> From: Luc Hermans <[EMAIL PROTECTED]>
> > Hi,
> > In my embedded linux application I've written a modbus interface (RS485)
> > on /dev/ttyS1 (cfmakeraw) the RTS is used to enable/disable the RS485
> > driver.
> >
> > I use the ioctl(fd[c], TIOCMSET, &status) fucntion.to change RTS'
> >
> > Before sending I enable RTS, send all data, (the RS485 echo's all)
> > wait until all data is received, and then disable RTS.
> >
> > The problem is that the RTS is sometimes dropped to slow, up to 5ms,
> > witch is to long (because I get an answer from the polled device withing
> > 3ms)
> > but sometimes also fas, lesser then 500us.
> >
> > What is the cause of this ?
> > I work with an i386 40MHz cpu at 9600 baud
>
> Interupt latency and kernel to application latency...
>
> This is the biggest pain in the butt with RS-485. After you send your
> data you need to flip the direction of the transceivers. The hardware
> that people try to use for RS-485, is pretty strapped to do the job
> correctly (ie in hardware).
>
> Basically if you want any chance of making this work you will need to
> handle the ISR yourself and you need to flip the direction control
> (aka RTS) in the ISR. Note: You can NOT use the transmit interrupt to
> do the flipping because the TX ISR will happen when the last byte
> begins to shift out of the UART, not when it's finished shifting
> out!! The way most people handle this is to listen to what you are
> transmitting, and flip when you receive your last byte.
>
> So we are to the point where you are going to rewrite the serial port
> driver. That shouldn't be too bad since it's just a minor mod.
>
> The good news here is it's a lot easier to write a driver for Linux
> than it is for say, Micros~1 windows.
>
> But you still have the problrm of interrupt latency. You're new
> driver, although it works a lot better, will still NOT be bullet
> proof. Why... because you are still at the mercy of ethernet and
> harddrive drivers that may disable interrupts for longer than you're
> maximum letency requirement of 500uS.
>
> To get around this you might want to look at the RT-Linux stuff. With
> that you can make you RS-485 ISRs top dog in the processing chain.
>
> http://www.rtlinux.org/~rtlinux/
>
> Good luck...
>
> Hey by the way, MODBUS ??? That sounds like industrial automation,
> PLC stuff right. I've been watching with anticipation for a while and
> so far your one of the first that I've seen playing with Linux in
> that arena. I'm happy to see that not everyone in the industrial
> automation world is in bed with WinCE. Have you ever done anything
> with DeviceNet?
>
> -_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_
> -_ Gary A James
> -_ Critical Link, LLC Voice: 315.425.4045 x218
> -_ 404 Oak St. Fax: 315.425.4048
> -_ Syracuse, NY 13203
> -_
> -_ E-Mail: [EMAIL PROTECTED]
> -_ WEB: http://www.criticallink.com
> -_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_
> -_ __ _
> -_ Try... / / (_)__ __ ____ __
> -_ / /__/ / _ \/ // /\ \/ /
> -_ /____/_/_//_/\_,_/ /_/\_\ http://www.linux.org
> -_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_
Hi Thanks all for your tips
I have solved this problem by modifying the driver serial.c,serial.h, ioctls.h
and add a new ioctl( TIOCMRTSOFF, x) call which tells serial.c to
swtich RTS off after receiving x chars. (done in the ISR)
This works the following way ex. sending a command of 10 chars.
- put RTS on (RS485 tx enable) ioctl( TIOCMSET)
- call ioctl( TIOCMRTSOFF, 10) if I had to send a command of 10 chars
- write ( the 10 data chars)
- after the 10 chars are echoed (because the RS485 receiver is allways
enabled)
the RTS will automaticly switch off.
I you want this modification I will be happy to give it.
Luc
--
To unsubscribe from this list, send a message to [EMAIL PROTECTED]
with the command "unsubscribe linux-embedded" in the message body.
For more information, see <http://waste.org/mail/linux-embedded>.
