Tuning Bellows

Recently I’ve been working on making myself a new set of tuning bellows. The first tool I made towards this goal was a new craft knife for cutting leather and card. I ground it from a piece of HSS machine hacksaw blade, because I had read that the steel is more wear-resistant than a simple carbon steel, and I wanted to test the theory.


I’m not convinced it holds an edge better than a well-tempered O1 blade, however it does work pretty well. Unfortunately the spalted beech scales (glued on with epoxy resin) popped off after a couple of days. It would have been better to have riveted them on, however drilling holes in hardened HSS is easier said than done. For the time being I have just bound them on with a length of thick elastic.

On a roll with the knife-making, I next made myself a leather skiving/paring knife from the same materials as the craft knife:


This video shows how I use it:

I do of course also have a Schärf-Fix 2000 skiving machine, which I wrote about previously:

The Schärf-Fix is good for long strips, whereas the knife is better for skiving small pieces (particularly gussets) and odd bits here and there. Both tools are tricky to use, and I wasted a frustrating amount of leather while getting to grips with them. The Schärf-Fix has a tendency while edge-skiving a long strip to suddenly dig in and cut a chunk out of the strip. The key is that the blades have to be absurdly sharp, because the thin leather tends to stretch badly if you have to use any significant force to pull it through the machine. I need to spend some more time figuring out a way to resharpen the disposable blades (stropping didn’t help much), or I’ll end up going through at least a couple of them per concertina.

I used my Käfer Dial Thickness Gauge to make sure I was paring it consistently:

I needed to make end frames for my bellows, so I first needed a way to accurately make the corner reinforcing blocks. Here’s the jig I came up with. There’s a bit of a knack to using it, but the results aren’t bad:


And here’s one of my new corner blocks next to an antique Lachenal one:


I cut the sides of my bellows frames from 9mm plywood using my Nobex Proman 110 mitre saw, and glued them together with hot hide glue:


After a bit of shaping, I checked that they fit in the scrap Jeffries bellows I was copying my dimensions from:


Next came the bellows mould. This proved quite a large sub-project in its own right. It has six forms (one of which is split in two to make it possible to remove the forms), screwed to a hexagonal core, suspended from a stand. Making the forms was the hardest part. I started by gluing blocks of pine to strips of plywood, with the grain running across the form, being careful to avoid including any large knots:


Then I used the bandsaw with the table tilted over at 45 degrees to relieve the under-sides of the forms:


Then I used the CNC milling machine to cut the valleys into the top sides of the forms (video sped way up: it actually took about an hour to machine each form):

After I had spent days making all six forms, I laid them out next to each other and realised I had made a silly mistake: five of them were spaced wrong, and in fact all of them were pitched slightly too tight to fit comfortably inside the Jeffries bellows:


I could have tried to unglue the blocks and glue them onto new plywood strips with the correct spacing, but I decided it was easier just to start again and remake them all. By the end, I was getting really tired of the noise the milling machine made as it cut the valleys, not to mention the dust everywhere!

I made the core of the bellows mould by mitring the edges of six pine boards on the bandsaw and gluing them together. I deliberately made it slightly oversize, then hand planed it to final shape/size (a good idea as it turned out slightly wonky, plus I wasn’t certain exactly how big it needed to be until I tried assembling it inside the Jeffries bellows):


The stand is a simple affair with the vertical ends roughly dovetailed to the base. A nice feature of this style of core is that if you turn the central bar one way up, it presents the sides uppermost, and if you turn it the other way up, it presents the corners instead. This picture also demonstrates that the scrap Jeffries bellows fit on the mould:


Finally time to start making the bellows! I cut the 108 individual cards out by hand, using a template I cut from a piece of scrap aluminium to match the shape of a card taken from the Jeffries bellows:


I hinged pairs of them together using strips of fine-woven linen cut on the bias and bookbinders’ maize paste (very similar to wheat starch paste but supplied pre-cooked and with some anti-fungal stuff mixed in):


Then I hinged the pairs together into six strips. Note that the hinges are both on what will become the inside of the bellows, and the valley hinges have to be pasted on with the hinge partly closed or they will tend to tear themselves apart when they close. There may be a less fiddly way to do this but it seemed to work well enough.


After the paste had initially dried, I noticed that the cards had all warped a bit, so I pressed them all tightly in a big wooden clamp for a couple of days (forgot to take a photo), which helped to flatten them out again.

I tied the strips of cards onto the mould with string, then hinged the corners together with more bias-cut linen, though at this point I switched to using hot rabbit-skin glue. It’s messier and more difficult to work with than paste, but in my tests it was the strongest of all the glues I tried (slightly stronger even than PVA wood glue, and significantly more flexible when dry).


I made a simple press to clamp the bellows shut, and whenever I had to let the glue dry before the next stage, I took the bellows off the mould and transferred them to the press. If I hadn’t, they would have dried in the fully-open position and possibly torn apart when I tried to force them closed. It’s also necessary to periodically take the bellows out of the press and exercise them to avoid them drying fully shut.


Next I glued on the valley leather strips. I used goatskin for all the leather on the bellows. I pared the valleys down to about 0.65mm and skived the edges for cosmetic reasons. They are simple rectangles rather than butterflies because that’s how they were on the Jeffries bellows I was copying:


Next the gussets:


Here’s a video of me gluing a couple of gussets on:

The top and end runs. In hindsight the end runs would have worked better if they were both narrower and thinner, and perhaps I need to work on my technique for gluing them on, because I was unable to get them to go round the corners without creasing, which caused the end sets of gussets to be stiffer than the rest. It probably didn’t help that I made the cards all the same size (the end ones probably should have been slightly taller because of the inset):


Pressing it all together:


To make them look a bit prettier, I made bellows papers from decoupage paper:


I screwed a plain piece of plywood onto the bottom end, with a sheet of black “funky foam” (closed cell EVA foam sold in thin sheets for craft purposes) as a gasket, and a couple of pieces of scrap lead to pull the bellows open with a consistent amount of force:


The top board has a reed holder next to one edge. It’s a simple design that doesn’t require any adjustment for different sizes of reed, though you do have to hold the reed in place with your thumb while sounding it. The top plate is slightly thinner than a reed frame so that it’s possible to file the reed in situ, and it has a slight undercut so as to hold the dovetailed reed frame more securely. It took some careful measurements and fiddling about to get the wind slot just right so that it works for the full range of sizes of reed I had available. It might require further adjustment if I ever want to use it with even bigger reeds from a bass instrument. The separate brass screw is used in conjunction with a specially shaped thin spring-steel shim (not pictured) to hold the reed tongue up above the frame while filing.


