Parity Bites Back at Thirty Kilometres

The APRS packet that came through at 28,400 metres looked perfect. The one at 800 metres was garbage—just enough corruption to pass the amateur radio network’s basic checks but not enough to actually decode. That’s the problem with telemetry from the edge of space: the atmosphere doesn’t corrupt data uniformly. It corrupts it almost correctly.

Checksums solve this, and they’ve been solving it since the 1970s when W. Wesley Peterson published his work on cyclic redundancy checks. The idea is beautifully simple: treat your data as a polynomial, divide it by a known generator polynomial, and append the remainder. If the remainder on the receiving end doesn’t match, something flipped.

Here’s a basic CRC-8 in Python, the kind you’d run on a microcontroller before keying up the transmitter:

def crc8(data: bytes, poly: int = 0x07) -> int:
    crc = 0x00
    for byte in data:
        crc ^= byte
        for _ in range(8):
            crc = (crc << 1) ^ poly if crc & 0x80 else crc << 1
            crc &= 0xFF
    return crc

packet = b"VE6SLP>APRS:28400m,-42C"
print(f"Packet: {packet}")
print(f"CRC-8:  0x{crc8(packet):02X}")

And the same logic in Zig, because when you’re writing firmware for a payload that might land in a frozen slough (ask me how I know), you want compile-time safety:

fn crc8(data: []const u8) u8 {
    var crc: u8 = 0;
    for (data) |byte| {
        crc ^= byte;
        for (0..8) |_| {
            crc = if (crc & 0x80 != 0) (crc << 1) ^ 0x07 else crc << 1;
        }
    }
    return crc;
}

The polynomial 0x07 is CRC-8-CCITT, common in embedded systems. Different polynomials catch different error patterns—some are better at burst errors (cosmic rays!), others at random bit flips (cold-induced oscillator drift).

The trade-off is overhead versus detection. CRC-8 adds one byte per packet but catches about 99.6% of errors. CRC-16 catches 99.998%. CRC-32 is what Ethernet uses. For a 300-baud APRS packet, eight extra bits per frame is barely noticeable. For a 1200-baud burst at -40°C when your crystal is drifting and the antenna is coated in ice, it’s the difference between knowing your payload landed in a field and knowing it landed in a slough.

I rebuilt the payload board last night with CRC-16. Next launch, if it goes silent at 800 metres again, at least I’ll know why I can’t decode it.