stabilizing a comparator

Nov 20, 2006 2 Replies

I was looking at this guy's voltage regulator



formatting link



T1 acts as a comparator with positive feedback through R4 for hysteresis. To eliminate oscillation, he adds the capacitor C1.



"Capacitor C1 stabilises the circuit at higher load. I have observed that the switching behaviour of the circuit without this capacitor is pretty "sharp" and stable if the load draws less than ca. 0.5 A (typically); however at higher current I observed a tendency to oscillate. This can be observed best with an oscilloscope, or a 12 V lamp as load: it will start to glow, instead of being switched completely on or completely off. C1 eliminates this instability"



I built a voltage regulator using that circuit and it works. It does oscillate without the cap, and installing the cap in the circuit prevents the oscillation.



I was wondering if I could use the same strategy with a comparator. I have a project where a comparator will have high impedance, slow-changing inputs, so I'm concerned about oscillation.



I was skimming through the National Semiconductor datasheet for LM*111 comparators and they showed a couple of strategies for preventing oscillation, but no cap from the output to the negative input. Is there some reason it wouldn't work on a comparator? I'm actually going to be using LM2903.



It is not so important to stop oscillations, (since this regulator always oscillates between on and off) but to control the frequency of oscillation. You want to keep the switching down in the frequency range where the transistors have lots of gain and the switching time of the darlington is short, compared to the on or off time. In other words, you want to stay away from high frequencies where the switching becomes linear, instead of digital, and the switching losses become very large.

You also don't want the regulator switching frequency to lock on to the rectifier ripple.

All that said, I am not sure the schematic shows the best way to accomplish those goals, but at least it rolls of the gain faster than without the cap.

I see no reason (except, possibly ruggedness) that you couldn't use an integrated comparator in place of the first transistor. Keep in mind that the zener reference tempco is compensating for the tempco if the transistors, so a different reference (more temperature stable) may give a more temperature stable result with a comparator.

If I could add to Mr. Popelish's observations, you need to be careful if you're using a cap as feedback to avoid oscillations. If you're using a cap for feedback, you might end up with the input voltage at the inverting input going below the voltage at pin 4. If that happens, you'll likely destroy the IC.

You might want to use a smaller value resistor in series with your cap, and have that in parallel with the higher value feedback resistor you're using to prevent oscillation.

But it really shouldn't be necessary to have very much hysteresis to prevent oscillation with a slowly changing input voltage. Other causes are usually the culprit of the bane of comparators (e.g. locallized power supply bypassing, glitches in Vcc due to driving either large loads or inductive loads). This oscillation *is* a problem, but the solutions usually are elsewhere. Try setting switching voltages with a voltage reference or something other than the Vcc, being very careful with driving inductive loads like relays, and providing good local bypassing of the comparator power (possibly including using a small resistor in series with the power supply, like this:

| | VCC | + | | | 22 ohm .-. | | | | | | | '-' 0.1uF | | || | o---||---. | | || | | | === | |\\| GND | -|-\\ | | >- | -|+/ | |/| | === | GND | (created by AACircuit v1.28.6 beta 04/19/05

formatting link

Another thing to remember is that some single supply op amps like the LM358 can also be used like a slow comparator, if they're driving a ground-referenced load like an NPN transistor base. It's not the most elegant solution, but in some cases where switching oscillation is driving you nuts and you can keep your input voltages more than a couple of volts below Vcc, they can provide something that works.

Good luck Chris

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required