What not to do

No. 1: Brun Part 3: End Boxes

Part 3 of the story of how I built my first instrument is going to cover the main wooden frames of the instrument ends; the bellows frames and the walls of the action boxes. I came up with what seemed like a clever plan on paper, but quite a few things went wrong along the way leading to a lot of fiddly corrective work and a couple of slight cosmetic issues in the finished instrument. The next instrument is definitely going to involve some significant changes in the way I build the end boxes.

I had read that good quality vintage instruments were commonly built from sycamore, typically with decorative veneers like ebony or rosewood on the outside, though for this instrument I decided to go with plain solid sycamore and try to use nicely figured pieces of it for the most visible parts of the walls. I struggled to find a timber supplier that stocked quarter sawn sycamore, though after much searching I found a place thirty miles from me that had a stack of roughly 2 ¾” thick slabs that had been plain sawn and kiln dried. I went and searched through them and picked out one taken from the middle of the tree, so it had two quarter-sawn (ish) sections either side of the pith.

I used my circular saw to cut off a section and remove the pith (the centre of the trunk, where the growth rings become very small). If you click on the next picture to zoom in and look at the growth rings, you should be able to see what I mean. The areas where the rings meet the surface at a right angle are known as quarter sawn, and on sycamore they will show pretty figuring after planing it smooth.

I next used my wimpy hobby-grade 9″ bandsaw to further break the chunks down into a set of thin boards approximately 3″ wide. It just barely succeeded, cutting very slowly with a great deal of groaning and stalling. I have since bought a much more powerful 12″ bandsaw that will make this sort of operation a breeze on future instruments.

I cut more of them than I needed so that I had a selection from which to work around knots and pick the prettiest faces to be visible on the outside of the instrument. After ripping them on the bandsaw I stacked them with air gaps around them and let them acclimatise for a couple of weeks; this proved to be a good idea because they definitely warped a little in the first few days.

After drying I picked out the nicest ones and planed the best face flat, revealing the medullary ray flecks and occasional ripples:

I mitred the ends, using the planed face as my reference:

Then I cleaned up the mitres and trimmed them to exact length using a special 45° shooting board I made for the purpose:

Trial assembly gave me my first real feel for how big this instrument was going to be relative to my hands:

Here’s where my clever plan begins. I had the idea to cut various rebates and steps and sound holes and things, and even the tapered area of the bellows frame using the CNC mill, before gluing the boxes together. This part actually went fairly well, apart from me messing up the first one due to a programming error. The main problem was that it took quite a lot of work to do the CAD and CAM to program the machine (there were actually six different programs due to the differences between the sides: left left, left right, right left, right right, tops, and bottoms). You might also have noticed that I used sycamore for both the action boxes and the bellows frames: I’ve since learned that it is common to use a lighter, cheaper wood for the bellows frames; if I’d done that I could have still used the same technique by gluing a strip of sycamore and a strip of whatever secondary wood together before machining them as if it was a single piece.

Trial assembly of all the sides before gluing:

Gluing. Here’s where things started to go wrong. The sides seemed to warp slightly when I put the glue on, making it extremely difficult to get them to line up perfectly. My first attempt was clearly a mess so I broke them apart before the glue had dried, washed the glue off, and tried again.

Second attempt didn’t look too bad at first:

But on closer examination, most of the joints were slightly squiffy, some more than others. It may not look like a lot, but it caused me all sorts of headaches further down the line.

Unusually, the instrument has two reed pans per side. This meant I had to make a pair of dividing walls for the bellows frames:

Next I cut the boxes apart into their three sections: the bellows frame and the two halves of the action box. I was a little nervous about this step but it went smoothly. Those slots routed in the walls are there specifically to make it easier to cut the boxes apart.

Incidentally, I didn’t know at this stage how deep the reed pans were going to be so I made the bellows frames fairly deep to be on the safe side. They probably could have been 5mm shallower in hindsight. This caused me a bit of a headache later on because I ended up having to accurately shim the bottom of the reed pan recesses on both sides because they were both deeper than the reed pans I actually made.

Next I had to inlay the captive nut plates in the bellows frames, and drill the bolt holes in the action boxes. All sorts of things went wrong during these operations, some of them because the frames weren’t perfectly square, some because my milling machine wasn’t big enough to get to all of the walls without turning the boxes 180 degrees. I’ll definitely be working on coming up with an easier and more reliable way to do this operation for the next instrument. The worst thing that went wrong was one of the bellows frames came loose from the (inadequate) clamps and the router bit tore a chunk out of the inside of the wall.

I had my finger on the emergency stop button and managed to hit it before it had done too much damage:

I repaired it by cutting a section out, gluing in a new piece, and planing it flush:

The repair is hidden on the inside of the instrument, underneath the chamois leather gasket.

