Jack Crossfire's blog View Details
Archive for February, 2021
Posted by Jack Crossfire | Feb 25, 2021 @ 03:00 AM | 9,579 Views
1 day left for the details, but having 5 panels assembled gave a good idea of how heavy it is. The 1.2mm container was 615g. The .4mm container is 384g. The coroplastic is unknown.

These large panels would be easier to print & more accurate on a bigger printer. Not sure if duct tape or gluing additional PLA sheets would be better. It's not very efficient to print sheets of material.

Washing the bed with dishwashing soap & heating it to 60C greatly improved the adhesion. Much less horizontal expansion is required & the bottom layer doesn't chip away if the nozzle is higher than normal.
Posted by Jack Crossfire | Feb 23, 2021 @ 07:10 PM | 9,661 Views
The .4mm nozzle made the dimensions loose enough to dismantle the metal wires connecting the panels. The PETG panels can be used until they break, then replaced with PLA. The other panels can be replaced when they break without printing a new container. The floor is expected to break 1st.

A larger printer which could print 1 panel in a single go would be money. The floor takes 4 separate printouts. Painting the X's with a sharpie didn't look as good as hoped. It's real hard to get a pen into the isogrid corners.

The whole thing is going to look rather transparent & flimsy. Lions only know the .4mm thickness is the same as coroplastic. Coroplastic has 2 .4mm thick sheets, while lions exchange 1 sheet for risers.
Posted by Jack Crossfire | Feb 22, 2021 @ 06:13 PM | 8,158 Views
The internet claims & the lion kingdom has found PETG to be weaker than PLA. Opinions vary, depending on the sponsor. Some of the internet says PETG has better layer adhesion. It definitely has better heat resistance, is easier to hot glue, & is harder.

PLA went back on the table for at least the .4mm panels. The PETG panels bended more easily. 1 broke when separating from the bed, but it may have been a leveling problem. The lion kingdom tightened the eccentric nuts under the bed, to get the corners more level with the center, but it's never going to be perfect.

New motor encoders were printed out of PETG. It seemed to hold the screws better than PLA. It was about as hard to shove a screwdriver through PETG as PLA, but the PETG shattered instead of making a round hole. PETG might be a better match for gears, sensor holders, & fine details which PLA is too soft to retain under compression. It has the consistency of glass or shellac. It's surprising anything printed in layers can have a shine, but PETG does have a shiny finish.

PETG required 250C, 65C, no cooling fan.
Posted by Jack Crossfire | Feb 21, 2021 @ 08:00 PM | 8,822 Views
If it looks like a water bottle, that's because it is a water bottle in a different form. Unfortunately, Cura failed to create a shell for a small upside down overhang. Layer adhesion is pretty bad, even at 250C.

The lion kingdom is still undecided about permanently retiring PLA. It's a question of what PLA can do that PETG can't.
Posted by Jack Crossfire | Feb 21, 2021 @ 04:01 AM | 8,928 Views
Improving the vlogging quality started with the motorized camera mount, but continued with the headset. The humble electret condenser microphone has improved greatly since 1990 & definitely can sound better than the logitech headset's default microphone. The Zoom H2 donated is 4 high fidelity electret condensers to wearable applications & now finally they have their 1st use. Many gootube videos with very high quality sound were recorded with the default zoom microphones.

This piece of .2mm nozzle .1mm layer height precision was originally concieved for PETG, but PLA proved good enough. The lion kingdom tried the arc welder plugin, but those flat edges are because of the stepper motor resolution.

It doesn't farsten very securely to the recycled USB cable/microphone tube. The wind screen no longer blocks wind from the larger microphone. Things could definitely be better. There is a longer term vision of 3D printing an entire headset. It definitely records better audio with the zoom microphone but the recording software is now the weakest link.
Posted by Jack Crossfire | Feb 19, 2021 @ 05:59 PM | 8,970 Views
In another bid to make a lighter container, a test panel was taken from 6 layers to just 2 layers, .4mm, with the isogrid lines 1mm wide. The printer made the isogrid lines 1.5mm wide. It was 30% lighter. It was harder to push a screwdriver through the 2 layer PLA than coroplastic, but the PLA was more brittle.

