Digest V4 #464 / William T Wilson comments

[Ken Tompson]

Hi All,
I have been following along for about a week and am finding I need to
get in 15 mins earlier each day in order to have a quick sniff through
the digest.
The question, "Will a PC be able to run an engine using Windows
operating system?"
I think not.
If you can get it running properly, I will take off my shoes and walk to
Sydney in bare feet, a trip of about 3000 miles. Look, I am sure the
engine will run, but so does a K-Jetronic or sprint car with its garden
hose sprinkler system.
We sample combustion pressure at rocket fast speeds using a PC. We make
an encoder for 360 pulse per rev, or 720 ppr (plus tdc, or other
reference), 10,000 rpm? on the 1/2 degree, no problem.
With a serial cable deal? you are really flogging a dead horse, man this
is one huge bottle neck.
Windows for real time control?
The main problem here is that there are only one million micro seconds
in a second. So if we give the serial cable the flick, we need to
digitize the analog data into RAM, so that the Intel can do something
with it. WINDOWS HAS A HUGE PROBLEM AND THAT IS WINDOWS, but other
operating systems are not a whole lot better at resolving our real time
PC nightmare. Windows interrupt latencies is a big drama, the engine
needs an answer sooner than the PC is prepared to give it. Data will
enter the PC memory via one of the following pathways:-

  1. A generated interrupt, telling the PC to stop what it is doing and
     go take a look at whatever. Very SLOW!
  2. The software can be set up to go poll whatever you want looked at.
     Again the poor old Intel is dragged to it's knees.
  3. Direct Memory Access. Dma is efficient at getting the data into
     memory, but it is still a slug, unless you can buffer your data on
     the hardware you are using, not going to give the Intel a real time
     picture, is it? The extra piece of hardware is triggering the
     sampling (not the Intel Pentium). A PCI add-in with a bus master
     controller, is a slightly different twist. So now we are not
     limited to ISA Dma availability, but this was not our problem.

The tightest code that you will write, will be with a WATCOM C/C++
version 11 compiler. You can set up your globals to act as a server for
the various events taking place. We have a virtual dyno controller set
up exactly like this, so I know the PID update rate that is possible.
The faster the Pentium, the greater your chances of course.
We are able to sample at 1 Mega HZ and even much faster, but then the PC
is merely a tool to display collected data, the add-in hardware with its
on board firmware and co-processor does all the work, not the PC.
I once had a little steam engine that ran. Just add water and methylated
spirits and it went round and round and up and down, but that is about
all it did. A bit like our 8 stroking PC controlled engine.
I had planned to stand back and watch the experiment unfold. As a newby
I was not sure whether I was supposed to shut up and observe, but after
reading and agreeing with the comments of  W. Wilson, I guessed that it
would be permissible for me to put in my 20c worth?
I think this is a great venue for exchange of ideas, but a lot of work
for the people doing the organizing.
Regards,
Kevin Jaeger.
email     kjaeger at superflow.com.au
PS Remember the engine is not going to want to stop for us to take it's
picture.