With the nuts installed, I was able to bolt the boxes together and lightly plane the outsides to get the walls to match up nicely (zoom in to see the nice figuring):

I slightly rounded all the corners so it feels comfortable to hold:

The outside of the bellows frame needed planing to taper it down slightly and round the corners where it meets the bellows, which are deliberately slightly smaller than the boxes and have rounded corners.

I then lightly French-polished the action boxes, leaving the bellows frames bare because I needed to glue the linen and leather to them. After all the difficulties I had polishing the end plates, I was starting to get the hang of the technique by this point and did a much more satisfactory job of polishing the action box sides. If you zoom in you can see how it really brought out the figure of the wood. The dye in the meths still made the finish look a little bit purple, though!

I made a rookie mistake at this stage: as you can see in the previous photo, the bellows frame walls perfectly meet the action box walls with no step. What I should have done was to plane the action boxes a tiny bit smaller, so that when I glued the leather on, the leather would have been able to come right up to the underside of the action box without the edge being visible. I tried to work around the problem by skiving the leather down to virtually nothing where the frame ends, but it didn’t work perfectly and the join between the two looks slightly crude. It probably also would have helped if I’d made the end run a couple of mm wider and wrapped the skived edge of the leather around the corner a little, so the edge of the leather end run was clamped inside the joint between the boxes, underneath the chamois gasket.

No. 1: Brun Part 2: End Plates and Hand Rails

Apologies for the long delay since part 1 of this short series about how I built my first instrument, Holden Concertinas No. 1, a small rectangular 30 button Hayden Duet. I have now finished the instrument and begun design and tooling work on No. 2, which will be a more conventional hexagonal 30 button Anglo.

First a caveat. This was my first instrument, built from scratch without plans or instructions, in some cases involving materials and techniques that were new to me, not under the direct guidance of an expert (though several experienced makers offered useful advice when I had a specific question). I was constantly figuring out how to do things and building tools as I went along, and sometimes my ideas didn’t work out perfectly, or by doing a process one way I learned something that helped me to see a better way to do it. In some cases I was limited by the tools available to me, and I made a point of making everything myself in my own workshop (apart from the stainless steel screws). I think overall the instrument turned out remarkably well, with a few mainly cosmetic flaws, but there is certainly plenty of room for improvement when I build No. 2. This article is the story of how I did it on my first attempt and what I learned along the way, not a definitive guide to building one yourself.

The wooden end plates were made from 3mm birch plywood (the best quality I could find), hammer-veneered using hot hide glue and a veneer hammer I made myself from oak with a brass blade. The top veneer is a very pretty fir burr (or burl for the Americans), with a cheaper beech veneer on the bottom (you always veneer both sides of a board at the same time to reduce warping; in this case it also helped to thicken and strengthen the board). Unfortunately on the finished instrument the beauty of the pattern is rather hidden by all the piercings, buttons and handles, though there are areas where it is still visible. I dried the board between paper in a simple press for a week or so. After taking it out, it warped slightly, but it was flexible enough that screwing the plates to the rigid end boxes pulled them flat again, and in fact I now find that after a few months being held flat by the boxes, they remain flat when I take them off.

I decided to try French polishing the instrument, and I thought it would be easier to polish the board before I cut out the fretwork (in hindsight this was probably a mistake). This was my first time using the technique. It took a very long time, I made a bunch of mistakes, and the results weren’t entirely satisfactory. I learned a great deal, though, and I’m sure the next one will be better. One of the problems I ran into is that in the UK the sort of alcohol you can buy in a DIY shop (methylated spirits) has a small amount of poison and purple dye added to it to prevent people drinking it. It turns out that if you dissolve blonde (clear) shellac in ordinary purple meths, the dye doesn’t evaporate along with the alcohol, it stays in the finish. The thicker the finish, the deeper the purple, and due to inexperience I accidentally made the finish on the end plates way too thick. Funnily enough the change was so gradual I barely noticed it happening, until I happened to place the polished board next to an offcut of the original veneer! For my second instrument I am planning to try dissolving the shellac in clear ethyl alcohol that I found being sold online for medicinal purposes; it’s quite a lot more expensive than meths per litre, but I don’t need huge amounts of it to polish a concertina.

Next I cut the board into two pieces and stuck my cutting templates to them, then drilled all the holes on the CNC mill. Ignore the fact that the mounting holes appear to be in the wrong places – there was a reason I had to do that, which I’ll talk about in the next instalment. The fretwork design is my own, inspired by Indian mandala designs. Circles within circles within a circle within a square. It may not be to everybody’s taste but I am rather proud of it. The space below the fretwork was left empty because I knew the handles needed to go down there, but I didn’t drill any mounting holes at this stage because I hadn’t yet figured out exactly how or where they were going to attach.

