Path: utzoo!utgpu!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!mailrus!ames!pasteur!ucbvax!hplabs!hp-pcd!hplsla!tomb
From: tomb@hplsla.HP.COM (Tom Bruhns)
Newsgroups: sci.electronics
Subject: Re: A Challenging PWM Design Problem
Message-ID: <5170022@hplsla.HP.COM>
Date: 7 Dec 88 21:01:41 GMT
References: <1663@maccs.McMaster.CA>
Organization: HP Lake Stevens, WA
Lines: 63

> 
> I have an interesting Pulse Width Modulation design problem;
> 
> (pulse 1-1.9ms; 18ms between pulses)
> 
>   The pulse width varies linearly with a transducer variance of
> 0 - 100%.  My question: is there a simple solution for the conversion
> of this PWM signal into a corresponding analog signal, ie. a simple
> circuit for the following black box?
> 
> where the analog output function looks like;
> 
> 
>    Vcc  |                  .
>         |               .
>         |            .
>         |         .  
>         |      . 
>         |   .
>         |.
>        -|-----------------,
>         1.0              1.9
> 
>           Pulse Width (mSec)
> 
> -- 
> Darel Mesher				...!uunet!mnetor!maccs!darel
> McMaster University			    darel@maccs.mcmaster.ca
> ----------
Depending on the accuracy and time response you want, maybe the
following "one-chip" solution will work (thought the FET switches
may be EASIER to do as a second chip -- same with the one-shot).

You get to work out the details; this is only an outline:

Use a quad opamp chip.  One section is configured as a one-shot
which is adjustable, and nominally 1 mS.  It is triggered by the
leading edge of the received pulse.  Its output is used below.

Second opamp is used as an integrator.  It ramps to the desired
fullscale output in 0.9 mS.  It has an FET switch to reset it;
that switch is closed ("RESET") when the 1mS one-shot is active.
It also has an FET switch on its input:  the input is taken from
a pot fed from a regulated negative voltage, so full-scale is
both stable and adjustable.  However, the input is switched off
(switch open; no current into integrator) whenever either the
one-shot is active or the input pulse is low.  Thus the integrator
is reset and ramps up for each pulse, and holds the value between
pulses.

The remaining two opamp sections are used as either a linear
lowpass filter to smooth the output (if needed), or as a simple
sample/hold sort of thing ...

Now, you didn't say anything about the number of R's, C's, D's
etc ...

Hope this gives you some food for thought.  It's only one of many
ways of doing it... but the parts I'd have in my junkbox.

Cheers,
Tom Bruhns
tomb%hplsla@hplabs.hp.com