Chair Rick Chinn opened this meeting with a brief report on the AES Amsterdam convention. Then, as is our custom, all attendees introduced themselves.
Bob Smith, Staff Scientist for Medtronic Physio-Control of Redmond, WA, (and owner of BS Studios by night) related the story of the audio development for one of his company's heart defibrillators. Heart defibrillators are electronic medical devices that shock an improperly beating heart back into a regular rhythm. A growing market for these devices is placement in public gathering areas such as airports, like fire extinguishers or first aid kits, where they can be used by untrained personnel in a heart attack emergency. The units are completely microprocessor controlled and cannot send a shock unless needed. The standard way to guide the user is with prerecorded voice prompts. The on-site sound is also recorded for later analysis. Not surprisingly, audio is often treated as an afterthought in the design, and regulatory, physical and electronic constraints can be severe.
Bob felt compelled to "sell" the need for some good audio engineering on this defibrillator to management. The voice prompts are critical in a noisy, tense situation. Bob played a security video synched to the defibrillator audio/heart data recording during an actual resuscitation to demonstrate the environmental situation. He also demonstrated the unit used in the video, and notably, the prompts were somewhat underwhelming and distorted, and the level of the logging audio fluctuated due to a poor AGC.
Some design changes were obvious - choose a decent speaker, and place it on top, pointed at the user, instead of inside the box at the end of a channel. Don't put the logging mic way inside at the end of a long plastic channel. Some things required some research and testing, such as codecs, bit depths and sampling rates, and compression. Noise measurements were made on real aid calls, in aid cars, a ferry boat and emergency rooms. Analyzing this helped determine the requirements for better playbacks and recordings.
The internal recording mics are the familiar little aluminum electret capsules. However, a waterproofing membrane on the case, electrical isolation requirements, and mounting on the PC board at the end of a plastic channel meant the old unit was actually a Kazoo. Its simple AGC made gross volume changes every time a prompt played. A new unit placed the mic near the case surface, away from the speaker, used a much improved AGC, a better codec, higher sample rate and greater bit depth - and works much better.
The audio system in such medical products must be higher grade than consumer products, which limits design and component selection. The power budget was about 1 watt, since this was a portable device. Bob demonstrated how a few intelligent design changes greatly improved the loudness and clarity of the playback system.
Voice prompts (in 21 languages) are voice actors, not synthetic. Great care was paid to recording of the voice prompts with proper urgency, cadence, delivery and diction. Recordings needed to be made with later compression and encoding in mind.
Bob ran the older unit (with a simulator) to hear all the prompts, then ran the newer unit, with very obviously improved loudness and clarity. It used the same audio amp, but had about 12dB difference, according to Bob. A modest additional amount of memory was needed.
A grab bag of door prizes was doled out to the 14 attendees, including some items from AES Amsterdam courtesy of Rick Chinn, and plenty of Medtronic swag such as keychains with CPR moisture barriers inside.
We were reminded that audio should not be an afterthought in such products, but a discipline applied to the initial design.
-Reported by Gary Louie, AES PNW Secretary