Because the board is not just a board. It is a way of turning vague “self-awareness” into an actual measurement-and-feedback loop, which is annoyingly more useful than vibes.
But the honest version is this: most people should not design their own human-connected PCB unless they need something specific. UART/I2C/SPI do not magically increase interoception. They are just plumbing. What helps is that a custom board lets you control the parts that actually matter:
-
You can trust the signal more. EEG is tiny, often in the tens of microvolts, and surface EMG is noisy even when its raw amplitude can reach the millivolt range. Signal quality depends heavily on input impedance, common-mode rejection, electrode impedance balance, cable motion, and line-noise handling. A custom PCB lets you choose the front end, grounding, filtering, connector layout, and electrode checks so your “neurofeedback” is less about your nervous system and less about your cable flopping around like a tragic noodle. (NCBI)
-
You can make the feedback loop fast and consistent. Neurofeedback and biofeedback depend on contingency. Timing matters. One review explicitly notes that the latency between behavior and feedback plays a crucial role in learning and should be as small as milliseconds. A custom embedded board can sample, timestamp, detect, and trigger haptics/audio/light locally instead of dragging everything through a laptop, OS scheduler, USB stack, browser tab, and whatever other modern indignities are in the loop. (Springer)
-
You can tailor the loop to your body and your task. Interoception is about sensing and interpreting internal bodily signals, but the evidence that biofeedback reliably improves interoception is mixed and task-dependent. Reviews on HRV-biofeedback and cardiac biofeedback find heterogeneous results, with protocol details and the measurement task making a big difference. That is precisely why custom hardware can matter: you can build a loop around the one signal that is actually meaningful for you, in the context you care about, with the feedback modality you can actually learn from. (PMC)
-
You can do real multimodal self-monitoring instead of fake omniscience. A custom board can combine a biopotential front end with IMU, respiration, PPG, pressure, temperature, buttons, haptics, and event markers. That matters because human signals are ambiguous. “I’m anxious,” “I blinked,” “my jaw tensed,” and “the electrode shifted” can all contaminate the same stream. Synchronizing multiple sensors on one board makes it easier to separate state from artifact. This is less mystical than people want, and therefore more useful.
-
You learn the signal by building the signal path. Designing the PCB forces you to learn what electrode impedance, lead-off detection, saturation, clipping, aliasing, 50/60 Hz contamination, reference choice, and motion artifact actually do. That can make self-monitoring better in a very unromantic way: you stop confusing measurement error with psychological truth. The board becomes a tutor in epistemic humility, which humans need in industrial quantities. ADS1299-class biopotential parts even include features like simultaneous sampling and lead-off detection precisely because these practical issues are central, not optional decorations. (Texas Instruments)
-
You can shrink the system until it becomes wearable enough to use consistently. A giant bench setup may be “accurate” but useless if it only gets used twice. Custom PCBs let you move toward battery-powered, body-near, lower-motion, application-specific form factors. For training, adherence often beats theoretical elegance. Humans, with their tiresome habit of not behaving like lab instruments, benefit from hardware that fits life.
The big caveat, because electricity attached to flesh is not a hobbycraft sticker project:
-
For EEG/EMG, safety is a first-class design requirement, not a later patch. Patient-connected devices require very low leakage current and proper isolation. Analog Devices notes that patient-connected equipment needs robust isolation, very low leakage, and isolation barriers on external connections to peripherals. Even seemingly simple home-health devices are subject to these concerns. (Analog Devices)
-
Do not treat ECoG like an enthusiast upgrade path. ECoG does have higher amplitude, higher spatial resolution, broader frequency range, and better SNR than scalp EEG, but it requires invasive surgery. That puts it firmly in clinical/research territory, not “I spun a PCB on a Saturday and now I’m cybernetic.” (PMC)
So the real answer is:
Design your own board when off-the-shelf hardware blocks one of these specific goals:
- lower latency,
- better artifact handling,
- tighter multimodal synchronization,
- better wearability,
- individualized feedback mapping,
- deeper understanding of what the signal actually is.
Do not design your own board just because “closer to the metal” feels spiritually advanced. Sometimes that path leads to insight. Sometimes it leads to a gorgeous four-layer artifact generator with a murderously bad ground return.
For most sane people, the best progression is:
off-the-shelf dev kit or proven biopotential front end first, custom PCB second, human-connected safety paranoia throughout.
For SPI/I2C/UART specifically: SPI is often the workhorse for high-rate biopotential ADCs, I2C is handy for slower auxiliary sensors and configuration, and UART is fine for simple streaming or debugging. The protocols are not the benefit. The benefit is that custom hardware lets you decide where computation happens, how synchronized everything is, and how much junk gets added before the feedback reaches you. (Texas Instruments)
If you want, I can turn this into a concrete design rubric for a safe EEG/EMG neurofeedback PCB stack, including what should stay off-board, what must be isolated, and what absolutely should not be connected to a human body.