Atmega16 adc question

Check the datasheet. AVcc limits: (Vcc - 0.3, Vcc + 0.3) and (2.7, 5.5). ARef limits: (2.0, AVcc) single-ended, (2.0, AVcc - 0.2) differential. A 5 V AVcc looks OK to me.

On the mega16, you can program the ADMUX register to connect AVcc to ARef internally, in which case a capacitor can be hung off of ARef to soak up noise.

The successive approximation DAC only cares about its voltage reference, not the chip's Vcc, so as long as none of the recommended limits are exceeded and the precision voltage reference can source enough current, you should be OK.

This is a very legal question. Of course, Atmel does not specify the maximum voltage between AVCC and VCC. One would expect it to be around 0,5V to avoid NP conduction.

An alternative may be to reference Vcc from your AVcc, using a series diode with your reference, and an emittor follower transistor as buffer. If you prefer using a regulator IC (lower output impedance than the emittor follower) you can use a LM337 (1.25V step) and reference it to your AVcc with a resistive divider.

Success Sander

The AtMega16L specs say AVcc must be within +/- 0.3 volts of digital Vcc. So you are just within spec at +/- 0.25 volts range. Therefore it should be ok. Personally I would use a closer tolerance device than the

Note also that if you go the other way, and elect to tie AVcc to Vcc, then Vref *cannot* exceed Vcc.

Although the AVR chips are broadly similar, there are differences in the details. On the mega16, the bits REFS1 and REFS0 in the ADMUX register control what voltage is used as the ADC reference.

REFS1 controls a mux that selects between AVcc and the 2.56 V bandgap reference. REFS0 controls a gate that connects the output of the mux to the approximation DAC.

If ADMUX is written with 01xx_xxxx then AVcc will be used.

See Figure 98 and Table 83.

Perhaps you can use the internal reference if you divide down the voltages to be measured - however, full scale will be equal to Vint, about 2.56 volts, not AVcc.

Actually, the LM78L05 drift is smaller than some regulators, and may be acceptable for your application - (check datasheet) - the issue is you can't predict the voltage of an individual device. In that case, you can put a calibration factor in EE memory and multiply your ADC reading by that. You'll probably need calibration anyway.

Roger

Thanks for all the info and help. I think I have a handle on it now.