Long mains cable leading to product filters noise out?, or brings more noise in?

[ocset]

Hello,
We have an open collector comparator (TS391ILT) in our non-isolated offline LED lamp.
It sometimes trips due to noise.    We wish to try and assess how to reduce the noise. 

The Lamp has no AC input filter, as it is mostly linear regulator based.

The comparator is connected as per the attached schematic.
(  The input signal comes from a 220R resistor in series with  the output of an opamp. The 8V3 reference is from the output of an opamp connected as a noninverting  amplifier  )

Even when the input signal  to the comparator is continuously high (ie 12V), the comparator still sometimes trips (presumably due to noise) and gives  a high pulse at its output.      

In fact, it’s a little more involved than this, because when we connect the product to the mains via an isolation transformer, the comparator does  **not**  trip and go output_high when its input signal is continuously high. So in other words, powering the product through a mains  isolation transformer makes the product less noisy. (and thus prevents noise tripping of the comparator)   
Do you believe that the leakage inductance of the mains isolation transformer,  and its parasitic Live/Neutral capacitance,  is actually filtering out the noise from the mains and stopping it from pervading around the LED lamp? Also, do you believe that a  very long length of mains cable to the lamp would have (to a degree) the same parasitic effect as the mains isolation transformer?.    ..ie, in terms of having a stray inductance and capacitance (albeit smaller than that of the isolation transformer) which acts as a filter?…  .or alternatively, do you think that a long length of mains cable leading to the lamp would actually bring more mains bourne noise to the product, by acting as an “antenna”?   
 

TS391ILT datasheet:
http://www.st.com/content/ccc/resource/technical/document/datasheet/de/4c/b3/3d/64/7d/48/8e/CD00001660.pdf/files/CD00001660.pdf/jcr:content/translations/en.CD00001660.pdf

[danadak]

Do you have a DSO so you can capture, one shot trigger mode, noise,
to ID it source ?

Also look at power rails to see if any transient occurs there.

Hysteresis around comparator would be appropriate.


Regards, Dana.

[dmills]

Where is the real circuit, not a spiced up generic one, but the real thing (All that spice shows me is a generic O/C comparator, which apart from the lack of any hysteresis tells me nothing) ?

Details matter so that means seeing at least a reasonable amount of the surrounding doings together with a reasonable amount of the layout.

This should however be an easy to identify sort of problem, you have an obvious event to trigger a scope on, and scopes with pre trigger capture are not the hens teeth they once were, a couple of high voltage differential probes and some probing and you will know what is going on in short order I would expect. 

As to the effect of the cable, get thee to a spectrum analyser and a LISN and just look, odds are it does a bit of common mode filtering and acts a bit as an antenna, but a spectrum analyser (And VNA if you want to see how the source impedance varies) will reveal all, your local EMC lab will have these available and a day or so of precompliance time is not that expensive.

Regards, Dan.

[capt bullshot]

If the isolation transformer has influence on the noise induced tripping of the comparator, your design is simply bad.

Add hysteresis to the comparator (as danadak suggests), add filtering to the reference (R/C), add filtering to the comparator input (you've already got the R, add a C), check the driving OpAmps input signal (maybe this is the source of the noise), filter at the input of the driving OpAmp.

The isolation transformer probably removes some higher dU/dt transients (noise) from the mains power, so my guess is your noise results from not having appropriately filtered the comparators input signal. Such kind of stuff finds many ways to couple into your circuit. You don't need to filter the mains input itself, but rather the signal you feed into the comparator or the OpAmp.

[capt bullshot]

Another guess (as you don't provide detail information on your circuit, we can't do anything but guess):

The isolation transformer has high output impedance in comparison with the plain mains. With many non-linear loads, this flattens the sine waveform, causing a lower peak value. Whatever your comparator senses, maybe this reduced peak voltage brings you some margin into the circuit so the noise cannot trigger the comparator anymore ...

Did you properly decouple the comparators supply?