Here’s the finished tuning bellows clamped to my bench. The flap of leather is a relief valve to let the air out when you raise the bellows. Not clearly visible in the picture, there are a pair of straps tacked to the sides of the frames that prevent the bellows opening too far.


Finally, here is a quick video of me showing them in action. The reeds are the highest and lowest reeds from my antique Lachenal 48-button English, plus the A4, which is at 444Hz because it is tuned in old pitch, meantone temperament.

I want to thank Geoffrey Crabb for all his advice on the construction of the bellows moulds and the tuning rig, and also Bob Tedrow for his bellows-making essay (although my technique is quite different, I picked up several good ideas from it).


Hand-Cut Maker’s Mark Punch

There is something magical about the ability to anneal1 high-carbon steel, work it into a useful tool, then harden and temper it so that it can hold a sharp cutting edge for a long time. It was one of the most important discoveries of the Iron Age, enabling the manufacture of tools that were far superior to those made of softer metals like copper, bronze and wrought iron.

I have made a few hardened steel tools of my own; wood carving knives and gouges, and simple punches. It is a wonderful and exciting feeling to use a tool that you made yourself.

Part of the process of hardening tool-steel involves heating it to something in the region of 760-800C2 and holding it at that temperature for a while. If the temperature is too low the steel won’t harden properly, and if it’s much too high you get a coarse grain forming inside the metal that will affect your ability to sharpen the tool. In the past I’ve heated my tools using either an open solid-fuel forge or a propane blowtorch, but in both cases it is difficult to accurately gauge and regulate the temperature. Because I need to make more tools for concertina production (mainly press dies), I’ve been looking out for a better way to heat them.

Recently I came across an old electric laboratory muffle furnace on eBay. Luckily I managed to get it very cheaply because it was described as faulty and it was near enough for me to collect it in person. It looked to be in good condition in the photos, and I figured that even if the heating element had burned out, it would be cheaper and easier to re-wind it than to build one from scratch.


The fault turned out to be very simple. It is supposed to have a fusible link inside the inner chamber that melts if you overheat it (a thermal fuse). This was probably quite easy to do because the original controller was a simple simmerstat3, and I suspect leaving it at 100% would cause it to reach melting point in about 45 minutes. The thermal fuse was missing. As a temporary measure I bypassed it and the oven fired straight up.


I learned nearly everything I know about heat-treatment of steel from Hardening, Tempering & Heat Treatment by Tubal Cain from the Model Engineer’s Workshop Practice series. It was only after I’d bought my oven that I happened to be flicking through the book and saw a picture of it: the author had the same model!

Tubal Cain had replaced the simmerstat in his heat treatment oven with an early computerised temperature controller. I wanted to do the same thing, and I could have simply bought a fairly cheap Chinese PID controller like the one I used on my glue pot (though high-temperature ones seem a bit harder to find and more expensive), but for various reasons I decided to build my own instead. I used a MAX31855 thermocouple interface from Adafruit, an Arduino Nano clone, a solid state relay, a 16×2 alphanumeric LCD (HD44780 compatible), a couple of push buttons and a rotary encoder stuffed into a plastic project box. Because I already had most of the parts other than the high-temperature K-type thermocouple (which I would have had to buy anyway) and the MAX31855, the project worked out pretty cheap.


I developed the firmware using the Arduino IDE and several off-the-shelf open source libraries. This is the first time I’ve used the Arduino system, though I’ve done quite a bit of embedded programming in the past. I must admit a decent C++ compiler and a large set of libraries made it pretty easy to quickly bolt something together, though the lack of a debugger is a bit of a pain, and some of the libraries are poorly documented and/or provided with example code that doesn’t work properly out of the box.

The most difficult part of the project turned out to be tuning the PID loop parameters. Get them wrong and the oven either never reaches the desired setpoint or it overshoots and oscillates around the setpoint. One advantage of developing my own Arduino-based controller was that it was easy to log the temperatures and power level at regular intervals to a laptop over USB, then plot a graph to figure out what was happening over time. To cut a long story short, after hours of test cycles and trying many different values, I eventually found a set of parameters that perform well enough for my purposes. It overshoots by a few degrees when it first gets up to temperature or after disturbing the system by opening the door, but I don’t believe that is enough to cause a problem. In this graph, the blip at 5200 seconds is the result of me opening the door for a few seconds:


For years I have been fascinated by how punches and dies were made prior to the invention of the rotary engraving machine. I’ve read what I’ve been able to find on the subject (not very much, to be honest) and studied some antique punches to try to work out how they were made. I decided to try making a punch from my maker’s mark to see if my ideas were practical.

I think positive punches and single-line name stamps usually used at least one counterpunch per letter to form the hollow spaces (the counters), and the outside waste was cut away with saws, files and probably a selection of gravers. There were also negative punches that were probably either entirely engraved or stamped with a set of reversed positive punches, but I’m not going to cover those today.

My maker’s mark is a fairly simple symbol (an upper case A with two arms added), though I wanted to challenge myself by making a serif version with multiple stroke widths.

