Firewheel PCB tracing

[esaj]

This is probably an exercise in futility, but nevertheless, I decided I "must" trace out the Firewheel mainboard. I have two dead boards, one has no lacquer-covering, but it has the 3.3V -line shorted to ground. The other one came with the wheel (along the working board), and the previous owner (EUC Extreme) said it had died pretty soon after he had got the wheel. I thought at first I'd fix the non-lacquered one, as I thought the problem was only with LM5007 -chip, but since the entire 3.3V line goes to ground, there's something bigger going on, and I don't have a suitable 3.3V regulator and such right now anyway...

So, I still mapped around the boards using multimeter, to figure out how things work (I found out how the battery voltage measurement goes, as well as found out that the power switch pulls the LM5007 R_on pin to ground, that's why it turns on if the switch is unconnected / open), but then figured that I have a new phone from work (OnePlus 3T+ or whatever), and it has a pretty good camera. I snapped a few pictures, and I noticed I can actually read the resistor codes and such from the pictures. Anyway, here's what I have so far (this is a small version, about 25% of the full size )

Notice that the smaller image on the left is the same part from the other board (which didn't have the heatsink on), match it with the C9-C12 caps and C21/R12/R13 to figure it out... Also on the top right, there's another smaller part from the other board shown, as this one had glue over it.

Here's a link to the full-sized one, it's a big picture, 4640x3480, several megabyte download, probably something you don't want to open over a slow mobile connection or such:  LINK

 

EDIT: Pictures & schematic from updates below:

 

 

 

 

 

 

 

FirewheelBoard.pdf

[Hunka Hunka Burning Love]

Can you take quadrant close ups?    Anything interesting on the backside?

[esaj]

4 hours ago, Hunka Hunka Burning Love said:

Can you take quadrant close ups?    Anything interesting on the backside?

I'll have to try to take better pictures with better lighting, but taking more pictures closer to the board won't likely work, as I tried that earlier, and the camera won't focus anymore if it's brought really close. 

I already had an old picture of the backside, it only has markings for the connectors and pins, no components there:

 

 

[lizardmech]

Are these 4 layer PCB? I studied pictures of these boards the most for my designs as they're from a more expensive servo controller. The L components near the gate driver are ferrite beads.

[esaj]

8 hours ago, lizardmech said:

Are these 4 layer PCB? I studied pictures of these boards the most for my designs as they're from a more expensive servo controller. The L components near the gate driver are ferrite beads.

Not sure on the layer count, but probably yeah... The bottom-most layer seems to be the GND, BAT+ and 3.3V go all around the board through vias, so at least there's one (or do they always come in pairs?) layer between, but they might also be partially tracked on same layers, don't know. Also the power-button runs from almost the other side of the board to the LM5007's Ron-pin, it's a bit weird placement (from just below the Q7-chip to the LM5007-chip in the upper left portion in the picture). Didn't get to try where the 5V goes yet, I was first thinking the ACS709's, but they should work with 3.3V also... I also just read about the ferrite-beads on an application note a day or two ago, they're for killing/preventing gate ringing?

I'll try to get the tracing complete and see if I can get better pictures. The GND-layer prevents using light-table to see the inner traces, so it's a bit of a hit-n-miss with continuity-mode on the meters. The other boards' lacquered, so sometimes it's hard to get the probes so that they actually make contact, and on the non-lacquered board, the 3.3V is shorted to ground, so I need to check on the other board whether it's actually 3.3V or GND, but other than that, it's much easier to probe   Might also try to desolder the 3.3V-regulator, with luck it's just that what's shorted, although if the D-PAK's soldered from underneath, it might prove tricky (lots of heat, big tip... eyes closed and rip it out? ). So far I haven't removed the heatsink on the non-lacquered board.

There's a small 6-legged SC-74 (I think) -packaged chip near where the power-button and other wiring goes, marked as Q7. Near it are unconnected pins on the header marked as "Output #1 (NMOS)" and "Output #2 (NMOS)" on the other side, so I'm guessing it's a dual n-channel mosfet. No idea what's it for though. There are also multiple (at least 4) 750k + 33k -voltage dividers around the board, at least two of those seem to be for measuring the battery voltage and the step-down output, not sure on the last two (they seem to go to unconnected pinheaders near the mode-switch), might be just some debugging/testing stuff... The buffer seems to mostly just connect traces in the SMT32, although at least two of them go into a resistor on both sides. I figured out the values and the circuit for the step-down (except of course the capacitors, since the SMDs are always unmarked), unless there are still hidden vias under the components, but didn't draw it in Kicad/LTSpice yet.

