Simpler Than A Thermostat?

Suppose a feedback control system does not have any way to determine the [+/-] sign or magnitude of the error as in a thermostat.
Instead the control system is 100% dependent on the feedback being reliable enough to kick in in time to never pass a reference point.
If the control system somehow screws up and slips past the reference point the control system will effectively be worthless.
A feed forwardish approach could make the system more reliable but still that's not good enough.
Is there some Carolina Charlie jerry rig solution that would work without actually determining the error and the sign of the output?
Bret Cahill
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Bret Cahill wrote:

What feedback? From where?

If sign is irrelevant, what does "slips past" mean?

I resent that characterization! The proper term is "jury rig". Look it up.

What can be prompting you to want to control a thing without knowing how big it is or even whether it's more or less than you want? Off hand, it sounds to me like the product of a fevered mind.
Jerry
--
Engineering is the art of making what you want from things you can get.

  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
At first glance this control system would seem even more primitive than a conventional simple thermostat control. It would work only if the initial conditions were the right sign because it only cuts on when the output is equal to or just above (or below) the reference temp.
For example, in the morning you set the ac to 80 degrees when it was 75 degrees. At noon the temperature reaches 80 degrees and the ac automatically cuts in as it should. The room cools down a degree or two and then the ac shuts off as it should. The room warms up again and the ac cuts on again.
The controller seems to be working.
If you waited until the afternoon, however, when it was already 85 degrees in the room to set the ac at 80 degrees, the ac would not cut on as it could not tell the difference between being above or below the reference temp.
One less than satisfying solution might be for the controller to start looking for the output. It could only do this by _changing_ the output until it reaches the reference. In the thermostat example the controller would wait a certain amount of time and if the output doesn't reach the reference point, the controller turns on the ac for x minutes. If that didn't find the output it shuts off and turns on the heat pump for 2x minutes. If that doesn't work then back to the ac for 4x minutes . . .
Obviously this isn't going to revolutionize the thermostatic controls industry, but has anyone heard of any solutions that could end run sign determination yet still be reliable?
Bret Cahill
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Bret Cahill wrote:

Most of us know about thermostats.

Why not? What kind of mechanism has a set point but no way to determine if it is exceeded or not reached? A built-in absolute-value function?

And that is better than an ordinary bi-metal thermostat?

Yes. Remove the sign suppressor.
Jerry
--
Engineering is the art of making what you want from things you can get.

  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

A thermostat-simple controller.

Absolute value is the best way to look at it.
The ref would be x = - b/m of the line y = |mx + b|.
The only difference is there isn't much of way to measure the magnitude error.
Of course, conventional thermostats don't really care much about the magnitude of the error either and they work just fine.

The thermostat was only proffered as an example. I'm not interested in temperature or ac, just controls solutions.

That would require a sign _recoverer_ which will complicate things. (I'm one of the greatest apostles of the KISS religion the world has ever known.)
There's some predictability to the time lags so it might be better to have the output dither just above and just below the reference. Once the system is up and running the controller would allow a certain amount of time where it is _guaranteed_ to pass the ref and then reverse direction. If the ref isn't found in that amount of time the controller knows something is wrong.
This would be good enough for a prototype anyway. Someone else can get fancy if that's necessary.
Bret Cahill
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Bret Cahill wrote:

Then KISS and KTS (Keep The Sign). Error signals come from subtractions -- desired minus actual -- and subtraction yields a sign. Keep it.

We might be better able to help if you describe the actual problem instead of stand-in systems for which existing trivial systems are the clear answer. I'm unlikely to think of a solution to your problem without understanding why it a problem.

Hmmm... Simpler than a thermostat, but fancy if necessary. Cryptic!
Jerry
--
Engineering is the art of making what you want from things you can get.

  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

It might not be possible to get a useful magnitude from the sensor. All you get is an indication that the output is where it should be.
That's why constantly dithering around the ref. pt. may be a good prototype solution.
Bret Cahill
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Bret Cahill wrote:

If you can't tell if it's not where it shouldn't be, how can you tell it is where it should be?
Jerry
--
Engineering is the art of making what you want from things you can get.

  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Mon, 13 Jul 2009 18:37:35 -0400, Jerry Avins wrote:

You may wish to check the Google Groups archives for sci.electronics.design for this guy's behavior; it may save you some time here.
--
www.wescottdesign.com

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Tim Wescott wrote:

Thanks, but I'm done anyway.
Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

A delta function is out there somewhere.
Anyway here's a couple approaches, which may or may not be in the literature.
1. Several reference points [sensors] are closely spaced around the desired ref. then simple diode logic determines status & +/- action. This would be pretty cheap and easy. You don't complicate your life with a lot of transfer functions, stability analysis or software.
2. Use two reference points sensors] just above and just below the desired reference point. The difference in the derivatives of both -- one should always be negative and one should always be positive -- to determine action. Maybe it's closer to conventional thermostat controls than originally thought.
3. Some combination of 1 & 2.
Are there any more?
Bret Cahill
"When compared to a genius, that is, he who begets or gives birth, the scholar comes off a bit like an old maid. He's not conversant with the two most valuable functions of man."
-- Nietzsche
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.