emc problem about radiation emission on GPIO at PUSH-PULL mode

[yundm2]

Posted on October 17, 2015 at 03:40

I use STM32F373, freq of PLL is 72MHz.

If one pin (PD1, PD2 or any pin) is configured in PUSH-PULL mode, then 144MHz harmonic is immediately detected at this pin by the near field probe with one spectrum analyzer.If this pin is configured in OPEN DRAIN mode with PULL UP, harmonic is disappeared.Does anybody can explain such  appearance and tell me how to disappear such harmonic when in PUSH-PULL mode?Any help is appreciated.

[Tesla DeLorean]

Posted on October 17, 2015 at 17:27

Not sure why specifically setting pin(s) in PP would cause more noise. I think you'd want to review the power and ground connectivity to all the supply pins, and the routes over which current would flow. Check also the analogue supplies, the PLL/VCO are powered there, and I'd suspect the VCO is pulsing at 144 MHz, the PLL clock is normally halved to ensure 50/50 duty of the output clock.

Is this a custom board? Is it a 4-layer PCB with ground/power planes? What kind of ripple currents/voltages are you seeing on the supply pins?

[yundm2]

Posted on October 18, 2015 at 07:43

Hi clive1. Thank you for reply.

It's a 4-layer PCB with ground/power planes designed by myself.I check carefully the radiation near the cpu step by step when program running.After the PLL is ready (72MHz) but before selected as system clock, no regular radiation on 72MHz or 144MHz near the CPU can be probed out by near-field probe.Once step one code that select PLL as system clock, radiation on 72MHz and 144MHz are immediately probed out on almost all corners of STM32F373VCT6 except one corner where the VDDSD12/VDDSD3/VREFSD+ pins are nearby. The radiation strength of 144MHz is about 50dBuv and 72MHz is a bit of lower. Before initiatlize any GPIO, there is no radiation on route connected to this GPIO pin.Hereafter, if a GPIO pin with longer route (120mm) is configured PUSH-PULL, radiation on 144MHz is immediately probed out (about 50dBuv) on such route. If I cut off this route at end nearby this GPIO pin and connect this route to this GPIO pin by a magnetic bead in serial, radiation is dissappered.I consider three group supply in my pcb.1st group is VDDA/VREF+ connected to +3.3V by a magnetic bead.2nd is VDDSD12/VDDSD3/VREFSD+ connected to +3.3V by another magnetic bead.3rd is all VDD directly connected to +3.3V.I will check ripple voltages on this supply pins tomorrow.Thanks again.Any idea is appreciated.