1 of 34

INPUT DEVICES RECITATION

HTMaA 2021 - Week 7

2 of 34

What we will talk about

What sort of things would you like to measure?
How to read a sensor

Reading analog inputs with an ADC

Example

Serial communication - more options, configurations

How to read and approach a sensor's datasheet
Example

Post processing

Noise
Filtering

Some basics of digital signal processing

3 of 34

What we will talk about

What sort of things would you like to measure?
How to read a sensor

Reading analog inputs with an ADC

Example

Serial communication - more options, configurations

How to read and approach a sensor's datasheet
Example

Post processing

Noise
Filtering

Some basics of digital signal processing

4 of 34

What sort of things would you like to measure?

Potentiometer - Rotary and linear displacement
Load Cell - Force and torque
Hall Effect Sensor - Magnetic field
Thermocouple / Thermistor - Temperature
Photoresistor - Light

Analog to Digital Converter (ADC) - Voltage* (an ADC is in the loop for most digital measurement). Here's how an ADC works.

Remember that you can mix and match! A light sensor can measure proximity and a hall effect sensor can measure displacement, etc.

5 of 34

What sort of things would you like to measure?

Cedric made this excellent HTMSTMaA writeup on sensors

https://fab.cba.mit.edu/classes/865.21/topics/metrology/03_sensors.html

6 of 34

What sort of things would you like to measure?

Analog vs. Digital (binary) inputs

7 of 34

What sort of things would you like to measure?

This recitation will use a hall effect sensor as an example, but everything here can be applied to inputs in general!

We are measuring magnetic flux (strength of a magnetic field)
Commonly used as displacement sensors

https://www.electronics-tutorials.ws/electromagnetism/hall-effect.html

8 of 34

Measuring an Analog Input

Let's look at some raw hall effect sensor data from the DRV5055

9 of 34

Measuring an Analog Input

An Arduino Uno has a 10 bit ADC. This means that we can resolve 2^10 = 1024 unique values for our measurement. Or 1/1024*100 =
Imagine you are measuring angle of a volume knob on a stereo:

2^8 = 256 possible values, so 360/1024 => 1.40 deg
2^10 = 1024 possible values, so 360/1024 => 0.35 deg
2^16 = 65536 possible values, so 360/65536 => 0.0055 deg

To measure at higher resolution, you typically need to take measurements slower
Some resolution will always be lost to noise
You probably won't be using the full range of your signal

So in practice, for a 10 bit resolution measurement, half might be lost to your signal range, and 10 counts to noise, leaving 1024*(0.5)/(10) = 51.2 true resolution.

10 of 34

Measuring an Analog Input

We are measuring a physical parameter represented as a voltage, then reading a value in bits, then back to some physical parameter

Do we need to calibrate with another sensor? Or will the datasheet just tell us?
Where is our zero point? Is our signal signed? If so where are we centered?

11 of 34

Measuring an Analog Input

Let's start by taring our measurement
We can take a measurement while it's in some known configuration
For a linear signal you technically only need to do this in two places

12 of 34

Measuring an Analog Input

Hysteresis is the dependance of our measurement on past measurements (usually from the immediate past)
Let's see if we can see some in real time

13 of 34

Filtering

Noise (unwanted disturbances) is easiest to see at steady state.
To address this, let's start by just averaging a bunch of measurements together

Moving Average Filter
Arduino Example

https://www.gaussianwaves.com/2010/11/moving-average-filter-ma-filter-2/

14 of 34

Filtering

We were able to reduce noise from 4 counts to 1 count!
Also oversampling is effectively getting us more resolution! (at the cost of bandwidth)

15 of 34

Filtering

What's the price of filtering?
We introduce a phase shift in our measurement

16 of 34

Filtering

We have to start to be careful if we are measuring a signal that contains a frequency close to (or above) our sampling frequency!
Nyquist limit is ½ the sampling rate

17 of 34

Filtering

We can actually engineer more sophisticated filters that target specific ranges of frequencies
If we overfilter we introduce a phase shift, so we have an incentive to design our filters carefully

https://www.electronics-tutorials.ws/filter/filter_2.html

18 of 34

Low Pass Filter

Step 1 - Determine your cutoff frequency and sampling frequency

Cuttoff frequency we want is 100Hz
We will sample at 1kHz using a delay(1)

Step 2 - Calculate your two coefficients

RC = 1 / (2pi * 100) = 0.00159
deltaT = 1/1000Hz = 0.001

Step 3 - Plug it into your loop!

