Guitar Pickup Winder

[tauzero]

4 hours ago, rwillett said:

Thats the intention. It needs some sort of directed randomisation, which does sound a contradiction.

 

I'm slightly nervous about how random things are on an Arduino. For those of you who play with IT, true randomisation is quite difficult. No idea how the randomisation works on this sort of low level kit. Suspect it's not random at all though.

 

The documentation for the random() function says that it's a pseudo-random sequence, and if you use the randomSeed() function with the same number, you'll get the same sequence (presumably it's always the same sequence, only the starting point differs). I suspect that it's a sequence intended to avoid repetition so it will give a pretty even distribution, which I presume is your desired outcome. You could assess its suitability just by generating a list of random numbers in your desired range and seeing if they would be suitable.

[tauzero]

1 hour ago, rwillett said:

I always liked this method of generating randomness

 

https://en.wikipedia.org/wiki/Lavarand

 

and it's successor

 

https://blog.cloudflare.com/lavarand-in-production-the-nitty-gritty-technical-details/

 

Rob

 

ERNIE's first iteration used the noise generated in a neon cold-cathode tube as the basis for its random numbers. https://ernie.virtualcolossus.co.uk/ernie.html

 

The lava lamp one featured in an episode of NCIS where a wall full of lava lamps gets dumped on the floor. Just a random fact.

[rwillett]

I haven't posted in a while, not because I've not done anything but I've been trying to understand the various menu libraries that Arduino systems can use.

 

I spent a lot of time trying to work through the ad-hoc, hand built, unsupported, poorly documented, over the top, over complicated, under specced, version tied and generally not fit for purpose menu libraries before giving up and writing my own. I'm sure there are good ones out there, but trying to find a lightweight library (Arduinos don't have much code space) that actually works, is properly documented and works on my simulator is harder than doing it yourself.

 

So I did it myself. I currently have nine options that need to fit on a 20x4 LCD screen. You can see below that the screen is showing options 1-4. As I am an ex C-developer, everything starts at row 0 which is right and proper. The green and yellow buttons will be end stops that I haven't wired yet. I have two digital inputs left but can't be bothered yet.

 

 

int noRootMenuItems = 9;
SimpleMenu_t rootMenu[9]  = {
                    { "0Run                " ,  { INTEGER , { 123 }} } ,
                    { "1Number of Loops    " ,  { INTEGER , { 123 }} } ,
                    { "2Winder1 Steps      " ,  { INTEGER , { 123 }} } ,
                    { "3Winder2 Steps      " ,  { INTEGER , { 123 }} } ,
                    { "4Bobbin Width       " ,  { DOUBLE , { 1.23 }} } ,
                    { "5Wire Width         " ,  { DOUBLE , { 3.21 }} } ,
                    { "6Set Left Edge      " ,  { DOUBLE , { 7.89 }} } ,
                    { "7Set Right Edge     " ,  { DOUBLE , { 9.87 }} } ,
                    { "8Stepper Direction  " ,  { DIRECTION , { CLOCKWISE }} } ,
}; 

 

The above is the rootMenu setup, the numbers at the start of the char* will go, it's just for me to check. Eventually this struct will be enlarged to handle functions, but I'm struggling with how Arduino compilers declare and pass arrays of structs around. A simple C compiler would let me pass a pointer and thats it, the Arduino compiler is not playing ball and keeps refusing to compile what I think is valid code, a pointer to an array of structs. I may bring out the big hammer shortly and declare a void * to teach it a lesson. I've written compilers so this should be easy, but I must have forgotten something important or the compiler is a lot stricter.

 

Anyway, each line has a type attached to it, INTEGER, DOUBLE and DIRECTION. When you select a line using the 'D' character, it moves to a sub-menu and allows you to edit the value thats passed. I've written the code to edit integers, I'll use that as a basis to input floats/doubles and the last one will be to change the wiring direction, CLOCKWISE or ANTI_CLOCKWISE. The aim is to parameterise everything so adding new features is just a new line in the menu declaration. We'll see how that goes

 

I have all the looping sorted out, keypad entries seem to work, but no idea if they debounce in the real world.

 

I also have printed most of the parts for the winder, including my first simple bobbin. I have most of the winder parts sitting on my desk, but am away skiing from Thurs to Monday so nothing is going to move forward for a few weeks yet.

