Skunkworks project ends up smelling like a rose

A fellow project team member encountered difficulty stabilizing a pneumatic valve developed from a skunkworks effort. His voice-coil-driven valve offered a new technology to our industry, and we were looking at the feasibility of using it as an exhalation valve in a new medical ventilator. The voice-coil motor would permit a fast lung-ventilator response to patient-initiated breaths, easing the patient's work of breathing. The fast prototype valve had the undesirable effect of overshooting the targeted patient pressures. The new valve not only had technical difficulties, but also plenty of detractors from management who wanted to use a slower, more proven valve. I have always pulled for the Sisyphus engineer, and I wanted to help him push his valve to the top of the mountain.

The mechanical-engineering group resisted my recommendations of attaching a transducer to the armature and feeding back the oscillations to the motor driver, correcting for the overshoot. They felt that it was too involved to attach a sensor and that a sensor might add too much weight to the motor's armature. The valve's poppet attached directly to the armature, and significant weight would affect performance. After designing the valve's power amplifier, I thought that a feedback signal would be a natural solution. The valve engineer attempted to stabilize the assembly with springs and pneumatic dampening devices, but the dampening devices deleteriously affected the valve's linear performance.

Using an oscilloscope to monitor the pressure sensor, I observed the airway-pressure changes when I applied a step function to the valve driver electronics. The velocity feedback immediately improved valve stability, but the system still needed some refinement. I placed a variable resistor in the velocity amplifier's feedback loop to allow control of the dampening, which I subsequently demonstrated to the valve designers by tuning in overdamped, critically damped, and underdamped responses. The stark performance improvement with the crude bar magnet and hand-wound coil convinced management that the approach was practical and that the voice-coil-driven valve was feasible.

Eventually, we refined the system to allow software-gain control of the feedback signal and usage of standard magnetic parts for the velocity signal. We found that controllable gain was desirable for the dampening because different airway-pressure levels required different dampening levels. After answering a bunch of "how-can-you-do-that?" questions from management, the valve quickly rose from a skunkworks project to full-development status.

This case was another one in my career in which I pursued an unpopular approach that all players endorsed once it proved successful. It's interesting how a few disparate parts, such as a paper magnet, pot-core bobbin, magnet wire, and super glue, can form a breakthrough. I eventually received a patent for this system, and the valve and electronics are in 23,000 ventilators across the globe.