GND and AGND short

Hi,

The answer is to help you to make a correct PCB for analog/digital measurements. There must be somewhere a common point, they cannot be separated completely and at one moment you must short them, otherwise you cannot make the measurement.

Also see this at page 1370 in data sheet:

I agree with you. They must be shorted at some point on the board. What is the point if they are shorted inside the micro-controller ? I took a micro-controller alone and checked the connectivity between GNDA and GND.They are short internally.

Can anyone clear my doubt.

Thanks in advance

Hello Sai,

Your point is that if you keep the uC on a bench and check continuity between the GNDA and GND pins, the beeper goes off. Right?

Can you let us know which two pins on the package did you check for?

Regards

Amit

Actually for any GND pin on the uC and GNDA are internally shorted. I have even tried with other uCs(new)

Any idea why is this happening?

Hello Sai

Thanks for the info. I will check this with the team and post an update in the next week.

However it will not be affecting your development work?

Regards

Amit

I am not sure. It might effect. We separated all analog and digital grounds on the PCB that we have designed. They are shorted at very far distance on the board. But now, because of the uC they are shorting at the micro-controller itself.

Hello Sai

What it would affect is the Analog module behavior (but marginally)

Also as you mentioned that you have a far end short. I hope there is a good copper pour and not just a thin via or short.

Regards

Amit

Like all previous responders here - I've no "insider knowledge" of these MCU's internals.

However - your measurement method may be suspect - and may color your findings.  MCU alone - as you describe - was not under normal/customary bias and thus its pins must be probed w/some care/consideration.  You do not want excessive currents to flow nor do you want to, "break down" pn junctions.  A crude meter may do both - your "curiosity/checking" may have injured the device(s)!

Indeed - at some well chosen point - both analog & digital grounds should join.  We were taught that the analog ground path should not carry high or variant currents - as either would necessarily reduce ADC accuracy.

From my knowledge of other MCU's "internals" - it may be that w/in these MCUs the Analog ground paths are routed strategically - usually with a directness and avoidance of likely noise sources.  W/in the MCU I'd bet that analog ground and digital ground paths are separate - and that efforts are made to prevent/reduce any "digital ground bounce" from impacting analog ground...

After what you have said, I am guessing the internal of uC to have GND and AGND separated by some kind of a ferrite bead which acts as a short for DC (When I check with multimeter, it shows short) but offer a high impedance for AC signals(which i did not check for).

Am I right with the above reason?

Best answer arrives from Amit's, "insider harvesting." 

Past performance MCUs my group designed likely did not include ferrite bead.  (was 10+ years past - I'm not 100%)  Great pains were taken to reduce current on analog ground and avoid nearness to noise sources as that analog ground routed internally.  (PLL, oscillators etc)

However - we note improved ADC behavior (these days - these/other ARM MCUs) by our use of 06-03 ferrite bead (as you mention) but placed in series w/3V3 (and adjacent) to Vdda - joined by normal/customary filter caps - on our pcb.  And - to restate - try to route your board so that digital ground currents are not directed along the analog ground path.