Don't know how much I get done this weekend, relatives coming to visit and all, but I'll see if I can squeeze some work somewhere in there

[lizardmech]

The only thing I can never figure out on these are the diodes across the allegro current sense. I have seen shunts or PCB traces used to modify sensitivity but a diode just seems strange.

[esaj]

33 minutes ago, lizardmech said:

The only thing I can never figure out on these are the diodes across the allegro current sense. I have seen shunts or PCB traces used to modify sensitivity but a diode just seems strange.

Freewheeling maybe, if the sensors have stray inductance? I don't really know, just guessing   From what I can tell so far, they sit between the phases and the low-side mosfets (the mosfets on the other sides of the diodes are the ones with ground-fills, so I'd at least expect them to be low-side) in parallel with the current sense-chips, with cathode pointing to the phases. Scratch that, measuring them I found out that they go from the phase to ground, in parallel with both the sensors and the mosfets... There are small vias behind them closer to the middle of the board, going directly to ground. So likely they're freewheeling diodes to prevent the body diodes of the mosfets from conducting and heating the mosfets up?

Btw, did you ever get your controller finished? Mostly I'm poking these just out of pure curiosity, but I was wondering if I'd manage to build a "good enough" driver for the Firewheel-motor and to learn motor control, although I don't dream of making anything like a self-balancing device, but maybe... something like a motorized wheelbarrow?   As in actual wheelbarrow for yard work, not "self-balancing wheelbarrow" like the Chinese call the wheels sometimes  Or a small trike with front-wheel -drive...

[lizardmech]

Almost, I just got PCBs with the MCU and IMU integrated the other day but then I got sick and shaky hands make soldering hard. The motor control hardware is pretty simple once you build a few, there's just no straight forward guides on designing them. Basically just get high powered gate drivers, minimize distance between the driver and mosfet, always use a diode for rapid turnoff and a ferrite bead to supress high frequency ringing. Always have capacitors right near the high side mosfets, I had one board where one phase was 10mm away from the caps, blew the same mosfet every time  when trying to run near it's maximum.

The majority of the work is actually the software even the stm32f4 from the vesc causes some headaches in controller design. What I have found is cpu core in the MCUs doesn't matter much, all the PWM, ADC and DSP functions are the biggest factor. The further you move away from the ideal MCU the more software you end up having to write as you try to come up with software hacks. My favourite motor control MCU is the newest TI c2000 delfino models, it has hardware to support the math needed for motor control. You can just have everything using floats with no work arounds using IQ numbers, and the hardware accelerators run the FOC algo 10 cycles or on a 200mhz mcu.

They have a very nice open source controller software stack but my programming still isn't good enough to complete it and I can never find anyone will to help work on coding so I'm using VESC and the old stm32f4 for now. 

 

[esaj]

Well, the non-lacquered board is officially toast... I desoldered the 3.3V regulator, still the 3.3V line is shorted to ground. Then removed the SMD-output filtering cap, and desoldered and pulled out the leg of the electrolytic on the input-side, still the same. I then remembered a (probably very old) trick I saw in Rossmann's videos: I soldered pieces of copper wire into the pads (+3.3V and GND), and hooked up the power supply there. Feeding 1A of current (at which point the voltage going into the board was about 1.88V or something like that), I started spreading isopropanol on the board to see where it starts to boil off fast... and the winner is: the STM32 itself. Yup, the CPU's gone, after a short while it became too hot to touch, so can't do much to the board anyway. On the "bright" side, now I can tear down without caring of damaging something else   Although probably I should wait until I get around to buy a hot air station, so I can salvage some of the chips (like the gate-driver, might come in handy).

Finished a couple of drawings on the side too:

As far as I can tell, the half-bridges are as simple as this (except the 3rd phase, which doesn't have the ACS709, but apparently is otherwise the same). The diodes on the board picture (taken from the non-lacquered board, I removed the heatsink) look a bit "funny", because they've soldered two on top of each other, the other board had only single diodes. There's no diode in parallel of the high-side mosfet. Left the component designations out from the schematic:

D? = D1, D2 and D12, depending which phase you're looking

High-side mosfets are Q1, Q3, Q5, low-sides are Q2, Q4, Q6. The ACS-designations are probably under the chips.

