CFAG320240C-FMI-T dots ON/OFF

We fitted a graphic LCD in a coffee vending machine. I work for the manufacturer as an electronic consultant/designer. A custom pcb was designed for this application. The "brain" inside this machine is a PIC 18F820 CPU (128K prog space) running at 19.6Mhz. I've been working for a year on this project. We assembled nearly 40 units to this date.

The problem i have is that sometimes, on some units, i get randomly dots on the screen that are turning on or off. There is no damage done. When the CPU clears the screen and fills it with new data everything is fine.

During prototyping, I would have this problem often. This appened when the flat cable between the pcb and LCD module was too long. 20 Inches was the maximum lenght. This is the lenght we are using right now on our units.
Do you think it still too long ? What do you recommend ?

On the picture attached, you can see the CFL driving PCB far from the LCD module. The problem appears more often when the high voltage white cable are near the LCD unit. During assembly we make sure the high voltage cable are "taped" away.

The LCD is held by 4 nylon screw and nuts. The front panel is ABS plastic. The other flat cable comming from under the LCD is the membrane keyboard. Low frequency/voltage signals are used to scan the keys.

I saw coffee machines worked perfecly in our factory and then got this dot on/off problem once the coffee machine was instaled in another environment. It's like if the LCD module is picking up electrical noise from other equipement near. Should i put some sort of sheilding behind the pcb of the LCD ?

I can supply you with the details of the electrical connection between the control pcb and LCD if needed.

I attached a picture to show you the way we hook the LCD. I would like to show you more of the project but can't find a way to attach more than one file to this post.

Thank you for any inputs to help me get rid of this annoying bug.


Yves Belzile

I have not personally used this display, but I can suggest some things to try:

1. Use a shorter ribbon cable if possible (or change the routing path).

2. Add pull-up or pull-down resistors at the display end of the cable. I know this is not easy to do conveniently, but reducing the impedance may reduce the noise susceptability.

3. Lengthen the timing in the write cycle. For example, use 500 nS instead of 150 nS. Also allow some delay between control and data transitions, to allow things to settle.

4. Perhaps using ribbon cable with a mesh ground plane would help.

I agree with cosmic's suggestions. I would also add:

It might be that there is some noise on Vcc at the display end. Try adding some capacitance between Vcc and Ground right at the display. For testing, put a 100uF electrolytic, a 10uF tantalum and a 0.1uF ceramic all in parallel and solder them between Vcc and Ground at the display's connector. Watch the polarity of course.

It might also be that there is ringing on the "E" signal. Try looking at "E" with your scope at the display. Ground your scope to the display's Ground pin at the display end of the cable. Then look at the edges of E for excessive ringing. Try putting a small resistor (25 ohms to 100 ohms) in series with the E pin at the microcontroller end of the cable. This would be an attempt to match the impedance of the driver to the cable.

Very nice looking machine by the way. I like the way the coffee on the graphic appears to be pouring into the cup

1. Use a shorter ribbon cable if possible (or change the routing path).

Click to expand...

Unfortunately, the control pcb can't be closer to the lcd. Other devices need to be connected to the control pcb. I would love to have a 4 inch long cable but it's just not possible in this project.

3. Lengthen the timing in the write cycle. For example, use 500 nS instead of 150 nS. Also allow some delay between control and data transitions, to allow things to settle.

Click to expand...

Right now the E pulse in a write data transmission is a little more than 203ms long. Modifying the software to make the delay longer would be an easy fix. I will definitely take a look at this possibility. Will do some tests.

I also will check the idea of adding a resistor for the E signal.

If all this does not work, i will then add some capacitors on the LCD for possible noise on Vcc.

The ideas a about a flat cable with mesh ground plane and pull up downs resistors are all good but will be used at last ressort. I have to keep things simple for production.

Thanks to Cosmic and CF Tech for the help. I don't have time to work on this right now but will keep you informed in the next weeks.

Thanks for the nice comments about our coffee machine. Here are some more views of the LCD :