Last week I got two beta versions of the Kantronics D4-10 UHF data radios. I've now had a chance to do some work with them and have some preliminary results. The hardware: 10W UHF (tuned for the 430 MHz range) with better receiver front-end than the 2M Kantronics radio. 2 channel, using a PLL locked to crystals in the 6MHz range. A single PLL is used for both TX and RX. Advertised T/R switch time is 2ms. The radio has an on-board data slicer and direct fsk keying, with TTL level inputs and outputs. It also has the same analog interface as the 2m radio. The receiver has a switchable bandwidth... the narrow position uses the same filter as the 2M 9600 baud radio (standard NBFM bandwidth, but with slightly broader skirts), while the wide position uses discrete components for a 60KHz bandwidth. The radio also has provision for AFC for satellite work. So far, most of my testing has been with a direct interface between the Ottawa PI board and the TTL input of the radio. After some problems with tcp parameters (thanks, Barry!), I got the two systems talking reliably to one another. Running at 19.2kb, I've gotten ftp throughputs of 1700-1800 cps on a 120 kilobyte file (copying from ramdisk to ramdisk to minimize i/o delays). I'm getting reliable results with a TXD of 5ms and a TXTail of 1ms. Using one of the systems on a switch machine ethernetted to my unix box, the performance is good enough to make "echo accept" in telnet mode livable. I've also experimented at 28.8kb, and the system works but there are quite a few more bad packets (and/or acks getting stomped on) and the best throughput I've seen has been 2200 cps. More typically, I'm getting 1900-2000 cps at this speed. Some of this may be due to the fact that the fsk is 20KHz, which may be on the narrow side for this speed. I'm also not sure if the 8530 in the PI card is able to reliably do clock recovery at this speed; I'm going to build a null modem between the two cards to test that theory. At 28.8, a TXD of 3ms seems to work... it appears that the limiting factor is the number of bits necessary to get the 8530 dpll clock recovery in sync (thus, a longer TXD at the lower bit rate). The next step is to try the radios with DataEngines and modified RUH modems (bypassing the analog portion; essentially using only the scrambler and clock recovery circuits -- a preliminary design from Kantronics). I've had some success with this combination, but heavy traffic like ftp causes missed packets and acks, increased backoff, and general crud that brings things to a standstill. I also plan to get a real-world path (instead of across the hamshack) running soon to see how things work with lower signal levels. The exciting part of this is that the PI card and radio worked together with absolutely no tuning or other diddling. All I had to do was build a cable (5 connections) and set up autoexec.net. A plug-and-play solution to higher speeds may finally be at hand. I'll publish a more detailed report once I've had a chance to do some more tweaking and to get the DataEngine/DataRadio combination working better. If anyone has any suggestions for tests to perform, or ways to improve throughput, I'd be glad to hear them. John Ackermann AG9V jra@lawday.daytonoh.ncr.com