EMC Immunity Issues with RS-232 to USB Converters

These little converters are super handy to interface between your modern PC or laptop and the simpler, lower technology RS-232 serial port used by many pieces of equipment for control or debug purposes. However, like any commodity item there are design compromises, including EMC ones, that you need to be aware of.

I was recently performing some Electrical Fast Transient (EFT) testing on a customers product and was surprised to observe it failing at quite a low level of injected transient of 200V. It appeared that the whole system crashed when the bursts were applied to any of the digital I/O ports.

Even more confusing was that I’d looked over the schematic and the port protection measures that they had implemented were very sensible with ferrite beads and diode clamps.

A pointer came from observing the front panel of the device with all of it’s indicator LEDs blinking away as if it was working properly. Yet the equipment under test (EUT) wasn’t responding to serial communications and the TeraTerm serial port software was still showing a connection.

Checking through the test setup, I theorised that the RS-232 to USB converter that I was using might be crashing or responding to the EFT pulse as a start bit to a frame. Despite being isolated with a Coupling/Decoupling Network (CDN), when a scope probe was added to the RXD line on the decoupled side of the CDN, a transient with 30V of pk-pk amplitude was visible when the EFT burst was applied.

I tried two other converters that I had in the lab and none of them were happy with this pulse and also refused to work correctly.

a selection of usb to serial converters

So I knocked up a small filter PCB with a pi filter on each line (RXD, TXD and 0V) consisting of 2 x 100pF capacitors and a ferrite bead. The non-line side of the caps was taken to the HF ground plane using some adhesive copper tape (the EMC scoundrel’s last resort!) to return the currents back to the generator and not into the converter.

EFT test setup showing flow of HF current and position of small filter

Success! No more interference and the converter works perfectly.

As an experiment (OK, I got slightly distracted by something interesting) I played around optimising the filter and managed to get it down to just two components – a 100pF capacitor on the TXD and RXD lines of the converter.

Now I know that these devices will be designed to the lowest price point but two 0402 capacitors is hardly breaking the bank! It does make you wonder how they managed to get through their own EMC testing, if at all.

Incidentally when this was later tested in the chamber it had some fairly strong 12MHz harmonics from the USB 1.2 data lines that only just squeaked under the limit line lending further weight to my suspicions of corner cutting and poor design!

So today’s lessons are:

  • Beware of cheap generic test adaptors and EMC issues caused by them – both immunity and radiated.
  • Consider your port filtering carefully. Many I/O interfaces can stand a small capacitor or filter adding to it and the benefits for EMC are significant. It gives a path for interfering signals to the local ground and will also improve your emissions too. The customer who’s product I was testing had such parts fitted; it passed the testing at 1kV EFT without issue (the spec is 0.5kV).
  • Using a fibre optic serial port adaptor would probably have helped here by increasing the common mode impedance of the connection (assuming of course it had been designed properly!)