linear EFI

[Darrell Norquay]

At 09:33 AM 8/21/96 -0500, Tom Cloud wrote:

>To answer your question, yes, I have such a system.  It is the
>Holley aftermarket EFI. 

Here's my $0.02 ($0.015 CDN) worth on the subject, then I'll shut up.

The basic Holley system is analog / digital, but no uP chip.  As I said,
essentially an electronic carburetor.  Nothing wrong with this, but not as
adaptable as a full blown EFI.  Sufficient for most applications, and
definitely easier to tune and just as reliable as a carburetor.

>I work with an engineer who designed an EFI for racing jet skis
>using hybrid analog / digital 
>He has supported me in thinking that, unless trying to achieve zero
>emissions, simplicity (i.e. no micro-controller) is the best.

Who says you can't have a simple microcontroller system?  This guy sounds
like an old school type, knowledgeable about analog but scared of
microprocessors.  In fact, digital systems can be (and most of the time,
are) less complex (hardware wise) than analog systems.  Imagine trying to
build a completely analog soundblaster card...

>I observe that analog, though the building block everything else
>is built upon, is not being taught our new engineers.  I guess

Here, I agree.  Unfortunate, but true.

>Like most, I love the 'neatness' of digital, but I've learned
>from hard knocks that sometimes a relay is preferable to an
>SSR (solid state relay) and digital is not necessarily all it's
>cracked up to be.

To all things, their application...  Digital works best for digital type
things, and analog works best for analog type things, and microprocessors do
both pretty well.  Remember, simple analog circuits can only do one simple
thing, while simple microcontrollers can do many amazing things.

>Analog can be unstable (temp and oscillation), but digital has
>sampling, aliasing, etc. errors and loss of resolution in conversion
>to and from digital.  (Besides if it's too cold or hot for an analog
>system, this old boy's staying inside!)

Analog electronics' cost, complexity, difficulty of calibration, etc.
increase almost exponentially with increasing the amount and type of things
that it has to do.  Every time you want to change the operation of some
circuit 'cause it doesn't do what you expected, you have to drag out the
soldering iron and rewire the damn thing.

With microprocessors, you basically can design the hardware with enough or
excess capability for what you want to do, and it's carved.  From then on,
changing anything is simply a matter of rewriting a little code.  You can
rip it all out and start over with a minimum of fuss.  You can reinvent it a
thousand times at minimal cost.  You can use the same hardware for many
different applications, by using modular CPU and I/O sections.

>The cost of doing a micro-
>controller system can be overwhelming, unless you have all the stuff
>at work.  One needs a computer, assemblers, dis-assemblers, high speed
>oscilloscopes, maybe logic analyzers and ICE's, logic probes, and
>lots and lots of printer paper.  (Programmers get rusty zippers and
>yellow sneakers, I hear -- just gotta compile one more time.)

Not true.  There are lots of developement systems for just about any
microcontroller flavor, for under $100, including software.  Lots more are
available for under $300, with more features and better software.  Lots of
free, shareware, or low cost software is also out there.  Up until a few
months ago, I did all my development with freeware tools and a serial port.
I did get an ICE, but only because it saves a little time, which is
necessary in the professional environment, but not for the hobbyist.  I
still prefer my $50 assembler to the $1000+ professional tools that I have
at work, except for the most complex of jobs.

>One can do the analog design with a 20 MHz scope, a triple output
>power supply, and a proto board.  Of course, a copy of Electronic
>Workbench or such to do a computer simulation would be helpful.

Hmmm, using a computer to design an analog system...  Ironic, wouldn't you say?

>An advantage of the digital:  it can do data logging and can even
>be taught to 'learn' standard system conditions, thereby sort of
>self calibrating or adjusting itself -- but this becomes a serious
>programming effort.  Been there, done that!  I'm getting old enough
>where I'd like to use it in this life.

Think of it as a hobby.  Get a basic system working and tweek it or add to
it as necessary afterwards.  I agree, the self learning and adaptive
algorithms are tough, and probably aren't worth it for the DIY'er.

>The digital system could be wire wrapped.  The analog system
>could be soldered.  Else one would need to lay out a PCB and get
>it etched, drilled and plated.

I've found that a PCB is absolutely necessary for analog circuits.
Think about the cost of several iterations of pcb's before you get it right...

>help but believe that the simple hybrid or pure analog path is 98% as
>effective as the pure digital with probably 10% the development time >and
cost! 

Possibly the former, but definitely not the latter.  Have you checked out
the cost of good analog chips lately?  An instrumentation amp costs $8,
while I can get a micro for $2.  I design both analog and microprocessor
systems for a living, and believe me, the hair pulling, nailbiting,
[expletive deleted] problems that you end up with in analog circuits
sometimes are a real pain.  The same can be said for programming, of course,
but IMHO it's easier + faster in the long run.



regards
dn
dnorquay at awinc.com