Joint Movement Sensors - Prototyping

[digitaldanny]

Describe the solution you'd like

• Build the flex sensors and measure their resistance ranges. Measuring the ranges will help decide if the signal needs to be amplified before low pass filtering.
• Basic SW to select mux channel and read ADC values.

Additional context Research for this issue was completed in #22.

[digitaldanny]

If the resistance range is low, the output voltage range the ADC can capture will be low. A small number of ADC steps between a straight joint and a bent joint will create inaccurate joint angle estimations. This problem can be solved using an op amp as a differential amplifier to subtract out the minimum voltage and amplify the area of interest to 0 - 3.3 V. For example, if my voltage range is 2.51 - 2.56 V, I can subtract 2.51V and amplify the remaining 0.00 - 0.05V range to 0 - 3.3 V. This would result in a larger number of ADC steps for the voltage range of interest. I can design the amplifier using the formula shown below to provide the max voltage range 0-3.3V.

[digitaldanny]

Copper tape and pressure sensitive fabric came in today (02/05). Picking up thin plastic sheets tomorrow.

[digitaldanny]

I made a ~1 inch long sensor to test that the change in resistance is high enough to sense. I taped the flex sensor to a glove under the pointer finger's MCP joint, and I put the flex sensor in a voltage divider with a potentiometer. I was seeing the largest voltage range with the potentiometer set at 1.2kOhms.

As I bent my finger, the voltage increased from ~2V to ~4.5V. This is a much larger voltage range than I was expecting, and I think it will be plenty to work with on the microcontroller. There was some very small noise, but a moving average filter might solve that problem without the need for low-pass filter hardware.

The minimum value was a little inconsistent (~1.8 - 2.2 V). I'm hoping adding a thin plastic sheet to the outside of the sensor will help return the sensor to a consistent starting point. Will test that out tomorrow when I pick up the plastic sheets.

[digitaldanny]

Was having trouble getting usable voltage readings from some sensors I made today before realizing that the adhesive side of the copper tape does NOT conduct like it was advertised that it does. I made a new sensor with this in mind, which is working much better.

• Currently seeing a resistance range of ~225 Ohms when fully bent to ~12 kOhms when straight.
• Due to the non-linear voltage divider formula R2 / (R1 + R2), a very small R1 means that the output will be basically 1 for a large portion of flex sensor's resistance range. This log curve is shown below.
• A possible solution would be to offset the resistance range of R1 using another resistor in serial. In the screenshot below, I added a 10k Roff resistance to push the resistance range up to ~10.225k - ~22k Ohms. This resistance range should work well with an R2 of 20kOhms.
  • The bad news is that this would be required for each of the 10 flex voltage dividers.
  • Unfortunately, the screenshot below still shows a very logarithmic curve.

Problematic curve using base resistance range

Better curve using offset resistance range

Some other thoughts..

• Try building a sensor using 2x as much plastic sheet to see if this help return to original form better.
• Try building a sensor using folded pressure fabric to see if I can boost the resistance range without using resistors in series.

[digitaldanny]

Seeing some weird behavior when using my flex sensor in the mux breadboard.

• Screenshot 1 shows that the flex sensor's voltage range is 0-0.7V when selected by the mux. There also looks to be some voltage drop on the output of the mux for some reason.
• Screenshot 2 shows that the flex sensor's voltage range is 0 - 2V when NOT selected.
• Don't think this will affect anything since the amplifier could always boost more and the microcontroller will still see 0 - 3.3V, but it's still worth noting for later reference.

0 - 0.7 when selected

0 - 2 when not selected

[digitaldanny]

GOAL CHANGE

The final demo will be utilizing only 3 joint sensors to control echo time, (possibly) reverb time, and toggling the effects lock. Because of this, I will most likely limit my project to 3 joints being captured. The mux will still be useful because all amplification and filtering hardware can use the single mux output as the input. This limits the amount of hardware on the glove PCB.

[digitaldanny]

Log and Anti-log Amplifiers - This video analyzes anti-log circuits in the later half and provides a formula for creating the output curve. The issue is that this circuit requires a negative supply voltage. Right now, the plan is to use a single positive voltage supply so I will be avoiding this solution for now.

I'm going to try amplifying the flex sensor's output voltage range to reach rail-to-rail voltages. On the MSP432, I'll create a LUT of ADC values that determine a specific joint position and interpolate between to estimate the joint's position. This might be the best way to linearize the joint data before sending out to the client.

[digitaldanny]

Completed the following goals for this issue today.

• Built a differential amplifier to boost the voltage range of interest from the flex sensors. I thought I would be able to boost from the current 0-1V to 0-3.3V, but I did not realize that the Vout swing for the LM324N is 0 to Vcc-1.5V. Because of this, I'm seeing a 2V max unfortunately so my gain is only around 2x rather than 3.3x. This should still be enough to take some good ADC readings. If not, I may be able to use one of my rail-to-rail op amps for higher gain.
• Also created a 2nd-order LPF to attenuate frequencies higher than 10Hz. This definitely smoothed out the curve and will make for more stable ADC readings, but it also means there will be some significant delay for discharging the capacitor. The screenshot below shows about 400 ms delay.

TODO:

• I'm going to focus on getting the ADC readings on the MSP432 before worrying about reducing this delay. Maybe I can try this design with smaller capacitors and larger resistors.
• Add a voltage divider and voltage follower to control the differential amplifier's subtractor voltage. Currently powering this input with the DAD board's waveform generator.

[digitaldanny]

Quick update - Reducing the caps and increasing the resistors decreased the final discharge time from >400ms to ~130ms. This should be a lot less noticeable for the end user. I think it may be worth trying to drop this down a little more if possible. Will post a schematic of the current design tomorrow after I add the offset volt divider / buffer to the circuit.

[digitaldanny]

Focusing on charger PCB before coming back to this.