I made both the counterpunch and the punch from 3/8″ silver steel, which is a commonly-available high-carbon water-quenched tool steel with a little chromium in it that is supplied as accurately ground round bar. Since the counterpunch was to be smaller than the punch I first tapered the end. There are several ways to do this, but I decided to use the compound slide on my Taig lathe. I also used the lathe to face both ends to make them square, and used a file to round over the hammer ends a little so that when you strike it you don’t hit a corner.


The facing process left the end that was to become the punch square and reasonably flat but very slightly rough. Because this might effect the performance of the finished punch I decided to lap it flat.


First I clamped the blank in a Vee block, using a piece of paper as a shim to make sure the punch protruded from the block by a tiny amount:


Then I lapped it on a cheap diamond plate in a figure-eight pattern:


The lapped face of the punch blank:


Next I started cutting the counterpunch using a jeweller’s saw to make the initial grooves:


I removed the waste from the outside using a fine flat hand file:

Then I used a graver to widen the grooves (you can see in this one it’s pretty small compared to my index finger):


The nearly-finished counterpunch:


To check it was the right shape, I got it sooty in a candle flame and pressed it onto a piece of paper (this is called a smoke proof). Actually I realised at this stage that I had made a mistake, but I opted to push ahead and modify the proportions of the punch to compensate rather than starting again. I could get away with this because the design of my mark is rather fluid anyway, and doesn’t have to match the style of a particular font.



The final part of cutting the counterpunch was to bevel the corners and do a test punch into a piece of end-grain hardwood. I’ve found that a punch doesn’t work very well if you don’t bevel it at all, but it’s not entirely clear to me how steep the bevel angle should be (the old punches I’ve looked at aren’t all the same).


Next I had to harden the counterpunch before I could use it. In order to reduce scale buildup and decarburising while soaking in the heat treat oven, I coat the business end of the tool in jeweller’s borax flux. Applying it isn’t an exact science. I warm the punch to not-quite boiling point, then smear some thick borax paste onto it and wait for it to dry. Once in the oven, it will bubble up a bit at first but then it should melt and flow out across the surface (you can see the difference after heat-treating; the areas that were protected by the flux are still bright underneath).


I put the counterpunch into the oven on top of a piece of bent stainless steel that prevents it directly touching the oven floor. After waiting something like 20 minutes for it to get up to temperature, I let it soak for another 20 to ensure that it was fully austenized all the way through. From what I’ve read, not soaking for long enough will significantly reduce the hardness you can achieve.

After soaking, I pulled it out with blacksmith’s tongs and immediately quenched it in a bucket of water. The shock actually causes most of the Borax flux to fall off, which is handy because it can be difficult to remove.


As a quick check to make sure it hardened, I see if a file will cut it. Generally it will scratch a tiny bit because of the surface decarburisation effect but it will be hard enough underneath that the file just skates off. Some people recommend using a good sharp file for this test, but I find that it tends to blunt the file so I prefer using a rubbish file and just press quite hard (I have experienced a tool that didn’t harden properly and the difference was pretty obvious).

After hardening, the steel is very hard but also very brittle and highly stressed. If you’re not careful you can shatter it just by rough handling (I’ve done that with a fancy spring clip that I had just spent half an hour forging). What we do to cure this is to temper it, which means re-heating it to a lower temperature and soaking it for a while. This reduces the hardness somewhat but also reduces the internal stress and greatly increases the toughness of the steel. If you temper it lightly you end up with quite a hard tool that might be at risk of chipping. Higher tempering yields a tougher but softer tool; if you go too far the edge may tend to roll over in use. The highest levels of tempering are used to make springs.

There are several ways to heat a tool for tempering. I could even use the heat treat oven itself, but it would take a few hours to cool down sufficiently and it’s not a good idea to leave the tool in its fully hard state for that long. I decided to use a small electric deep fat fryer with sunflower oil in it. The highest temperature the thermostat will go up to is about 200C (and the oil probably wouldn’t be happy going much higher than that), which is towards the low end for tempering silver steel.


After tempering I used a rotary wire brush in the pillar drill to remove the scale and any remaining flux.


Finally I used a spirit burner to re-heat the blunt end up to a spring temper so that it can better resist the shock of hitting it with a hammer. It will probably mushroom eventually but it is easy to cure that by grinding. After it reached a blue colour I quenched it again to stop the heat travelling too far up the shank.


A quick test in a piece of scrap aluminium proved it had worked as planned:


The first stage of cutting the real punch was to strike the counterpunch into it. I think this is possibly the most difficult part of the entire process. You have to line it up perfectly, then hit it multiple times very hard to drive it as deeply as possible into the steel. Although the silver steel is supplied annealed, it is still a pretty tough material and it takes a lot of force to counterpunch it. The counterpunch bounces out of the indentations after every strike and you have to be very careful to line it back up perfectly before the next strike – I had a bit of an accident and made an extra small dent but luckily it was in a waste area. The counterpunching causes the surface to raise up a bit around each indentation, which I opted to get rid of by lapping it flat again. Here is the result after counterpunching and lapping:


I roughly removed the majority of the waste from the outside with a jeweller’s saw and files again:


Mostly done, though not yet adjusted the width of the strokes:


After a bit more fine fettling, and adding the bevels:


I didn’t bother photographing the hardening process as it was exactly the same as for the counterpunch. The final set of photos shows the result of testing it in several different materials. First end-grain beech:


Aluminium, both a light strike and a heavy one:


A light strike into side-grain pine tended to crush the soft fibres:


End-grain pine worked fairly well though. The bottom one was an experiment in soot blackening the punch first:


Brass. I had to hit it pretty hard to get a decent impression but it came out quite nicely:


Thick leather. This looks sharper in reality than the photo appears to show:

Mild steel. The top mark was a moderately hard strike onto the wooden bench and it’s barely visible. The bottom mark was a very hard strike backed by a steel anvil. I’m not sure I’d want to risk striking the punch that hard very often as I’m concerned the tiny serifs might be at risk of breaking off, though it didn’t suffer any visible damage as a result.