 

And also didn't draw the gate-drives and the caps and power inputs on the ACS709, since I haven't studied them much yet. The ACS709's seem to be powered from the 3.3V -line.

 

Power-stage:

If you wonder why the LM5007 VCC -pin goes to ground through C13 (I did), from the LM5007 -datasheet:

7    VCC   Output

Output from the internal high voltage bias regulator. VCC is nominally regulated to 7 V.    If an auxiliary voltage is available to raise the voltage on this pin, above the regulation set point (7V), the internal series pass regulator will shutdown, reducing the IC power dissipation. Do not exceed 14V. This output provides gate drive power for the internal Buck switch. An internal diode is provided between this pin and the BST pin. A local 0.1uF decoupling capacitor is recommended. Series pass regulator is current limited to 10mA.

All the SMD capacitor values are marked just as "C", because I don't know the values. Also the 1K (01B)-resistor designation is missing, it's apparently under the LM5007 -chip. Power-button trace must run in through a via and then in a hidden layer, I found it just by probing earlier. Not 100% sure I haven't missed anything (in case there's things like hidden vias or such). Of course the step-down output is also run elsewhere in addition to the Vin of the 3.3V regulator (like the gate driver-chip and the 5V regulator, which as far as I know right now, only powers the hall-sensors).

The 3.3V -regulator is missing on the photo, as well as C24, as I had already removed them at this point.

[Rehab1]

I assume these are the motor wires. Are they 14 AWG?

On 5/12/2017 at 5:59 PM, esaj said:

I then remembered a (probably very old) trick I saw in Rossmann's videos: I soldered pieces of copper wire into the pads (+3.3V and GND), and hooked up the power supply there. Feeding 1A of current (at which point the voltage going into the board was about 1.88V or something like that), I started spreading isopropanol on the board to see where it starts to boil off fast... and the winner is: the STM32 itself. Yup, the CPU's gone, after a short while it became too hot to touch, so can't do much to the board anyway. 

You need to shoot some video of your awesome experiments!

 

On 5/12/2017 at 5:59 PM, esaj said:

On the "bright" side, now I can tear down without caring of damaging something else   Although probably I should wait until I get around to buy a hot air station, so I can salvage some of the chips (like the gate-driver, might come in handy).

Curious, what model air station are you considering?

[esaj]

15 hours ago, Rehab1 said:

I assume these are the motor wires. Are they 14 AWG?

Yeah, those are the motor phases, 14AWG on the board-side, but seems they've used something thinner in the actual motor-side (16-18AWG?).

15 hours ago, Rehab1 said:

You need to shoot some video of your awesome experiments!

Didn't bother with trying to setup a camera to record it... Louis has a good (although long) video of the technique, jumping pretty close to where the action happens:

 

 

15 hours ago, Rehab1 said:

Curious, what model air station are you considering?

I've been looking mostly at the cheaper Chinese combination-stations (hot-air + soldering iron) sold in EU, like Aoyue ( http://www.aoyue.eu/aoyue-soldering-hotair-rework-desoldering-station-preheater-repairing/aoyue-rework-station-multitools-hotair-smt-hotair-soldering-iron-pcb.html?dir=asc&order=price ) and Atten ( http://www.batronix.com/shop/soldering/hot-air-stations/atten/index.html ). Not bad prices, but it's hard to tell whether they're "good enough" or just shit.  Especially since I don't know which ones are better, the ones with a pump or a small fan blowing the air. Buying within EU, there should be less hassle with warranty replacements, in case problems occur...

[lizardmech]

I just bought a cheap chinese one off ebay, lasted 12+ months so far.

[Rehab1]

2 hours ago, esaj said:

Yeah, those are the motor phases, 14AWG on the board-side, but seems they've used something thinner in the actual motor-side (16-18AWG?).

That would be the same as the current Gotways and I assume other wheels. Hopefully 14 awg will soon be used on the motors.

 

2 hours ago, esaj said:

Didn't bother with trying to setup a camera to record it... Louis has a good (although long) video of the technique, jumping pretty close to where the action happens:

Awesome technique! Amazing how fast the alcohol evaporates! BTW do you own one of these microscopes?

