EMI Interference

Abstract

   With our current plan of transmitting data over long wire, the possibility arises that they are susceptible to noise generated by Electromagnetic Interference. Since our solar car will be running 4 hi-power brush less motors, their huge AC power spikes can induce heavy amounts of noise into our system. This introduced noise can wreck havoc into our wires, causing data loss and/or inaccurate sensor measurements.Thus this objective is to measure the effects of EMI and determine ways to eliminate it from our system.

The setup

    Our experimental setup follows: an arbitrary voltage DC source that acts as our "signal" is connected on one end of a long wire. We then induce an EMI on the long wire and measure the voltage output from the other end of the long wire. The following image shows the setup on breadboard:  

The "signal" source is a basic DC power supply that will supply 3.3V to our long wire. This will function as an equivalent of a digital "high" signal. A 2.2M resistor is placed in series with the long wire the limit the current that passes through the wire, this will simulate the signal line.

The oscilloscope probe is placed on the node connecting the long wire and the 2.2M resistor. This will measure the resulting "signal" after it traveled over long wire.

To induce the EMI, another wire is wrapped around the long "signal" wire. It is then connected to the output of a function generator. I should be noted that this wire is left floating and it is not connected to anything else but the function generator output. By providing and AC voltage to the output of the function generator, the change in voltage will induce current into the long "signal" wire. The purpose of wrapping the wire is the maximize the induction of current onto the long signal wire.

On the function generator, the output was set to a 16 MHz 10Vpp sine wave with the output load set to high-Z. The image shows the following output parameters:

Measurements

Without the interference applied, by turning off the output of the function generator, the image shows the measurement from the oscilloscope. The DC power supply is on and set to 3.3V output.

 This will form as our basis for the measurements.

The output from the function generator is also measured. This was achieved by attaching a 2.2M resistor in series with the function generator output and probing the voltage across the resistor with the oscilloscope. The image shows the resulting waveform:

The measured Vpk-pk is not exactly 20Vpk-pk as desired, but this error is not significant for the analysis.

With the EMI applied to the long "signal" wire, the measured output follows:

The output result is probably grossly exaggerated, but this should be appropriate enough to form a comparison with the filtered signal output. 

Common Mode filters

One possible solution to this problem is using a common mode filter, by taking two wires and removing voltages that a common between then, it can be used as a way of elimination EMI. So far two types of common mode filters are tested: Differential Amplifier, and bypass capacitors.

Differential Amplifier

With the use of a differential amplifier, the signals that common to both inputs (V1 and V2) can be suppressed from the output Vout while the difference between V2 and V1 are amplified. In this case we only want to eliminate noise that are induced in our signal wires, if R1=R2, Rf=Rg and the op-amp is ideal, the theoretical output follows the equation: Vout=Rf/R1*(V2-V1).

By setting R1=R2=Rf=Rg, we will have a unity gain amplifier and thus have Vout=V2-V1. We can make use this feature to eliminate EMI by taking two wires and only passing the signal voltage on one of the wires. By inducing EMI on the two wires, they can be suppressed by the unity gain amplifier and have only have our desired signal on Vout.

Testing setup

Using a Texas Instruments TLV314 op-amp, the unity-gain amplifier was configured using 2.2M resistors with Rout being a 10M resistor. The Op-amp was powered using 3.3V and -3.3V sources respectively. The oscilloscope is probed across Rout.What is not shown in the image is that the function generator's output is wrapped around the long "Signal" wire to act as our EMI source. We desire the EMI to be induced equally along the two wires.

I also tested the amplifier using 100k resistors and with 0.1uf capacitors coupling the Vsig inputs to GND.

Before I began the measurements, I tested the approximate resonant frequency of the long wire to see at which frequency will the highest Vpk-pk will be induced into the long wire. The function generator was set to a sine wave with 20Vpk-pk output with output load set to High-Z. While these measurements are perhaps over-exaggerated compared to more practical measurements, but it does help minimize other uncontrollable noise factors such as 120V AC noise. For now this will be considered the worst case scenario. From what I found a 27 MHz frequency gave me the highest Vpk-pk that was induced onto the "signal" wire.

Additionally, the 3.3v "signal" was connected to Vsig+. This will represent our signal which should show up in our Vout.

Measurements

With 2.2M resistors

Unfiltered

Filtered

WIth 100k Resistors

Unfiltered

Filtered

With 100k Resistors + decoupling capacitors to GND

Unfiltered

The screenshot appears to be missing. But its measured Vpk-pk was 7.5Vpk-pk

Filtered

Bypass Capacitors

This circuit also makes use of the difference between two voltages. The noise that is induced into both Vin+ and Vin- will be shorted to the GND by the capacitors, additionally, any voltages common to both Vin- and Vin+ will be cancelled out at Vout if the voltage is measured about the resistor. It should be noted however that this GND is an Earth GND, which is separate from the common reference GND.

The Setup

In this setup Rout=10M and the capcitance for the bypass capacitors are 0.1 uF. A 3.3v "signal" is passed into Vin+ with EMI induced onto both Vin+ and Vin- wires. The oscilloscope was placed across Rout (Not shown in diagram). The Earth GND was connected to the green "Earth" terminal on the power supply.

Measurements

Unfiltered

Filtered

Results

Button Debouncing

Abstract

Switches are not prefect components. Sometimes they "bounce" and their state will not be a single transition from one to the other, but transition multiple times before settling to the new state. This is undesired as they are unpredictable and can cause incorrect readings into our driver input. Thus the objective is the measure the behavior of the button bouncing and develop hardware solutions that will prevent the button readings from bouncing. Software solutions will not be a focus for now.

The Setup

Two types of switches will be used for the measurements: a button switch and a toggle switch. As the basis for the measurements, we will first measure the unfiltered button signal. The schematic shows the initial setup:

Vout is where we connect our oscilloscope probe to provide the measurements.

After the measurements are made, two circuits are tested that are used in attempt to provide debouncing, the schematics show the following solutions:

For this measurement, R1 is a 10k resistor while R2 is a 380r resistor.

The switches that were used are in the following image:

The one on the left is a toggle switch while the one on the right is the button switch.

The capacitor used in this circuit is a 10000Pf capacitor:

Measurements

Circuit 1: Unfiltered Switches

Button

Toggle

These images show the cases when the switches bounced. It should be noted that multiple measurements were made but not all of the measurements showed cases of button bouncing. However the frequency of those bounces are not being recorded for this study.

Circuit 2: Capacitor in parallel with the resistor

Button

Toggle

The addition of a capacitor created a second order RC circuit here. It happened that the system is Underdamped and it is creating these voltage oscillations (this is further studied in ece 240 and in calculus). This is not a desirable result since these voltage oscillations can also cause unwanted readings into our digital circuits or in worse case: damage the input pins by providing a voltage that is too high.

Circuit 3: Additional discharging diode

Toggle

The button measurements for this test are missing. However this measurement shows an issue with this circuit: the bouncing is still prevalent. Perhaps by allowing more current to flow into the capacitor and reducing the current that discharge the capacitor, the results would improve.

Circuit 4: Circuit 3 but with the resistors swapped

Button

Toggle

By having R1=10k and R2=380k the results have definitely improved. The result is a much smoother curve, but we some oscillations are still present in the curve. The circuit could be further improved with the use of a schmitt trigger.