Small Roubo-inspired Pine Workbench

I’ve just finished building a new workbench for my concertina-making workshop. bench1

I designed it myself, loosely based on an 18th century French design described by Andre Roubo. Although mine is much smaller, made from construction-grade pine, and has modern (ish: they are probably older than I am) cast iron quick release vices instead of a wooden leg vice, they do have several important elements in common. The joinery is very similar, particularly the unusual dovetailed tenons that allow the legs to come flush with the edge of the top. It has a planing stop, a front vice, a holdfast, and a tool tray between the stretchers.

I made it 1m tall, which is on the high side for workbenches, because I mostly do fine detail work on small components, often under magnification, and in these circumstances a high bench reduces back ache from constantly bending over. The top is 400mm deep by about 1.2m long, which I felt was the biggest I could comfortably fit in my small garden shed workshop while also fitting in a coal stove to the left of the bench and two machine tool stands along the back wall.

The next few pictures show a few of the many workholding options available to me on this bench. The Woden quick-release front vice with leather-lined jaws is likely to be my main workhorse:workholding1

I also have a very similar Record quick-release vice, which I installed as a central tail vice with leather-lined wooden jaws that span the full width of the end of the bench. This idea is a bit of an experiment on my part, but I have a feeling it is going to come in handy (obviously I will have to be careful to avoid clamping things too far off-centre as excessive racking force could damage the vice).workholding2

For light metalwork, I welded up a steel bracket that allows me to hold a Record No. 3 in the jaws of one of the woodworking vices (I’ve since painted the bracket the same colour as the vices):workholding3

The holdfast is an important feature of Roubo’s workbench. Instead of the traditional forged iron bar that wedges into a hole when you hit it with a mallet, I went for a Woden hold-down clamp that performs pretty much the same function but with a screw to control the clamping force. It fits into a special tall cast iron collar with ridges inside that mate with notches on the back of the vertical rod, and it only came with one collar, so I put a lot of thought into the most versatile place for it to go on the bench.

Another important feature of Roubo’s bench is the adjustable-height planing stop. Mine is a significantly different design, though it performs the same function. I made it from a limewood board without any metal teeth on the top, partly to avoid marring the work but mainly because I don’t want to risk accidentally running into it and damaging the plane, particularly when planing very thin stock. There is a wing nut under the bench that allows it to be clamped at the desired height.workholding5

A feature of Roubo’s bench that I didn’t copy is the crochet; the hook on the front that you would use to wedge the end of a long board when jointing. It didn’t really make any sense given the fact that I can’t easily remove my front vice, which in any case can easily be used to clamp the end of a long board.

I also have various clamps and small vices that can be held in the bigger vices to hold things in creative ways:workholding7



I don’t yet have any bench hooks or shooting boards, though I plan to make some soon.

The idea of the tool tray under the bench isn’t permanent storage, it’s just somewhere convenient to put down tools that you aren’t currently using, other than putting them on the bench top itself where they get in the way and risk being knocked onto the floor. I modified the idea a bit though, by putting mine on hinges with a storage box underneath it:bench2

My plan is to store heavy things in here, thus lowering the bench’s centre of gravity and making it more stable. Currently it contains five fire bricks (!), a bunch of metal clamps, and a couple of small anvils. Combined with the fact that the bench structure itself is very rigid, the two other things I did to stop the bench moving around in use were that I glued rubber pads to the bottom of each leg and I attached the tops of the legs to the wall of the shed with angle-iron brackets, after bolting a steel channel to the wall to make it more rigid. I’m very pleased by how well this has worked.

Light is extremely important for fine detail work of the sort found in concertinas. Of course I put the bench next to the window to pick up what natural light I can, but that isn’t much use when it’s dark outside! The ceiling is too low for a conventional fluorescent light tube above the bench, and spotlights are problematic because they cast light and dark areas. After a bit of research I found out about LED light panels. I bought a 300 x 1200mm 40W daylight panel and fitted it directly above the bench, and I was amazed by the quality of light it produces. It is like being directly under a skylight on a bright day.ledpanel(On the negative side, I also replaced the bulbs in the two batten light fittings with newfangled LED filament bulbs, and I found those to be a bit disappointing: slightly dimmer than the claimed equivalent tungsten bulb, with a very dim patch directly below the bulb).

To finish up, here is a selection of pictures from the construction process.

The wood I bought to construct the bench, mainly 3″ x 3″ planed “redwood” (i.e. Scandinavian pine). IIRC it cost about £40 including the T&G on the left that I used for the bottom of the tool box (the lid was a piece of scrap plywood I had lying around).benchbuild1

First I had to restore my Record No. 7 jointer plane, which I had never really had a use for before now, not being a furniture-maker.benchbuild2


Then I used it to joint the boards that I glued together to form the long stretchers and the top:benchbuild4

I glued one joint at a time using hot hide glue and three sash cramps (it might have been preferable to use a couple more, but I was horrified when I found out how much decent ones cost). My vintage electric iron was used to preheat the wood to extend the glue’s open time.benchbuild5

After the top was fully glued up, I roughly trued and squared it with a wooden jack plane:benchbuild6

Then I flattened the top surface with the jointer plane. After final assembly I re-flattened it, and was surprised to discover that in the intervening couple of weeks it had cupped by about 2mm.benchbuild7

Cutting the joinery:benchbuild8







First assembly of the bench, with Pippin the pushme-pullyou dog (this was an accidental trick photo caused by my dad activating the panorama feature on my iPhone).benchbuild15

All three vices needed a good cleanup and repaint. The Woden vice also turned out to have had the quick release mechanism modified in an inexplicable way that prevented it functioning – after reversing the bodge it worked fine.

Annoyingly the new paint didn’t dry very well at all, and after several days I got fed up waiting for it to harden and fitted them anyway, resulting in the very soft paint rubbing off in a few areas.

