Correct guard trace implementation.

[implor]

I get the concept of guard traces and it's easy to do around OP amps when you have a source to create a trace from (+ or - input dependent on configuration).

But how do you do it when not a suitable source for a reference voltage exist?

Exampel TPS3780. We have 3 resistors in series and a total resistance of 2Mohm (low power application).

Guard trace potential should be around/close to 1.134v (negative going threshold is more critical then positive going)

So how would you create that 1.134V? Voltage divider? Voltage divider + Driver?

Maybe it's overkill with a guard trace here but just assume it's needed for some weird certification or customer requirements.

[T3sl4co1l]

Only 2M?  Don't worry about it.  You need to clean that assembly well anyway, if you want to maintain the low supply consumption you're expecting.  If you've got gigohms or ppm accuracy on the table, and less pressure for low supply consumption, give guarding some consideration again.

But say for academic purposes we wanted to consider it anyway: what we'd do is of course place the guard trace, with removed soldermask, around the nodes in question (mind this is pretty much impossible on DFN/SON and CSP types; should be doable with the SOT version), then bias it to something nearby, which would indeed have to be an external reference source or voltage divider in this case.  You could do a TLV431 (or there are lower current references in similar voltages), another voltage divider (with probably around the same resistances), or add on a buffer op-amp to bootstrap it (the priciest but most precise option).

Tim

[implor]

Thank you. Very helpful.

The circuit is very sensitive. When you get close to the trigger level (within 30mV or so) it "jerks" low 6-7 times to then go high again.  Really annoying when it's purpose is to turn of a regulator. Now it just turn the regulator on and off with massive VCC dips.

I placed some capacitors between the resistors to filter out transients and it seems to solve it. But it's enough that i'm poking with my finger in that area and we will get false trigger.

[T3sl4co1l]

Yeah, filter it.  That device claims a few percent of hysteresis (depending on which version you have), it will not chatter on a stable DC input.  Any AC on the input reduces that hysteresis band accordingly, and too much causes chatter like there's no hysteresis at all (or you can get the next higher hysteresis version I suppose).

"Between the resistors"?  Shouldn't they be from the resistors to ground..?

Tim

[implor]

Yeah, filter it.  That device claims a few percent of hysteresis (depending on which version you have), it will not chatter on a stable DC input.  Any AC on the input reduces that hysteresis band accordingly, and too much causes chatter like there's no hysteresis at all (or you can get the next higher hysteresis version I suppose).

"Between the resistors"?  Shouldn't they be from the resistors to ground..?

Tim

Bad wording from my part. The net between the resistors to GND ofc .

Swapped my bodge capacitor to some nice smd parts and it worked perfect. Can touch the inputs with my finger and just occasionally get a false trigger. 

I just have to add a line in the manual, "please don't disassemble the product and lick the pcb!"   

[AE7OO]

That last line got me rolling...

I needed a laugh badly, hence the thank you....