OBD2 Pwr+Gnd causes poor GPS reception in device plugged into it, looking for causes & solutions

Question

Myself and a few others have built a custom device (PCB & components) which is designed to run while plugged into an OBD2 connector in a vehicle and capture GPS location data and transmit it via cellular to a cloud server. General functionality works well. However, we have noticed that we are consistently seeing poor GPS quality while the device is plugged in (e.g. a common scenario is the location is shifted to one side or another by about 3-4 meters or more and just generally drifts away from the true location by several meters in random ways). This phenomenon does not occur if we power the device directly from a separate battery (the reported GPS locations in this case form a much tighter group within about 1-2m of the true location). We've been back and forth with the GPS module & antenna vendor on the GPS reception aspect and believe that is all configured and tuned correctly (and indeed when we provide nice smooth power to the device we generally get great GPS reception).

When the power enters the device it goes through:

• an eFUSE/protection module: TPS26620DRCR
• a switching regulator: AP64352SP-13
• and it has appropriate decoupling caps near each of the relevant modules (MCU, GPS module)

We're looking for ideas at this point of how to narrow this down. We're not sure if the problem is possibly related to ground bounce (and btw, we're using chassis ground pin 4 on the OBD connector, and when we tried pin 5 signal ground the problem oddly enough gets worse). Or if it could just be ripple on the main input power somehow still making its way into the system, or both.

Any ideas of things to test, or common ways of protecting or isolating against this type of thing would be greatly welcome.

Answer 1

For the sake of posterity, I'll mention as a follow up here that additional decoupling capacitors at various frequencies ended up making a huge improvement in GPS signal quality. In terms of decoupling, the power system was more or less set up as:

12Vin -> eFuse -> 5v swreg -----> [each module with it's own decoupling caps]

And between myself and the team working on this we found that carefully going through the power line and adding more decoupling at every stage made a significant difference in GPS quality (we observed typical margins of error were cut in half or more). Particularly the output of the 5V switching regulator had only one or two caps, and the very front main input originally had none - and we focused on adding capacitors to and trying to smooth out the power as much as possible before it was even delivered to the various modules. That made a big difference.