Making riven beech pegs to hold the joints together. I didn’t use any glue, or even bother to drawbore them.benchbuild18

Setting the hold-down collar into the benchtop using my router plane.

The fixed jaw of the end vice was morticed into the benchtop to avoid having a gap between the end of the bench and the wooden jaw.benchbuild20

Unlike the Record vice, the Woden one had a really rather wonky top surface where it bolts to the underside of the bench.
benchbuild21As a result, I had to spend a couple of hours painstakingly carving the underside of the bench to match the top of the vice in order to get it to fit snugly and not rock:

Before I finally attached the top to the legs, I whittled my maker’s mark and the year of construction on the underside:benchbuild23

The long stretchers were held into the legs with both a peg and a pair of wedges (which open the tenon out into a deliberately tapered mortice). This is a very strong joint, even without glue. Getting the wedges out if I ever want to disassemble it will probably be a bit of a struggle though.benchbuild24

Simply screwing the hold-down collar to the bench top didn’t work very well; when I tightened the clamp the screws started to pull out, so I fixed the problem by making this steel clamp plate that goes on the underside of the bench. The screws are M5, stainless steel.benchbuild25

The short section of benchtop behind the planing stop shrunk and split after a few days, so I cut it out and scarfed in a chunk of dry oak instead.benchbuild26


New Beginnings

I have moved back to Burnley! Soon I should be able to start making progress on building my first instrument. Currently my tools and materials are piled up in crates and boxes in my bedroom and garage, and my milling machine is broken down into several pieces, so my first task is to convert this damp 6×8 feet garden shed into a cosy little workshop:


At some point I hope to save up enough money to replace it with something a bit bigger and more substantial that is designed from the ground up to be dry, insulated and secure. For now, my plan is to install a tiny coal stove (which can later be transferred to the new workshop), repaint the outside walls, add more ventilation, and store as little as possible in there when I’m not actively using it.

The first thing I intend to make in the new shop is a workbench. Nothing massive or fancy, just a solid pine worktop (made from a spare kitchen table I found in my parents’ loft) at a convenient height, secured to the floor/wall, with some old cast iron vices and a planing stop. After years of working on flimsy trestle tables and Black and Decker Workmates, it will be great to have a sturdy fixed bench that doesn’t skitter across the floor when I try to plane something. In a stroke of good luck, just as I was starting to think about workbench design, I happened to stumble across a copy of The Workbench Book by Scott Landis for a couple of pounds in a local charity shop. It’s a hefty book filled with inspiring ideas. My first self-built workbench probably won’t be my last, but seeing what other woodworkers are working on has helped me to identify more clearly what features I think I will find most useful for the type of work I do and avoid some things I might have regretted later (e.g. I don’t want any dog holes in the top: too easy to drop small parts down them).

Making concertinas from scratch requires a lot of tools. Metalworking, woodworking, leatherwork, and some very specialised things like a tuning rig. In addition, there are opportunities to speed up production if you invest in special tooling and machines. In order to try to avoid falling into the trap of Tool Acquisition Syndrome (gotta collect them all!), I’ve started categorising tools like this before I decide whether I need to buy one or not:

  1. Tool that is the bare-minimum cheapest way to perform an essential operation. For example, it’s perfectly possible to cut out fretwork ends using a cheap and cheerful bent steel fretsaw you can pick up for a quid or two at most car boot sales. I have a few of these in various sizes. They are a bit slow and tiring to use for long periods, but they are perfectly capable of producing excellent results.
  2. Tool that speeds up an essential operation. For example, a scroll saw is a machine that does essentially the same job as a fretsaw but can cut perhaps four times as fast with less physical effort. The difference is actually probably less significant with more complex designs because you spend so much of your time stopping and moving the blade to a new piercing. I have both a treadle and an electric scroll saw. I should probably get rid of one of them to save space, but I need to do some proper side-by-side tests before deciding which of them I prefer to use.
  3. Tool that is a higher quality version of something I already have. For example, my scroll saws are both hobby-grade machines: either will do the job for now, but certain features are inaccurate, fiddly and annoying to use, or not built to last. If in the future I find myself with the budget to upgrade, it would be nice to step up to a higher quality saw like a Hegner, or maybe even a vintage industrial machine.
  4. Tool that would be nice to have, but isn’t essential to concertina-making at all. I’m struggling a bit to continue with the fretwork analogy here. Let’s say an electric bandsaw: one could perhaps argue that it could be used to rough out the blank more quickly than with a handsaw, but it isn’t of any use for cutting the fretwork itself because you can’t get the blade inside a closed piercing (and they can’t cut very fine details and tight curves).

Milled Boxes

It has become a tradition for me to make several Christmas presents each year using whatever tools/skills I have picked up most recently. This year my latest tool acquisition is the CNC milling machine, so I decided to do a project that would show off some of what it can do, as well as giving me some useful experience in programming it. A friend suggested trinket boxes. After considering various possible construction methods, I settled on milling them from blocks of limewood with thin plywood bases and lids.

I needed to make presents for three people. The nature of one-off CNC manufacture is such that most of your time is spent at the CAD/CAM stage, because you have to plan every detail in advance and figure out how to tell the machine what to do (being careful to avoid anything that is physically impossible, for example asking it to cut a 3mm wide slot when your smallest cutter is 3.175mm in diameter). It would have been easy to design one box and produce three exact copies, but where would be the fun in that? I instead decided to do three different designs. I learned new things from each one and I feel the third design is the best, so it wasn’t a wasted effort.

The CAM workflow was basically the same as with the maker’s mark stamp except the designs and machining operations were far more complicated. An additional step was deriving the profiles of the lids from the shapes of the rebates they sit in; this was complicated slightly by the fact that I needed to round off all the points because the internal corners of the rebates can’t be any sharper than the diameter of the router bit.

The milling operations themselves mostly went OK. I suffered one stepper motor stall while milling the first box, but I heard it happen and managed to hit the e-stop in the nick of time before it did any noticeable damage.