 

2 hours ago, esaj said:

I've been looking mostly at the cheaper Chinese combination-stations (hot-air + soldering iron) sold in EU, like Aoyue

I read some of the reviews. One customer was worried about overheating though the hole components.

[esaj]

1 hour ago, Rehab1 said:

That would be the same as the current Gotways and I assume other wheels. Hopefully 14 awg will soon be used on the motors.

Yeah, it's a bit weird that they use thicker gauge on the board when the motor wiring is much smaller. Let's hope they'll change it in the future, although with the Firewheel, that never caused any problems (then again, the motor is 550W nominal).

Quote

Awesome technique! Amazing how fast the alcohol evaporates! BTW do you own one of these microscopes?

The alcohol-trick is very useful, have to remember it next time I need to figure out which component is shorted on a larger board. The IPA (isopropanol / isopropyl alcohol) is available in different sized containers from electronics component distributors and from pharmacies, but at least here, the pharmacy-stuff costs something like 80€ per liter, whereas I got a liter container for something like 5.90€ from TME

Unfortunately I haven't got any type of microscope, I just use a handheld small loop and the camera pictures. The microscope would of course help a lot when soldering very small components (like 0402's or 0201's), but so far I've managed without one (and I mostly work with larger SMDs, like 0805's and 1206's).

Quote

I read some of the reviews. One customer was worried about overheating though the hole components.

The heat and airflow are adjustable in most models, so that shouldn't be a problem as long as you try it out first to find correct settings. Mostly I'm worried how fast they breakdown, and sometimes badly built cheapo-stations can be downright dangerous (fire / shock hazard).

 

3 hours ago, lizardmech said:

I just bought a cheap chinese one off ebay, lasted 12+ months so far.

That sounds pretty good, although based on what I've read, the noname-stations can be a bit of a hit-or-miss (ie. even if you buy exactly the same stations, one unit might be just fine whereas other has all sorts of problems... poor quality control?).

Earlier this evening, I corrected the picture of the power-stage in one of the above posts... I rushed the measuring and didn't take into account that the inductor L1 has very low resistance, so I made a mistake in how the inductor is placed in the circuit (it's of course in series with the load, not in parallel).

Now I poked around the gate-driver, and managed to get the circuit down. It's pretty basic, almost straight from the datasheets, although I don't know the capacitor values:

Example circuit from datasheets:

 

Circuit from the board:

Forgot C17 and C20 (bypasses for the VDD-input) from the schematic... Each of the gate drive outputs have similar circuit, made from (probably) a ferrite bead, and a 75ohm resistor in parallel with (reversed, cathode pointing to the gate-driver chip) schottky-diode. The pins marked "STM32" go to the CPU, the actual legs are above C48 (if you look at the earlier pictures), but I didn't try to pinpoint exactly which input goes to which pin there.

Notice that R10, R8, D8 and L8 are missing from the board. I desoldered them to make some measurements. R10 was removed so I can measure which way around the diode in parallel is (or if it's a TVS or something). It seems to be a schottky diode, with the cathode pointing towards the driver-chip pin, presumable to speed up the turn off of the mosfet by bypassing the 75ohm resistor.

I also took off all the components on the R8/D8/L8 -set. I soldered the L8 to an DIP/SOP/TSSOP -adapter board, between two legs so I can place it on a breadboard. 

There's also the R8/R10 resistors and the diode there, but I was mostly interested in the inductor L8. I first tried to measure the inductance with my cheapo LC-meter, but it was so low that it couldn't read much anything. The lowest setting is 2µH, but it probably used fairly low frequency (never measured the output), and the reactance of the (presumable) ferrite bead is probably very, very low at those frequencies.

Instead I then placed the L8 in series with 47 ohm resistor (the through hole on the board), and fed different frequencies of sine-waves from my DIY-signal generator. Didn't get that much out of it, except, yes, clearly it is an inductor, probably a ferrite bead  The voltage-divider drops more voltage over the inductor with high frequencies, and less at low frequencies. Here's one of the scope screenshots

20MHz signal is fed through the inductor and the 47-ohm resistor. The yellow (channel 1) signal is the input from the generator (very messy, I had it hooked up with jumper wires and it picks up a lot of noise, some of it probably originating from the generator itself), the greenish/bluish signal is the voltage over the resistor. The "MATH"-signal is the difference between the signals. Anyway, with the input VPP average around 404mV, and the drop over the resistor at 115mV (average), it would seem that at this frequency the reactance of the inductor is about