I couldn’t drill the button holes to final diameter on the CNC mill because the drill bit I had was too long for the Z height of the machine, so I instead drilled small pilot holes and then manually enlarged them on the pillar drill. I was a bit worried the face veneer would splinter out when the drill bit entered the board, but I used a sharp brad point bit, fed it in gently, and they all turned out perfectly.

The end plates fit into rebates in the end boxes (in order to hide the cut edge of the plywood), which meant I had to do some fiddly manual trimming of the edges using a block plane and shooting board to get them to fit in perfectly without a gap.

I cut the fretwork by hand with a vintage fretsaw. It may look tedious but I actually really enjoy this part of the job!

Here is a little video clip of me working on the piercing while listening to Lady Maisery on the stereo (you have to click it again to stop it playing or it will automatically repeat):

https://www.instagram.com/p/BOo5Y5mACys/

After I’d finished piercing and peeled the sticky paper templates off, I found that the thick shellac finish had gone rough, so I decided to sand it down and try polishing it again. 

The result was a significant improvement, though not totally perfect. In hindsight I probably should have kept going, taken all the shellac off, and completely started from scratch, but I was loath to do that because of the many hours it had taken me to build it up in the first place.

I reamed the button holes out a little from the back using a taper hand reamer, then glued in bushes made from 0.85mm piano bushing cloth. This is a fiddly job that is difficult to do neatly, though I found that I got better and quicker at it with practice. Here’s a quick video clip showing how I did it using hot hide glue. I’m working from the back but aiming to get the front edge of the cloth to sit flush with the front face of the end plate because it’s not possible to neatly trim it flush afterwards.

https://www.instagram.com/p/BO_-jBWAfd_/

After the glue had dried, the buttons fit in the bushes but there was too much friction, so I turned a tapered polished brass rod the same size as the buttons and used it in the milling machine spindle at 10,000 RPM to burnish the inside of each bush. This had almost no effect! Second attempt, I made another tapered rod that was slightly larger than the buttons, and waggled it around a bit in each hole. This made the holes just right so the buttons went up and down with hardly any friction or noise. After a couple of weeks I found that a few of the buttons were sticking because the bushing cloth had bounced back, so I repeated the burnishing operation. This second burnishing seems to have done the trick; months later the buttons are still working smoothly.

The extra bit of felt sticking out of the back of each button hole was in some cases interfering with the cross holes in the buttons and preventing them closing fully, so I trimmed them all flush with the back of the plate.

On to the hand rails (I know I’m skipping over the casework and action; I’ll cover those in another instalment). I had an idea to try making more ergonomic handles and spent a while experimenting with a set. Unfortunately they didn’t really work out for various reasons that are outside the scope of this article, but I’ll include a couple of pictures because I think they looked rather pretty.

Back to the drawing board, I decided to make a pair of simple Anglo-style hand rails, but with the ability to adjust them backwards and forwards, and a spacer block that allows the height to be raised or lowered.

The top parts of the rails were made from slightly spalted apple wood from a tree that was taken down in my best friend’s orchard:


The spacers below the main rails were made from contrasting sycamore wood:

While planing the two pieces to match each other, I clamped them together using the same bolts and captive nut plates that would later be used to hold them onto the end plates:

Because the handle could be adjusted back and forth, it didn’t make sense to attach the strap to a fixed point on the side of the action box, which meant I instead had to put the strap clamp screw on the end of the rail. This meant making a captive nut that fit into the end of the rail, and was held very securely by one of the main attachment bolts passing through a cross hole drilled through the nut:

This wasn’t my first attempt at making thumb screws: the earlier ergonomic handles had four larger-diameter screws. I think the final ones I put on the instrument turned out reasonably well, though the knurling isn’t as neat as I’d like due to the knurling tool I bought being a bit too sloppy. I am planning to make a stronger tool with finer wheels before I knurl the screws for the second instrument.

I lightly French-polished the handles and cut the straps from 2mm full grain veg tanned calf skin. I think the shape worked out OK, though I did the decorative creasing around the edge using a tool improvised from an old soldering iron and it went a bit wonky in the curved areas (I’m planning to get or make a better creasing iron before I make the next set of straps).

I found that after playing the instrument for a few minutes the thumb screws started to unscrew themselves. The solution came in a photo I found of a Dipper instrument that had the thumb screws in the same location: a clamp plate like a large washer that is bent around the end of the rail. This prevents the screw turning when the strap moves and looks quite snazzy too! I have since seen exactly the same arrangement on instruments made by Geoffrey Crabb. I put the two bends in the clamp by clamping it in the middle of a toolmakers’ clamp that happened to be the same width as the rail (1/2″).

The final touch on the end plates was to add my maker’s label. The design was hand lettered by my talented friend, Oliver Densham, in a style inspired by the labels used on vintage Victorian instruments. I laser-printed it onto archival grade paper and sprayed it with a special protective lacquer intended for decoupage art, so it should hopefully last a long time.

