Velostat Pressure Sensing (January 2025)
========================================

In late 2024, into early 2025, an exploration was made into the possibility of using velostat as
a means for cost effective load sensing.

The resistance of velostat changes with pressure. When sandwiched between conductive material, it
can be used as a pressure sensor.

A First Attempt
---------------
Here is a load cell built from copper tape, washers and velostat. It is then sealed with packing
tape:

.. image:: binary/washer_sensor.jpg
    :alt: A sandwich of packing tape, copper coated washer, velostat, copper coated washer,
          packing tape.

These were glued to the bottom of a gardening cup. The purple lettuce plant was then placed into
the cup.

.. image:: binary/washer_sensor_cup.jpg
    :alt: Velostat washer sandwiches installed on the bottom of a gardening cup. A plant that
          will go into the cup is also shown.

Data was collected via Arduino.

.. image:: binary/washer_sensor_data.png

Other sensors were investigated.

.. image:: binary/sensor_misc.jpg

Using Machine Learning to Take Data
-----------------------------------
This one is 12 gauge copper wire, wrapped in two layers of velostat strips. The velostat layers are
wrapped in a copper strip layer. This copper velostat ring sandwich is compressed between two
regular mouth canning lids in a canning ring. The sandwich is then compressed with a disc of
particle board. There is a 220 ohm resistor in line.

Expanded form:

.. image:: binary/ring_sensor_expanded.jpg

Assembled form:

.. image:: binary/ring_sensor_assembled.jpg

The above velostate ring sensor was placed on a kitchen scale. 5 volts was applied to the following
resistor divider.

.. image:: binary/velostat_circuit.png

A photographic time lapse was taken of 1. the kitchen scale display and 2. a voltmeter displaying
Vo in the circuit above, using the above velostat ring sensor.

The data was read from the pictures via a trained
`Tesseract <https://github.com/tesseract-ocr/tesseract>`_ models.

    - `mm.traineddata <https://github.com/highvelcty/growbies/blob/main/tesseract/mm.traineddata>`_
      # For reading the multi-meter (voltmeter) display.
    - `scale.traineddata <https://github.com/highvelcty/growbies/blob/main/tesseract/scale.traineddata>`_
      # For reading the kitchen scale output

Here is the fixture. Water was added while a timelapse was taken of the data.

.. image:: binary/data_timelapse.jpg

|

Here is a sample picture of the voltmeter provided to the model

.. image:: binary/mm_data.png

Here is a sample picture of the kitchen scale:

.. image:: binary/scale_data.png

Here is the collected data plotted:

.. image:: binary/ring_sensor_vs_kitchen_scale.png

The bowl was filled with water, then siphoned close to empty, filled with water and siphoned once
more.

The arrows show the progression of the data over time.

It is interesting that there are two paths the data follows when weight is being added. There are
also two paths followed when weight is removed, although less pronounced.

Other Data
----------
Here is another sensor being testing with a test without machine learning. Only a few samples of
weight are available due to the laborious nature of gathering the data. Yay machine learning!

An R1 value of 100 ohms was in place for the duration of this testing.

.. image:: binary/ring_sensor_sweep.png

|

Another rendition of the ring sensor hardware.

.. image:: binary/ring_sensor_sweep_2.png