115mV = 404mV * (47ohm / (47ohm + X_L)) <=>  X_L = (404mV/115mV - 1) * 47ohm  => X_L = around 118.113ohm

Knowing the reactance and the frequency, inductance should be around

X_L = 2 * pi * F * L  <=>  L = X_L / (2*pi*F)

For 20MHz, that should be about

L = 118.113ohm / (2*pi*20*10⁶ Hz) = 9.3991... * 10^-7 henrys, or about 940 nanohenrys.  (Assuming I got my math right and didn't make any "off-by-decade"-errors or such )

Not that the actual inductance even really matters... from what I've read, ferrite beads have (relatively) low inductances and tend to have more meaningful reactances in much higher frequencies (datasheets could give the reactance at 100MHz frequency, for example). With a sweeping signal generator (and one that doesn't add that much noise at high frequencies ), more accurate values could maybe be obtained, but I don't have such a signal source, and like said, probably it won't matter that much.

 

EDIT: And reading further on the subject, you can't even calculate the inductance at higher frequencies like this:

Could I technically calculate the inductance knowing that the impedance of the inductor is (ZL = j*2*pi*freq*L)?

Yes, but only as a low frequency estimation. At high frequency, the ferrite works as a lossy inductor which can't be represented by a series resistor.

 

This is the application note on ferrite beads and gate ringing I read before:  https://www.microsemi.com/document-portal/doc_download/14693-eliminating-parasitic-oscillation-between-parallel-mosfets  

Although it mostly speaks of paralleling mosfet gates and preventing ringing with ferrite beads, it also mentions they work with single mosfets too: 

"A ferrite bead can also be used with a single MOSFET that is not connected in parallel with any other MOSFETs. The effect is the same; high frequency noise on the gate is blocked, eliminating any tendency for oscillations."

 

[Rehab1]

2 hours ago, esaj said:

Earlier this evening, I corrected the picture of the power-stage in one of the above posts... I rushed the measuring and didn't take into account that the inductor L1 has very low resistance, so I made a mistake in how the inductor is placed in the circuit (it's of course in series with the load, not in parallel).

So glad it's fixed! I was going to bring that to your attention. You are a wealth of information! Wished I had your talent!

[Hunka Hunka Burning Love]

My hot air rework station has a fan in the blower handle.  I've seen the other type with the air tube and blower in the main unit itself.  I don't know which is better.  I would lean towards the fan in the handle one myself as I don't know how much air pressure the other type can create through the tubing.  I haven't been electrocuted yet, but it might be prudent to open up the main box to double check everything is secure.

[esaj]

Enough tracing for one evening... if someone want's the Kicad-files, let me know and I'll look somewhere to host them (Dropbox or Google drive or something)...

Can't guarantee that I've made no mistakes with the STM32-pins (or otherwise).   Also I'm not sure if the FAULT/FAULT_EN/VOC -pins in the ACS709's don't go anywhere, marked them as unconnected for now. Still missing the two pinheaders near the bottom of the board, related resistors & connections to the STM32, the gyro/accelerometer chip and the weird loops going from the STM32 through the buffer back to STM32... 

EDIT: Crappy resolution in the clipboard-plot, attached a PDF-version (much better, SVG would take up about 1MB...)

FirewheelBoard.pdf

[Hunka Hunka Burning Love]

Do you think replacing the CPU with a new one would get things working, or is there a damaged component frying the CPU?  They must sell these CPU's and MPU's on AliExpress or somewhere?  Looks like $11 at Digikey.

https://www.digikey.ca/product-detail/en/stmicroelectronics/STM32F103VBT6/497-6070-ND/1646345

[esaj]

2 minutes ago, Hunka Hunka Burning Love said:

Do you think replacing the CPU with a new one would get things working, or is there a damaged component frying the CPU?  They must sell these CPU's and MPU's on AliExpress or somewhere?  Looks like $11 at Digikey.

https://www.digikey.ca/product-detail/en/stmicroelectronics/STM32F103VBT6/497-6070-ND/1646345

I think it's just the CPU itself, but can't be 100% sure of course... the chip alone won't help much without the original firmware, though

[Hunka Hunka Burning Love]

Firmware smirmware!  Just make it work!  

I wonder whether this guy still is able to get replacement boards... what model Firewheel do you have?

