Decode 2-byte floats using nominal min/max scaling.

Map fixed-point CAN payloads back to physical units so cell_group voltages and other 2-byte signals no longer read as raw uint16 garbage.

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: sbasu107

src/decoder.c

@@ -85,12 +85,22 @@ int decoder_extract(const signal_def_t *sig,
                 double v; memcpy(&v, raw, 8);
                 out->value = v;
             } else if (nb == 2) {
-                /* Some entries in format.json declare 'float' with 2 bytes
-                 * (e.g. pack_voltage). Treat as unsigned 16-bit scalar. */
+                /* 2-byte floats are fixed-point values scaled across the
+                 * signal's nominal min/max range (see encode_signal.py). */
                 uint16_t v; memcpy(&v, raw, 2);
-                out->value = (double)v;
+                double span = sig->nom_max - sig->nom_min;
+                if (span > 0.0) {
+                    out->value = sig->nom_min + ((double)v / 65535.0) * span;
+                } else {
+                    out->value = (double)v;
+                }
             } else { /* nb == 1 */
-                out->value = (double)raw[0];
+                double span = sig->nom_max - sig->nom_min;
+                if (span > 0.0) {
+                    out->value = sig->nom_min + ((double)raw[0] / 255.0) * span;
+                } else {
+                    out->value = (double)raw[0];
+                }
             }
             return 0;
         }