Getting line width below 1.5mm from a 1mm nozzle has proven challenging. In order for the isogrid lines to fuse together, the printer needs to make 2 passes for each line. A .4mm nozzle would drop the weight to 55% of the 1.2mm piece but take 4 hours.

There might be a way to print the flat layers out of TPU, then switch filament to print the isogrid out of PLA. The maximum strength is going to be from printing it all out of PETG.
Posted by Jack Crossfire | Feb 18, 2021 @ 10:43 PM | 11,350 Views
The tires have so far not spun out, since being covered in rubber cement. The rubber cement appeared to rub off immediately, but enough of it might have stayed attached to do the job. Not sure of the the tread pattern made any difference in retaining adhesive. The grooves might impinge just enough on the road for the retained rubber cement to provide traction without getting worn off. There's a definite wear pattern where the laminations have been worn smooth.

Rubber cement is a much cheaper & easier enhancement than waterproof silicone. Waterproof silicone would definitely grip, but wear off quickly & be expensive.
Posted by Jack Crossfire | Feb 16, 2021 @ 04:11 PM | 12,637 Views
The lion kingdom finally made a command line interface for the firmware, so the user can generate the encoder table without enabling the TEST_MOTORS macro. It still requires manually converting the command line output to a header file & recompiling though.

There's a growing need for smarter persistent storage on the STM32, allowing it to save config files, encoder tables, & debugging logs individually. It has only ever saved a single config file. This would require some heavy duty flash management. The general idea with flash is to only write until all the erase blocks fill up, then start rewriting the erase blocks without the deleted files in order to free up space.

In its simplest form, it would simply limit the filesystem to a single erase block. When the erase block filled up, it would copy the current files to RAM & rewrite them to the erase block. A more complicated version would copy each erase block to RAM, erase the source erase block, & relocate files back until enough space was freed. Any system would require some inode or FAT system, maybe an inode at the beginning of each file with an offset to the next file. It would set 1 bit to 0 to indicate the area is in use & another bit to 0 to indicate the area has been erased. It could walk through every inode to discover the most recent files. Porting the entire ubifs system or adding an SD card would be overkill.

This is of minimal value. It would be used to fix maybe 1 bug & put away. The other wish is an open loop mode to turn the motors without an encoder.

The real need is bigger motors & lions can't afford to put $110 into another set of hobby motors when hoverboard motors that are too big would be only $100. The truck finally reached a point of being out of money.
Posted by Jack Crossfire | Feb 14, 2021 @ 11:13 PM | 8,232 Views
So a motor loosened from its motor plate, thus leading to a 3 mile walk. Not sure if the motor broke free or if it was never securely farstened in the 1st place. When they become lose, they stop aligning with their resolver & stop turning. The only workaround would be driving them with some kind of manual control of current & commutation speed. Lions considered this originally & now it's a bit higher in priority. They're too slow to drive with a standard back EMF algorithm.

The steering suspension has proven too soft. It was already so hard, the only reason for making it out of TPU now would be just for the strength. TPU is indestructible.

Anyways, lions devoured the http://www.litterbuggies.com/ link. It's intriguing whenever RC cars are used for practical transportation. He has 4 functioning ships, made from a Traxxas TRX-4, TRX-6, & custom aluminum chassis along with 3D printed parts. He uses powertool batteries which are much easier to charge, but bigger. He has the same problem of wheels getting flat in storage so has a stand. His containers are twice the size. He doesn't have any automation, just the stock controller. He has better headlights. He tried PLA tires, but still uses stock tires.

As a practical aid for cleaning beaches while walking 3 miles, it's probably easier just to pull a wagon. A motorized ship would be useful if he was running long distances.

