Triangle wave oscillator switching noise

[DaveHardy]

I'd like to employ this circuit in a design I am working on, but it makes too much noise on the power rails and it radiates into another part of my circuit.  I have found that slowing down the switching speed of the comparator removes the radiation and power rail noise, but I'm not sure about the best way to do this.   Does anyone have suggestions or experience with this circuit?

Can I replace R2 with a small cap or put a cap on the output of the comparator?

[nugglix]

What about proper decoupling?

[Zero999]

The TL082 doesn't make a very good comparator. Try the LM393 or LM311 for the comparator section.

[T3sl4co1l]

TL082 isn't going to "radiate"...

What are you actually seeing?  And as above, do you have supply bypassing in place?

Note that an op-amp integrator is not an ideal integrator, but has limited high frequency gain.  The result is feed-forward on rapid transitions: instead of a clean triangle wave, you get a triangle with, er, "nipples" on the peaks.  This is not a supply bypass problem, or a "radiating" problem, but fundamental to the design of the circuit shown.

Tim

[danadak]

Bypassing on rails should consist of a MLC type cap, say .01 to ,1 uF,
and a polymer tantalum cap, size depends on current drain/load. Rep-
resentative values 12 - 10 uF typical.

Both these technologies, MLC and Polymer, have excellent ESR vs frequency
characteristics. Read the cap datasheets, not all caps from vendors, for the
same values, behave the same electrically.


http://www.vishay.com/docs/40076/polymerguide.pdf

What is your load ? Scope probes compensated properly ? Using 10X to minimize
C loading on comparator output ?


Rewgards, Dana.

[T3sl4co1l]

Well gee, even a 10uF aluminum electrolytic is enough.  It's a TL082, not a GPU core switching regulator.

Tim

[danadak]

Here is a presentation discussing the various tehnologies -


https://training.ti.com/system/files/docs/SPSCS_7-1b_CapSelect_TypesOfCapacitors.pptx

https://www.pa.msu.edu/hep/d0/ftp/run2b/l1cal/hardware/component_information/kemet_tant_and_ceramic_caps.pdf


Regards, Dana.

[Audioguru]

I agree that a TL082 is not a comparator because its output slews rather slowly. If its output is overloaded or is driving a capacitor to ground then it will cause a little noise on the power rails and will cause interference by capacitive coupling between all the messy wires and rows of contacts on a solderless breadboard.

[DaveHardy]

My supply is very sound and I've taken extra measures to make sure it is properly damped.  The problem is when the comparator transitions.  The edge is extremely sharp and radiates EMI into a high impedance part of my circuit.  I have done experiments to slow it down and the EMI disappears if I insert a 4.7uf cap between the comparator output  and ground.  I was looking to see if there are other solutions to this problem because I don't like putting a huge cap on the output.

[RandallMcRee]

Instead of tying the TL082 V+,V- directly to your power rail, isolate it using two (say) 100 ohm resistors. 47uf caps after that, as well as bypass caps as advised by others.

Does this help? If not, the problem is not in the power coupling. Then I would suspect ground layout.

After bypass caps and supply decoupling, if problem persists, we would need details, photos, schematic, etc.

[Kleinstein]

To slow done the OP one could use a capacitor in feedback. So add a resistor to the GND connection of the inverting input and than add a capacitor from the output to that input.

For high impedance circuit parts, there could be direct capacitive coupling. So add a grounded shield.

[Ian.M]

I've modelled inserting a 2K2 resistor between the comparator -in and ground and adding a 1nF cap between the comparator's output and its -in to limit the slew rate, (using the T.I. TL082 PSPICE model under LTspice) as has also just been suggested by Kleinstein, and the results look promising.  Caution: too big a cap and/or resistor will cause excessive waveform distortion.

You should also decouple the supplies to the OPAMP as suggested by Randall, though personally, I'd start with 10 Ohm resistors

[T3sl4co1l]

My supply is very sound and I've taken extra measures to make sure it is properly damped.

Okay, fine.  How damped is it?  If you've "taken measures", show us the measurements!

The problem is when the comparator transitions.  The edge is extremely sharp and radiates EMI into a high impedance part of my circuit.

Show the measurements.

There is nothing about a TL082's capability that I would call "extremely sharp", or "radiating", or "EMI".

I have to assume what you really mean is: the voltage is capacitively coupled into a nearby high impedance node.

These terms (EMI, radiating, sharp, coupling) all assume something about the nature of the circuit that created the signal.  EMI is usually a lower impedance, common mode signal.  Radiation is a high frequency, modest impedance, propagating signal.  "Extremely sharp" suggests, at most, tens of nanoseconds risetime.

"Capacitively coupled" simply means you have an effective C-R (high pass) network, where the C might be the accidental capacitance between wires or traces, and R is the impedance of the affected node.

As the origins differ, so do the solutions: for this case, simply reduce or eliminate the coupling capacitance.  Or reduce the impedance of the affected node, or both.

A simple way to reduce coupling capacitance is to reroute wires/traces to greater separation distances, and add ground traces or fill between them (shielding).  A simple way to reduce node impedance, at the frequencies where it matters, is by adding a bypass capacitor.  At high frequencies, this leads to a capacitive divider effect, so that the error signal will not be eliminated, but will be significantly reduced.

