I made some tests yesterday with Analog inputs and had some strange results regarding to the VREF input of the Cortex, that is not supplied on the G120, and that would be useful for analog sampling stability.
In effect, when providing a stable Voltage in input of an anlog channel, I can see that the ReadRaw value can make a gap from 20 to 80 around the value between 2 measurments (0 to 4096 at 12 bits resolution…).
At VCC 3V, it means a fluctuation of about 2% or even 60 mV…
Can you confirm that the VREF stability can increase the stability of analog sampling, and thus, do you plan to add such an LM4040-N in a future realease of the G120, instead of simply making it common to the VCC Source ?
The goal is for G120 to be made to be small, not to separate analog and digital circuits. It would be difficult (and expensive) to achieve both. This causes noise on the analog circuit. If the noise you see on analog is more than your application allows then adding a separate ADC is the way to go. VREF is simply connected to VCC internally.
My idea is not to seperate ADC from the rest, but only to cut the link between the VREF pin and VCC, to interlace an LM4040-N (which costs about half a dollar) to provide a 0.2% stability 3V voltage…
@ LouisCpro - Actually what you are suggesting, whilst being a good idea, would actually break part of our existing application, which relies on the VRef being tied to the 3.3V rail!
Firstly, and as a separate issue, we use a separate ADC (MCP3422 on I2C bus) for our high performance analog readings.
But for other analogs we use the on-board converter. One of these is to read the battery voltage, so we know how much life we have left, and whether we need to shut down.
Now if you just read the battery directly, a strange thing happens. As the battery drops below the drop-out capability of the regulator, the value read appears to go up! Why? Because the 3.3V rail is dropping below 3.3V, and so the VRef value is falling, and the apparent battery level increases.
To get around this, we have a 2.5V reference (MAX6102) attached to one of the AD inputs. Now we can watch the 2.5V level, and if it appears to rise, we know that the 3.3V rail is falling. We can use this to calculate the actual voltage at the 3.3V rail, and to then correct the other analog reading, such as the battery voltage.
Of course knowing that the 3.3V rail is dropping and out of regulation tells us that we better think about shutting down soon!
This technique is better than just using battery voltage, as it is independent of the actual dropout value of the LDO regulator used, and automatically compensates for current draw of your device.
I suggest to use one of the existing 3,3v pins of the G120 to délicate to the LPC VREF pin, so that each Will use it as needed : simply connected to the 3,3v supply (your case), or thru a regulation component (as mine).
@ LouisCpro - Works for me, in that it is almost completely backwards compatible, no board layout change required, and being able to stabilize Vref, even just with a capacitor, is a good thing.
I doubt it will be high on GHI’s agenda, though, as it would require a layout change to the module and re-testing, whilst they still have plenty of SW problems do deal with.