He makes money by selling the robots. A lion burning $2100 on rent couldn't afford to make them for less than a reasonable selling price.
Posted by Jack Crossfire | Feb 13, 2021 @ 04:14 PM | 8,425 Views
The hall effect sensors were reoriented, which allowed them to finally sense the full range of motion. They have to point along the circumference rather than radially. The resolver in this configuration is thus the easiest way to detect motor position.
Posted by Jack Crossfire | Feb 12, 2021 @ 01:31 PM | 8,602 Views
Traction module 3 was integrated. Printing deformation made the motor seals bind if they didn't have the right side up. Front & back panels needed an extra mm of clearance for horizontal expansion. The motor encoders were off, despite being an unchanged design.

The hall effect sensors don't match the traditional design of a resolver.
Posted by Jack Crossfire | Feb 11, 2021 @ 04:11 PM | 9,361 Views
It turns out the cheapest source of large diameter motors for a direct drive vehicle is inside hoverboards. They're far cheaper than hobby motors of the same size. They're sealed from rocks. Hoverboards have long ago faded from popularity & their cheap motors might be going away. The motors replacing them are much smaller skateboard motors, with much less torque, at a much higher price.

The media was focused on the batteries in hoverboards, so the public became singularly focused on tearing down just the batteries while ignoring the motors. There are very few teardowns of the motors.

Getting inside a Hoverboard Wheel (2 min 1 sec)

These are sadly too big for the truck in mind. They're too big even for a brushless gimbal. It's remarkable how economies of scale during the height of the hoverboard craze made such large motors almost free, while tiny hobby motors are a fortune.

The hall effect sensors are embedded inside the stator magnets. They must not have driven the stator magnets while reading the hall effect sensors or the flux in that position must not have been affected by the electromagnets.
Posted by Jack Crossfire | Feb 10, 2021 @ 01:13 AM | 9,696 Views
The lion kingdom's 3 year old Qi charger was proving a disaster for charging remote controls. Putting a magnet in any controller would take up a lot of space, so there was no easy way to align them on the charger. Then, the unusually shaped controllers weren't held in position. The decision was made to finally make a new charging enclosure that would fix the controllers in the right position. It's a very loose position, but good enough. The status LED in the charger, once completely covered by the case, is now visible.

Printed it out with the .4mm nozzle & .4mm line width. It has only .8mm thick panels. Called it quits after 2 prototypes.
Posted by Jack Crossfire | Feb 09, 2021 @ 02:12 AM | 9,299 Views
The only thing left is spending a full day transferring the electronicals from the last traction module. It's the most solid module so far, despite having only 1mm thick walls. A flash goes through them like paper. It used a total of 286g of filament. The single piece panels take slightly more filament & are uglier than panels glued from many smaller pieces, but it comes together much faster & more accurately.

How far lions have come in 33 years, from gluing together rough balsa & plywood kits to designing intricate plastic models from scratch.
Posted by Jack Crossfire | Feb 08, 2021 @ 12:06 AM | 6,119 Views
Another day of modeling yielded all the changes to the next traction module. This one is printed in only 6 parts & farstened without glue. Most of the previous traction module survived intact. To reduce the side panels to 1 part, the isogrids were flipped to face inside. There are a lot of bevels to handle the terrain & run over animals. It has openings for the H bridges, but deliberately requires duct tape. Sometimes, duct tape is the best solution. Space vehicles design in captain tape & lions can't afford captain tape.

A new motor mount traps the o-ring without a wheel. It will need some o-ring tweeks to get the best seal. There's no way to print a circular trap with enough accuracy. It was limited to just gluing 4 clamps on.

The motor mount farstens to the traction module with only 2 bolts & no alignment tabs. It remanes to be seen if it stays in place.
Posted by Jack Crossfire | Feb 07, 2021 @ 04:11 AM | 5,127 Views
Added another 470uF of filter capacitance & the servo didn't malfunction for 13 miles, on the 5Ah battery. It still has dropouts to 3.8V. The Turnigy 5A is the problem. It's never been up to powering brushless servos.


