2FAC – Two Factor Alarm Clock

This week, I played with 7-segment displays and designed the shape for the clock and base station. I did not deviate much from my initial sketch because the pressure to learn FreeCAD and Blender left no room for further creativity.

This week, we played with microcontrollers, and I developed strong feelings for the RP2040. My final project will have a base station and the clock itself. One of them will probably run a RP2040 with Internet access to keep time using NTP.

This week, I learned about Charlieplexing. If I use 7-segment digits on the clock, it will need it to drive many LEDs (4 times 7 per digit and plus 2 for the dots). I am still undecided on whether to use OLED or super fancy e-ink instead.

Wood is very pretty and a good candidate for the base station or clock, or both.

This week, I made a working 7-segment LED digit. I am still not sure if I am going this route but it is very cool!

I experimented with all the inputs for my project this week:

DCF77

to get the current time

Rotary Encoder

to change the wake up time

Step Response

to detect a hand on the clock for snoozing

I experimented with making NeoPixel components for the clock digits, and using a piezo for sound. The latter was disappointing, I have decided not to use it.

More playing with NeoPixels; not applicable to my final project but I learned a lot about them.

I finally started designing the entire project in this week.

I printed a stand that will be slotted into the housing which holds the clock connector to connect to the dock.

While connecting the bottom part with the clock connector to the main board, I ripped off one of the connectors. I immediately re-milled the board and resoldered it.

While soldering, I printed some diffusers for the display to try out.

Last night, I found out the DFPlayer Mini does not work when on battery power. The 5V pin does not supply 5V when the battery is attached. I destroyed the RP2350 while testing the voltage with a multimeter.

I also spent some time with V-Carve and OpenSCAD to create toolpaths for milling the housing. I am planning on doing that next Friday.

I received the DFRobot MAX98357 I2S Amplifier Module - 2.5W in the mail today and immediately started testing it. Works great on 3V3!

The pin names on the board do not correspond to the I2S names in the MicroPython library. Here is the translation:

Also interesting: not all pin combinations are allowed. I got a ValueError: invalid ws (must be sck+1) error on the first try. Eventually I was able to produce a tone and play a WAV file I uploaded to the RP2350 using mpremote. After confirming it worked, I spent (a lot of) time making a footprint and adjusting the symbol in KiCad to update the main PCB.

During class, I decided to drop using the Shopbot for the housing because the multiplex I have is too uneven to get reliable pockets for the acrylic windows and PCB frame. I discussed it with Henk and will go with fully laser-cut acrylic.

I milled a new main board with the MAX98357 on it today but alas, the footprint I made yesterday is 2.54mm too wide. Next time, I will print the PCB on paper first when including self-made footprints to see if they actually fit. So, I corrected the footprint and milled yet another board.

Today was "LightBurn and burn-smell day"! It was more fun than I imagined. I did a cardboard test first, then the faces in wood (very pretty!), then the windows in clear acrylic, and finally the rest of the shell in white because the cardboard was not sturdy enough.

The first cardboard test failed. Somehow, DeepNest messed up the scale of the SVGs, so I dropped DeepNest and nested the parts myself in LightBurn.

Settings used:

I mismeasured the thickness of the acrylic, so the intermediate slices are too thick. I can probably fix the interior to make it fit (new frame and base), so I'll look into that this weekend.

Today I cleaned up the acrylic slices for the body (removing the protective foil and clean them with soap to get rid of the smell left by laser cutting). Although the slices are too thick, they are already usable without reprinting the internals. I was afraid the white of these rings would not be as pretty as the cardboard, but it looks very nice.

The rest of the day I spent on designing a dock. Extra points for being able to print it on my Prusa Mini (180x180x180mm).

Printed the model I designed yesterday and updated the connector boards to have a more logical pin order, making it harder to flip power and ground when docking it the wrong way.

Got some threaded 5mm rod to bolt the housing slices together (unfortunately, the shop was out of bolts long enough). I also worked on the 3D models of the clock and dock base because I need USB access without taking it apart all the time. I milled the dock connectors to get a better pin layout and created new wires to match.