Real-Time Software Implementation of Analog Filters

19 of 34

High Pass + Band Pass Filters

What if we want to get rid of low frequency noise (60Hz is a common example)? Or only pass data at a narrow range of frequencies?

High Pass

Band Pass

https://www.electronics-tutorials.ws/filter/filter_2.html

20 of 34

Hardware Filtering

Remember that filters can (and often are) be implemented in hardware!
Useful for DAC approximation with PWM

https://www.electronics-tutorials.ws/filter/filter_2.html

Our capacitor is 0.1uF
Our resistor is 15kΩ
RC is 0.0015 (same as last time)
F_cutoff ~= 1kHz

Teensy PWMs at either 3.6kHz or 4.5kHz

What do we see on the scope?

21 of 34

Linearity

Why might a hall effect sensor be a poor distance sensor?
If the magnet moves 1mm 10mm away from the sensor vs 2mm from the sensor

22 of 34

Linearity

Hall-effect example
Zero-point is at roughly 2000 ADC counts - note that our derivative is zero here!

If we move 1mm at point B, our signal will change only slightly
If we move 1mm at point A our signal will change a lot!

23 of 34

Linearity

To deal with nonlinearities you can:

Ignore and accept some amount of error (see straight line in last slide)
Fit an analytical model to your data and solve it every measurement (slow but memory efficient)
Make a lookup table (fast but memory demanding)

https://en.wikipedia.org/wiki/Lookup_table

24 of 34

Configuring Advanced Input Devices

Let's change gears and talk about configuring a more sophisticated sensor with its own logic and communication protocol.

We might want to do this because:

Analog signals are susceptible to noise

If our sensor is far from the ADC or there's noisy stuff near it...

Limited number of ADC channels to measure lots of sensors
Maybe our sensor allows us to configure/optimize it for our application?

25 of 34

Configuring Advanced Input Devices

Let's use TI's TMAG5170-Q1 3-Axis Linear Hall Effect Sensor With SPI Interface as a case study for the rest of recitation.

It measures flux in 3 different axes!
It can perform temperature compensation!
It can be configured to measure smaller flux with higher accuracy or higher flux with lower accuracy!

26 of 34

Configuring Advanced Input Devices

27 of 34

Configuring Advanced Input Devices

Here's how we communicate with this device, it happens to use SPI, but more on that in networking and communications week...

28 of 34

Configuring Advanced Input Devices

Look for a register map in the datasheet

29 of 34

Configuring Advanced Input Devices

Think of configuring a register as flipping groups of switches
Each register is composed of 16 switches
One switch could give us two states for one parameter (2^1 = 1)
Or 3 switches could give us 8 states for one parameter (2^3 = 8)

30 of 34

Configuring Advanced Input Devices

When we have everything configured, we will just ask the device what's in a result register

31 of 34

Configuring Advanced Input Devices

Let's look at some code and an example

32 of 34

Configuring Advanced Input Devices

If this feels overwhelming you can generally find an input device that fits your requirements sold by Sparkfun or Adafruit with a breakout board and a library already written for it.
For example if you want to measure force, the HX711 Load Cell Amplifier is a great place to start, and it has a very simple arduino library.
Let's look at their guide briefly

33 of 34

Configuring Advanced Input Devices

#include "HX711.h"

#define calibration_factor -7050.0 // This value is obtained using the SparkFun_HX711_Calibration sketch

#define DOUT 3

#define CLK 2

HX711 scale;

void setup() {

Serial.begin(9600);

Serial.println("HX711 scale demo");

scale.begin(DOUT, CLK);

scale.set_scale(calibration_factor); //This value is obtained by using the SparkFun_HX711_Calibration sketch

scale.tare(); //Assuming there is no weight on the scale at start up, reset the scale to 0

Serial.println("Readings:");

}

void loop() {

Serial.print("Reading: ");

Serial.print(scale.get_units(), 1); //scale.get_units() returns a float

Serial.print(" lbs"); //You can change this to kg but you'll need to refactor the calibration_factor

Serial.println();

}

34 of 34

Questions