When you build your own tools, you always have to be thinking one step ahead (a talent I have yet to master). My CNC router has an adequate spoil board on it now but before I get carried away and rip it to shreds with future projects, the 1st job off the ranks is to make a new spoil board. So rather than just talk about the fun and games, sit back and relax as you get a chuckle from this video.

I'm sure many men with sympathise with my plight for there are many times when my my wife causes a rapid and unexpected constriction in my pants. I find my wallet tightens every time she purchases the exorbitantly priced Kuih Lapis - the traditional Indonesian layer cake.

After some research, I now understand the high price - it is extremely labour intensive and consists of innumerable egg yokes and blocks of butter. A person's arteries harden just reading the ingredient list.

To ease my wallet tension, I thought about home baking the layer cake. A bit of recipe hunting showed that it is technically doable ingredient-wise but it is horrible when it comes to the baking process - it is built up one 3mm layer at a time. No way would I stand there loading and unloading an oven for an hour just to make one cake. Has to be a better way.

Further unsatisfactory Google-foo led me to believe that there is no quick and easy way to produce such a cake so in my mind, it is a perfect opportunity to implement some automation. As the sad TeslaBot is not available, it falls to me to come up with something myself. A Malaysian University research paper (view here) gave me the lowdown on how this is actually a difficult problem to solve. Sounds like the perfect sort of project for me to add to my "Incomplete Project" pool.

My goal is to produce a multistage rotating cake layering system. It automaticallly builds up the cake layer by layer just as is done in the manual process. The main feature is the baking side is hands free. The down side is that even making the batter is quite a job so definitely not worth the effort for one cake.

In my mind's eye, everything is doable and the one tricky bit is the baking. A failure here is a blocker for the entire project. To that end, it was time to open the constricted wallet and fork out for some quartz heat lamps to see what they can do. I purchased 2 of these 400W lamps:

That should be good enough for a start. I envisage needing 4 of them to get an even cake heat coverage. Now I am not sure how dangerous reflected IR light is so I do feel a tad uncomfortable testing it out (I did find one research paper where a woman suffered eye damage by continually looking into a heat lamp but that seemed an extreme situation [view here])

But still, I was somewhat cautious but not so cautious as to connect the earth wire to all the exposed metal bits... I could not find a ring terminal for the life of me. Sure, in a pinch I could have used an alligator clip but that has only just occurred to me...

It does get nice and toasty very quickly say about 20 seconds. I did wonder if the meter was reading the reflected light (ie the light source temp) and not the actual metal pan tempurature. You need to be aware of reflected vs absorbed light. So just to be sure, I tried again but took the temp with the lamp off and proved the meter reading was correct.

So far so good, now time to take it up a notch and get something a little more practical.

I'm still somewhat concerned about escaping light/thermal conductance causing unexpected/expensive damage so it's stacked layer on layer with some dark pans in the middle to absorb the light. Shiny stainless steel is not always one's friend.

And now for a temperature test with 2 lamps in the baking test rig.

Works a treat! That was enough for one day - not motivated to go shopping for heaps of eggs and butter. The batter test can wait for tomorrow... (and we all know tomorrow never comes)

A somewhat underwhelming result from a few days of designing the drive chain for the dust rollers. Gears that roll nicely when held together in your hands behave completely different when driven in their final location.

The main problems appear to be:
1 - Gears twisting because they are supported only on one side.
2 - Screws do not properly align parts as I assume the thread digs into the hole sides asymmetrically.
3 - A lack of rigidity holding the gear train in place. A millimetre makes the difference between gears slipping and them moving smoothly.

So back to the design phase. Not sure if my printed gears will work - they look like they could do the job. The next iteration should tell.

After many days of fiddling around trying to get bits and pieces all the right size, I finally have the 1st assembly designed and printed. Seems this project is going to take a long, long time.

It always takes a few attempts to get the bits to fit together - 2d printing is never exact enough

Eventually it all come together in a satisfactory manner. Now to ponder how to mount the gears and motor. The problem is I have no idea how fast the rollers need to rotate and motor selection is very limited.

It's not as though it's been a Ruslan Gatash "Strange Sunday" as I'm not a fan of what I can only assume is "jazz", but it still has been a Strange Vibe Sunday. One of those days when there are so many tasks to do that even achieving one of them leave you feeling empty inside.

