Correct treatment of Analog Ground

Designing a custom Apalis SOM carrier based on the Ixora Reference schemaitc1.0 and Apalis Carrier Board Design Guide. The MXM3 SOM connector defines the area at pin numbers 303-321 for analog ADC, Touch and audio ports along with the related analog Vcc power and AGND. The Apalis Design Guide Figure 51 depicts the AGND pins directly connected to the regular GND plane, despite the AVcc LDO regulator referred to AGND in the same figure. The Ixora carrier 1.0 schematic shows the MXM connector X1A AGND pins 303,304,308 and 313 tied to AGND, The bead L21 is on another page near the AVcc LDO IC8 with a note that states the bead should be located close to the MXM connector AGND pins.

Is the direct connection to GND in fig 51 of the board design guide really a typo? According to the TK1 SOM datasheet, the AGND SOM pins are tied directly to the internal GND, thus rendering the external GND separation useless and possibly risky for EMI. Generally, great care is required when defining separate ground regions in a mixed-signal design, and is often discouraged by chip producers in order to avoid creating ground current loops that could generate EMI and other problems.


thank you very much for your interest in our products and for using the Toradex Developer Community.
You are definitely right, the figure 51 of the Apalis Carrier Board Design Guide is not correct, there is a typo. The module pins should be connected to the AGND.

Thank you very much for your valuable feedback!

Nevertheless, the other information related to the AGND pins on TK1 connected to the digital GND is also correct.

In general, our experience says that it is always good to have a single digital GND but virtually divide it to have an area which is not crossed by high-speed signals.

Also, the other approach using a separate AGND plane works, but care must be taken in order to ensure that this plane is thick enough to avoid voltages across two sides of it.

I really hope that this helps, please don’t hesitate to write again if needed.

I suggest that since the planes are combined on the SOM module, they should remain combined on the carrier. The plane on the SOM module will carry both digital and analog GND return currents, resulting in digital GND noise present across the SoC chip. Analog signals entering the SoC will not see noise differential if they remain referenced to the common plane that is also the Vss of the analog circuitry in the SoC. If unimpeded plane is extended out to the point of external connection, this will prevent conversion of common-mode noise into differential. It’s important in this case, and actually in all cases, that digital signals do not traverse the immediate GND plane where the analog signals are run, hence the analog end of the MXM connector right out to the analog connection points should be a “no-fly-zone” for digital signals.

The only case where an RF-separated plane works as intended is if the separation carries right into the SoC analog section, since on-chip digital return currents will not be present on the AVss. There are two major risks using separated plane when AVss is locally tried to a common GND. First, the digital GND noise on the SoC plane will appear as differential on an incoming signal that is referred to a separated plane. The design is then dependent on the stray coupling of the signal to the local common GND plane to reconcile the noise difference. Second, an RF-separated plane extending from the common GND on the SOM module appears as a moderate impedance antenna, and can exhibit weird resonances depending on its shape.

you explained it perfectly, wìthat’s exactly what I meant in my previous comment. Thanks a lot for your interesting and accurate analysis, that is really valuable for us and for other customers which will be wondering on the same problem.
Thanks again, I wish you a nice day.