bolts down. After trying many different things, I eventually worked out that there wasn’t quite enough clearance between the reed tongue and frame on one side. Somehow, tightening the end bolts down was bending the sound board and applying a force to the reed frame that distorted it just enough to cause the tongue to slightly graze the vent side, which made it sound flat and slightly buzzy.
The finished instrument ready to go back to work, playing traditional English dance music:
Recently I was asked to try to come up with a concertina reed that works in both directions, or at least to figure out why it hasn’t been done before. Ordinarily, a free reed only speaks when you suck air down past the tongue, into and through the vent in the frame. Most concertinas have a pair of reeds controlled by each button; one inside the reed chamber that only sounds on the pull stroke and another mounted on the underside of the reed pan that only sounds on the push stroke. Anglo instruments take advantage of this to play different notes on pull and push (this is known as bisonoricity), whereas English and Duet instruments play the same note in both directions so they need a pair of identical reeds for every button. If it was possible to make a bidirectional reed that worked as well as two standard reeds, it could potentially enable unisonoric instruments to be made smaller, lighter, and cheaper.
The way I went about solving the problem was to first build what was pretty much a standard reed in an oversized frame and check that it sounded normally in the suck direction, then screwed on a roughly horseshoe-shaped plate that fit around the tongue. I didn’t expect this to work because there was no way for a significant amount of air to get past the tongue to start the oscillation cycle, and indeed it didn’t.
I also modified my bellows bench a little to allow me to block up the standard dovetail socket and screw the new oversized frame to it elsewhere, and provided a means to block the one-way valve that normally releases air when I raise the bellows so that the rig only works in the suck direction.
Next I took the horseshoe back off and started experimenting with filing away various parts of the bottom of the horseshoe vent around the tongue, to provide some space for air to get to and past the tongue and allow it to start. Eventually I got it to sound, albeit poorly, in the suck direction, and it even made a tiny bit of sound in the push direction.
I had a theory that the triangular profile resulting from filing the underside of the horseshoe vent was causing the airflow to be cut off too gradually in the blow direction, so I next made a new horseshoe piece, this time with a square-sided recess milled into the underside so there was air space all around the tongue. This was supposed to cut the flow off more cleanly when the tongue swung up into the horseshoe vent.
This did work a little better, however it was very inefficient, very quiet, and worked much better in the suck direction than the blow direction. I figured that the reason it worked unequally was because the reed tongue was profiled only on the top surface, so when it passed into the bottom vent it cut the airflow cleanly and suddenly whereas when it passed into the top vent it cut it progressively from the tip towards the root.
In order to try to solve this asymmetry, I built a second, more complicated, reed. On this one the reed tongue is set into the bottom frame by half the thickness of the reed stock, it is profiled equally on top and bottom of the tongue, and I also restricted the air pocket to the last third of the tongue, which I tried to profile fairly flat so that it cuts the airflow fairly cleanly in both directions.
The second reed was the most successful prototype I built, however it revealed the biggest flaw with the idea. When set up carefully it works pretty equally in both directions, however the amplitude is very limited compared to a standard reed:
I believe I now understand the reason for this, however it is a little tricky to explain. Before starting my experiments I had observed that with a standard reed playing at normal volume, the tongue swings well above and below the restriction point at the entrance to the vent. I imagined that with the bidirectional reed, it would swing past both restriction points and generate a similar amplitude level, perhaps with a different tone. This was based on a couple of misunderstandings about how reeds work.
My current understanding of what happens with a standard reed when it first starts up is that the tongue gets drawn down towards the frame (it needs to be set such that at rest there is a slight gap between the tongue and frame so air can start flowing). I don’t fully understand the physics behind why this happens, but it seems to me that the faster the airflow into the vent, the harder the tongue gets pulled down. The tongue descending towards the vent opening restricts the airflow into the vent, the force pulling the tongue down reduces, and the tongue springs back up, eventually peaking slightly higher than its rest position. Because it is higher, the gap between the tongue and the frame is larger and more air is able to flow through it than on the first cycle, so it gets drawn down a bit further, and springs back a bit higher than before. Over the course of a number of cycles, the amplitude builds up and up until the tongue is swinging a long way below the top of the vent. In order for this build-up to work, it’s important that every time the tongue swings a bit higher, it results in more air flowing through the vent, which causes the tongue to be pulled down harder and the amplitude of the oscillation to increase. Eventually the oscillation reaches an equilibrium level that depends on the pressure differential between the top and bottom of the reed frame. If you squeeze the bellows harder, the tongue oscillates to a greater height and the ‘packets’ of air being chopped up by the tongue passing through the frame are larger and more energetic, which results in a greater volume of sound from the instrument.
What goes wrong with my bidirectional reed that prevents it developing a decent amplitude at normal bellows pressure is that the second vent restriction, the one ‘above’ the tongue (whichever direction that happens to be), cuts off the air supply whenever the tongue tries to swing higher than the opening into the second frame. It’s impossible for the amplitude of the oscillation to ever build up any higher than the second frame, because it restricts the air supply as the tongue swings higher instead of allowing more air through. It is a lot like the governor on an engine, which throttles the fuel supply whenever it tries to exceed a certain speed.
It is possible to increase the amplitude at which the limiting occurs by increasing the distance between the two vent openings inside the reed, however there is a limit to how far you can take this. If the distance is too wide, the reed oscillations take several seconds to build up to an audible level, or never start up at all. It also becomes impossible to deliberately play the reed very quietly: you end up with a reed that has essentially no dynamic range.
To make matters worse, as well as the limited volume issue, there are several other disadvantages to this type of reed:
They seem to be less efficient (i.e. they use more air than a standard reed operating at a similarly low amplitude), possibly because the way it is constructed to allow it to perform equally in both directions has the side effect of not cutting the airflow very cleanly in either direction.
They are considerably more difficult to make than a unidirectional reed, probably something like 75% of the work of making a pair of standard reeds. A lot of the extra work has to do with making both frames a tight fit around the tongue without catching on the sides. Because nearly all of the cost of a hand-made reed like this is in labour time, it wouldn’t be a large cost saving to make an instrument with half the number of bidirectional reeds.
They are significantly bigger and heavier than a standard reed because of the need to be able to screw the two parts together; you would save a little compared to a pair of standard reeds but not as much as you might think.
There are a bunch of issues around the fact that what you would call the ‘set’ on a standard reed is fixed at manufacture-time by the relationship between the height of the recess and the thickness of the tip of the tongue. You can’t easily alter it deliberately, and it is possible to alter it accidentally as a side-effect of tuning the reed. It’s also important for the tongue to be set precisely central between the two frames, otherwise it starts poorly or not at all in one direction or the other.
I don’t know for sure, but I suspect this design would be more susceptible than a standard reed to getting dust and fluff caught inside it and impeding its operation, because the air gets forced through a narrow recess inside the reed.
It goes without saying that this type of reed is useless for an Anglo instrument because it produces the same note in both directions.
Following up on a slightly different line of inquiry, I made two final experimental reeds, one which only had a rectangular recess right at the tip, and another which was very similar but with a triangular recess instead. Neither of these worked as well as the second reed, I suspect because they only have a tiny amount of space for air to squeeze past the tongue. They sound in both directions (just about), but are very inefficient and quiet.
Here is an audio recording I made of the four experimental reeds plus a standard reed for comparison. The first reed had the second horseshoe fitted.
Although this work didn’t lead to a usable product, it was still a useful exercise for me in that I believe I now have a significantly better understanding of how concertina reeds actually work.