As it seems to be such a day today. In the arvo I decided to knock up a quick small PCB to distribute 5 and 12V power to a number of terminals ie nothing more than a power bus board. Fired up the CNC router to do the PCB milling and betterfingers me dropped the 0.4mm milling bit a whole 30mm but that was enough to snap off the tip. Well it seems to be like that but I need to check later as my eyesight is just not that good.

While it was busy milling away with an expensive new bit, I decided I really needed to do something about the way the circuitry is all spread over the work bench.

It is a painful and depressing short circuit just waiting to happen. Now, I could go an build some sort of.. hang on, what is the word???... "en...clo...sure" for it but that is definitely some time off. For now, I should just attach the few loose compnents and put the board up against the wall and drape something in front of it to stop stray metal shard ingress. To do that, I have to move the peg board that is currently occupying that space...

No matter how long I think about a it, I just cant get away from the fact there is a fuse box right in the middle of the board. Oh well.

Unloading all the tools off the peg board was a risky time - just one dropped tool/hook would instantly release the magic smoke from the CNC countroller as it was busy milling the PCB. But in the end, no tradgety and the board is in place.

By then it was time to change the routing bit for a drill bit. All went well, the bit change homing procedure worked a treat. Hit the button for it to continue drilling and the head moved off to its 1st hole position. All good. Then is started lowering... waiting for the spindle to power up... no need to panic yet, it may just be getting to it's drill homing height... that position reach, still no spinning... come on, you can do it... almost on the board... come on... NOPE... crack.. crack crunch... head moves up for the next drill hole... Hit the abort button and that's me done for the day.

You can make out the fragment of the drill bit left in the board to the right of the shaft where it completely failed to press out a 1mm diameter cylinder of PCB.

Thinking back, this is the 1st time I've tried to mill a PCB since I made the toolchange code compatible with Fusion 360. Obviously I do not have a regression test suite... In no mood to debug gcode so </blog>

...Time passes...

Later that evening. Looks like the PCB tool change algorithm does not start up the spindle before it does the drill operation. Bugger.

G01[MOVE] Z-0.2000 F[SET_SPEED]508.0
G00[RAPID_POSITIONING] Z1.0000
M05[SPINDLE_OFF]
G04[DWELL] P3.000
(File:power-expansion.bot.drill.gcode)
(TOOLCHANGE: 1.0mm drill bit[0.0])
G00[RAPID_POSITIONING] F400.000 X38.903 Y68.114
G01[MOVE] F200.000 Z1.000
G01[MOVE] F75.000 Z-2.000
G01[MOVE] F200.000 Z1.000

Should not be too hard to fix

Finally after several month, the orders of critical parts have finally come in. Time to start on the next incomplete project...

Probably more truthful that any other title I could come up with.

It was time to test out the fan brushless motor with a home-designed 3d printed impeller. Checked out a few YouTube videos nad it seemed the design process was not too tricky. Of course, when it came to do it myself, I really didn't know what I was trying to achieve. Need to do more research and figure out what is the ideal form for an impeller to provide strong dirt suction. I know it is not going to be anything like a mains powered vacuum cleaner, but it still needs to get every bit of suction that is possible.

I was quite surprised how it worked out in the end. Only needed one attempt to print and it all fitted as designed. I will take a win whenever it arises.

The first power up test worked quite well. The RPM was around what it should be. The current draw was a lot less than the maximum so the motor was lightly loaded. That means I should be able to make a bigger impeller and improve the airflow.

And there ended the testing of the impeller. Sometimes you are a winner, other times a looser. I was aware of the warning on the ebay listing: (Attention: Do not change the rotation direction when the motor is working or it will damage the drive). From now on I will make sure the motor and controller are powered from the same power source. I can't rely on the separate 5v battery supply that powers off when the battery goes flat - as what happened in this case.

Have to wait now for the delivery of motor number 2 on the slow boat from China. I can use the time researching impellers and get a propery design modelled up now that I know it technically works.

Suffering badly from a lack of motivation. I see numerous YouTuber's building excellent robotic mechanisms and they have their 3d printer so finely tuned that it makes awesome results (of course, there could be editing in there as well).

