Custom Open Melt PCB Hardware Changes

[TJMakes]

Hello!

Thank you so much for all the hard work that has gone into your firmware, testing and documentation. I read that your goal is "Make translational drift accessible to more combat robot builders – and try to keep it that way into the future." It certainly achieved this for me! This firmware is far beyond what I am capable of myself.

I am allergic to wiring and so I am designing a PCB to support Open Melt and hopefully make it easier for people to use (+have a unified hardware architecture to support).

SCH_OpenMelt2 Schematic_2024-01-26 (1).pdf

Key Features:

• 6-32V Input voltage directly from Battery (2-6S)
• Buck Regulator with 5V 1A output
• 2 x 30A Brushed Motor Switches.
• Integrated H3LIS331 IMU
• Pro Micro Headers
• 5V Header LED Output (no current regulation onboard)

I have made the following design choices and I would like to hear if you have any feedback. This is your platform so I am very keen to design it around your priorities as well. I will hopefully be able to contribute via pull requests to firmware changes needed to support the PCB.

Design Updates

Swapped from Arduino Micro to Sparkfun Pro Micro, but kept as external plug in module
The "Arduino micro" is not the most common atmega32u4 platform, and the "Pro Micro" originally released by sparkfun (but cloned to oblivion) is the ubiquitous board. Even searching Arduino micro results in more results for Pro Micro than it does the original Arduino board, which I feel is confusing to potential users that are interested in Open Melt. The Pro Micro is much cheaper and readily available for me in Australia. I chose not to integrate the Atmega32u4 directly, as then users will have to mess around with burning bootloaders onto the chip which is quite a hurdle if you do not have that experience.

Integrated H3LIS331 & Level Shifting to OpenMelt PCB
The H3LIS331 is quite an expensive niche chip, and I wasn't able to source the Adafruit board easily, so I needed to use the sparkfun board for initial testing with an external level shifter. This was a bit of a pain and I'm a bit suspicious of the long-term availability of these breakout boards. So I have integrated the chip and the level shifting circuitry onto the PCB itself, with a dedicated 3V3 LDO to drive it.

Integrated Buck Regulator Module
I build primarily featherweight robots, here 6S is the most common voltage at that level. So I have inserted an off the shelf 6-32V Input, 5V 1A output regulator module. I did this as a very lazy and quick option. I would like to have a more powerful 3A module in future using an integrated version of the MP1584 modules that are available everywhere.

Things I am not yet happy with

• The Power and Motor Connectors are uninsulated, and I would like to find better solutions. Ideally Insulated PCB mount connectors that are compatible with 3,4 or 6mm RC bannana conncetors.
• I will improve the buck regulator in the next version.
• I would like to move away from Mosfets to proper motor driver ICs, but it might make sense to keep the power stage off this control board entirely and require users to use standard ESCs (brushed or brushless).
• The current size (60x60mm) is not small enough for antweight. I can make another design that is 2/3S and 5A power level which can be much smaller. I have taken care to separate raw battery power traces from potential sources of short circuit on the PCB for safety but this is not the most compact.

Future Design Goals

• I would like to design and 3D print a nice case.
• I would like to be able to switch between OpenMelt and Traditional Drive modes (simple passthrough of RC signals), but there is a lot that goes into this.
• I would like finished modules to be available for people on Tindie or similar. You can have any sales revenue if you like to support your development.

[nothinglabs]

Thanks for putting this together! I think this is a great idea. I do have a bunch of thoughts - and will follow-up with them in the next couple days.

-Rich

[nothinglabs]

Sorry for the delay (I was out of town on a bike tour for a bit).

This is all great!

I had thought about doing a PCB board to go with the project - but wanted to keep the project focused on the firmware / minimal test-build. (I think success on a lot of projects is about keeping the scope down to something you can execute on)

I went with the Arduino Micro mostly for name-recognition (to help make it clear "this is easy - It's just Arduino!"). If all the needed pins are exposed - should be no problem using any other atmega32u4 board. I think including the "Pro Micro" on the board directly makes a lot of sense (no reason to re-invent the wheel). Preserving the USB programmer is great as it makes firmware modification easy.

The MOSFET to drive the LED is a nice addition. Driving the LED from the AVR at ~20ma is "just bright enough." Earlier builds of Open Melt based on the Pololu Baby-O boards could drive a 1-watt LED (blinding!) - which is much better for visibility in a brightly lit arena.

Having motor drivers on-board / keeping off-board is a good question. One option may be to have motor drivers onboard - but also have the motor signal pins broken out on connectors.

One thing to think about is accelerometer placement. At a radius of 5cm - you end-up around 400g at around 2800rpm. Being able to keep the accelerometer close to the center of rotation is a consideration - and if the board is too big - might cause challenges (rotating the board off-axis can help). This is an upside of a separate breakout board for the accelerometer (of course doing a custom breakout board would be an option).

https://www.sigmaaldrich.com/US/en/support/calculators-and-apps/g-force-calculator

I went with the H3LIS331 due to availability on breakout boards (wanted to assure the project was buildable without a custom PCB). If it becomes an issue in the future - the H3LIS331 could probably be swapped with a different part fairly easily (from a firmware perspective).

A long-time ago I built a version of Open Melt that supported "normal" drive mode. It worked with some bidirectional motor controllers from Pololu. It's something I've left out of the code for simplicity. Doing different behavior when the throttle is at 0% is pretty easy - but currently moving the stick to the front / back do some things in the current code (config-mode and RPM flash-out). There probably some smart / easy ways to deal with this (10% to 40% throttle = normal drive).