During the long 3D print, I looked at why the DCF77 does not work on RP2350 (it does on the RP2040!). I found some hints on an old blog post (hurrah for the wayback machine!). But, after adding a (real) pull up resistor and a capacitor as in the example, still nothing happened. I tried switching to 5V, nothing. More pull up by enabling the onboard pull too, nothing. I'm giving up for now, printing a new clock base with USB access should allow me to set the time (many with Web Serial page to do so?).

Annoying, low-progress day..

It has been a long and busy day. I printed (and reprinted) the dock base for the new pin layout and connector position. In the second print, I made the footprint of the dock slightly narrower to avoid pins bridging each other when connecting (I probably need diodes on the V and GND pins).

I sawed the threaded rod to match the clock thickness and filmed putting the clock together in a couple of takes. Kdenlive is a bit harder to use than I imagined, but I can probably make a short video.

For music I found Ambient Cinematic by AtlasAudio on pixabay.

The USB worked awesomely but broke after putting the clock together for the 5th (?) time…

I spent a lot of time writing the firmware today. All the logic is in place, but when it's packaged up and the sound is playing, the step response is very unreliable. It is unworkable because it is impossible to detect a hand on the clock. Need sleep, will fix tomorrow.

I fixed the step response by extending the pads to the outside of the housing using aluminum foil. A hand on the clock is now a lot easier to detect!

Plenty more tweak of the firmware and I downloaded some alarm sounds from Pixabay:

• Rooster
• Generic Alarm Clock
• City Alarm

In the evening (low light), I filmed my son using the alarm clock, and edited it into the presentation video.

The following parts are used to build this project:

The SBOM (Software Bill of Materials) is not that interesting. I used:

• MicroPython 1.28
• steptime.py by Neil Gershenfeld
• micropython-dcf77 (currently not used due to RP2350 DCF77 issues)

The clock contains all the "live" components: display time, sound alarm, and run on battery.

The dock is a simple 3D-printed box with acrylic screwed onto the sides to guide the clock into the right position. I wanted it to be bigger, but I printed it at home on my Prusa Mini, so I was constrained by the 18x18cm print bed.

Having spent much time on this project, I am pretty happy with the result, but there is still a lot of room for improvement. I finished it a couple of days before the deadline but have no energy left to work on these issues. Here are a couple of things I would like to improve:

The DCF77 does not work on the RP2350. I spent hours trying to get it to work but nothing success. I may try a XIAO ESP32 I have laying around, which also supports battery charging, to see if that works, or use an RP2040 and add some charging circuitry. This is really bugging me.

The step-response approach I envisioned for the snooze function is now "MacGyvered" using aluminum foil. Experimenting with step response in the input devices week already gave me an uncertain feeling about this technique, and having it in a box packed with electronics confirmed that step response can be quite challenging.

Making something pretty requires more care in material selection than I gave this project. The PCB copper is on full display because I think it looks nice, but there is a lot of color variation and fingerprints on it. Also, the PCBs and the rest of the interior are stained by superglue fumes; it is very ugly, and it is still outputting this white coating after several days.

The NeoPixels vary in color and brightness because they come from three different batches. They are also too bright, so I should have experimented with adding a resistor to the power line. I did introduce a 0Ω resistor in the schematic and the PCB to do that, but did not try changing it because it works now and I don't want to break it. Also, the diffusion parts I printed don't work very well. It is hard to see what time it is in a bright room. I need more experimentation on that, too.

USB access to the microcontroller is very convenient for debugging but at least as convenient is the ability to reset and put it in boot mode (to allow flashing it with a new UF2-file) without taking it about would be really nice too. In a next iteration that would be very nice.

The dock is too lightweight and should be mounted on something like wood to make it more stable when operating the knob.

Fab Academy was awesome! It's been a wonderful experience in which I met amazing people, and learned an incredible amount of new skills. A big thank you hug to everybody involved!