Strange/Poor Response with LM339 Comparator. Help Requested. / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

Strange/Poor Response with LM339 Comparator. Help Requested.

1 2 next

Move Thread

LAN_403

Sigurthr

Fri Feb 13 2015, 03:18AM

Registered Member #4463 Joined: Wed Apr 18 2012, 08:08AM
Location: MI's Upper Peninsula
Posts: 597

Hello everyone!

I'm working on a SMPS and I grabbed a comparator for the feedback as I'm doing this one sans uC. I've never used a comparator in such a demanding application before (previous uses were just logic signals or light sensors, etc) and I've observed strange behavior in my circuit. So, I took the comparator out and made a test jig on breadboard to examine the chip's operation.

Now, I realize it is a REAL component, and not an IDEAL one, but I was expecting behavior to more closely match what an ADC channel on a uC would do; a fixed offset range of indeterminate output (in a uC it's related to the ADC's bit depth and input range).

Here's my test jig and results; 5V is regulated, decoupled, and filtered. Hysteresis was added on both inputs and output in second test, no change in behavior noted from zero hysteresis.
Hkc9V5S

As you can see there is a rather large voltage offset between inputs and switching point (+-20mV or +-50mV), and the offset is not identical both ways. Between the reference voltage and the input switching point the output oscillates. With Vref on the Non-Inverting Input it oscillates at (Vref - 20mV) to (Vref + 20mV). That's 40mV of feedback range which equates to a 5.6V range on the SMPS output. If I use the Inverting Output for Vref it is 100mV of feedback range, which is 14V of range on the SMPS output. (140:1 divider) This makes certain loads really throw the smps out of regulation.

With a 10bit 5V ADC There's only ~24mV of indeterminate range, and you can code it to largely rule that out and always output a certain way in that case, negating the problem. I thought going analog would offer better precision!

I was expecting a couple mV of this indeterminate state range, not 40mV or 100mV! Is my chip damaged? Did I make a simple mistake in application here?

Carbon_Rod

Fri Feb 13 2015, 03:49AM

Registered Member #65 Joined: Thu Feb 09 2006, 06:43AM
Location:
Posts: 1155

IIRC the Schmitt trigger comparator needs a feedback resister to your Vref.

Sigurthr

Fri Feb 13 2015, 06:53AM

Registered Member #4463 Joined: Wed Apr 18 2012, 08:08AM
Location: MI's Upper Peninsula
Posts: 597

Hmm, didn't see that in any of the app notes in the datasheet or searched circuits, not doubting you though as something sure is wonky. Any more detail on application of said feedback resistor? Appropriate values, formulae, etc?

Carbon_Rod

Fri Feb 13 2015, 07:04AM

Registered Member #65 Joined: Thu Feb 09 2006, 06:43AM
Location:
Posts: 1155

Try 33k from pin 1 to pin 7 of cicuit 2 to help set the hysteresis...
so switching time will be sharper.

LGTFY:

Cheers,
Rod

Shrad

Fri Feb 13 2015, 08:14AM

Registered Member #3215 Joined: Sun Sept 19 2010, 08:42PM
Location:
Posts: 780

those which are not schmidt trigger will usually require a capacitor in parallel to that resistor to integrate over the acceptable range of variation to prevent oscillation or spurious events if you use it in a control loop

also keep in mind that you're usually better working in a zone around +2V/3 and -2V/3 for future part replacement (not all accept the same ranges) and that comparators usually behave unexpectedly when supply voltage is lower than a safe value so always use a power-on delay for the supply to reach its nominal value

Sigurthr

Fri Feb 13 2015, 08:40AM

Registered Member #4463 Joined: Wed Apr 18 2012, 08:08AM
Location: MI's Upper Peninsula
Posts: 597

Thanks!

Yep I had tried the parallel capacitance on I/O pins, didn't make a difference. Didn't know about the positive feedback resistor though, will give that a shot tomorrow.

mister_rf

Fri Feb 13 2015, 08:49AM

Registered Member #4465 Joined: Wed Apr 18 2012, 08:37AM
Location: Bucharest, Romania
Posts: 145

I have done my tests with an LM2901, and the input offset voltage was normal, as expected. (less than 5mV).
Thatâ€™s a chance for an out of specs circuit.

Wolfram

Fri Feb 13 2015, 01:38PM

Registered Member #33 Joined: Sat Feb 04 2006, 01:31PM
Location: Norway
Posts: 971

It sounds like your circuit is operating correctly, but oscillating when the input voltages are close. This is probably caused by capacitance between the output and the inverting input, or missing decoupling capacitors. Do you have proper decoupling caps directly between VCC and GND of the comparator? If the decoupling is fine, you could try adding some capacitance across C3 to decouple the reference voltage for the comparator.

You don't want to use a comparator as a feedback amplifier, unless you're doing pulse skipping or something like that. A regular op-amp is more suited for the job.

Sigurthr

Sat Feb 14 2015, 01:13AM

Registered Member #4463 Joined: Wed Apr 18 2012, 08:08AM
Location: MI's Upper Peninsula
Posts: 597

@Wolfram The VCO I'm using has nasty instability at the upper edge of acceptable Vset range, so I'm using pulse skipping via the enable line and manual frequency adjustment. It won't be seeing dynamic loads, it just has to be relatively stable throughout the wide operable range. On a fixed voltage converter or one with less input/output range I'll have to look at using a regular op-amp and controlling the output via oscillator frequency.

I was careful with stray capacitances and I always decouple all of my ICs.

@Carbon Rod that 33k between Vref and Output did the trick! Thanks a ton!

I made a reference schematic for future readers and google searchers so that they won't wind up with the same issue. Credit was given where due! =)

How to Use a LM339 Comparator:
ViG45P3

Picture of the SMPS in action: 12V IN; 177V OUT.
HWQiRMl

mister_rf

Sat Feb 14 2015, 09:35AM

Registered Member #4465 Joined: Wed Apr 18 2012, 08:37AM
Location: Bucharest, Romania
Posts: 145

In my opinion if you use the 33k resistor thatâ€™s too much hysteresis added.

Iâ€™ll prefer to use at least ten time the actual value, let's say a minimum value of 470k.

1 2 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob