It’s finished!

The nano scared me! I have plugged it after like two weeks in and it worked wonderfully. Then I’ve re flashed it and it started bootlooping IDK why but one more flash fixed it.
The code worked previously but with the help of AI we were able to make the code cleaner at least hopefully.
And in between I did code reviews of the generated code and fixed some issues myself. Mainly a unused state machine lol.

After that I went on to write the readme and an image of the finished project shouldn’t hurt. :)

There it is after a lot of hours I managed to learn more about the IC, Meshtastic, T9 keyboards, chineese nanos, retro fitting…

I have finished the panel!.
The PCBs got revised a bit and it looks super neat.

IDK what to say more about it. I’ll order and We’ll see.

TLDR; finished the three versions and then realized that palatalization would be cheapest.

SO today I have started on making the 1/4WL stub filter footprint.
It was pretty painfull as this is super illegal in PCB design because we are connecting two diferent nets. So yeah took inpiration from a netie and it didn’t give me much. Well because I missed one little tiny small config that allows connecting specified pads together. Then I fought why it wouldn’t copper fill the needed area after that why it would fill too much, partly and at the end I was able to make it.
Unfortunatly HackClub lapse decided to delete my 2hr suffering neat now I don’t have to look at it.

After that I have inplemented it and this step was pretty simple thanks for the previous experience.

And the last one was pretty easy as I had all the right ingredients to cook it :)

Then I had the briliant idea to panelize it as it would reduce the price per board.
So that’s going to be for nextime. Because I’m fighting with lapse now.
like ALL OF MY 3HR PROGRESS GOT LOST!

Today I have finished the filter PCB. Next I’ll do the 1/4 wavelength short stub.

I know the dimensions of the SMA connector now.

I began researching on impedance matching in KiCAD to have the best reception as we are dealing with super small power levels and we can’t afford to loose any power.
Well I still have no clue on impedance matching but it’s hopefully 50ohm.

Then I began calculating with KiCADs calculator the target impedance on a planar waveguide with a GP. After a bit of fidling I think that I got the right calculation.

So started moving the footprints. And the I have made the 1mm traces as well as the 0.18mm coper fills to match impedance.
The researched on how to do via stitching in KiCAD and found that there is a RF tools plugin.

Added graphics and a description to the silkscreen.

And that’s it. It probably sounds pretty easy but it’s super intensive and RF is magic.

Decided on the filter and the SMA connctor

So I have searched a lot to even find one that is in the right band. LCSC doesn’t have the right data and aliexpress is horible in this. So I have asked a community member that had already made it.
I have settled on a HDF6861 because it’s the chapest and it’s ok.

And for the SMA I went with the one that has pins because it’s more flexible.

At first I was fighting with vibecoding. And copilot wasn’t able to do anything that would work. And I wasted a hour going back and fourth debuging and etc.
I was trying to do too much at once . At first I was going to do just the TCA implementation but then I got feature sniped into doing the whole code. But It quickly turned into a tangled mess.

So I’ve started thinking with my own brain. Should’ve done it a lot earlier. And have used copilot as a tool.
And I have just focused on the basics. So I striped down the I2C code as it was only spaming the terminal. And focused on just reading the keyboard. And I was able to make a super reliable code.
It setups, enters loop, turns built LED in off, sleeps, on interupt it set’s up the pins for reading, reads, holds until released and loop continues.

So I have finished it! In this screenshot the PCB is intended to be plugged into a raspberry pi to leach 5V and 3V3 :)
And the nrf is flashed as it was powered by 5V. The only thing was the 3V3 rail as it powers the SX126 Lora. And now it’s plugged into a PC PSU as our club leader didn’t think that using RPI for this is a good idea.
I also wanted to do an observer with it but it doesn’t seem to be working.
I’ve ran the script and nothing. But It’s probalbly a linux problem .

So this project is working!

I’ve wired everything tohether.
The modules were mde before starting this HC project.
I could have just made a nice board that would plug in the IDE connector neatly. But I chose to make a mess with the wires.
It was pretty hard finding the right GPIO in the mess. But I managed to get everything working. Atlest appart from the “Canned message” three button navigation.
Also I have vibecoded a funky implementation of the TCA8418 IC on I2C. but the system just hang’s there so I need to put more effort into it.

Next I’ll do a proper implementation of the IC. Also this is the same one I’m using on travel buddy. neat :)

Allright so, the SX126 doesn’t work. well it’s because I don’t know much about powering the thing.
I’ll try to fix it in our electronics club.

I have found the issue! It’s in the way I’m powering it. The nRF52 is fine because of it’s internal DC-DC convertor. And the SX is connected to the 3V3 line which isn’t connected anywhere. So the fix should be to have an LDO powering the 3V3 line. There’s also an onboard LDO or DC-DC but these go to 3.7V. No frie today

https://forum.rakwireless.com/t/rak4630-schematic-vdd-nrf-question/13897/8
https://semtech.my.salesforce.com/sfc/p/#E0000000JelG/a/RQ000008nKCH/hp2iKwMDKWl34g1D3LBf_zC7TGBRIo2ff5LMnS8r19s
https://docs.rakwireless.com/product-categories/wisduo/rak4630-module/overview/
https://docs.rakwireless.com/product-categories/wisduo/rak4630-module/datasheet/#schematic-diagram
https://docs.rakwireless.com/product-categories/wisduo/rak4630-module/quickstart/
https://images.docs.rakwireless.com/wisblock/rak4631/datasheet/schematic.png
https://docs.rakwireless.com/product-categories/wisblock/rak19007/datasheet/#schematic-diagram

With my frined we have beem able to flash a bootloader and meshtastic FW.

The first thing we did is to folow this tutorial https://docs.rakwireless.com/product-categories/wisblock/rak4631-r/dfu/#converting-rak4631-r-to-rak4631 we used windows for this, it’s after my atempts at folowing the meshtastic tutorial on meshtastic docks on Linux. Then we folowed this tutorial https://meshtastic.org/docs/getting-started/flashing-firmware/nrf52/convert-rak4631r/. And installed the RAK4631 firmware wersion.

We were able to connect wia the mobile app, but we weren’t able to send any messages.
The radio couldn’t initiate. I think that it’s an HW issu,e but I could be wrong.

//\ E S H T /\ S T / C

....

[0m[34mDEBUG [0m| ??:??:?? 1 [34mSX126xInterface(cs=42, irq=47, rst=38, busy=46)
[0m[34mDEBUG [0m| ??:??:?? 1 [34mSX126X_DIO3_TCXO_VOLTAGE defined, using DIO3 as TCXO reference voltage at 1.800000 V
[0m[32mINFO  [0m| ??:??:?? 1 [32mStart meshradio init
[0m[32mINFO  [0m| ??:??:?? 1 [32mRadio freq=869.525, config.lora.frequency_offset=0.000
[0m[32mINFO  [0m| ??:??:?? 1 [32mSet radio: region=EU_868, name=MediumFast, config=4, ch=0, power=27
[0m[32mINFO  [0m| ??:??:?? 1 [32mmyRegion->freqStart -> myRegion->freqEnd: 869.400024 -> 869.650024 (0.250000 MHz)
[0m[32mINFO  [0m| ??:??:?? 1 [32mnumChannels: 1 x 250.000kHz
[0m[32mINFO  [0m| ??:??:?? 1 [32mchannel_num: 1
[0m[32mINFO  [0m| ??:??:?? 1 [32mfrequency: 869.525024
[0m[32mINFO  [0m| ??:??:?? 1 [32mSlot time: 12 msec, preamble time: 32 msec
[0m[32mINFO  [0m| ??:??:?? 1 [32mFinal Tx power: 22 dBm
[0m[32mINFO  [0m| ??:??:?? 1 [32mSX126x init result -707
[0m[33mWARN  [0m| ??:??:?? 1 [33mNo SX1262 radio

So I made an angled holder to hold the Vision master E290 with its case.

The design is pretty simple.
Although the fillet is way too big and it doesn’t fit too well.

After printing it I brought it into my workshop where I’ve drilled holes and screwed it all together.

I’ve assembled it and integrated it into the case.

Firstly I tried to test fit it, but the PCB was a bit bigger. So I’ve shortened it to size. Used the holder thingi as a guide to drill holes, while avoiding traces, but I was lucky as I don’t have traces in the area.
Then everything got screwed up into the case.

Next I’ll do the code and wire it up to the VisionMaster, which I haven’t had at the time.