Swanky Switch

I’ve spent what seems like an inordinate amount of time getting the bellpush switch right. This photo shows my first attempt:

switch1

The long contact strip is made from a roughly T shaped strip of 0.7mm brass. It’s bent into a zig-zag at one end to make it more flexible (I also had to file it narrower for the same reason). One of the problems I ran into was I initially didn’t recess the contacts deep enough to ensure the terminal screws can’t press against the underside of the top. The top of the flexible strip is slightly curved so that when the lip of the button pushes down on it, it flexes a little and ‘wipes’ the contacts against each other to (in theory) break through any oxide film that may have developed since the last time it was used. The contacts are made of short sections of flattened 4mm diameter sterling silver rod (for good corrosion resistance), silver-soldered onto the brass parts. The screw terminals are salvaged from a UK 13A mains plug, filed a bit narrower due to the limited space available. I did all the soldering with the Eclipse spirit blowpipe described in an earlier article. They are held down by M2.5 stainless steel machine screws, mated to square S.S. nuts morticed into the back of the box.

When I got to this stage I thought the switch was done. The button action felt nice and the contacts closed when I pressed it. Unfortunately there was a problem, as I discovered when I proudly showed it off to my friend Juliet. She pressed it normally a couple of times and it worked fine, then she tried pressing it really gently. It didn’t work. She tried the same thing a few more times with intermittent results and proclaimed it faulty.

It was a user interface problem: press the button down firmly all the way to the bottom and it worked fine. Press it very gently and it was possible to feel the slight increase in resistance as the contacts began to close and stop pressing too soon. Result: the bell doesn’t sound, and you might not realise if you were outside and the bell was inside. I suppose some people seeing such an ornate bellpush might think that it looks delicate and press the button gently in fear of damaging it (in fact you’ve probably got more chance of breaking your finger than the button). This graph illustrates the problem (figures are estimates):

buttonfeel

The problem is that first step when the button lip touched the flexible contact strip and the switch began to close. If you were pressing gently enough, it (wrongly) felt like that was the bottom of the button travel when you actually needed to press a tiny bit harder to close the contacts.

I won’t go through the list of ways I tried to solve this problem. I now have quite a collection of discarded springs! What I eventually settled on was a second helical torsion spring attached to the top of the flexible contact that applies gradually increasing pressure to it over the  full length of the button travel. The result is that the switch closes smoothly without any detectable step increase in force, and the contacts are fully closed at about 80% of full travel. If you press the button firmly enough, it bottoms out against the top of the contact strip and wipes the contacts as originally intended.

I also had to remake the first spring with more and bigger coils to weaken it (I could also have used thinner wire but I didn’t have any in stock), because the combination of the two springs made the button force uncomfortably high. I didn’t want to remove the first spring and just use the second one to return the button because it’s set such that the pressure on it is zero at the top of the button travel so as to ensure the contacts release properly, which means the button wouldn’t return to the top as cleanly with that spring alone.

Here’s a cute picture of the little bracket I made to attach the new spring. I filed it from one of the brass 13A plug pins that I got the screw terminals from and silver-soldered it onto the contact end of the flexible strip (that’s a 0.8mm diameter hole):

switch2

This photo shows the final setup. The brass peg to the left of the flexible contact (a screw with the head filed down) is there to ensure the torsion spring can’t swing to the left and disengage from under the button:

switch3

 

Now that I had a working bellpush, I wanted to make a little video to show it off. I don’t have an electric bell here and the continuity buzzer in my multimeter doesn’t sound impressive enough, so I hooked it up to something else instead:

 

 

Captive Nut Plates

One of my intentions for this blog is to document techniques that I learn don’t work as well as those that do. Here’s one of my ideas that didn’t go as planned.

I decided to use a pair of M2.5 stainless screws for the end bolts of the bell push (I bought them with slotted cheese heads, then re-shaped them a little and polished them in the lathe). M2.5 was the closest modern metric equivalent to the size of the end bolts on my antique Lachenal.

This meant I needed to make a pair of little metal captive nut plates to screw to the wooden back box with an M2.5 threaded hole in each one. Unfortunately I don’t have any M2.5 taps (M2 and M3 but not M2.5) to thread the hole with, so I thought I would instead try silver-soldering commercially made square stainless steel nuts onto thin brass plates.

Bad idea. The solder wicked into the threads of the nuts, pulled them out of position due to capillary action, and generally made a bit of a mess. I could probably clean the solder out of the nuts – if I had an M2.5 tap. But if I have to buy a tap anyway then I might as well buy a full set of them and then I can just make one-piece nut plates from thicker brass and not have to mess about with soldering and cleaning the flux and oxide off afterwards.

failed_nut_plates