I made one design mistake: one of the lids was cut to the wrong outline (you can see it in the next photo). I didn’t spot it at the simulation stage because it was approximately the right shape for the compartment, just slightly too big.


When I cut the replacement lid I took the opportunity to make a little Christmas tree decoration at the same time:


The boxes looked pretty plain with ordinary birch plywood lids, so I bought a selection of patterned decoupage paper on eBay and pasted various combinations to the lids. I also pasted plain coloured paper to the bottoms.


I turned the knobs by hand from beech. Because there was quite a bit of variation I made more than I needed and picked out sets that went well together.



Here are all three finished boxes after lacquering:


The ‘S’ box is my favourite (made for my mum, Sandra):


I came up with the basis for the design while playing around with mathematical knots. It is based on the 8 18 knot.

Incidentally I now have an Instagram account and I’m using it to post pictures of things I’m working on.


New Stamp

I’ve been spending quite a bit of time recently working on my CNC milling machine and learning how to program it.

The main improvement I’ve made is to replace the spindle drive motor. The one that came with the machine was a 1/4HP single phase motor that ran at a fixed speed that gave me, on the top pulley ratio, something like 4000RPM at the spindle.


The new one is pretty much the same physical size and weight (maybe a little bit lighter) but it is a 1/3HP three phase motor that, paired with a used Mitsubishi Variable Frequency Drive (VFD) from eBay, can run at any speed from almost 0RPM up to higher than I’ve dared to take it (it’s comfortably fast enough to run the spindle at its rated maximum speed with the standard bearings: 10,000RPM). It’s a fully enclosed industrial induction motor, so it should be robust and have a long lifespan in the presence of dust and swarf. I’m surprised more people don’t go down this route to put a variable speed motor on their Taig mills; it seems almost ideal if you’re happy with the standard Taig spindle. Eventually I’ll probably put a similar setup on my lathe. The main drawback is that it’s fairly complicated to configure the VFD to get it to perform optimally.

The VFD replaces the contactor (the glorified on/off switch bolted to the old motor) and I won’t be mounting the VFD on the side of the new motor, so that has reduced the weight on my Z axis a bit more.

The drive shaft key had been removed from the old motor to avoid having to cut a keyway in the pulley (not me: it was like that when I bought it). Although I’ve never had a problem with it, that felt like a bit of a bodge so I used a square needle file to cut a keyway in the pulley for the new motor:




The VFD can take an analogue voltage input to set the speed, and the machine controller I’m using can output a PWM signal when the G-Code commands it to switch the spindle on, so at some point I’ll build an interface board to connect the two together. For the time being I’m just manually controlling the motor using the VFD’s front panel. I wrote a little Python script to figure out what pulley ratio and motor frequency to use to get a particular spindle speed.

If there’s anything in the new motor setup I’m not totally happy with, it’s the Gates miniature V belt drive. When it’s working well it’s adequate for the task, but the belts are stupidly expensive and seem to be quite easily damaged. After having a couple go bad recently, I have taken to slackening the motor mounting bolts and releasing the belt tension every time I change pulley ratios and when I’ve finished using the machine for the day.

The new motor did show up a problem with the CNC controller. It seems that VFDs put out electrical interference. Lots of it. I had previously found with the old motor that I would sometimes get a spurious E-stop input when I switched it on or off. With the VFD I could get them at any time the motor was running, and also the steppers sounded ‘lumpy’ and kept randomly stalling. I found that if the spindle motor was running while an axis was doing a rapid move and I picked up the VFD and moved it close to the CNC controller box, the stepper would inevitably stall. Conversely if I moved the VFD and controller as far away from each other as the cables would allow and placed a metal bucket over the VFD, the problem went away. The permanent cure turned out to be that I needed to connect the 0V rail of the Arduino in the controller to the chassis earth star point. Since then it’s been behaving itself.

I’ve fixed quite a few bugs in Handwheel and implemented a few new features, and it’s running pretty well for me now. I’ve been doing the development on an old dual-core Macbook Pro and my workshop machine is a quad-core Raspberry Pi 2, and it performs great on both of those. Handwheel is divided into several threads, structured in such a way that as long as your computer has at least two processor cores (ideally four), the overhead of updating the GUI shouldn’t slow down a file send. I decided to check that it worked OK on a standard (not overclocked) single-core Raspberry Pi 1: nope, the experience was dreadful because the rapid GUI updates were chewing up all the CPU and slowing the whole system to a crawl. Several hours of optimising later, the experience on the Pi 1 is now acceptable and faster machines like the Pi 2 are even snappier. I think it will run pretty well on the new super-cheap Pi Zero too, which has the same processor as the Pi 1 but is clocked something like 40% faster. I’m hoping to finally put out an initial public release over the Christmas break.

I also got side-tracked into working on a problem that was causing data corruption between the computer and the Arduino at 250000 baud: I tracked it down to the firmware on the Atmega 16U2 processor they use as a USB to Serial bridge, and came up with a fix.

I’ve started learning how to take a design and make it into a set of instructions to control the machine. This is called Computer Aided Machining, or CAM for short, not to be confused with Computer Aided Design (CAD), which is basically a way to do engineering drawings on a computer. There are a huge number of CAM programs available, with a wide range of capability and maturity at prices from free up to thousands of pounds per year. After quite a bit of research I settled on a program called CamBam, which seems to be the most capable option I could afford. Awkwardly, because it’s a Windows program and I’m a Mac user I ended up also buying a second hand Windows XP license and installing it on a virtual machine. A little bit clunky, but it seems to be working OK.