 

If anybody has any spare broken pickups they are happy to provide, it would be great to see how they are made and understand more about each of them. I don't really want to buy lots of working ones to break apart.

 

Thanks

Rob

[benh]

Just found this topic a week too late, just slapped together a much more, shall we say, rustic version of a pickup winder - really looking forward to seeing how it progresses!

[tauzero]

ESP32s give you much more space to code in, plus they're probably cheaper than Arduino Unos and also let you have wireless and Bluetooth connectivity. I love them.

[rwillett]

I've got a Mega256 sitting doing nothing. More pins so rewired my code to use that. Stuck a CNC shield on top with some decent drivers. 

 

Still working on the menu code. As expected it's 10x the effort of the actual stepper motor code. 

Edited January 17 by rwillett

[Beer of the Bass]

This is all so much more sophisticated than the pickup winding setup I cobbled together a few years ago. I can't find the pictures, but it was a little hobby drill lashed to a plank with cable ties, a speed controller board off ebay, hand fed wire and a couple of chunks of wine cork over a dowel on the side of the plank to set the limits of the wire travel. I'm still using a couple of the pickups I did, on my fretless bass and a electric guitar, so I guess it got me there!

[tauzero]

18 hours ago, rwillett said:

I've got a Mega256 sitting doing nothing. More pins so rewired my code to use that. Stuck a CNC shield on top with some decent drivers. 

 

Still working on the menu code. As expected it's 10x the effort of the actual stepper motor code. 

 

I've just knocked up a CV sequencer (awaiting a breadboard to accommodate the ESP32 a bit better than the lashup I currently use) which uses a rotary encoder for parameter selection and input. That might be an alternative to your menuing system using a keypad.

[rwillett]

Since my last update, I have now written most of the software and it works (for some value of 'works') with the simulator.

 

 

I've now just about reached as far as I can go with the simulator, there's probably going to be some hardware issues to resolve now and the only way to do that, is to get the hardware built and the code uploaded.

 

I'm now going to use a Mega 2560 Arduino rather than a UNO as it has more pin outs. I'd literally run out of pins on the Uno and was struggling to put all the bits I wanted. I had a Mega 2560 sitting around doing nothing so thats now the base. It also gives me far more code space to program in and save stuff in.

 

 

I've got a CNC Shield as that sits on top of the Mega 2560 which saves vauable real estate, I'll only use two stepper drivers at the moment, though I can use four. I might use a 3rd for tension control of the wire but not yet. The CNC shield takes 12-36V and a bench power supply is on order (60V/10A). Stepper motors can take a lot of amps, I suspect I won't need this sort of output power, but it's nice to have a lot of head room just-in-case.

 

 

The menu interface is quite good (even though I do say so myself), I can create sub-menus but so far haven't needed it. I can use a 4x4 membrane keypad to enter integers, floats and select from lists.  I'd love to be able to find a blank 4x4 keypad, or one where I can change the keys to be what I want them to be. I'll have to use sticky labels at the moment, but as these keypads are about £2 each, not a problem.

 

So next steps are:

 

1. Working out how to do microsteps on the stepper motors. The reason for this is that there is a ratio between the number of loops that the main pickup winder motor makes and the number of turns that the lead screw turns to ensure that the pickup wire moves across the bobbin so that the pickup wire is evenly placed. I know some people might want randomisation but thats V2 of the firmware.  These microsteps appear to be controlled by the CNC shield and setting jumpers (!).  if I can move from 400 steps per singe revolution to 800 or 1600 steps, then I can get the ratio between the two motors to be higher and closer to an integer. Thats good

 

2. Putting everything onto a thick A3 plywood base. I wanted to use a larger base as that reduces the angle that pickup wire moves through as the bobbin is turned. If the bobbin is very close, then the angle subtended gets closer to 90 degrees. Imaging your finger wagging at 10 times a second going through 90 degrees vs at 45 or 30 degrees. I might be worrying about nothing but lets see. I have printed most things including guides to position everything accurately.  I have all the hardware to do it, just been away and its bloody cold in the garage here in North Yorkshire.

 

3. Work out how to build an Arduino dev environment on a M series Mac. I have a simulator but not a clue how to actually prgram an Arduino on a Mac. Last time I did this, I used a Windows laptop. I might still have to yet. I really want to built it around Emacs but suspect thats not going to fly.

