The characterization report
One page that says what your instrument does, with numbers, methods, and a plot per number. The document that turns a project into an instrument and a student into an engineer.
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.
Why one page
A datasheet is a characterization report written by the manufacturer. Yours is the same thing, written by you, about the whole system: chip, board, cables, electrodes. It is one page because that is what anyone will read, and because if you cannot state what your instrument does in one page, you do not yet know. Every row is a number, the conditions under which you measured it, and a pointer to the plot.
The template
INSTRUMENT: 8-channel EEG amplifier, rev B DATE: 2026-11-02 AUTHOR: ...
Configuration: ADS1299, gain 24, 250 SPS, internal reference, bias drive on,
inputs via 2.2 kΩ series + touch-proof connectors, Ag/AgCl cups with Ten20.
Metric Value Conditions / method Plot
Input-referred noise 0.92 µV RMS inputs shorted at connector, 60 s, 0.5–40 Hz fig1
Noise with electrodes 2.1 µV RMS in phantom, dipoles off, 60 s, 0.5–40 Hz fig1
Noise spectrum shape 1/f below 3 Hz Welch, 4 s windows fig2
Bandwidth (−3 dB) 0.3 – 64 Hz phantom dipole sweep, 10 µA fig3
Gain accuracy +1.8 % 10 µA at 10 Hz vs. forward-model prediction fig3
CMRR, bias drive off 78 dB 100 mV 60 Hz on phantom vs. reference fig4
CMRR, bias drive on 104 dB same fig4
Crosstalk, adjacent channel −68 dB 20 Hz on dipole 1, ch2 vs ch1 fig5
Input range (linear) ±180 mV amplitude sweep to 1 % deviation fig6
Sample drop rate 0 in 150,000 10 min, frame counter n/a
Battery life 9.5 h 3×AA, streaming over Wi-Fi n/a
Known limitations: 1/f noise dominates below 2 Hz; CMRR falls to 84 dB at 180 Hz;
channel 6 shows 0.4 µV more noise than others (suspect input resistor).Rules for each row
A number with units. “Low noise” is not a row. “0.92 µV RMS” is.
Conditions sufficient to repeat it. Bandwidth, duration, what the inputs were connected to, the gain and rate. Someone with your board and your notes should get your number.
A plot for every number that has a shape. Noise gets a trace and a spectrum. Bandwidth gets the sweep. CMRR gets the injected and recorded spectra on one axis. Plots go on a second page or in an appendix; the first page stays one page.
Comparison to the spec. Where the datasheet claims a number, put it in a column and compare. Your CMRR will be lower than the chip’s because the electrodes are in the loop; say so.
Known limitations, honestly. The row that impresses a reviewer is the one where you found something wrong and characterized it instead of hiding it.
What it is for
It is what you send a professor when you ask to join their lab, instead of “I’m interested in BCI.” It is what you bring to Blackrock or Ripple when you ask for a tour. It is the verification section of the mock design history file in Phase 5. It is the appendix of your first poster. And it is the habit that, more than any other, marks the difference between people who make things that work on the bench and people who make instruments.