I'm current using an Arduino as standalone with the ATMEGA328 chip. If in this project I'm not going to use the ADC do I have to connect AREF, AVCC and AGND? Not connecting these can make my PCB simpler.
ze.gmonteiro:
I'm current using an Arduino as standalone with the ATMEGA328 chip. If in this project I'm not going to use the ADC do I have to connect AREF, AVCC and AGND? Not connecting these can make my PCB simpler.
You must connect Avcc and Agnd even if not using the ADC. Aref could and should always be left unconnected unless you are going to use an external voltage reference for the ADC.
Lefty
I got into a bit of a tiffy because of this before, but Ill say it again, no not really, only if you have no use for analog input
AGND according to the stuffy shirt bunch should be connected, though I have had an atmega328p running though all sorts of "tapping the limits" abuse with only one ground connected, in a 85 degree C 85% humidity environmental chamber for over 2000 hours with no performance issues as of last sample.
Ivory tower answer is, yes you need them connected if you wish to meet atmel's (ideal situation) bell curves.
Osgeld:
I got into a bit of a tiffy because of this before, but Ill say it again, no, only if you have no use for analog inputAGND according to the stuffy shirt bunch should be connected, though I have had an atmega328p running though all sorts of "tapping the limits" abuse with only one ground connected, in a 85 degree C 85% humidity environmental chamber for over 2000 hours with no performance issues as of last sample.
Not surprised you got into a bit of a tiffy over the subject. Usually the bible for such questions should be taken to the applicable AVR datasheet for the device. Assuming we are talking about a mega328P chip the datasheet says:
1.1.7 AVCC
AVCC is the supply voltage pin for the A/D Converter, PC3:0, and ADC7:6. It should be externally
connected to VCC, even if the ADC is not used. If the ADC is used, it should be connected to VCC
through a low-pass filter. Note that PC6..4 use digital supply voltage, VCC.
yea its grey, the safe awnser is in the datasheet, the real answer requires pushing it, somewhere in the middle is the truth.
take the eprom for example, we have seen it go past a million write cycles, its only rated for 100k ... do you feel lucky?
if its something that absolutley needs to be that way in a PCB design do you feel lucky? or do you feel like going for the average over a very small sample during the data gathering phase (maybe what 30 units in early design verification vs. over multimillion runs)?
I say yes, you say no, its somewhere in the middle, and probably dependent on application.
If you do something that the data sheet says you should not then all bets are off. It might function but it won't work. If you have any problems with it you will get no help from anyone.
Why do you think these recommendations are made?
Do you think it is is people being stupid?
Remember Challenger and the O rings? Engineers were ignored then as well.
If you have any problems with it you will get no help from anyone.
Obviously you have never talked to anyone at Atmel about support
hell our company was buying 40k units of a part a month and got a 2 month EOL notice from them with a 4000 unit backstock...
Why do you think these recommendations are made?
CYA
Do you think it is is people being stupid?
no its being safe, I do the same thing in production. Do you want to have a bad day and deal with 6 months of reaming?
Remember Challenger and the O rings? Engineers were ignored then as well.
Really?
Ok I am out, basics stand, you want to be 110% sure go by the datasheet, but keep in mind the data sheet was based on a very limited sample during prototyping and may not reflect the exact truth.
I am trying to find out that truth based on my limited testing, but with the MCU pegged at 245ma load under thermal stress running since the first of June, I dont totally buy the datasheets. Though I am not saying that you can count on a part to run at hells kitchen for over 2000 hours with no faults. Do you feel lucky?
For the record I also recently spent a small pile of money on fets that were ESD rated for 35KV on the datasheet, and they puked on themselves at 10KV, 94% (and some change) of the time, in that case the datasheet was worth less than paper one would read or use in the bathroom ... so datasheets are not golden gods 100% of the time
just sayin
Thermal measurements are only one factor. Have you tested the noise immunity?
Obviously you have never talked to anyone at Atmel about support
I had extensive contact with Atmel in the past and they were singularly useless. They quoted life on some flash memory as 100,000 our testing showed that this was closer to 80. So much for the data sheet being conservitave.
The correct treatment of grounds and power supplies is essential if you want your device to work RELIABLY. You might find that things seem to be OK so long as you have got some power getting in, but digital electronics is sensitive to noise on the supply lines (particularly ground).
You should absolutely follow the data sheet instructions with respect to these things, and also ensure that you provide decoupling capacitors close to the supply pins (0.1uF ceramic).
The likely result of not maintaining proper power supply 'hygeine' is that your device works some of the time, but falls over and crashes intermittently.
David
One thing I noticed where AVCC makes a difference. I bootloaded 8MHz internal 8MHz osc onto an ATMEGA328. I run it bare (without even a pullup on RESET, with only VCC and GND connect on the low pin side), and blink runs fine on 5V, but it fails to run on 3.3V. As you can guess, it ran fine once I connected the GND and AVCC on the high pin number side. I hope this saves time for other who try this 