 

4. Get it installed and get the firmware properly debugged.

 

5. I have a couple of cheap bass pickups I've brought, so will gently take them apart and see if I can replicate how they are made. My knowledge of magnets is limited but I'll start reading up over the next week or so.

 

6. Buy some 42AWG wire. I'll start with the thicker stuff first.

 

7. Work out how to fit bump stops in case everything goes out of control. There are pins on the CNC shield for this, its mounting them thats the challenge.

 

Thats probably enough to go on for the next week or so. I still have to start work on my 3d printed bass but thats on hold whilst I get this working.

 

Rob

Edited January 24 by rwillett

[choob.squeemer]

This is great - but I don't think I would want  any Random stuff - cos if on the 29th go you 

wind the pickup that is utterly magnificent  - hot - mellow - rounded - just perfect , 

well how do you duplicate it .  I reckon with this much control you can experiment where 

you want the "scatter" winds - 3 layers then a scatter then 4 layers and 2 scatters .......

 

Might take a long time , but sooner or later you will arrive at a formula which you know 

will sound Brill - and make the pyramids rotate and Machu Pichu  vibrate 

[sandy_r]

+1 for starting with regular wind version(s) (which can then provide a performance  'reference' for variants, eg scatter-wound)

 

AFAIA scatter wind is a method which reduces the self-capacitance of a coil - the magnetic properties are likely to be  largely unaffected (apart from reducing the no. turns which will fit in the same enclosing volume, and hence the signal output level - so compensate with extra turns, but bulkier coil?)

 

The possible benefit of using an alternative wind pattern, such as scatter or basket-weave etc, would be to reduce top-end signal loss due to the self-capacitance. Each layer of the whole coil would need to be wound using this new scheme, to make significant difference in the capacitance - only experimentation will tell if the change in capacitance is worth a bulkier coil

 

Another experiment to try, which might make a difference to the coil windings from an electrical field PoV, is to wind each layer in the same forward direction (ie not just recirocating windings), leaving a gap between turns then returning to 'start' end in one stretched turn and offsetting the next layer to align with the gaps between turns in the previous layer

 

...just my €0.02

 

Edited January 25 by sandy_r

[rwillett]

4 hours ago, choob.squeemer said:

This is great - but I don't think I would want  any Random stuff - cos if on the 29th go you 

wind the pickup that is utterly magnificent  - hot - mellow - rounded - just perfect , 

well how do you duplicate it .  I reckon with this much control you can experiment where 

you want the "scatter" winds - 3 layers then a scatter then 4 layers and 2 scatters .......

 

Might take a long time , but sooner or later you will arrive at a formula which you know 

will sound Brill - and make the pyramids rotate and Machu Pichu  vibrate 

 

Randomisation is not in V1. Scatter winding appears to be how Fender wired these things, basically hand wound trying to make it near but with imperfections. At the moment my focus is getting anything down, neat is a bonus.

 

3 minutes ago, sandy_r said:

+1 for starting with regular wind version(s) (which can then provide a performance  'reference' for variants, eg scatter-wound)

 

AFAIA scatter wind is a method which reduces the self-capacitance of a coil - the magnetic properties are likely to be  largely unaffected (apart from reducing the no. turns in the same enclosing volume, and hence the signal output level - so compensate with extra turns?)

 

The possible benefit of using an alternative wind pattern, such as scatter or basket-weave etc, would be to reduce top-end signal loss due to the self-capacitance. Each layer of he whole coil would need to be wound using this new scheme, to make significant difference in the capacitance - only experimentation will tell if the change in capacitance is worth a bulkier coil

 

Another experiment to try, which might make a difference to the coil windings from an electrical field PoV, is to wind each layer in the same forward direction (ie not just recirocating windings), leaving a gap between turns then returning to 'start' end in one stretched turn and offsetting the next layer to align with the gaps between turns in the previous layer

 

...just my €0.02

 

 

If the winding pattern can be reduced to a function, then everything is easly possible. What you are describing above is what I would call typewriter winding (just made that up), you wind to the end and then instead of winding back carefully and slowly, you move the wire back quickly to the start and then wind across.

 

There are loads of ways to skin this cat, I could do randomisation and capture this in a stream of data that is sent off the Arduino. I could inject random data into the Arduino from an external source that is precaptured. I have a TrueNas file server here with a lot of free space, so capturing  data and storing it is not a problem. There's a number of options to consider but I'm not ready to do that yet.