Today I forced ChatGPT to write code for me. Then I ran into free limitations. Time to switch to copilot with my free GH pro. :)
I’m doing it in class so psst. ;)
It’s just blinking D13 when an interrupt is triggered.
I need to think about the layout and how to make better code for it. It’s pretty good for now as it sleeps and wakes on a button press. The only thing that’s bugging me is that I have put one row on another set os interrupts.

After a lot of back and forth from and to my workshop, I managed to get a serial enamurated. A big leap forward!
The problems:

• Bridged pads
• USB was incorrectly wired

Now I need to check which bootloader is on there (it needs to be the Arduino one) and flash the MeshTastic FW onto it. Charm it with I2C environmental sensors. And also build the other one.

After soldering it, I’ve tried to debug it. So I checked the resistance between the capacitor and I noticed that it was around 2M ohms. So I replaced it with another one this time 100nF.
And now it went back to 9mA! And after that I realized that I forgot to hook up the Vbat_NRF to the USB VBUS. yikes 😅

And now it seems to be working!
So I began the search for antennas as I was worried about the PA without antennas. (It took me around 15 mins which I didn’t log)
When I went through my FPV box the second time it hit me. I have PFV drone ELRS antennas with the right ufl connector! BT is fine but LoRa is sub GHz and I was worried that i still may blow the PA, so I asked ChatGPT and it told me that the SX LoRa chips PA is pretty resilient. Oh nice

After that I plugged it in into my Linux laptop and it seemed to kinda work so I ran dmesg and there it was although it wasn’t properly enumerated. But that’s progress.
OFC I was the whole time frantically checking the voltage and my cheapo multimeter tripped me up as it was showing 6.7V on the USB which scared me. Luckily it’s just a drained battery.
So I needed to run back down to the shop multiple times in -10C. But I love saunas and cold hardening. ;)

I’ll debug the USB tomorrow, it’s getting late, and I can’t think straight.

I thought about soldering the PCB in an oven but this will be messy.
I’ve made a similar breakout PCB for a different module with pads underneath.
This one has castellated holes thankfully.
I’ve soldered on a pin and a pin on the other side on the shorter side.
Then I put a ball of solder on my iron and went over all the pins. Also a lot of flux helped.
It was so cumbersome as the ball of solder was sticking so much. I also don’t have the smallest soldering iron tip so heating up the small castellated holes wasn’t ideal.
So after it looked finished I heated it up with a heat gun. Which was a mistake as the substrate isn’t made for high temperatures and it started to delaminate, so I stopped of course.
After doing that I checked with a multimeter if the traces weren’t bridged. Which thankfully wasn’t.
So I plugged it into my PSU. And I saw no signs of life. I thought that the current draw was below mA, So I checked the voltage on the trace, but there wasn’t any voltage on the 3V3 trace. And then I found out that I managed to desolder the pins from the PCB presumably when using the hot air.
So I soldered back on and tested it. It worked! I saw 15mA.
Then I soldered up the RST circuitry. And now I don’t see signs of life.

Let’s debug. Probably going to desolder the pullup and cap.

In DDM I also drilled holes and put all the buttons in. But I realized that I didn’t drill a few holes so I didn’t bother.
But when I went to solder all the diodes, I noticed that I didn’t drill a lot of holes. So I took a rotary tool in my workshop and finished it. I was pretty scared that I’ll break the drill bit. Luckily I didn’t
i don’t have the nano here, so I’ll try to get it working later.

At our shool we use the good old school method of etching our PCBs. I etched on tuesday after shool club.
The steps were:

1. Print out the design on paper
2. Cut the cuprexit to size and sand it
3. Apply fotoresist and cure it in a oven
4. Make the paper transludcent using oil and expose it in UV
5. Develop it in a developer agent
6. Find that it didn’t develop propertly and scribble all over using a marker
7. Remove the Copper in a acid foam bath - A teacher kindly etched for me as the after school club was ending.

Evhery step is so time consuming. I’ve made a batch of three PCB’s in total which took me like 2.5 hours after school to make. Also you need to babysit evhery step in the process, as it can go downhill really fast.
It’s cumbersome to log the full 2.5 hours and it wouldn’t be honest as I done them in paralel so thats why I’ll log around a halve a hour. Please feel free to make adjusments

And there it is pretty good PCBs.

After that I cleaned it using Líh. Drilled holes in DDM as I hadn’t had more time after shool. It took halve a hour.
We normally put a Kalafuna (Flux) on there to provent it from oxidizing and it makes it better to solder.