There’s a pretty steep learning curve to CAM, particularly as I haven’t got a great deal of experience or any formal training on manual milling machines. For example when milling manually I would guess at a spindle speed and depth of cut, then adjust the feed rate by gut feel based on the sound of the tool (and perhaps how much smoke/steam was coming off the cutting lubricant!). With CNC you calculate the parameters in advance, configure them in your CAM program, double and triple check everything, and hope you didn’t put a decimal point in the wrong place and accidentally command the machine to stab the end mill into the work at 500 miles per hour. Actually, I’ve found with my initial experiments with routing wood using carbide bits, the limiting factor in how fast I can cut is my machine’s top speed (1400 mm/min, 55″/min), which is set by how fast the Arduino running grbl can emit step pulses (30KHz). As a result I’ve been turning down the spindle speed to something like 5000-7000 RPM in order to get a decent chip load and avoid making a lot of dust and heat. It isn’t really a problem for what I’m doing with the machine, but I must admit the part of me that wants to optimise everything has been considering overclocking the CPU in the Arduino to increase the maximum pulse rate!

The first thing I milled under CNC control was to flatten the top of a plywood spoilboard I made that bolts to the bed of the machine and allows me to easily screw down work and to cut through it without damaging anything important (when it gets too chewed up I can re-surface it, and when it gets too thin I can glue a new piece of plywood on to build the height back up).

The second thing I made is a wooden stamp of my maker’s mark. I started out by designing it in Inkscape. It began as a Futura ‘A’, then I tweaked the proportions a bit and added the arms and the outer border:


Next I exported it to DXF and imported it into CamBam. This screenshot shows the finished design including all the generated toolpaths. Figuring out how to do this is what took the bulk of the time spent on the project. Now I have a better idea what I’m doing I could do a second one fairly quickly:


While developing the CamBam design, it was extremely useful to be able to be able to simulate what would happen if I was to machine the part. For this I used an open source program called CAMotics (formerly OpenSCAM). It’s still a little rough around the edges but I found it super useful. As soon as I actually make some money from CNC I’ll be sending a donation their way.


Before making the real part, I did a test run using some scrap pine. Good job I did, because it turned out I’d set the stepper motor acceleration parameters a bit too high and part way through the Y axis started losing steps:


After dropping the acceleration by 20% I completed a second, successful, test run, then made the real part from end-grain limewood:


I’m pretty pleased with the end result. The wood soaked up the ink like a sponge; in hindsight it might have been a good idea to seal it with a coat of shellac or something before using it.


Octagonal English Earrings

I made a new design of silver concertina earrings; an octagonal 12-button English layout, slightly larger than the previous two at 25mm across flats. The reason I reduced the number of buttons and also increased the overall size a little was so that I could fit in a more complex fretwork design than on my earlier hexagonal English design:

(Don’t try to make sense of the layout from the picture; I accidentally posed them the wrong way round.)

The pair in the picture was a birthday present for my mother. If anyone wants to commission a pair, I’ll do them for £75 + postage. Drop me an email via the contact address to reserve some time in my schedule, the sooner the better because I’m very busy with other projects.

As you might have spotted from the video in my previous post, this is the first design where I drilled the pilot and button holes using CNC. I still cut all the piercings by hand with a very fine jeweller’s saw because the details are extremely small with lots of sharp corners. Obviously all the soldering, polishing, and making the ear-wires are done by hand too.


CNC Mill Progress

A little over ten years ago I bought a manual Taig milling machine with the intention of converting it to CNC. For reasons too boring to go into here, the conversion got postponed indefinitely. It remained in its manual configuration and saw little use (probably a couple of hours a year, mostly as a glorified drill press with XY table).

About a year ago I finally started collecting parts to do the conversion: Chinese stepper motors, switch mode power supply, driver boards and an Arduino Uno clone via eBay, and motor mounting kits from Lester Caine at Model Engineer’s Digital Workshop.

I was happy with most of the components, except for the Chinese driver boards which turned out to be a hopeless waste of money (basically, a design flaw means they only ever run at 30% power level, and poor quality components mean they would probably quickly burn out if you hacked them to work at full power). If money was no object I would have bought Gecko drivers, but they are very pricey here in the UK so I searched for an alternative and found these THB6064AH kits. I went for the ‘special’ model because it’s easier to mount to a heatsink. It took me a couple of hours to assemble three of them, and I’m very pleased with the design and performance of them.


I mounted them on a couple of large P3 Xeon heatsinks salvaged from old servers. I’ve left space for a fourth driver board in case I ever decide to add a rotary axis for gearcutting or something. They barely get lukewarm.


The stepper motor controller is a cheap Arduino Uno clone running grbl, which interprets simple g-code and generates step and direction pulses for up to three simultaneous axes. I needed a way to connect it up to the drivers and various switches so I knocked something up using a (poorly designed) prototyping shield and a bit of stripboard:


This is currently working fine for interfacing to the drivers, but I think I need to add opto-isolation to the inputs because I’m currently getting an occasional spurious e-stop when I switch my spindle on or off. Luckily it hasn’t happened at any other time and the program execution is always paused when I start/stop the motor, so it hasn’t caused an issue yet. It’s more likely to cause a problem once I get around to fitting limit switches. At some point I’ll probably switch to using an Arduino Nano soldered to a custom PCB with full isolation.

I mounted the power supply, drivers and Arduino inside a steel box that started out as a dead fan heater (sorry for the punny name):


I’ve never liked the sound of the standard Taig stepper motor couplers so I decided to go with the MEDW Oldham coupler kit instead. When I took off the handles to install them, I soon realised it was going to take a bit more work than I was expecting. First I had to carefully shorten the stepper motor shafts so the couplers fit as close as possible to the motor:




Then I had to make something to fill the space left between the nut and dial when you remove the ball handles (which you have to do because they won’t fit inside the coupler tubes). Theoretically I could have just sawn and filed the arms off the handles, but that seemed like a bodge too far for me, so I turned up some spacers from 1″ aluminium bar instead:


Installation and adjustment of the couplers was slightly fiddly, but I’m very happy with the results: no backlash and more than strong enough to handle the motor torque.


