I had trouble reading the schematic, but assumed it was the micro that enabled the LCD based on some signal it got from the switches. Maybe that's not the case. I would think the grounds within the case would be close to the same (short leads), but maybe not. Might need to drag out the scope.

Sorry if my schematic or description was unclear. The switches are for UI menu navigation, so they're related to the display in that sense, but none have any direct relationship to the enable pin in question.

Also, I may be using a misleading name for that pin. I believe there are two enable pins: backlight enable and write enable. The latter determines when a command or character to display (defined by the states of the data pins) is acted upon (executing the command or displaying the new character.) So the process as I understand it is set all the data pins to the values you want, then toggle write enable pin on/off (or off/on, can't remember) to enact the change, then set data pins to new values and repeat.

Although that enable signal definitely normally comes from the micro, my suspicion is that either the enable pin/trace within the display housing is cycling high/low in response to noise, or more likely that the ground reference within the housing is moving such that the enable pin appears to the display to be moving because of the changing ground reference. Does this make any sense or am I just talking crazy?!

So to put it another way, I *think* that everything is perfectly fine on the micro side, and that the effect of the noise is just within the display housing.

 

Have you tried putting a ferrite core around the ribbon cable?

 

And yes, I think it is time for the scope. I just haven't had time. We got swamped with orders so I'm too busy building and testing our normal production to spend enough time on this right now. My mind is still spinning on it, trying to make sense of things, but I've had almost no time to experiment.

 

Have you tried putting a ferrite core around the ribbon cable?

Interesting thought. No, I tried ferrite around the ZIF tail where I've assumed the noise was coming in, but not on the ribbon cable. Worth a shot. I'll try it next chance I get. Thanks!

 

Sorry if my schematic or description was unclear. The switches are for UI menu navigation, so they're related to the display in that sense, but none have any direct relationship to the enable pin in question.

Also, I may be using a misleading name for that pin. I believe there are two enable pins: backlight enable and write enable. The latter determines when a command or character to display (defined by the states of the data pins) is acted upon (executing the command or displaying the new character.) So the process as I understand it is set all the data pins to the values you want, then toggle write enable pin on/off (or off/on, can't remember) to enact the change, then set data pins to new values and repeat.

Although that enable signal definitely normally comes from the micro, my suspicion is that either the enable pin/trace within the display housing is cycling high/low in response to noise, or more likely that the ground reference within the housing is moving such that the enable pin appears to the display to be moving because of the changing ground reference. Does this make any sense or am I just talking crazy?!

So to put it another way, I *think* that everything is perfectly fine on the micro side, and that the effect of the noise is just within the display housing.

No, could be right.
So you can reverse the same thing on the enable line. If it has a pull up put it on the box end. If it doesn't need a pull up you could try a small cap to the ground at the box end.
The ferrite is a good try as well. If you have any spare pins in the cable make them ground.

 

No, could be right.
So you can reverse the same thing on the enable line. If it has a pull up put it on the box end. If it doesn't need a pull up you could try a small cap to the ground at the box end.
The ferrite is a good try as well. If you have any spare pins in the cable make them ground.

Yeah, I thought of the spare pins as ground idea too, but none to spare.

The capacitor from Enable pin to display ground (dc common) works! I tried both 0.01uF and 0.1uF ceramic caps and either one does the trick with no apparent adverse side effects.

It feels like a kludgy fix - I'd rather truly understand how the noise is getting in and address it there, but I'm pretty happy to have at least one possible solution finally.

So, any advice on how to pick the best capacitor value here? The two I tried just happened to be what I have around for other projects. I'm guessing there's no way to pick an ideal value without knowing more about the nature of the noise, but I thought I'd ask anyway.

Thanks for all the help!

 

So I tried a range of cap values and used my oscilloscope to see the effect on both noise and good signals, and finally settled on 0.047uF. It's 5 times the lowest tested value that blocked the noise, but one seventh the highest tested value that didn't shut down the display. Feels like comfortably wide margins on both ends.

Still wish I understood the source of our problem better (presumably grounding) but I think this solution will suffice for now. Thanks again for the help!

 

So I tried a range of cap values and used my oscilloscope to see the effect on both noise and good signals, and finally settled on 0.047uF. It's 5 times the lowest tested value that blocked the noise, but one seventh the highest tested value that didn't shut down the display. Feels like comfortably wide margins on both ends.

Still wish I understood the source of our problem better (presumably grounding) but I think this solution will suffice for now. Thanks again for the help!

Sounds good.
I think what happens is the noise on the frame gets coupled to the box then by capacitance to the flex. So now when there is noise on the frame the ground on the box can move with respect to ground back at the controller because of the inductance of the cable. Now to the display it looks like enable moved, but in fact it was ground that moved. Adding the cap between the box ground and enable makes both move together.