At our shool we use the good old school method of etching our PCBs. I etched on tuesday after shool club.
The steps were:

1. Print out the design on paper
2. Cut the cuprexit to size and sand it
3. Apply fotoresist and cure it in a oven
4. Make the paper transludcent using oil and expose it in UV
5. Develop it in a developer agent
6. Find that it didn’t develop propertly and scribble all over using a marker
7. Remove the Copper in a acid foam bath - A teacher kindly etched for me as the after school club was ending.

Evhery step is so time consuming. I’ve made a batch of three PCB’s in total which took me like 2.5 hours after school to make. Also you need to babysit evhery step in the process, as it can go downhill really fast. Like on this etch. ;-)
It’s cumbersome to log the full 2.5 hours and it wouldn’t be honest so thats why I’ll log around a halve a hour. Please feel free to make adjusments

And there it is perfectly imprerfect PCB.

After that I cleaned it using Líh (Ethanol or rubbing alcohol?! I have no idea what it is in english). Drilled holes in DDM as I hadn’t had more time after shool. Which was about halve a hour.

Made a super short readme for this project.

Made a super short readme for this project.

Today I’ve routed the PCB.
It started with measuring the pitch between switches.
Next I set a custom grid to the measured pitch, which is 17.5mm.
It was routed without too much thought on MCU pins as all digital pins have the functions that I need. I firstly ordered them in order, which was a mistake.
Next was assingning the pins. And adding the two headers. I also had one header with spare GPIO but that would be too much hassle.

I’ll bake it with my other PCB. I mean, etch it.
Then I’ll code the code and integrate it.

Made a KiCAD footprint for TESLA 3 FK 573 00, whatever that means.
It’s super simple as I only need the pitch, hole diameter and dimensions.
I first began by placing one pad, changing it to THT added the measured hole diameter of 1 + 0.1, and the pad sie is pin diameter + 5.
Then added the outline or f. margin, which is 14x12mm. And with a 0.1mm offset I added a silkscreen.

I’ve finished the board. Also thought about making the PCB, and I decided to etch it in school.
That’s why there are THT components. Also one clever thing I’m quite proud of is the USB-C connector. As we would only do the back side, and we don’t need to care about the front.

I didn’t know how far apart the headers should be. So I thought about making it like an ESP dev board, but that would force the board to be a lot longer than needed. So I took the components and roughly layout them out to their corresponding positions, like top or bottom. and that way I knew that almost all components would be at the top aside of the USB-C connector. The USB-C was positioned on top of the module. After routing, I’ve noticed that it’s interfering with the module.
After that, I made pretty labels and that’s it. At least for the PCB.
When I’ll return from winter break I’ll manufacture the PCB. I’m also going to log that. :)

I’ve started working on this project the day prior to Flavortown. 😅
I’ve looked at the meshtastic RAK4631 definition, as I won’t need to make my own board definition.
I have in mind that most likely I’ll be CNC milling the PCB. That’s why I also added a USB2.0 header to be able to hack in a USB cable.
Ehm… many blank stares into the RAK datasheet…
I came up with a minimal schematic seen in the second screenshot.
Then I looked into the pin definition and named the pins with their functions, which was a mistake, as I wouldn’t really know which pin is which and I would need to reference the pin definition which is quite cumbersome.
So I went with “{PX.Y}/{Relevant Fn}”
Added pin headers, as Dev boards would be useless without it. ;)
And today I added the menu button and LED. Noted: the nrf52 chip can’t provide enough power to drive the LED by itself.
And that should be it for schematic!

Today I’ve managed to finish the schematic along with footprints.

I started researching button matrices three months ago for my project called Travell Buddy.
So I took inspiration from there. I mean that I stole my beautifully drawn matrix.
After that, I decided on MCU. In the original design, there was an ATtiny with which I don’t want to bother, as it’s meant to be a rough prototype. So I went with a simple Arduino Nano. And after consulting with ChatGPT it seems to have everything that I need!
I’ll be CNC milling the PCB, so I chose a 1N4148 diode with 1N4148 footprint. As for the switches I’ll be using the TESLA ones. I have a whole bag of them.

All that work done under a hour!?, what a beauty of having prepared things in advance.

Next I’ll make my own TESLA button footprint and route the PCB which needs to be one sided.

Removed the keyboard from the original box. And snipped unnecessary electronics.