Packaging, power, and telemetry
Sealing electronics against body fluid for decades, getting a hundred channels through the seal, powering an implant without wires, and sending the data out. The unglamorous problems that decide whether a brain implant is a product.
You are skimming: the title, the first figure, and the short version. Switch to Read in the header for the full page, or Deep to open every deep dive.
An electrode array is the visible part of an implant. Behind it is a sealed can with amplifiers, a radio, and a way to get power, connected to the electrodes through a seal that has to pass a hundred or a thousand wires without leaking for thirty years in warm salt water. Pacemakers solved this for a handful of wires in titanium cans decades ago. High-channel-count neural implants have not fully solved it, and the companies that do will define the field. This is the part of the field with the fewest students and the most open problems.
Hermeticity
Body fluid is warm, salty, and patient. Water vapour that leaks into an electronics package condenses, corrodes, and shorts. HermeticHermetic packagingSealing implanted electronics against body fluid for decades, one of the hardest unsolved engineering problems in the field. Glossary entry packaging keeps it out: a titanium or ceramic housing, welded or brazed shut, with a leak rate so low that the interior stays dry for the implant’s life. Pacemakers achieve this with a few feedthroughs (insulated pins through the wall). The test is helium leak detection; the standard is a leak rate below about 10⁻⁹ atmospheric cubic centimetres per second for a package of that size.
Feedthroughs, the hard part
Each channel needs a conductor through the hermetic wall. Pacemakers have four to eight. A Utah Array has 96, a modern array 1024 or more, and each feedthrough is a potential leak. Ceramic feedthroughs with many platinum vias, fired as one piece, are the current approach, at densities of hundreds per square centimetre. Alternatives put the electronics on the electrode itself in thin-film encapsulation (silicon carbide, alumina, layered polymers) instead of a can, trading proven hermeticity for scale, and measuring lifetime in accelerated soak tests at elevated temperature that stand in for years.
A thin-film polymer encapsulation passes a one-year soak test at 37 °C. How confident should you be that it lasts ten years in a patient?
Not very. Failure processes in encapsulation (water permeation, delamination, ion migration) accelerate with temperature roughly following Arrhenius kinetics, so soak tests at 60 to 90 °C are used to compress decades into months. The extrapolation depends on the failure mechanism staying the same at the higher temperature, and it has been wrong: coatings that passed accelerated tests have failed in animals within a year. This is why the pacemaker industry stayed with titanium and why every thin-film company’s central risk is the encapsulation.
Power
Primary batteries last five to ten years in pacemakers and DBS generators, then the generator is replaced surgically. Rechargeable implants (many DBS and SCS generators now) are recharged through the skin by inductive coupling, weekly or so. Fully wireless power to a skull-mounted or brain-surface implant uses inductive links at a few megahertz through a few millimetres of tissue, limited by tissue heating (the specific absorption rate limits) and by the power budget of the electronics, which is why implant amplifier chips are designed for microwatts per channel. A thousand-channel recorder streaming raw data needs tens of milliwatts, near the limit of what tissue heating allows.
Telemetry
Getting data out. A thousand channels at 30 kHz and 12 bits is 360 megabits per second, far beyond any safe implanted radio. So implants compress: transmit threshold crossings and spike-band power rather than raw waveforms, or decode on the implant and send only the result. Radios at 2.4 GHz through tissue are lossy and heat it; the medical implant band around 400 MHz and ultra-wideband links are used. Optical links through the skin exist. The data-rate budget shapes what a BCI can be, and the decision to decode on the implant is why the field talks about ultra-low-power neural signal processors.
Connectors and leads
Between the array and the can, wires flex millions of times a year and pass through the skull. Pacemaker leads fail at a few percent per decade. Neural leads with a hundred conductors are proportionally harder. Some designs eliminate the lead by putting the can on the array (Neuralink’s approach) or by going fully distributed. Every connector is a failure point; the fewer, the better; the original Utah Array’s percutaneous pedestal, a plug on the skull, is what the field has been trying to get rid of for twenty years.
MRI
An implanted patient who cannot have an MRI is a patient who loses access to the standard diagnostic tool for the rest of their life. Metal leads heat and displace in the scanner; MRI-conditional labelling, which specifies scanners and settings under which the implant is safe, is now expected of any new device and is a significant engineering constraint on materials and geometry.
Deep dive Testing an implant's lifetime 2 min
Soak tests in saline at elevated temperature with periodic electrical checks; helium leak tests; thermal cycling; mechanical fatigue of leads; biocompatibility per ISO 10993; and, before humans, months to years in large animals. A company’s test lab for these is as large as its design team, and the people who run it are engineers with exactly the characterization habits Phase 2 taught.
Explain what this page was about to your roommate in three sentences. No jargon they would not know.