I have noticed some unexpected behaviour of the ADC cicuits.

A bit of investigation shows that the measured voltage, with constant input voltgae, appears to depend on the total "load" on the Vu line. The main load is an LCD display which draws 18mA (measured).

I am powering my mbed from the USB line.

Am I overloading the power line or is soemthing more subtle happening?

Allan,

I suspect that the A:D converter has a fairly low input impedance and may require buffering.  See my notebook:

http://mbed.org/users/loopsva/notebook/pressure-sensor-and-lpc1768-ad-converter/

I am still struggling with a noise problem, but I was able to get fairly repeatable A:D readbacks (counts within a +-10 margin)

...kevin

Yes, ADC seems to have low impedance input - for audio I'm currently driving it from a headphone out, an audio line out doesn't work great at all. Also using a resistor divide for dc bias and a capacitor to couple the ac circuits.

From resetting mbed the results take a good while to settle. I'm looking at adding an active buffer.

Would be useful if the mbed guys to recommend a suitable front end. The 1768 application notes and schematics are patchy, and there's also additional circuitry on the mbed board (LC circuit on the ADV supply voltage and positive ref for instance).

The ADC supply is tied to the ADC +ve ref on the schematic - so I guess if you're pulling enough power for the Vu voltage to drop at all, the ADC results will indeed vary.

I am just measuring the voltage from an LM35DT temeprature sensor so the low input impedance should not be a worry.

There are some aspects of the signal variation that just cannot be put down to noise however.

Since my original post I have made a few more experiments and it seems clear that drawing more than a few mA on the 5V line risks compromising the accuracy. Perhaps the USB port I am using is not happy supplying a lot of current and so drops.

I am going to add an external power supply for the auxiliary devices and see how this affects things.

I can now see that using a digital sensor has advantages.

I have now set up an external power supply that drives just my temperature sensor and the LCD display. This now gives temperatures that are consistent both an IR thermometer pointing at the sensor and the measured voltage on the sensor.

What is worth noting is that many small power supplies give very high open circuit voltages, my record was a 5V supply that measured 14.5V open circuit, I imagine this will drop as soon as any load is applied but not knowing the times involved and the robustness of the devices I was attaching I chosse to ignore that and found one rated at 5V which measures 5.17V both open circuit and when wired up.

If I connect this 5V supply to VIN will everything work correctly even when connected to a PC via a USB lead ? Any comments will be welcome.