Before 1925, listening to a radio or phonograph was miserable. Sound came from horn-loaded reproducers or magnetic "reed" units driving cardboard cones — narrow-band, honky, distorted, and violently colored by mechanical resonances. Playing music meant tolerating whatever frequency range the horn happened to pass.

Then two engineers in the General Electric Research Laboratory in Schenectady, New York — Chester W. Rice and Edward W. Kellogg — did something almost nobody had done for a loudspeaker: they treated it as a physics problem.

Their July 1925 patent application, granted in 1929 as US Patent 1,707,570 ("Sound Translating Device"), and their accompanying AIEE paper "Notes on the Development of a New Type of Hornless Loud Speaker," described what we now just call a speaker:

• A lightweight paper cone attached at its apex to a small cylindrical voice coil.
• The coil suspended in the annular gap of a strong permanent (or field-coil) magnet.
• An audio current through the coil producing a Lorentz force that pushes the cone as a rigid piston.
• A soft, compliant surround so the cone's mechanical resonance sits below the audio band — meaning the driver operates in the "mass-controlled" region where response is inherently flat.
• A large flat baffle (later, the sealed cabinet) to separate front and back radiation and preserve bass.

They also worked out the uncomfortable corollary: this design is inefficient. Roughly 1% of electrical input becomes acoustic output. That was fine only because vacuum-tube amplifiers had just become powerful enough (see de Forest's Audion) to shove watts into a coil. Rice and Kellogg essentially married two 1920s technologies — the permanent-magnet motor and the audio power amplifier — and produced the template every listening device would use for the next century.

How modern is this patent? Open any speaker made in 2026 — a $30,000 studio monitor, a $2 earbud driver, the woofer in a Tesla, the ceramic-magnet cone in an Amazon Echo, the ~6 mm micro-speaker in an iPhone — and you find the same four parts in the same arrangement: cone, voice coil, magnet gap, compliant surround. The resonance-below-band rule Rice and Kellogg derived is still how loudspeaker engineers pick a driver's Thiele/Small parameters.

The reach goes further than audio:

• Haptic actuators — the "Taptic Engine" in iPhones, the LRAs in Apple Watch and game controllers, the linear vibration motors in Android phones — are voice-coil motors driving a suspended mass. Direct descendants of the 1925 topology, retuned to a single resonance.
• Hard-drive head actuators that swing read/write heads across a platter are miniature voice-coil motors.
• MEMS microspeakers now emerging in true-wireless earbuds (xMEMS, USound) still model their radiators against the mass-controlled piston equations Kellogg published.
• Active noise cancellation in headphones only works because the driver's response is predictable in the mass-controlled band — a property engineered in by Rice and Kellogg.

The most manufactured transducer in history — an estimated tens of billions of units — is essentially their 1925 sketch, shrunk and stamped in polymer instead of paper.