 

I just want to get something working and try it out. I am going to have to make a test bench guitar for these pickups (6 and 4 string) so might quickly take the 3d printed guitar design and just print the absolute essentials (bridge, pickup and neck modules) and attach it to a plank of wood. Another thing to add to the list

 

Rob

[BigRedX]

When you get around to introducing scatter-winding would it be possible to save the stepper motor movements for each "random" option as they are used? That way if you do come up with a pickup that is noticeably better than the others due to the winding pattern it would be possible to recreate it. Or at least see how much the winding pattern affects the sound on two supposedly identical pickups.

[rwillett]

10 minutes ago, BigRedX said:

When you get around to introducing scatter-winding would it be possible to save the stepper motor movements for each "random" option as they are used? That way if you do come up with a pickup that is noticeably better than the others due to the winding pattern it would be possible to recreate it. Or at least see how much the winding pattern affects the sound on two supposedly identical pickups.

 

Everything is possible. The easiest way is probably to create a file externally that has all the motor actions, basically the same as a gcode file for 3d printing. The looping motor just loops, an external source with a lot of memory (Raspberry Pi comes to mind), connects to the USB port and sends down the stepper motor commands for the lead screw motor to move it along the X axis, perhaps a step or two extra left or right or take a step out. The most difficult thing would be the connection from Pi to Arduino and keeping the timing right.

 

Other options include making the bobbin with angles so that the top of the bobbin could have a wider diameter to the bottom or the middle is thinner or anything really. As I'm 3d printing the bobbins, that fairly easy. No idea if it makes a difference though. Perhaps this is V3.

 

Rob

[rwillett]

Managed to get the IDE working on my Mac. Didn;t have to download any drivers from an obscure Chinese web site. That was never going to happen

 

Copied all the code across to the IDE, compiled it, uploaded it and blow me down with a feather it works.

 

No drivers setup, but I now have working keypad, screen and Mega 2560. That was rather painless. Suspect the CNC shield and stepper motors will be a lot more hassle.

 

[Jackroadkill]

21 hours ago, rwillett said:

Since my last update, I have now written most of the software and it works (for some value of 'works') with the simulator.

 

 

I've now just about reached as far as I can go with the simulator, there's probably going to be some hardware issues to resolve now and the only way to do that, is to get the hardware built and the code uploaded.

 

I'm now going to use a Mega 2560 Arduino rather than a UNO as it has more pin outs. I'd literally run out of pins on the Uno and was struggling to put all the bits I wanted. I had a Mega 2560 sitting around doing nothing so thats now the base. It also gives me far more code space to program in and save stuff in.

 

 

I've got a CNC Shield as that sits on top of the Mega 2560 which saves vauable real estate, I'll only use two stepper drivers at the moment, though I can use four. I might use a 3rd for tension control of the wire but not yet. The CNC shield takes 12-36V and a bench power supply is on order (60V/10A). Stepper motors can take a lot of amps, I suspect I won't need this sort of output power, but it's nice to have a lot of head room just-in-case.

 

 

The menu interface is quite good (even though I do say so myself), I can create sub-menus but so far haven't needed it. I can use a 4x4 membrane keypad to enter integers, floats and select from lists.  I'd love to be able to find a blank 4x4 keypad, or one where I can change the keys to be what I want them to be. I'll have to use sticky labels at the moment, but as these keypads are about £2 each, not a problem.

 

So next steps are:

 

1. Working out how to do microsteps on the stepper motors. The reason for this is that there is a ratio between the number of loops that the main pickup winder motor makes and the number of turns that the lead screw turns to ensure that the pickup wire moves across the bobbin so that the pickup wire is evenly placed. I know some people might want randomisation but thats V2 of the firmware.  These microsteps appear to be controlled by the CNC shield and setting jumpers (!).  if I can move from 400 steps per singe revolution to 800 or 1600 steps, then I can get the ratio between the two motors to be higher and closer to an integer. Thats good

 

2. Putting everything onto a thick A3 plywood base. I wanted to use a larger base as that reduces the angle that pickup wire moves through as the bobbin is turned. If the bobbin is very close, then the angle subtended gets closer to 90 degrees. Imaging your finger wagging at 10 times a second going through 90 degrees vs at 45 or 30 degrees. I might be worrying about nothing but lets see. I have printed most things including guides to position everything accurately.  I have all the hardware to do it, just been away and its bloody cold in the garage here in North Yorkshire.

