AnalogRead unstable with ColorLCD

Hello,
I just installed ColorLCD from Sparkfun on my Arduino Mega. It works Fine.
But when I put a 10 kOhm potentiometer in an analog input (with two border pins connected to ground and 5V), the value is VERY unstable (+30 / -30).
And Even if I'm connecting the analog input to the ground (with no potentiometer), the value is varying between 0 and 30 (on 1023), with some kind of cycles.
If I disconnect the LCD, the analog values become stable again.

So I think maybe there are some kinds of capacitors (the varying has... constent cycles) in the LCD which are perturbating the inside of Arduino ??
Someone did have the same problem ?
Can I add some regulation in the analog input of Arduino ? (like capacitors)

Thank you in advance, and sorry for my poor english.

So I think maybe there are some kinds of capacitors (the varying has... constent cycles) in the LCD which are perturbating the inside of Arduino ??

No I would not think so.
It is the very lack of capacitors that is causing you the problem.

First make sure your power supply can delver the expected current. Then add a 10 to 100uF capacitor across power and ground of the LCD.

Yes, you're right, I'm powering with my computer and USB cable, maybe it's not enough. I will try to power with external power supply and add capacitors.

Thanks

gaith,
I noticed that when using digital- and/or analog Outputs and Analog inputs, using the same Gnd-pin the Analog input values are lifted about 30, like you mentioned. Even grounded analog inputs show so.
When using a separate Gnd-pin for analog-inputs the problem is gone. When measuring the voltage between the two Gnd-pins you'll see a value raising to >100mV, depending of the number of digital outputs.
It seems it has to do with the differential measurement at the analog input cicuits.

This is classic issue - any ground wire that carries an appreciable current (such as powering LCD backlights) will have a voltage
across it (in fact the bulk of this voltage can be due to the contact resistance of the header pins).

You want any measurement signal to use a separate, non-current-carrying, ground wire as its reference - then you are not
adding/subtracting any IR-voltages to your signal. (V = IR). [ well more accurately you are keeping the current really small
so the error is less ]