// Package kpa talks to the Elecraft KPA500 and KPA1500 amplifiers.
//
// One package for both: they share the Elecraft command set — ASCII, a caret
// prefix, a semicolon terminator, case-insensitive on the way in and upper case
// on the way back — which is the same family as the K3/K4 panel in
// internal/cat. What differs between the two models is the transport and which
// commands exist, not the grammar.
//
// # Transports
//
// KPA500: serial only.
//
// KPA1500: serial, and a network server. Four things may be connected AT ONCE,
// which is unusual enough to design around — the Host PC USB port, the XCVR
// SERIAL connector when repurposed as a second host, ONE TCP client, and any
// number of UDP clients:
//
//   - TCP on port 1500 (changed with ^CP). Single client. If the operator
//     already has the Elecraft utility or another program on TCP, OpsLog will
//     not get in, and the failure is a refused connection rather than anything
//     the amplifier says.
//   - UDP on the same port. Many clients, one command per packet and at most
//     one response, and packets may be dropped under congestion — so it is the
//     right choice for sharing the amplifier and the wrong one for a command
//     that must not be missed.
//
// # Pacing
//
// There is NO flow control. The reference is explicit: pace commands by waiting
// for the response to the previous one. So this client is strictly
// question-and-answer on one connection, like the ACOM and SPE clients, rather
// than firing a poll cycle and sorting out the replies afterwards.
//
// # Serial speed
//
// 4800 to 230400, 8N1, set on the amplifier (^BR / SERIAL SPEED HOST) and not
// negotiated. Elecraft's own utility finds it by sending bare semicolons at
// each speed until something answers — worth copying if operators turn up with
// amplifiers whose speed they do not know.
//
// # What is settled, and how
//
// From the KPA1500 Programming Reference:
//
//   - ^SW is the SWR IN TENTHS. "123 is 12.3:1", so ^SW015 is 1.5:1. This was
//     first taken from Hamlib's backend and is now confirmed by the document,
//     which matters more than it sounds: a wrongly scaled SWR bar reports a
//     good match on a bad antenna.
//   - ^WS returns forward power AND SWR together, ^VI returns PA voltage AND
//     current together. Two round trips instead of four on the link the display
//     depends on while the operator is transmitting.
//   - ^SF returns the fault log: index, fault code, a short name in quotes, a
//     timestamp, and fault-specific values. ^FC describes the codes.
//
// # Not touched
//
// ^TX simulates a KEY IN — it makes the amplifier transmit from software — and
// ^ON0 switches the main supplies off. Neither belongs on a poll loop or behind
// a button that can be pressed by accident.
package kpa