| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
I've
read the paper by John Siemon and Dean Stoddart on the subject of RJ-45
reliability for PoE, and I have some comments:
1. I agree, the damage appears to be from arcing. But the
paper doesn't seem to distinguish between the two causes of
arcing:
a. On insertion, the arcing is due to inrush into the
capacitor.
b. On extraction, the arcing is due to kick-back from
the inductor
2. The experiment used a DC power supply in place of a PSE.
This causes arcing on connector insertion that wouldn't happen with a PSE,
because the PSE in detection mode has a minimum of 45kohms source
impedance. I did a quick SPICE simulation of the RLC circuit with a step
voltage input to estimate the peak inrush current. I assumed 0.1 Ohms
series resistance to approximate the wiring (something the Siemon/Stoddart paper
doesn't seem to mention). I observed a peak of about 13A. This is
probably where the damage is occurring, and it wouldn't happen with a
PSE.
3. The arcing on extraction probably causes little or no damage
since it's energy is so low. The test was at 57V, with 105 Ohms
load. This means 543mA thru the 100uH inductor just before
extraction. The energy stored in the inductor is 0.5(0.543^2)(1e-4)=14.7
micro Joules. (By contract, the energy that the 5uF cap sucks up after
insertion is 8.1mJ, or 552 times higher.)
For
these reasons, I think the results were invalid. I recommend the
experiment be repeated using an actual PSE instead of a bench supply. I
predict the amount of damage observed will be dramatically reduced, perhaps even
imperceptible.
|