 

3. Work out how to build an Arduino dev environment on a M series Mac. I have a simulator but not a clue how to actually prgram an Arduino on a Mac. Last time I did this, I used a Windows laptop. I might still have to yet. I really want to built it around Emacs but suspect thats not going to fly.

 

4. Get it installed and get the firmware properly debugged.

 

5. I have a couple of cheap bass pickups I've brought, so will gently take them apart and see if I can replicate how they are made. My knowledge of magnets is limited but I'll start reading up over the next week or so.

 

6. Buy some 42AWG wire. I'll start with the thicker stuff first.

 

7. Work out how to fit bump stops in case everything goes out of control. There are pins on the CNC shield for this, its mounting them thats the challenge.

 

Thats probably enough to go on for the next week or so. I still have to start work on my 3d printed bass but thats on hold whilst I get this working.

 

Rob

 

I don't know what any of this means, but if you can make some decent pickups I'll give you a shout for future projects!

[rwillett]

 

16 minutes ago, Jackroadkill said:

 

I don't know what any of this means, but if you can make some decent pickups I'll give you a shout for future projects!

 

It means that I have the basis of the controlling software for a winder. As far as I know, there is no code around for doing this, so I wrote my own, I know that people use electric drills and similar to do windings, but I wanted anything I did to be defined. I'm not very good mechanically so Id rather write software that does things for me, as thats easier for me.I have nothing but admiration for people who can make things creatively.

 

So if I can make this work (note the word if) and they sound OK, it's repeatable. If I want to make a pickup with 10,000 loops of AWG 44 wire or 5,000 loops of AWG 42 wire (thicker than AWG 44), it's a matter of changing the Wire Width parameter and the number of loops. The software should do the rest automatically. If I want to make a thicker bobbin or thinner bobbin, thats just a parameter change.

 

I now have the CNC shield setup, added the stepper drivers, DRV8225, set up 1/8 microstepping, I have a 60V/10A bench power supply but now I just realised I don't have any jumpers to fit the stepper motor wiring to the CNC shield. These are DuPont connectors. I can make them, I even have a kit, but I hate making them, Sod it, Mr Amazon to the rescue. £10 for 6 of them, I only need two (at the moment), so a bargain.

 

Everybody uses their own connectors

 

 

Rob

[Richard R]

Excellent!

Next step, to remove all variables and uncertainty in tone, is 3D printed plucker and fretters  driven by fast motors under computer control responding to MIDI.

That takes out that pesky player who keeps saying  "the tone is in the fingers"

 

Seriously though - great stuff and very, very  impressive!

Edited January 26 by Richard R
Typo

[rwillett]

59 minutes ago, Richard R said:

Excellent!

Next step, to remove all variables and uncertainty in tone, is 3D printed plucker and fretters under driven by fast motors under computer control responding to MIDI.

That takes out that pesky player who keeps saying  "the tone is in the fingers"

 

Seriously though - great stuff and very, very  impressive!

Odfly enough I did think about that but a simpler model. I have rejected the idea though

[rwillett]

So after putting up the bathroom mirror, changing lights,  sorting out the clock in my daughters bedroom, taking dog for walk, paying my tax to HMRC, I finally got around to doing some work on the winder.

 

I now have a case for the Arduino, screen and keyboard. I have ordered a mechnical 4x4 keyboard to replace the membrane one but got a 3x4 by mistake so a new one is on it's way. Whilst the membrane keyboard works, it's vile to work with. The gorilla tape just holds it down as I have no idea what the glue would do to it, probably dissolve it.

 

I've attached the lead screw mechanism to the plywood base and am working out how to get the motor that loops the bobbin as square as possible, hence the set square.

 

I would have this all powered up, but the bench power supply came with no cables apart from the mains supply, it needs 4mm banana plugs and do I have any? No!

 

The lead screw mechanism was a little tight as the screws that mount the linear bearing on each arm, were probably out by 0.25mm. Not a lot but thats enough to put friction in the system. So I'm reprinting that with a fraction more leeway in the screw holes to allow it to settle down. This lead screw mechanism is critical and it needs to be as smooth as possible as it moves back and forth as the bobbin rotates.

