I replied to this tweet and Espressif sent me this nifty “ESP32 Meshkit Button” to play with with MicroPython, let’s have a look …
Inside the ESP32 MeshKit Button
The device consists of a little snap-together clamshell holding a C-shaped board with four mechanical buttons and four RGB LEDs. The big cutout in the center is to hold a 200 mAh LiPo with JST-SHR-2P connector.
There’s lots more information about this device available on github:
I’ve used some excerpts from the schematic here to illustrate my investigations of the circuit.
I’m quite surprised, disappointed even, that this device doesn’t use the ESP32’s excellent capacitive touch sensing to detect touch through the plastic front panel. Instead, we’ve just got plain tactile buttons.
Tactile Buttons Circuit
(Note: the schematic shows BUT4_A
connected to GPIO33
but it is actually
connected to GPIO39
… according to the silkscreen, the button_driver.c
code
and experiment …)
There’s little symbols molded into the top, although they’re quite hard to see:
position | icon | GPIO |
---|---|---|
top left | IEC power | 39 |
top right | light bulb on | 34 |
bottom left | Sleep | 32 |
bottom right | light bulb off | 35 |
There’s some interesting circuitry around those buttons though: they’re powered directly
from VBAT
and there’s a little diode-OR circuit to power up the LDO if any of the buttons
are pressed. So if the CPU pulls PWR_ON
low, and there’s no VUSB
, and there’s no button pressed,
the LDO will turn off and the CPU too. This has got to be the ultimate in power saving.
Diode OR circuit
There is some interesting “Dickson charge pump” circuitry in there to drive the G and B channels at a higher voltage than would otherwise be available. Perhaps they’ll be brighter at 50% PWM than at 100%!
Charge Pumps
The R channel doesn’t get a voltage doubler, but then again red LEDs typically have a smaller voltage drop so it doesn’t need it. And there’s a set/reset flip-flop arrangement on the red channel, presumably so it can be locked on while the CPU is unpowered.
Flip-flop circuit
There’s a charging circuit to charge the little LiPo from a USB port. The USB port doesn’t carry data though, just charge. I have a couple of batteries around the right size but without the right connector, so I’ll just run it from external power for the moment.
Back of the ESP32 Meshkit Button
On the back is a pair of holes for a wall mount plate. These don’t really clip in, just sit there. There’s also a tiny programming port.
Programming Port
Here it is without the case. This is a mini header, 2 x 3 pins at 0.05” / 1.27mm pitch. I can totally understand why manufacturers want to go to a connector an eighth the volume of the more common 0.1” headers, and I’d love to see a standardized header for serial programming ports. But these connectors are pretty rare in hobbyist land still, so it’d be great to include one in the box.
Programming Port Header
I ended up mutilating a 10P x 1.27mm cable for now, and connecting the other end to an 8P 0.1” male header like an ESP-01 so I could use a serial converter I already had. If you’ve been messing with microcontrollers for a while you’ll probably have a million of these lying around.
Port | ESP-01 Pin | Diode? |
---|---|---|
2V8 | K | |
GND | GND | |
IO0 | GPIO0 | |
EN | RST | |
TXD | TX | |
RXD | RX | |
GPIO2 | ||
CH_PD | ||
VCC | A |
Blanks mean “no connection”. It’s a pity the power pin is to VDD_2V8 … it’s useful
to have the device powered for programming but the 3.3V out of the serial converter is
dragged down to 2.8V by the output side of the LDO.
I’m thinking of adding a diode in there to drop the voltage so the device can be
powered from the programming cable, but in the meantime it is sufficient to connect
power through the main charging port as well.
Programming Cable (very nasty)
With this cable in place, and power applied through the built-in USB port, I can run
esptool
, which identifies the device as a ESP32D0WDQ5 (rev 1) with 4MB flash.
I can also connect to the USB port at 115200 baud and see logging messages as the device
initializes and as it logs button presses. The auto-reset circuit even works!
This is MIT Licensed, implemented in quite readable C and available at the github repo above.
I’ve talked about the Internet of (Not Shit) Things and about L2IoT so I’m very excited to see what Espressif have made of ESP-MESH
The next step is to load MicroPython onto the device in place of the shipped firmware, and see what we can come up with. I just grabbed the latest master and build it and here’s the familiar banner:
MicroPython v1.9.4-788-gf874e8184 on 2019-01-25; ESP32 module with ESP32
Type "help()" for more information.
>>>
The buttons on GPIOs 32, 34, 35 and 39 are quite easy to read:
>>> import machine
>>> p = machine.Pin(32, machine.Pin.IN)
>>> p.value()
0
>>> p.value()
1
When it boots, the blue LED channel is slightly on, because GPIO14 is on. How about if we try out those voltage doublers?
>>> blue_pin = machine.Pin(14, machine.Pin.OUT)
>>> blue_pwm = machine.PWM(blue_pin, freq=1000)
>>> blue_pwm.duty(512)
… as predicted, the blue LEDs are brighter with a 50% duty cycle than they were when solidly ‘on’, ditto the green channel on GPIO26.
The red channel is a bit different, as it has no voltage doubler circuit and therefore the brightness is roughly linear with PWM duty cycle.
The flipflop behaviour is a little more mysterious, perhaps because I’ve not got a battery hooked up yet. It looks like it can be flipped on to turn the red LED channel on while the CPU is shut down.
Further work is needed!