Valves and Manifold Design

The Team:

• MarcStrauss?
• Ken

Current Work:

Collaborate with Ken

Proportional / Digital design

• Research possibilities with weight and rigidity

Plumbing

• Thru joint plumbing

Electronics

Valve Driver Board

This PCB is embedded in the robot arm. It serves a number of functions:

• Condition analog driver signal from interface board (split to either pressurizing OR exhausting valve)
• Use the conditioned signal to drive current to open and close the valves
• Amplify load cell signal, pass signal back to PC
• Pass pressure sensor signal back to PC

The PCB lives pressed right up to the valve assembly, and communicates to the interface board via a 10-pin Molex-terminated ribbon cable. Use the freeware from ExpressPCB to view the layout and schematic. Valve Driver BOM

The command sent to the Valve Driver Board is distributed to EITHER the pressurizing (if > 5V) or exhausting (if < 5V) valve. The magnitude of the (conditioned) signal sent to the valve is the (absolute value of) the difference between the command voltage and 5V. In this way, for example, if the command is 6V it is condiitoned to a 6 - 5 = 1V signal to the pressurizing valve, and if the command is 3V, it conditioned to a 5 - 3 = 2V signal to the exhausting valve.

Interface Board

This PCB powers and communicates with the Valve Driver PCB in the robot arm through the 10-pin Molex-terminated ribbon cable. There is no active circuitry on the interface board: essentially it merges the PC I/O with the power supply rails. There are four SPST switches which are used to turn the computer control to each muscle ON or OFF. The OFF position connects the control signal to +5V, which is the neutral (both valves closed) command voltage for a muscle.

Use the freeware from ExpressPCB to view the layout and schematic.

Valve Characterization

With the flexibility of computer control, a multitude of possible experimentation with the valves is possible. The refined experiment determined the minimum voltage threshold to open the pressurizing valve as a function of muscle pressure. The threshold data recorded was the lowest voltage (resolution 0.01V) to raise the pressure in the muscle by 10 PSI within 10 seconds. This was run on a range from 0 to 50 psi, and a linear model resulted: Vcommand - 5V = -.0117857 * Pmuscle + 1.906

The same experiment was performed for the exhausting valve. In this case, the model turned out to be: 5V - Vcommand = .0282857 * Pmuscle + 1.38 Note that both models are expressed in terms of the magnitude of the voltage signal away from 5V (neutral). The slopes have differing signs, which makes sense - for different values of muscle pressure, the pressure differential moves in opposite directions for the valves. This is because the pressurizing valve is connected to a compressed air reservoir (90 PSI) and the exhausting valve is connected to atmospheric (15 PSI).

I also measured the pressure rise time as a function of valve command voltage for a variety of starting and ending voltages. These curves look much like a decaying exponentials. This means that there is a certain voltage command above which the flow rate is essentially the same. According to Dongjun's analysis, this maximum voltage command appears to remain a constant value above threshold voltage, regardless of muscle pressure. For the pressurizing valve, the maximum voltage command is about 1.1V above threshold, and for the exhausting valve, the maximum voltage command is about 1.5V above threshold. These can be considered as the size of the active range of the valve. Below this range, the valve is closed; above this range the valve is fully open.

Load Cell Characterization

The load cell signal is amplified by an INA118 instrumentation amplifier with a gain of 500 (Rg = 100.2 Ohm). Our new load cells have a different response characterstic, so I've removed the old model from this page. The response of the load cells was measured using a handheld force meter. The load cell output was amplified, sampled, and displayed on the computer. Accounting for our running the load cell at half-voltage (5V) and a 500X gain from the INA118 instrumentation amplifier, I was able to build a linear model for the load cell. Extrapolating this model to the 300 lb. limit, we roughly confirmed the manufacturer's calibration data for each cell. Since the data was taken from human-induced transient forces (and taking each datum when the puller said "now!"), I am not comfortable enough with its accuracy to publish a new model.

Proportional Control

A gain of 0.5 is appropriate for the proportional control of the pressure. A gain of 0.5 is appropriate for the proportional control of the force.

-- MarcStrauss? - 31 Mar 2008