On the other hand, I have several projects that are at the "useful working stage" rather than than complete, so as usual, the interest has waned. Think CNC router, 3D printer pimping and SpaceMouse keypad combo. All of them do the job I need but not in a polished/optimal/convenient way. Enclosures, end stops, calibration, circuit boards are all things that need to be done to round out the projects so until a shard of metal flies the 40cm from the drill press to the 500W power supply or exposed PCBs, the project management has gone into a hiatus.

Time now for a project just for the fun of it. Did a quick pricing of 9225 brushless motors and useful drivers resulting in a bout of depression. There is no way I can justify swapping my bank account balance for that class of components - and to make something impressive requires a lot of those components.

I lowered my expectations and have moved to a project that I've always wanted to do. Sure, these days the commercial products are quite good but I have ideas to improve the functionality in completely impracticable, ugly and inefficient ways - and is that not how we roll in the amateur shire?

Last night it was time to do some quick planning and order some components so they will eventually arrive on the slow boat from China. I'm not going to reinvent the wheel here so I've ordered components/modules that I know will be a pain to manufacture or just take up precious time. Watching a couple of YouTube teardowns of commercial products has initially saved me as I was going to completely over-engineer this project. Sure, it is going to need upgrading as time goes on but a good modular design should not make that a hassle.

Last night's pre-design spur of the moment purchases:

I won't question anybody's intelligence by asking them to guess what I am building. Sure I can buy something off the shelf but I want to seriously pimp this thing with mopping/scrubbing facilities, extendible boom to get right into corners, flexible brushing/vacuuming system to clean skirtings etc. This is an industrial beast, not a pretty consummer item - not a smooth shiny faring in sight. I think the groinal attachment will appropriately set people's expectations.

I find project development enjoyable right up until the point of having to connect things to a PCB. Wiring connections is a tedious chore. It's all easy when using a breadboard and cheap 20cm dupont ribbon cable jumpers. Unfortunately I've been burnt many times using those same jumper wires on "production" systems - the connectors quickly become loose and the wires themselves have lousy inflexible plastic coating what must be the finest of copper threads. Just downright terrible reliability.

The solution is to build custom cables with decent connectors and wires. It's a lot of effort though. It's even more difficult when the connectors are getting smaller, the eyesight  is getting less acute and the fingers loosing their dexterity. Even with a high quality Japanese crimping tool, every connector is a struggle.

When it comes to dupont connectors, the problems really amounts to nothing other than orienting the connector in the crimping tool. Fat fingers always seem to get in the way. Here comes the 3d printer to the rescue!

A little jig that can be loaded easily with 8 conectors at one time, all with the correct orrientation and position ready to be inserted into the crimping tool

The crimping tool presses up to the jig and the connector is loaded into the tool at the correct depth. Then it is a simple matter to insert the wire and squeeze to crimp. A perfect crimp job every time without the need to struggle and fumble. A far more satisfying result.

Now that the ProMicro boards have finally arrived from China after several months on what must be one of the slowest boats (or lost off the back of a conveyor belt in an airport distribution centre), I completed furnishing the SpaceMouse lair. The microcontroller emulates a USB keyboard and allows me to feed in keystrokes to the cad program. Saves having to move my hand back and forwards to the keyboard just to press a single menu key.

I think I'm going to have to colour the keys or something because I am having trouble knowing which key is mapped to what Fusion 360 function. Time will tell if I can use to it.

As is usual for any half-complete project, it is a visual mess. The next stop is to print a decent organic curved keypad enclosure.

Maybe this is a fail or not. Can go either way. The goal is to use an Ardunio micro to provide a USB keyboard HID interface for the Spacemouse keypad dock I'm working on. As I am exceedingly lazy and did not want to solder some 20 jumper wires onto a PCB protoboard, I forked out the dollars to manufacture a custom PCB from a fab house. Of course, the frieght charges dwarfed the few bucks for the board, but at least it was a satisfying design process. See if you can identify the issue I discovered when I went to program the board...

Ignore the tactile switch hanging off the board - that is just to reset it for programming and will be later removed.

Yep, that's correct - no space at all to plug in the USB cable!

How could I make such a mistake? Also seemed odd that the board did not perfectly fit the footprint I had expected, but it was close enough that it will do the job by leaving out a few pins.

Seems like the board is actually correctly designed and I had a BOM ordering issue. It is designed for an Ardunio Micro but had ordered a handful of Ardunio Promicros. Reasonably compatible pinouts but the usb port is on the other end of the board.