Whereas, EMI usually needs a common mode choke, and local filtering; and radiation needs filtering and shielding.  A TL082 might be capable of producing signals of this nature if it's badly abused, say by placing a resonant circuit in the power supply pins, or something; but this should be prevented by your "properly damped" supply.  But again, you haven't told us a lick about your system, so for all we know, it might well be resonating!

I have done experiments to slow it down and the EMI disappears if I insert a 4.7uf cap between the comparator output  and ground.  I was looking to see if there are other solutions to this problem because I don't like putting a huge cap on the output.

Well, slowing down the output is one way, but this directly impairs the performance of the circuit in the first place (and, as you note, isn't friendly to the op-amp!).

But now you need to tell us more about your circuit: its application (what, precisely, do you actually need from it?), what the nearby circuit is, and so on.

Tim

[danadak]

You mention loading output with 4.7 uF. That would be considered normally
a poor practice due to excessive phase shift, reduced phase margin. Its G
runs out of gas at 4 Mhz.  Not to exclude the fact significant currents would
be flowing. Even though amp has current limited outputs one does not know
what their AC response look like, you could be generating some significant
current spikes.

What is the spectrum look like of the EMI ? Meaning is the comparator oscillating ?


Regards, Dana.

[eblc1388]

The TL082 output cannot drive much loading and integator R1C1 (2K2 / 0.1uF) might be too heavy a load for it.

I would use 22k/0.01uF instead which gives the same output frequency and avoid possible overloading its output. Just a thought.
 

[Audioguru]

A capacitor to ground at the output of any opamp causes it to ring or oscillate. A comparator has no frequency compensation like an opamp has so it is simply slowed with a capacitor to ground on its output.

You have a comparator circuit then why are you using an opamp instead??

[David Hess]

I have difficulty imagining a TL082 used as a comparator switching quickly enough to cause significant EMI except in a design which is already excessively sensitive do to for instance poorly situation high impedance nodes or inadequate decoupling.

Operational amplifiers driven into saturation like when used as a comparator sometimes behave oddly and this goes double for some dual and quad parts where saturation in one amplifier may affect the others on the same die.

[StillTrying]

I have difficulty imagining a TL082 used as a comparator switching quickly enough to cause significant EMI

So do I, imagine what a fast comparator in there would do!
A very poor gnd somewhere seems more likely, or a very high impedance or OC on the sensitive analogue circuit.

[DaveHardy]

Radiation is probably the most appropriate term.  When I get a scope probe too close to it, I can see the transitions.

Here is a better diagram of my circuit.  A diode is present in my circuit and there is also a 100k potentiometer mounted on a 8 inch connector to control the speed.  The comparator has a 100K  and a 50K resistor to get it to saturate.

This is an audio amplifier and has a 500 megohm input impedance.

[DaveHardy]

I have difficulty imagining a TL082 used as a comparator switching quickly enough to cause significant EMI except in a design which is already excessively sensitive do to for instance poorly situation high impedance nodes or inadequate decoupling.

Operational amplifiers driven into saturation like when used as a comparator sometimes behave oddly and this goes double for some dual and quad parts where saturation in one amplifier may affect the others on the same die.

This is my conclusion too.  I'm using a TLO74 and could probably replace it with a different quad package.  I'm still scared to put a cap on the output because I've been taught that it will be unstable; this is why I posted the question.  In practice, it behaves quite stable but who knows what will happen if I build a few of them. 

[David Hess]

A capacitor to ground at the output of any opamp causes it to ring or oscillate. A comparator has no frequency compensation like an opamp has so it is simply slowed with a capacitor to ground on its output.

Operational amplifiers can be frequency compensated with a capacitor from their output to ground.  Usually this takes the form of a relatively small series RC circuit but brute force with a large capacitor also works if very low bandwidth is acceptable.

Here is a better diagram of my circuit.  A diode is present in my circuit and there is also a 100k potentiometer mounted on a 8 inch connector to control the speed.  The comparator has a 100K  and a 50K resistor to get it to saturate.

I do not know if the text mentions it but the 301 is more suitable than most operational amplifiers for use as a comparator; it has a wide differential input range and uses external frequency compensation.

What specific diode did you use?  A switching diode would be suitable but a standard recovery rectifier like a 1N4001 could conceivably cause EMI problems by acting as a snap-off diode.

[tggzzz]

I'd like to employ this circuit in a design I am working on, but it makes too much noise on the power rails and it radiates into another part of my circuit.  I have found that slowing down the switching speed of the comparator removes the radiation and power rail noise, but I'm not sure about the best way to do this.   Does anyone have suggestions or experience with this circuit?

You haven't built it on a solderless breadboard have you?

If you have, stop now and build it "properly" on a PCB using manhattan or dead-bug techniques. Pay attention to minimising wire lengths and to decoupling capacitors.

If you haven't, post a picture of the circuit and its connections.

Whenever someone asks me to debug a circuit built on a solderless breadboard, we spend more time debugging the breadboard than their circuit.