comparator circuit

[ramuna]

MrAl said:
Maybe 100M or 150M might start to show this 'other' problem, but the question is then we need a very large resistor value which im not sure we can get in real life. Maybe some reversed biased si diodes in parallel ?
Even if you achieved such high resistance values, wouldn't measuring them accurately be hard to do ?

 

[MrAl]

MrAl said:
Maybe 100M or 150M might start to show this 'other' problem, but the question is then we need a very large resistor value which im not sure we can get in real life. Maybe some reversed biased si diodes in parallel ?
Even if you achieved such high resistance values, wouldn't measuring them accurately be hard to do ?

Hi,

Yes measuring them accurately would be hard, but not impossible. Luckily though we dont really need to know the exact resistance. All we need to do is be able to *vary* it. Very it until we get the effect we think we might see.

So i suggested maybe reverse biased diodes in parallel, maybe one, maybe two, three, four, etc., and see if we can ever get the output of the circuit to change state. That would tell us that in real life a bad resistor within some range could really look *almost* like the original problem, although it wont be exact because we are pretty sure of that from theory.

We really just might find it interesting to see if we can *ever* get this to happen. From theory though we know we need a very large resistance, if it will ever happen at all.

The largest actual resistors i own at the present time are 20 megohms, 1/4 watt. Ten of these in series might start to show the effect, or something like that. We have to get the current OUT of the input of the op amp to form a voltage that is sufficient to trip the comparator state, and that current is small so we need a large resistance.

There's no guarantee that this will work, but it would be interesting to try higher values 80Megohms and higher, up to maybe 200Megohms, as long as we want to try it anyway. But we do need a high enough resistance to do the test right. 20Megohms is not high enough to test this idea.

Did you by any chance try an open circuit? That might not work either because then there is no current to the transistor.

 

[Zabb Csaba]

Hi,

To measure low (bias) current you only need a capacitor and a stopwatch.
I used a 220nF capacitor at the bottom of the post #30.

1. Reset the capacitor.

2. Measure the time until the LED starts lighting (the voltage of the capacitor will be 6V then).
Two comparator bias currents will charge the capacitor during this time.
The charging time is approximately 19 sec (you can see it in the video).
The value of the capacitor charging current:
Uc=t*i/C (if the current is constant) > i=Uc*C/t ; Uc=6V, C=220E-9, i=1.32E-6/t (A, V, s, F).
The minimal value of the resistor of the actual example (LM339):
Rmin=Uc/i ; (Uc=6V) Rmin=6/1.32E-6*t (Ohm).
By me approximately (I didn't measure it exactly, only based on the video) t=19s i=69.4nA Rmin=86.5M.


Have a nice weekend!

Csaba

 

[JoeJester]

Csaba,

You can buy resistors "higher" than 90 M, but what are the chances that a 5k resistor increasing in resistance to over 90 M?

During corrective maintenance, most resistors I've replaced were open. During preventative maintenance, some were slightly out of tolerance on the upper side. The PM was designed to forestall failures rather than us just waiting for one to happen.

The problem statement hinted that you lost a single comparator in the window by the symptoms. Clients would only identify the problem as "it's broke."

 

[MrAl]

Hi Joe,

Yes we have already established the fact that it is very improbable that the resistor would blow out "to within the exact range of values needed to see this condition", but we were now turning to see if this would really happen anyway in the real life circuit.

So it's not that we expect this to really happen, but if it did happen could it really do what we think it might do (like flip the output state) in the real life circuit? We know it can in the simulation circuit, so we wanted to see if it can really happen in the real life circuit too. If it doesnt work *at all*, then we can completely eliminate this possibility from *ever* happening, regardless what the resistor value becomes. So it would move from a "remote possibility" to a "complete literal impossibility". If we cant prove this, then it still has to remain a remote possibility.

 

[JoeJester]

I agree AL. I just didn't see Csaba's point that you could "buy" such a resistor as being germane to the topic at hand.