[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Releasing RTS in RS485 mode
----- Original Message -----
From: "Steen Jansdal" <email@example.com>
Sent: Friday, February 14, 2003 13:14
Subject: Releasing RTS in RS485 mode
> I have problems getting /dev/ttyS3 to release the RTS soon enough in
> RS485 mode. The RTS are used to control the direction of the
> communication on the RS485 driver.
> Linux 2.4.19 with axis patches
> Mark/Space parity patches from Tamas Szabo ( Thank you very much Tamas,
> your patches are working great! )
> 115200 baud
> For enabling rs485 mode I've done the following:
> struct rs485_ctrl ctrl485;
> memset(&ctrl485, 0, sizeof(ctrl485));
> ctrl485.rts_on_send = 0;
> ctrl485.rts_after_send = 1;
> ctrl485.delay_rts_before_send = 0;
> status = ioctl(cb->fd, TIOCSERSETRS485, &ctrl485);
> And when I want to transmit I do the following:
> struct rs485_wrt io485;
> io485.outc_size = len;
> io485.outc = cmd;
> status = ioctl(cb->fd, IOCSERWRRS485, &io485);
> Using an oscilloscope I've measured some timings.
> On the first picture (twobytes.gif) you see two bytes (55h and AAh)
> with no parity. The time from the last stopbit until the RTS line
> is released is more than 120 us.
> Then I connected my devboard_lx to another device and sometimes
> I was able to read the proper data from the device but sometimes
> it failed, because the direction of the RS485 driver was changes
> too late.
> I captured a good (ok.gif) and a bad (fail.gif) situation.
> Is there anything I can do? Am I doing something wrong?
Make sure you have enabled CONFIG_ETRAX_FAST_TIMER
in the kernel config - that should make the time smaller and
If that is already enabled, it's a little trickier,
I guess that the interrupts are turned off for to long
sometimes - perhaps using RTAI and doing the RTS flipping
in a realtime task is the only safe way (or finding the place
where interrupts are turned off for to long)
I beleive the masters thesis workers that ported RTAI to ETRAX100LX
soon will publish the code (there were some review remarks that
Johan Adolfsson, Embedded Platform
Axis Communications AB