The most time-consuming piece of the project by far has been that I got side-tracked into writing my own user interface software called Handwheel. Turns out it’s a huge amount of work to implement a good CNC UI, but I’m pretty pleased with what I’ve ended up with. I’m going to release it as Open Source once I’ve cleaned up a few more rough edges and written some sort of user manual. More on that later.

To finish off this rather long and rambling post, here is a video of the machine in action. This was slightly nerve-wracking for me, as it was the first time I’ve programmed a CNC machine, the first time I’ve used the mill to cut metal since the CNC conversion, and the first time I’ve used Handwheel on a real part. Just to up the stakes a little, the workpiece was £12 worth of sterling silver! Fortunately everything worked perfectly.


Star Punch

I made an eight-pointed star punch today for decorating an item of silver jewellery that I have been commissioned to make. Although it is not directly related to concertina-making, I thought it might make an interesting article for the blog anyway.

I made the punch from silver steel, which is a high-carbon tool steel with some chromium in it that comes as precisely-ground round bar stock in a range of standard diameters. It’s pre-annealed so it’s pretty easy to work with hand tools and machines before hardening. The bar I happened to have in stock was 3/8″ diameter.

I started by grinding a double-angled cone on the end:

Then I filed the facets of the punch using square and triangular jewellers’ files under magnification and plenty of light, resting the punch in a corner of my bench peg.

I must admit this took me two attempts. The first time I completely messed up the relief angles (it produced a circle of eight triangles with no centre), so I had to file it back to a blank cone and start again!

I lightly punched  a piece of softwood to check how it looked prior to hardening:

I didn’t want to have to try to grind firescale off the working end of the punch after hardening, so I coated it in a thick paste made from a stick of chalk mixed with a drop of water. The idea is that it prevents oxygen getting to the surface of the steel so it doesn’t corrode despite the extreme heat. My research on what substance to use for this purpose turned up a wide range of possibilities from specially-formulated industrial coatings through cockroach poison (boric acid) to something that sounded like a recipe for white bread. I had some chalk on hand and I saw it recommended in more than one place, so I thought it was worth a try.


It wasn’t worth firing up the forge for such a small job, so I simply placed the punch in a (metal) bucket of dry coke and hit it with a propane torch. The coke quickly heats up and reflects heat back at the work.


Here’s where I made my second mistake. In the heat of the moment (literally) I forgot that you are supposed to quench the tool by lowering it gently into the water tip-first so as to minimise stress and risk of cracking. Instead I thought “must cool it as quickly as possible,” grabbed it with the tongs, and randomly dunked it into the bucket side-on. This resulted in a crack along the length of the shank, luckily not reaching all the way to either end.

I tempered it by heating the shank in a spirit flame until the straw colour reached the sharp end. This differential tempering makes the end you hit with a hammer much softer and tougher than the end that cuts into the work, which is a desirable quality in a punch.

The finished punch. The anti-scale chalk paste did a reasonable job I think; all I did after hardening was to clean it off with a wire brush:

Here you can see the crack most of the way along the shank. The tool seems to be working OK regardless though:

Finally, the proof of the punch is in the marks it makes. I haven’t tried it on silver yet, this is a piece of scrap aluminium. I rather like the slight unevenness of the points, and it’s nice how you can vary the size of the star by the strength of the hammer blow:


Glue Pot

I’m a fan of hot hide glue for musical instrument work. There’s no need to rehash the pros and cons of HHG versus liquid hide glue and modern synthetic glues like PVA; suffice it to say that it has been used successfully for millennia and I think there are very good reasons to continue using it for certain things including high-quality instruments.

To use HHG, you first dissolve it in water to make a gel, then heat it to about 60C (140F) to melt it. Too cold and the open time is reduced; too hot and it ‘cooks’, compromising the strength of the glue joint. Traditionally cast iron or brass double-boilers were used on a stove, an alcohol burner, or a charcoal brazier. Around the turn of the previous century somebody invented an electric glue pot, which used a thermostat and a heating element to maintain the correct temperature with much less fuss and risk of overheating the glue.

I know of two manufacturers still making electric glue pots. Hold Heet in the USA makes fairly large pots that are probably best suited for antique and reproduction furniture work. Herdim in Germany make smaller pots that seem to be targeted mainly at luthiers. From my research it seems that the Hold Heet pots are expensive in the US and very expensive in Europe, while the Herdim pots are expensive in Europe and very expensive in the US. Second hand electric glue pots of either brand never seem to come up on eBay in the UK, and it wouldn’t have made financial sense to import a used Hold Heet from the US and buy a 240V-110V transformer to power it. I strongly considered buying a new Herdim, and if money was no object that’s probably what I would have done.

If you search Google for alternatives to commercial electric glue pots, people have made them from various kinds of electric coffee pots, baby bottle warmers, old cast iron glue pots on electric hotplates, etc.

My new one is made from a mini deep fat fryer (0.5 litre oil capacity). I wasn’t happy with its built-in mechanical thermostat (it had about 15C of hysteresis and would probably have needed frequent adjustment), so I have instead hooked it up to a cheap Chinese PID temperature controller (a Rex C100 clone) with solid state relay output. This works remarkably well, regulating the temperature of the water bath to within a degree of the set temperature by pulsing a little bit of power into the heating element about once a second. I found I needed to set the water bath temperature several degrees higher than the desired glue temperature.


I knocked together a simple wooden box to hold the controller and the SSR, and a Perspex cover to hold the inner pot in position and prevent the water in the outer bath evaporating away. The light blue gaskets are made from two-part mouldable silicone rubber. The inner pot is a 0.25 litre Stewart Sealfresh screw-top food container. I have several of them and I cut a brush-sized hole in one of the lids to reduce the rate at which water evaporates from the glue in the pot.

I also made several glue brushes in various sizes by whittling the handles from green wood and binding hog bristles to them with string. The smallest brush I made by boiling the end of the stick in water for five minutes, then smashing the fibres apart with a hammer: