The signal from the vibration sensor is displayed by the built-in AD convertor of ESP32. The program is a very simple one that I found on the Internet, and it processes the signals imported into the ESP32 board with ArduinoIDE, and displays them graphically on a serial monitor. It is a short program, isn't it? This is all you need to create a simple seismograph! However, displaying the data on the LCD monitor on the board and saving the data to the SD card are still to be done. This work will be done tomorrow or later.

Everything went well, and strangely enough. Naturally, the ADC input of the ESP32 is only for positive voltage, so I thought of using a battery (1.5V) to bias the output of the vibration sensor (i.e., the input of the ADC). A simple circuit that forcibly solders a battery holder on a cheap amp I have on hand, extracts +1.5V, and raises the output of the vibration sensor. With this simple circuit, the output of the vibration sensor is within the range of 0 to 3.3V, and the output waveform is properly displayed. I am quite proud of this idea. Even the original Raspberry Shake was surprised at this simple specification.

I found the ADC input of ESP32 from the Internet, and found it is IO35 as you see in the figure. I made a circuit to insert a signal between this and the GND next to it. The socket and cable are 1.25mm pitch 4P and somewhat special, so I got them from AliExpress.

This is the finished product. Signals from the board are inserted into the board using a 1.25mm pitch 4-pin socket. The signal from the vibration sensor is inserted into the two bare pins on the other side of the board using worm clips. Super simple design. The vibration sensor does not require a power supply, so the other extension terminal on the board (far right in the photo) is not used. The signal goes through the yellow (GPIO35 pin) and green (GND pin) jumper wires on the auxiliary board. Red and black lines are wired but not used. The signals can be displayed as shown in the movie by specifying the pin numbers in ArduinoIDE. (2024-02-23. Note that the Japanese technical certification mark "Giteki技適"is printed on the ESP32's IC.)

However, I anticipated that the sensitivity of the vibration sensor by itself would probably be insufficient, so I soldered an amplifier in parallel with the construction. This is a circuit I wrote for a newsletter of the Seismological Society of Japan in the past.

I sent the CAD data to a cheap Chinese circuit board maker and outsourced the work, which cost less than 200 yen per piece. One of the two channels will be used, and although it is an OP amp, it should be usable with a 5V single power supply (from USB). This work will be done after tomorrow.

For the time being, the signal is displayed on the LCD. The sensitivity is still small, but periodic small electrical noises can be confirmed. There are still many issues to be solved, such as checking time marks and sampling speed, connecting to an amplifier, and saving data to an inner SD card.

Here is another video. I reduced the magnification rate of the signal and added a time mark (in seconds). It currently takes about 20 seconds to cross 480 pixels horizontally. 24Hz sampling is used. Free-run mode can go up to 48Hz. I found that ADC alone can sample at high speed while writing on the screen. In the previous microbarometer, the sampling rate was suppressed because of the slower sensor reaction speed of microbarometer IC tip.
http://www.yossi-okamoto.net/2022_Matsue/GeoSciEd9_Barometer_Poster_E_v1.pdf

We checked the sampling rate by free-running without delay() in the program. To my surprise, I found that I could sample at about 480 Hz, the full width of the LCD. With this, high-speed sampling of microseismic data (higher than 10 Hz) is no problem! I wonder how much the sampling rate will drop while writing data to SD.

近くの南海高野線の列車の走行音を2回記録している．記録時間は1秒．ラズベリー・シェイクと同等の地震を記録できる世界最小のデジタル地震計でありながら、それ以上に安価であると自負している！

This is a full view of the ESP32 seismometer assembled on the desk in my room. From left to right: seismic sensor (natural frequency 4.5Hz), amplifier board (modified from the following page), auxiliary board, and board with ESP32 LCD. The amplifier circuit is available at https://www.zisin.jp/publications/pdf/nf-vol103.pdf#page=6

The running noises of a Nankai Koya Line train nearby was recorded twice. The time mark is 1 second. I am proud to say that this is the world's smallest digital seismograph capable of recording earthquakes as same as Raspberry Shake yet more than cheaper!

Desk in my room with seismographs. This is a very common environment. Both the amplifier and the board are powered by a 5V USB power supply. The board seems to be powered down to 3.3V inside the board. The maximum output voltage of the amplifier is set to 3V by resistor voltage divider to meet the ADC standard of the board.

This is the circuit (slightly modified version of the "Naihuru" article) that I made for a seismometer prototype with the students of the attached high school, which was made by a cheap Chinese board manufacturer. I used the extra pins for the output resistor voltage divider and LEDs. It is comfortable that I prepared some extra pins in the blank space.

This is the raw data in the first AD conversion. It is so noisy that it is useless. The time marks are also buried and not visible. Each signal line is about 20 seconds long, and there are 30 signal lines in total, which is equivalent to 10 minutes of recording per screen.

This is a moving average of three data of AD conversions. It is a little better, but still not enough. After this, we tried averaging 5, 7, and finally increased the number of averages to 15. The speed of this AD conversion is extremely fast, and it is very surprising that the sampling speed was not affected much by the averaging operation even a cheap CPU circuit 2400JPY.

Finally, the moving average of the 15 data is displayed on the screen, and is also saved on the SD card. After about a month of struggle, I managed to save this screen capture as a BMP file. However, the address is still slightly wrong, and about 6 pixels wide on the left side of the screen is wrapped around to the right side of the screen. Furthermore, I tried to upload this screen image to my FaceBook, but FB rejected it. It seems that there is something wrong with the BMP file. The image has been converted to PNG and uploaded. I lifted the sensor and vibrated the floor in the room to check the sensitivity.

This is the numerical data stored on the SD card at the same time. The time and the numerical data of 12-bit AD conversion (originally from 0 to 4096) are displayed in [mV], so they are a little smaller.

This is the part where the sensor was moved by hand to see the min and max of the data. This data shows the AD conversion value in [mV]. Theoretically, the average should be around 1500[mV], falling between 0 and 3000[mv], so the output voltage divider or amplifier adjustment may still be wrong. The fact that the lowest value is not 0 seems to be a quirk of the board's AD.

This is a part of the ArduinoIDE program for controlling the seismometer. It is based on some of the source code that has been published on the Internet by various pioneers. I posted a raw capture of the program a while ago, but I deleted it as soon as I realized it, because my room wifi ID and PW is written on the raw capture image!

This is the part where 15 ADC data are averaged, and the moving average data are displayed and written. The absolute time is written to the SD card along with the signal level. Because this ESP32 chip can get the Japan Standard Time from their wifi.