Could you post a picture of what you are getting with 5V?
Also, what are you connecting to A3? An external power supply of 3.3V or are you connecting the VCC pin from the Base Shield?
How are you powering the board?
Have you tried to do the same measurement without the BaseShield?
Here is the image when 5V is used (below).
To A3, I connect the Grove "rotary angle sensor" which is just a rotary potentiometer which provides a voltage divider.
I am powering the Arduino breakout board off of a 12VDC power supply.
It would be difficult to try measuring without the base shield, since it provides my connector interface. This board contains passive components, mostly connectors, so I would not expect it to give this sort of behavior due to a voltage change.
During the test, I rotate the rotary angle sensor by hand and then stop the data capture. As you can see, it is smooth as expected for 5V, but has issues with 3.3V even though it moves through the full scale (0-1023) for both.
Graph showing 5V operation of A3:
This is still an issue.
It looks like the bits in the digitized ADC reading are swapped somehow. I can inject a smooth voltage change on the input, and I get jumps of 50+ counts on the reading. These always occur at certain points on the voltage scale, as you can see in my graph above.
mraa apparently has issues when the Edison and Arduino breakout I/O is set to 3.3V. This is a show stopper for those trying to use 3.3V.
Has anyone had success using mraa for analog inputs with a 3.3V supply?
I have been doing some tests to replicate your issue. I'm having the same behavior than you when I use 3.3V as AREF I tried using the Arduino IDE and MRAA, also I tried to use an external REF, with this I mean not using the 3.3V that the board has. I always get the same issue. I'm going to do a couple of tests more and I will also going to investigate for a workaround or if there is something that we could do to avoid this behavior. I will post here more information and results when I get them.
Unfortunately; we don't have updates in this. We have ran some tests and changing parameters to try to get a smooth signal but without success. As soon as we get improvements in this, I will post the results here.