LM393 hysteresis calculation vs real circuit

[getchar]

Hi all,
I am newbie here and I would like to ask you for help.
I am working on thermal protection module but comparator which I used doesnt work properly.
Calculations are from this literature:
http://saaubi.people.wm.edu/TeachingWebPages/Physics252_Spring2009/Week13/LabManual_Chpt13.pdf

My circuit:


Calculations:
Vref = 6V
Vp+ = 8V
Vp- = 4V
-> hysteresis 4V

I measured points when it was toggling:
Vp+ = 4,46V
Vp- = 3,86V
-> hysteresis 0,6V

Where did i make mistake? Thank you very much for any hint...

[jwatte]

When looking at your circuit, consider the equilibrium states when output is high, and output is low.

High: Effective 50 kOhm pull-up; 100 kOhm pull-down, so 8V input to +. - would have to go above 8V to turn off.
Low: Effective 100 kOhm pull-up; 50 kOhm pull-down, so 4V input to +. - would have to go below 4V to turn on.
Thus, your calculations seem accurate.

Why would this particular circuit not behave correctly? I don't know exaclty, but maybe the current drive provided by the 100 kOhm resistors is not sufficient. Try 10 kOhm resistors instead. That may still not be enough for properly driving BJT transistors -- 5 mA seems to be a magic place, so perhaps you should actually try 2 kOhm resistors instead of 100 kOhm resistors.

From what I've found, a comparator is just like an opamp, except with worse performance. I'd use a rail-to-rail single-ended opamp rather than a comparator.

[getchar]

Thank you for your reply jwatte.

I changed values of resistors to each 2kohm, but it still doesnt work correctly. Hysteresis is still about 600mV. I havent got any rail-to-rail opamps now, so i have used TL072 (it is pin-compatible) - hysteresis is still about 600mV.
I am confused...

[jwatte]

Re-looking at the diagram: The output is connected to a resistive divider. This likely back-feeds through feedback resistor.

How about you measure the actual voltage at the various nodes in your circuit in "on" and "off" states?

Also, you could either re-buffer the output with another opamp (yay!) or make the output drive the gate of a MOSFET without any pulling resistors. A 2N7000 or BS170 as a low-side switch would be a fine thing to drive from the opamp, and the resistor+LED would be tied to the source of that, up to VCC.