So much for this board - far too hard to unsolder everything so in the bin it goes I'd say. Luckily I have a few spares so I'll solder up another one but this time flip the headers to the other side of the board. As they are connecting to individual switches, the exact order does not matter the slightest.

It was really an ordering fail, not a design or build fail but with a bit of tweaking, it should turn out fine. Of course, we will ignore the fact that as I was soldering up the headers I thought to myself that I should be able to squeeze in a USB cable. Obviously was way short of firing neurons...

No matter how much effort is put into a project, it seems to always be a sequence of various degrees of fail until finally success is achieved. Fails should not be forgotten so here's a progressive record of their history...

Tevo Z Axis Upgrade

A lot of time has been spent to model the Tevo in Fusion 360 in the vain attempt to minimise mistakes. Started to implement parts of the upgrade just to see if things fit - according to the model it should work.

Raising up the Z axis to its limit to test the extreme limit showed an error that I had never thought of - the Z Axis drive I installed was fouling the bowden cable. It was a design that was just not going to work. Just goes to show that models have to be complete down to the very last item otherwise. It's always the bits left out that cause problems..

Back to the drawing board to try a few other layouts. Lucky I did because I thought the belt I purchased was 950mm long but in reality it was 852mm. The design was never going to work! Once again, an assumption leads to grief.

The new stepper position seems to work OK and correctly fits the belt length and is actually a simpler design although it's a little gangly.

Sure is....

The Tevo's standard cantilevered X axis has always concerned me.  Usually I'd position the print so that any skew in the Z axis did not matter - after all, I'm not making production-ready products. Usually 2 orthogonal axis out of 3 is good enough - Do the egss in the refigerator egg tray really care if they lean a few degrees?

I crossed a threshold a few months back and attacked the problem. I have automatic bed leveling but still had z axis skew. What was up with that? Eventually the light bulb moment occured and I realised that having the X axis parallel to the bed and the frame being perpendiclar to the printer base, does not mean that prints have an orthoganl Z axis.

The  bed was parallel with the droopy X axis so all the layers were nice and levelled but the z axis was off. After a day of printing test prints and using grbl's firmware skew correction, I finally got rather good orthoganal prints. A few percent out is easily ignored.

Pondering what should be the next Tevo upgrade - multiple print heads, automated server driven print head clear, but settled on a stable Z axis as everythign else would probably cause greater skew issues so dealing with the problem now will reduce future issues.

I finally received all the parts I need for an upgrade after a very long time (who would have thought to check the number of lead screw starts when ordering? Ugh - 2 more months later...)

My Tevo is rather constrained space-wise with all the mods so I have forced myself to model the important parts ensure it all fits. So far, so good. The model even highlighted a problem that was all sweet in my mind but was a physical impossibility.

Now I could go and finish the model before printing parts but that is no fun. It looks good to me so lets print a part or two and get some weekend's satisfaction. As I have only one printer, I do have to make sure I have all the bits needed before physically upgrading - once it is in bits, there is no chance to make alterations.

First rough test print for the bottom mounting bracket proves the concept works. A shaft collar pressing on a thrust bearing to take the load and a normal radial bearing to centre the shaft. Each bearing doing the job it's designed for. A definite overkill but it moves so very smoothly under load...

A few days later.....

The model seems to be prety good so now it is time to "go all in" and print some parts with proper densities. I pondered if it was worth risking an 18 hour print job for all the bits needed especially when the printer had frozen on me for the very first time only a couple days ago. But all seemed good after an evening's printing and I was keen to see what awaited me in the morning....

So many nice pieces but unfortunately the printer had frozen. Some low parts were salvageable. After a few more smaller print runs, it was time to try and assemble one axis that did not affect the printer operation.

Ugh!!! How can it be out of alignment when it is generated from the cad model? Seems I dimensioned the model to the bottom of the acrylic sheet rather than the top so I was 4.3mm out. Bugger! Back to the cad drawing board....

Bit still, the oldham coupler seems to do its job - not obvious when I wiggle both the shaft and the camera, but you get the gist.

I also discovered that the X axis hits the Z axis bottom mounting bracket. That will teach me for using approximate model measurements and assuming everything will be fine. Time to model the real world accurately rather than picture it all in my mind. More components to be printed....