I think the biggest question is what "pass-through" actually means. Ie - is that literal RC-signal pass-through to motor controllers taking RC-signal (in which case - they also need to be able to handle high-speed PWM to do translational drift). Or - is it generating some other kind of logic-level signal to a motor driver board? I don't think either of these would be that hard to implement (but if it ends up being supporting 4x different configurations - that's another matter).

Open Melt has a "non-commerical license" - but I'm good you selling supporting hardware like this (it's more I just don't want BattleBots using it to make a toy without including me).

Again - this is a great project! Let me know if you get to the point you need help with any firmware tweaks to get things working (or anything else).

[TJMakes]

Thanks Mate!

I just received the prototype PCBs so I will validate all of my schematics and then work on another iteration factoring in your feedback.

I am increasingly thinking the most performant and versatile platform will probably be inline with just Receiver PWM channels in, Motor PWM channels out, as I am not that confident on the power electronics and this way it can be equally applicable to all weight classes and more easily positioned near the axis of rotation.

[nothinglabs]

Looking good!

Doing 1:1 passthrough on the RX's PWM signal should not be that hard. I think the trick will be is finding a motor driver that will handle both RX-style PWM - and then be able to handle the refresh rates needed for transnational drift. It may be that any RC's PWM signal (probably50hz) would need to get "upconverted" to 490hz.

If you get specific set of hardware you want to try with something with for the pasthrough - let me know. Probably a few hours work to get something testable.

[nevsie]

It may not be important, or relevant - but I am doing my own build currently and have thoughts...
1 - Power management. On board is handy, however, my current board, being a USB C versions allows me to push in 3v-15v. Reduces my need there.
2 - Motor Switching - I was hoping to use PWM out to some small brushed esc's. Maybe eventually move to brushless, hence the approach. A pad for this would be good. Although, comment above about rates has me worrying now!
3 - Heading LED through shotty is good, as people may want to run multiple LEDs, either for when it inverts, or flashy effects.
4 - Accel on board. I would not. It massively limits position and design. In my case I am doing a UK Ant (US Fairy), and I am hoping to use the H3lis200dl and have the gyro just off centre.
5 - Likewise radio off board, as people have preferences.

I think the board should do the core - but also "not make assumptions". Assumptions about power, motor, esc, take away peoples flair!

But to you both. Amazing work. I've a long way to go. But just some thoughts.

[TJMakes]

Hi Nevsie,

Thanks for the input and definitely valuable! Also thank you for reminding me to get back to this. There are not many events in Aus so I get a bit distracted from my robotics projects in the off-season.

I will spin up a layout that has the motor drivers and power circuitry offboard (or at least as separate modules). I think this is the only way to accomodate all weight classes. You are right that I shouldn't have tried to make it do everything. If someone is running 12S in their robot they can regulate to 5V themselves haha.

My desire on integrating the accelerometer is that, with motor drivers removed, the entire openmelt pcb would be able to be not much bigger than the gyro breakout itself. I have a particular loathing for any extra wires in my smaller robots as they end up taking up a lot of the available space in the chassis.

I can do something where the the accelerometer can be snapped off and used remotely, and can be a lot smaller than the off-the-shelf modules, as it won't need all the pullup resistors, level shifting etc.

I will mock up a layout and send through, and we can discuss. There's 3 of us in the chain now and so I am happy to be outvoted!

[nevsie]

I think the question is - who is this for?
Aus Ants, they are 150g like the UK right? Therefore, motors drivers are tiny. But if this is for US Ants or higher weights, then the whole picture changes.
But if the board is tiny, with just the main core, and the accel, then I cannot see a need for it to be broken off.

In my case, hopefully pics attached. I have all the insides on the top, but then the accelerometer on the underside, beneath the motors. That keeps it really central, and so the low G readings mean a cheaper sensor! But... I am guessing accuracy improves if you move just enough away from centre.

Also 1 wheel bots effectively self balance.
But 2 wheel bots need better weight balance on the insides - therefore battery position is an issue with 150g Ants, less so bigger bots.

[TJMakes]

Unfortunately I can't see your pictures Nevsie. I compete in the Featherweight class so 13.6kg here.

See attached for mockup of a simplified PCB with just RC channels in, PWM out, with accelerometer onboard. The PCB is 31x18mm and is a bit smaller than the pro micro into which it plugs in (which is smaller again than the arduino micro). Can't get smaller than this without integrating the MCU and moving away from standard RC connectors, which would be a headache.

I have centered the HIS331 exactly in the centre of the PCB (marked on the underside by the white dot. I would use this module by soldering futaba pigtails onto one end to plug into receiver, then with a 3x3p pin header on the other side to plug in ESC/s and LED.

Main constraint with this approach is that you need to make sure your BEC can drive the current of whatever header LED you use. I would imagine that it would be very rare that a robot does not have 100mA or similar excess current capacity on its 5V rail.

Here is 3D model of PCB if you would like to play around with how it could fit into your bot. Note that it would have the sparkfun pro micro on top of it as well.

OpenMeltSimple.zip

[nevsie]

That is looking really nice.
I know what you mean about integrating radio (don't, I've been discussing how little fun this is with mark, malenki man!). I am currently using FS2A board, but also have also done ibus over UART. And tonight have been playing with bluepad32, connecting gaming controllers with Bluetooth. I might make a ESP32-pico-d4 board if I can bring myself to learning kicad, and so forth.

I forgot the images previously if my test ant chassis.
Note it's n20 motor 3000rpm, but next stop brushless... Maybe!

[nothinglabs]

nice looking bot!

[Imdad3456]

Is there a way I could get a gerber file of this pcb so I can make my own