2016-10-08, 13:54
Back again...
Bought another TL-POE10R and capacitors (2 x 0.1 uF 50 V ceramic, 1 x 100uF 25 V electrolytic) from Maplin, the caps take the purchase into the free P&P range. Chokes from Amazon but an add on item that has gone up 50p each since I bought mine, 220 uH, 4A and made made up a small PCB(*) to mount them on. Works a dream.
'Scope images, input has just over 100 mV of noise as before, output less than 10 mV of ripple. Both images taken with the 'scope set at 20 mV/div and 2 ms/div. Controller is now reporting a much healthier 4.9 Volts input and is quite happy. No noticeable noise being received either. There are still some quite large spikes/dips that follow the GPS LED and RJ45 data LED flashes but nothing seems to be objecting. I did try adding another 100 uF of capacitance on the filter output, made no real difference. Circuit is chokes in series with each power line, 0.1 uF across the input and output sides of the chokes and the 100 uF across the output side.
Checked for voltages between house supply earth and the 0 V rail, 38 V 50 Hz. Very variable and joining the two together through an ammeter gave a whole 20 uA of current flow and the volts disappeared, not going to worry about that. Worth bearing in mind and checking out if the PoE injector used is earthed (mine isn't). Adding the ground connection didn't make any difference to received noise levels, whether it improves signals is for later.
(*) I actually made two boards. It was going to be four as that is how many fit across the bottom of a sheet of "press 'n peel" but the guillotine found it harder to cut than I expected... A PDF of the layout is attached. The board is symmetrical, there is provision for 2 components across each side of the chokes. When fitting the chokes try and work out which wire is the "start" and put that in the same shaped pad for each choke. In theory any stray magnetic fields from the chokes cancel rather than couple.
The second board is drilled but not cleaned and up for grabs.
Bought another TL-POE10R and capacitors (2 x 0.1 uF 50 V ceramic, 1 x 100uF 25 V electrolytic) from Maplin, the caps take the purchase into the free P&P range. Chokes from Amazon but an add on item that has gone up 50p each since I bought mine, 220 uH, 4A and made made up a small PCB(*) to mount them on. Works a dream.
'Scope images, input has just over 100 mV of noise as before, output less than 10 mV of ripple. Both images taken with the 'scope set at 20 mV/div and 2 ms/div. Controller is now reporting a much healthier 4.9 Volts input and is quite happy. No noticeable noise being received either. There are still some quite large spikes/dips that follow the GPS LED and RJ45 data LED flashes but nothing seems to be objecting. I did try adding another 100 uF of capacitance on the filter output, made no real difference. Circuit is chokes in series with each power line, 0.1 uF across the input and output sides of the chokes and the 100 uF across the output side.
Checked for voltages between house supply earth and the 0 V rail, 38 V 50 Hz. Very variable and joining the two together through an ammeter gave a whole 20 uA of current flow and the volts disappeared, not going to worry about that. Worth bearing in mind and checking out if the PoE injector used is earthed (mine isn't). Adding the ground connection didn't make any difference to received noise levels, whether it improves signals is for later.
(*) I actually made two boards. It was going to be four as that is how many fit across the bottom of a sheet of "press 'n peel" but the guillotine found it harder to cut than I expected... A PDF of the layout is attached. The board is symmetrical, there is provision for 2 components across each side of the chokes. When fitting the chokes try and work out which wire is the "start" and put that in the same shaped pad for each choke. In theory any stray magnetic fields from the chokes cancel rather than couple.
The second board is drilled but not cleaned and up for grabs.