 

[esaj]

10 minutes ago, Hunka Hunka Burning Love said:

Firmware smirmware!  Just make it work!  

Oh yeah, that's easy, let me just write a couple of lines of code...

Quote

I wonder whether this guy still is able to get replacement boards... what model Firewheel do you have?

 

F260, but AFAIK, they all use the same mainboard (+ a small auxiliary board that handles the lights, battery display, recorded voice-messages and such). I probably could get new board from 1Rad too, but the shells are pretty damaged anyway (a gap is left between the shell-halves, allowing water to enter straight into the wiring, boards and battery compartments)... and I've still got one working board, that's bolted to the custom-frame chassis, which could (carefully) be transplanted to another backplate. The shells might be harder to come by, unless some sellers still carry them. I'll see if I can get around to try fixing the damn thing over the summer, but I've been thinking about using the motor for trying to build a controller on my own (not self-balancing, but just driving the motor in general, with FOC or something). I've already got a KS16S enroute, and a second-hand KS16B coming in later on...  

[Hunka Hunka Burning Love]

 Homebrew Drift Trike for the missus!

I wonder if these guys have any spare replacement shells...

https://fire-wheel.eu/en/6-f-serie

Hey so you bit the bullet on the KS16S hey?  Nice!  Did Bjorn end up selling you his wheel?

[esaj]

32 minutes ago, Hunka Hunka Burning Love said:

wonder if these guys have any spare replacement shells...

https://fire-wheel.eu/en/6-f-serie

Interesting, didn't even know a dedicated European dealer exists. Although they've listed everything as "liquidation sale", so probably the actual company has stopped producing the wheels or at least they're not going to buy more... Not bad prices, but I don't think I'm going to sink more money into the FW   Even fully restored, I wouldn't have much use for it after I get the new wheels, and probably would end up selling it at a considerable loss

Quote

Hey so you bit the bullet on the KS16S hey?  Nice!  

Yeah, mostly it was just waiting for them to come into stock... should be here by the end of the week, unless the courier fucks something up  It's -1C still outside, but should get warmer by the end of the week, so right on time.

 

Quote

Did Bjorn end up selling you his wheel?

Not from Bjorn, I actually got another offer in private after that, and since the price wasn't bad, I thought having a spare-wheel wouldn't hurt (I don't think I can convince my GF to learn to ride ). Nobody knows yet what issues there might still be with something as new as KS16S, especially since the motor and the mainboard are completely new and it even runs different firmware than the older models (the last I checked, KS16S firmware was still at version 1.0, B's and olders use V1.25, and apparently they aren't compatible). And even if there are no issues, I have another wheel to play around and maybe test some mods (yeah, and soon after that I'm going to need a new KS16B -board )...  Or probably could sell it at a small loss or same price, if it looks like I won't be needing it.

[esaj]

Easier way to figure out where the traces go:

Added U6 + connections, all the connectors on the lower side of the board (J6, J9, J11), all related resistors & capacitors, marked ACS709 FAULT_EN-pins to ground (faults disabled, apparently FAULT and VOC -pins are floating?).

Now U6 makes sense at least on a couple of things: the hall-sensors use 5V voltage, and it buffers the STM32-input pins that are 3.3V logic, should have been pretty obvious . Also it seems that the power button LED is powered through the buffer too. 

U6's also connected to the Q7, I've traced the pins but I don't know the exact chip-type. Probably some kind of dual N-channel MOSFET, couldn't find the code in any SMD-code books (it's something like 03h or O3h or upside down, HEO, HE0, 4E0, HEC...). I've given up on it for now, as it goes to unused pins in the header, which are the two empty holes in the header under U6, markes as OUTPUT #1 (NMOS) and OUTPUT #2 (NMOS) behind the board.

The casing's very small, SC-74 or something?

 

 

 

FirewheelBoard.pdf

[electric_vehicle_lover]

@esaj a lot of value here!!!!

I suggest you to copy-paste all this information to a CherryTree notes file, that after you can export to PDF file or HTML file, and host free on github or bitbucket -- see what I am being doing:
- https://eggelectricunicycle.bitbucket.io/
- https://opensourceebikefirmware.bitbucket.io/About_the_project.html

On CherryTree you can attach files like PDF files or others, like source files or KiCad or something. And you can keep growing your notes file and copy-move them to be organized and export the full or just some parts...