Two steps forward, one backward

Last week we discovered that our PoE for Raspberry Pi eventually breaks the Ethernet port of connected Raspberry Pi. Since we have a few units which have been running constantly for almost two weeks, we started to suspect that the problem is in connection/disconnection transients. The first step was to find out if these connection/disconnections really were the cause of the problem, so we powered a Raspberry Pi separately and plugged the Ethernet connection a few tens of times. The Ethernet port of Pi stopped working quickly, and we had the isolated the cause of failure.

Time for some oscilloscope shots: the oscilloscope was connected on a differential pair after the transformer. Red and blue are differential wires, pink is the voltage differential between the two.

Spike03

 

This was one of the cleaner events on the cable, with only one clear spike on it. Our initial suspicions are confirmed, the ~57V on the unloaded PoE line gets coupled to the data lines. The recommended protection against these electrical transients is to add clamping diodes across the transformer pins. We had some suitable zener diodes lying around, so it was a quick matter of connecting them in antiseries and soldering them down on pins of the transformer.

026

Time to take an another look with the oscilloscope. This time I caught a real monster of a cable transient on a tape:

zener03

Even while the supply lines bounce around relative to ground, they stay nicely within 3 V swing relative to one another. This means that the attached Raspberry Pi should receive a lot less electrical stress from these events.

The attached Pi did survive a long series of cable connections, but it still could not form a link with the switch. On a hunch I forced the data rate down to 10 Mbit/s, and the Raspberry Pi was happy to communicate at that rate.

The degradation of speed down to 10Mbit/s is not acceptable, but hopefully this will not be a problem when we use lower-capacitance surface mount diodes close to transformer pins. There always is a option of using specialty ultra-fast, low capacitance transient voltage suppressor arrays if zeners seem to cause problems.

 

Share your comments

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s