The Raspberry Pi (RPi) Pico offers a convenient and cost efficient way to create a portable resonance measurement device that can be used across different Klipper printers.
ADXL345/343 boards
There are different ADXL345 boards on the market and also the ADXL343 board is a valid alternative.
Depending on the board layout and used components a few things should be considered when connecting the board.
General
The boards can be mainly distinguished by potentially having a Voltage Regulator or a Level Shifter.
Voltage Regulator
The Voltage Regulator allows the board to be connected to either 3.3V or to 5V
The ADXL chip itself requires 3.3V. The voltage regulator ensures that only 3.3V actually reach the chip
In the following images, a Voltage Regulator is marked with a green rectangle
WARNING: Boards without such a Voltage Regulator must not be connected to 5V or the board will be destroyed
Level Shifter
Digital electronics, like MCUs (aka printer boards) or RPis require a certain voltage on their pins
Typical voltages in 3D printing are either 3.3V (most common) or 5V
A Level Shifter allows to translate between 3.3V and 5V or vice versa
The ADXL family is in the 3.3V domain by default. This means that it outputs only 3.3V even when connected to 5V (see voltage regulator).
A Level Shifter changes this: When the ADXL is connected to 5V (see Voltage Regulator) it will also output 5V on the board’s connecting pins
In the following images, a Level Shifter is marked with a yellow rectangle
WARNING: RPIs and many MCU boards are designed for 3.3V only. A board with Level Shifter must not be connected to 5V and to a partner tolerant only to 3.3V. This will destroy the partner.
I2C Boards
Most boards use the SPI as communication bus. SPI is fast enough to deliver up to 3200 samples/s, as it is needed for resonance measurement
Some boards seem to be forced into I2C mode (another bus type) with a pull-down resistor
Such boards do not work for resonance measurement but can potentially be reworked by removing the pull-down resistor
After un-mounting, the Pico should automatically reboot with the new firmware
Configuring Klipper
Get the correct serial path with ls /dev/serial/by-id/*
Add the following to the printer.cfg file:
[mcu pico]
serial: /dev/serial/by-id/usb-Klipper_rp2040_mySerial
[adxl345]
spi_bus: spi0a
cs_pin: pico:gpio1
[resonance_tester]
accel_chip: adxl345
probe_points:
100,100,20 # an example
Restart Klipper with the RESTART command.
Important: The first measurement, after freshly connecting the RPi Pico or restarting Klipper, will fail. This is due to some issues with the SPI initialization. Subsequent measurements will work.
Available SPI buses on the RPi Pico
It is possible to connect multiple ADXL boards to one RPi Pico:
ADXL345 pin
RPi Pico pin name
spi0a
spi0b
spi0c
spi1a
spi1b
VIN (or VCC)
3V3 Out
36
36
36
36
36
GND
GND
38
38
38
38
38
CS
CSn
2 (GP1)
7 (GP5)
22 (GP17)
12 (GP9)
17 (GP13)
SDO
SPI RX
1 (GP0)
6 (GP4)
21 (GP16)
11 (GP8)
16 (GP12)
SDA
SPI TX
5 (GP3)
10 (GP7)
25 (GP19)
15 (GP11)
20 (GP15)
SCL
SPI SCK
4 (GP2)
9 (GP6)
24 (GP18)
14 (GP10)
19 (GP14)
Notes:
Header is the name of the bus, values given as pin-number (GPIO number)
@glabifrons thanks for sharing.
One comment: There are many different ADXL boards out there. Some are only 3.3V tolerant.
Your wiring will potentially kill such boards as you are using a 5V pin
I configured everything as per instructions, however I’m getting an error when querying the ADXL345. ‘got f2 vs e5’
I know that the PICO and ADXL345 are working because I can access it via microPython (via SPI) just fine.
Few questions:
Where does spi0a come from? Given gpio notation, it seems to me it should be pico:spi0? Could it be that my other mcu SPI is being used?
EDIT: as per reddit linked above CS pin determines mcu being used, so I suppose the correct SPI port is being opened…
How can I bypass device ID detection?
Is there anything else I can try to troubleshoot this?
EDIT2: turns out I just needed to give it some time before spi starts working.
For reference it appears it took about 30 seconds for SPI port to be available for some reason:
00:30:02 !! Lost communication with MCU 'pico'
00:30:09 $ FIRMWARE_RESTART
00:30:34 // Klipper state: Ready
00:30:42 $ ACCELEROMETER_QUERY
00:30:42 // Invalid adxl345 id (got f2 vs e5).
// This is generally indicative of connection problems
// (e.g. faulty wiring) or a faulty adxl345 chip.
00:30:42 !! Invalid adxl345 id (got f2 vs e5).
00:31:04 $ ACCELEROMETER_QUERY
00:31:05 // adxl345 values (x, y, z): 2677.215450, -1988.788620, 15910.308960
Heh, I take it back. It appears that first access after reboot results in an error. No matter the time that have passed.
@glabifrons: For your guide: There is another potential pitfall with connecting to 5V. Some boards have a Level Shifter which will bring the output to 5V. In this case not the ADXL might get destroyed but the Pico.
To connect 2 ADXLs I use the following pins. For the second ADXL you will need to use GPIO8, GPIO9, GPIO10 and GPIO11 according to the colors. Red and Black are the same.
@Sineos , I’d like to take a stab at updating the Klipper Measuring Resonances documentation to reflect the information you provided here as it was really helpful to me. I’d love to use some of the images you posted in this thread.
I believe that means you’d have to release the images either to the public domain or under Klipper’s license, the GPL3. Do I have your consent? You are certainly under no obligation to release your images and even if you don’t, I very sincerely thank you for posting this really helpful information.
I’m a big fan of the WTFPL - Do What the Fuck You Want To Public License
But of course. As the official support board of Klipper, all my contributions here are automatically considered as common to the project and as such I would consider them under the same GPL v3.
Removed original response since my problem did not have anything to do with this post.
With the help of this info, I was able to do this with a MPU9250 sensor and PICO as well. I used a short run of twisted pair wire taken from a cat5e cable. It’s important to note that most spooled ethernet wire is solid core wire that is not meant to be flexed repeatedly. It will cause problems in this kind of application. Patch cables (like those you get with new cable modems) typically have stranded wire.
Sineos asked for confirmation on using the other available SPI buses. I managed to mess up spi0a and spi0b by getting the wires wrong and filling adjacent holes with solder. Anyway, I got spi0c working just fine. Config change was obvious -
I am adding my notes on the errors returned by the Pico based on what I learned in making so many mistakes in the hope that it may help others -
Send: ACCELEROMETER_QUERY
Recv: !! Invalid adxl345 id (got ff vs e5)
This is the error I got when I did not have any wires connected to the correct GP pins.
Send: ACCELEROMETER_QUERY
Recv: !! Invalid adxl345 id (got 00 vs e5)
This is the error I got when I did not have the correct cs_pin identified in the config file.
Thanks for sharing this. I’m not sure though, if there is a real system behind this. I have seen various error messages all across for various reasons.
Especially nasty are these where basically everything is fine but the sensor is unhappy because the cable is not made of pure gold with a triple screen made of Kryptonite
Very true - correlation is not causation. If nothing else it is a list of almost everything that can be done wrong. The only thing I may have done right is have a relatively short Cat 5e cable between the accelerometer and the Pico ( its about 300 mm long). The USB cable is plenty long to reach the keystone on the back of the printer. I got about half the noise readings compared to the long cable that I originally used to connect directly to the Pi GPIO. Although again I have not done enough research to know why the noise levels changed and it could be something else.
WARNING: Boards without such a Voltage Regulator must not be connected to 5V or the board will be destroyed
Important: The first measurement, after freshly connecting the RPi Pico or restarting Klipper, will fail. This is due to some issues with the SPI initialization. Subsequent measurements will work.