It's based on the XL2576 which is only rated for 3A, so only slightly better than a $1 USB charger. Not sure how the lion kingdom bought 2 of them, believing the 5A rating. Now, they're almost free.

The motor seal prototype appeared. The whole thing is 1mm nozzle, .32mm layer height. Lions aren't convinced the accuracy of a .4mm nozzle .2mm layer height is worth the weaker material.

The mane problem is it relies on the wheel to trap the O ring. Lions would prefer if the O ring was trapped entirely in the motor mount side. The seal takes a lot of vertical clearance. Clamps for the O ring wouldn't fit on the bottom. There would be 4 clamps glued on each corner to trap the O-ring. The motor mount also needs beveled corners to deal with rocks & body parts. Right angle parts are already breaking off the underside.

Anyways, it's a real game changer to be able to print a flexible O ring of any dimensions, even if it's not as precise as a genuine Chinese one.
Posted by Jack Crossfire | Feb 06, 2021 @ 12:59 AM | 8,018 Views
The Trackstar TS-940HG started intermittently dying at mile 8, so out came the oscilloscope. These servos worked reliably for years in the lunchbox & several weeks in the 3D printed truck before dying. They showed Vdd glitches down to 3V & it may have been switching to a battery that discharged to a lower voltage that caused the failures. There was no reason they should die in this vehicle instead of the lunchbox.

Anyways, out came a plan for a new shroud that completely seals the motor, except for the 1 wire. The entire traction module has to be reprinted. The vertical isogrids are finally going to die, since they're not very visible & they make the alignments worse. The vertical panels are going to be too cut up with motor mounts & H bridge clearances for an isogrid to be very useful.
Posted by Jack Crossfire | Feb 04, 2021 @ 09:52 PM | 10,188 Views
So the motors were spinning fine until suddenly they could no longer go faster than 5mph. Bluetooth showed battery voltage of 10.5V. Later found out the left motor picked up a rock & wouldn't turn anymore, while the right wheel didn't have enough torque to drag it along.

Also, tried improving traction by coating the tires in rubber cement. The rubber cement merely picked up rocks but didn't improve traction. Doubt rocks picked up by rubber cement were the cause of the motor contamination. The shrouds were clearly a failure.

The bright spot in this disaster is a new bumper proved better at absorbing impacts. It could do better if it was a single piece spanning the angle rods & the springs were parallel.
Posted by Jack Crossfire | Feb 03, 2021 @ 06:54 PM | 8,049 Views
Torque limitations gave way to TPU issues, as the ship was a lot noisier. Upon further inspection, the front tires had developed a flat part. The flat part remaned for an entire 9 mile drive. Most disappointing if TPU gets permanently deformed from being held in the same position. It's not a very good camera platform in its newly acquired bouncy form. The rear tires being slightly stiffer didn't have the deformation. There wasn't any heat wave when they were in storage. Maybe the ship could be stored on blocks.

The flat part tended to be heavier. They may always end up stored with the same part on the bottom, causing a flat part to grow. Tires are only going to get more flexible & prone to deformation.
Posted by Jack Crossfire | Feb 01, 2021 @ 07:15 PM | 9,050 Views
A quick test of a motor revealed their current usage at 12V is 3A. The L6234 only goes up to 4A if they're lucky, so a boost converter would also require custom H bridges. The polulu boost converter itself would only output 3A at 20V for a short time. A custom boost converter & custom H bridges would be needed, many lbs of copper.

There might be marginal losses from battery resistance, wire resistance & connectors, bringing the practical current at 12V to 2A. The leading theory these days is now to get some 1S 5Ah batteries & wire them in series with the 12V batteries. It's the cheapest way to test a marginal increase in torque. It might also extend the range enough to not have to carry around another 5Ah.

There's a 5060 motor which might also give marginal improvement, but above 50mm, hobby motors go to $300 & weigh 5lbs. Skateboard motors become cheaper beyond 5060, but also weigh 5lbs. For a new vehicle, the 5060 would definitely be the way to go.

Anyways, with the levitator overheating, the starship model is manely a refrigerator magnet.