[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-rt
Subject:    FW: [rtl] rtl & robots
From:       "Prescott, Richard" <richard.prescott () lmco ! com>
Date:       1998-10-26 17:34:05
[Download RAW message or body]

>Prescott, Richard wrote:
>
>> Just one thing:
>>         DO NOT GENERATE PWM WITH A COMPUTER. It`s take too much cpu for
>> nothing.  Use electronic for that.  Calculate them but do not generate
>> them. I was using a PMD chipset for that (very cool!)
>>
>> I use to have 60 fps (2 x 30 fps) with a i486 66 MHz chipset Cirrus in a
>> known enviroment .  So...
>
>I do not agree, it depends on how fast is your PWM basic frequency and what
>resolution and noise you are willing to accept. An on/off fire requires about
>8 us on a 200 Mhz Pentium. So when that resolution is enough, let's say up to
>a PWM frequency of 5000 hz, and you can accept some noise it costs nothing,
>can be done on the parallel port. In such case putting it in hardware is
>needed only for production solutions not in the development phase, and you
>get
>a lot of flexibility, think of managing many PWMs simultaneously.  That is at
>the extreme but remeber that there are many slower apllication of on/off
>controlles, that's what PWM is, and in those cases doing it in software is
>far
>far simple and easier. Clearly I would not use Linux for a PWM of the type
>adopted in modern brushless or switched  DC motor power amplifier at 10-30
>Khz. The official RTL is at stakes in many low frequency PWM applications, if
>you have to do substantial calculation to determine the firing times,
>because
>of the use of the 8254 timer to set the time. That is another matter and we
>have improved it a lot.  Another point is the jitter you get if you insist in
>loading the control application with disk and net I/O and that is another
>point still. But saying "DO NOT GENERATE PWM WITH A COMPUTER" as an absolute
>statement is too much in many cases "IT CAN BE DONE".
>
>For sure it can be done. But: 
>
>time to set the 8254 = 40 x 1.7 us = 68 us
>time to output to the pp = 1.7 us 
>
>	5000 Hz x 70 us = 35 % cpu time
>
>Too much time for a robot using a framegrabber but perfect for a prototype
>testing fuzzy-logic stuff.
>
>Ciao, Paolo.
>
>
>See you.
>
>Richard
>
>
--- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail majordomo@rtlinux.cs.nmt.edu OR
echo "unsubscribe rtl <Your_email>" | mail majordomo@rtlinux.cs.nmt.edu
----
For more information on Real-Time Linux see:
http://www.rtlinux.org/~rtlinux/

[prev in list] [next in list] [prev in thread] [next in thread]