 

The area at top left is reserved for a dedicated PSU once I work out what the power consumption is.

 

 

Thanks

Rob

[rwillett]

Whilst I've not made a lot of progress over the weekend, beyond what I put above, I have started to understand more about the Vref potentionmeter on the little driver sticks.

 

These drivers slot into the CNC board and controll the stepper motors. You can control the stepper motors directly from an Arduino but its better to control the drivers and let them control the stepper motors as thats what they are designed to do.

 

The Arduino was designed to allow physical components called shields to slot on top of the board. These shields provide additional functionality such as Ethernet, WifI, Real Time Clocks, stepper motor drivers, connections to sensors and so on and so on.

 

I'm using a sheild called a CNC shield, see below which is really just a simple way of adding in stepper motor drivers which are the four chips around the side.  CNC Shields are about £2-£3 each and are made in bulk. CNC shields have no programming capability, they just present a simple and uniform way to get stepper motors connected.  You access the stepper motors using a standard and uniform coding libraty as well on the Arduino. So instead of making complex calls to pins and timers, all of that complexity is abstracted away. Thats the theory anyway

 

 

However to get the stepper drivers to work, you have to do some low level coding to enable the stepper motors which appears in documentation that is hidden in a file cabinet in a cellar in the local planning office with the stairs removed, the lightbulb out and a sign saying "Beware of the Panther" on the door.

 

It's only two lines of code, but it's an important two lines of code as nothing works without it.

 

Once thats done, you can connect your PSU up to the blue terminals on the bottoms left. The input is 12V to 36V which is way beyond USB so a dedicated PSU is needed. I used this as an excuse to buy a decent bench power supply

 

 

This little beauty has one key feature. I can restrict the amount of current that can be drawn.  So I set it to 12V and 1A plugged it in and immediately the CNC board and stepper motors started to whine. It turns out that the DRV8826 stepper motor sticks have a little potentionmeter on the top that determines the maximum current draw. I initially set this to 1A so I didn't blow anything up. I didn't understand how this worked so decided to keep increasing the allowed current to see what happened. I got to 12V 4A and still nothing worked. Turns out that the potentionmeter was allowing around 8A per driver to be pulled and if I didn't have a decent bench power supply, I would have fried both driver sticks and the stepper motors. I only really noticed I had a problem when I touched the heat sink and promptly burnt my fingers.

 

So I put a multimeter on the potentionmeter centre point through a screwdriver and a crocodile clip, and the black to ground and turned Vref down to 0.6V from 3V. The whining stopped and the stepper motors started to work. Also the temperature dropped to non burning temperature.

 

The moral of this story is "Dont mess with things that you think you understand but really don't, because it could be expensive and you literaly do get your fingers burnt". Thankfully I didn't blow anything up, and the only damage was my singed fingers. I can see the burnt marks now.

 

Now I have the pots turned down, time to see if my Arduino coding actually will work and drive real stepper motors as opposed to simulators.

Rob

[Richard R]

27 minutes ago, rwillett said:

is hidden in a file cabinet in a cellar in the local planning office with the stairs removed, the lightbulb out and a sign saying "Beware of the Panther" on the door.

👍

That's the documentation department. Just next to the display department ( which is in a disused lavatory with a sign on the door saying 'Beware of the Leopard')

 

[Richard R]

31 minutes ago, rwillett said:

 I used this as an excuse to buy a decent bench power supply

 I do like your style!

[SpondonBassed]

38 minutes ago, rwillett said:

documentation that is hidden in a file cabinet in a cellar in the local planning office with the stairs removed, the lightbulb out and a sign saying "Beware of the Panther" on the door.

 

Nice nod to Douglas Adams there.  Cool.

[rwillett]

5 minutes ago, Richard R said:

👍

That's the documentation department. Just next to the display department ( which is in a disused lavatory with a sign on the door saying 'Beware of the Leopard')

 

Damm, misremembered that, though technically a panther is just a black leopard.

 

However I will claim that I didn't want to get sued by Douglas Adams estate rather than my memory is failing me

 

6 minutes ago, Richard R said:

 I do like your style!

 

I can now say that it saved me two £15 stepper motors, two DRV8825 drivers, a CNC board and an Arduino (though the latter is more difficult as it's more isolated), total of about £30. It's almost paid for itself.