HP Integral Modem -- trials and tribulations

Does anyone know anything about using the HP Integral modem, or have software that's _known_ to work with it? Today I tried the modem board (HP 82915) in my Integral. The machines powers up OK, and the load_modem command finds it and loads the driver. But any attempt to access /dev/modem00 (using stty, kermit (on the super-utilities disk), etc) hangs the machine. I decieded to do some investigating. Probing points on the I/O cards is difficult with the machine assembled, and I don't have an extender board. There is, however, a way to get good access to the boards, but it involves a little work. Remove the rear cover and rear screening plate over the logic assembly. Unplug the 2 cables from the back of the floppy drive, remove the 3 screws and the drive itself. Remove the eject button and spring before you lose them. Undo the screws holding the logic chassis in place. Unplug the power cables, HPIB cable, expansion backplane cable and HP-HIL cable from the logic boards,Leave the other (printer) cables connected, move the logic chssis towards the top of the machine. Remove the HP-HIL grounding nut (through a hole in the expansion backplane), then the 4 screws holding the PSU/expansion box in place. Take that out Undo all the screws and nuts holding the cover on this unit. Move the cover out of the way Undo the nut on the backplane earthing bracket, remove the 4 screws, the bracket, and the backplane. Reassmble and refit the PSU box (you don't need all the screws), plug the cables back into the logic assembly (you can leave off the HPIB cable if you're not using any HPIB peripherals at the time), plug in the expansion backplane cable with the backplane hanging behind th PSU box. Fit a couple of the logic chassis screws. Refit the floppy drive, hold it in place with at least 1 of the top screws. Recoonect its cables. The machine will run like that, and you can plug the expansion board into the backplane and probe pins on it. Now, to understand the next bit, you need to know a little about the design of the modem. The address decoder is simple. It responds to the slot-specific chip select line from a decoder on the CPU board, and further decodes lines A6 and A5. It provides 4 chip-select signals A6, A5 = 00 : Command/ID register. Bits 4-7 (only) used on write, to select interrupt line, force and enable interrupts. On read, bits 4-7 are much the same, bits 0-3 are a fixed ID value that identifies this as a modem. A6, A5 = 01 : Select 8250 serial chip. Address lines A1-A3 select the 8 registers in this chip in the usual way A6, A5 = 10 or 11. One selects the modem Tx chip (16 internal registers selected by A1-A4), the other selects the modem Rx chip (also 16 registers, selected by A1-A4). Anyway, after probing pins on the address decoder circuit and repeateded powering down/up the machine and running load_modem, I discovered that the only 'register' that this command accesses is to read the ID register, presumably to find a modem board. It does not attempt to initialise the serial chip or the modem ICs. I have the customer diagnostic disk. This also finds a modem, and I can run the diagnostics on it. It pases. It also seems to write things to various registers on the modem board (in particular, at the end of the test, the 8250 is left set to 1200 baud, the baud_out pin (wired to the Rclk input as usual) is running at 19200Hx). And it energises the line relay a couple of times. Any thoughts? Do I have an obscure hardware problem or just not the right software? -tony