back to index

Ethernet Keychain Tester

Problem

Sometimes it is needed to see if an Ethernet switch port or a wall socket is alive, or if a Power-over-Ethernet power feed is present.

The instruments on the market are large and expensive. A simpler device for everyday wear, with cost so low it can be lost or gifted with impunity, is required.

Method

Earlier it was found that the autonegotiation signals the switches are sending are powerful enough to light up some LEDs. The LED has to have suitably low forward voltage drop (so blue ones cannot be used), match high eye sensitivity because the energy of the pulses is low (so blue and red are out), and emit light, even if weakly, at very low current.

For the activity, two antiparallel LED pairs were connected at the Tx and Rx pairs (orangewhite/orange at 1,2, and green/greenwhite at 3,6). No serial resistor is used as the signal is inherently low-current.

Strictly said, the antiparallel pair may not be needed. It however can indicate a badly crimped cable where the pair wires are swapped. The LEDs also provide protection for each other against too high voltage across the diode if the polarity is wrong, as LEDs have pretty low reverse breakdown voltage.

The PoE sensing is done by a green LED, with a 4k7 resistor in series. This limits the LED current to a sensible value from 12 to 48 volts, while the glow is visible from 3 volts up.

Earlier version of the device had a DPDT pushbutton that connected the Rx/Tx pairs into a loopback, sending the autonegotiation frames back to the switch port and lighting up its LINK light. This however was dropped from this version to keep it simple and cheap, inter alia because the gigabit switches do not appear to work with such loopbacks.

Device

Design

The schematics was drawn in Eagle and manually panelized to a 5x5 cm board. The milling lines were drawn to separate the boards. The lines weren't made entirely accurate as it was not expected to get the boards milled apart.

Twenty boards were ordered from Dirty PCBs for the cost of


Boards, as received

Boards, as received

Boards, as received

Boards, as received

Boards, as received

Boards, as received

The boards have to be broken apart, then the hole edges ground away with a Dremel wheel.


Boards, broken up

The fit was tested with a RJ45 connector, an inexpensive kind.


Board, dry fit

Board, dry fit

Board, dry fit

Board, dry fit

After finding everything fits correctly, the connector was removed, the SMD parts were soldered on the board, and the connector was replaced and soldered in place.


Soldered assembly

Soldered assembly

Soldered assembly

Soldered assembly

Board lid

The board is exposed to the elements and abuse. To protect it, a lid is needed. For the prototypes it was decided to 3d-print one.

Several iterations with minor differences were designed in OpenSCAD. Some suffered small problems - being too thick, having too thin layer, or having a gap between compartments (to be important once transparent resin is poured in).

First iterations were printed from PLA and spray-painted black (one left natural transparent for comparison), the final ones from flame-retardant ABS (should be black, is dark-grey).

The print took about 6 minutes to run and consumed about 20 cm of filament (0.5 cm3 of material). The layer thickness was 0.3 mm. The printout had to have its top side filed smooth, as it was slightly uneven and would not mate well with the flat circuitboard.

The lid will be attached in place with epoxy, which will also act as a sealant for the LED viewports to avoid accumulation of dirt inside - the device should be maintenance-free.


Lid draft

Lid draft

3D-printed lid, top

3D-printed lid, bottom

3D-printed lids, different versions, tops

3D-printed lids, different versions, bottoms

3D-printed lids, different versions

The lids were placed on the boards and press-fitted onto the snap-fit pins of the connector. A small ring was added for easier attachment to the keyring.


Lids in place

Lids in place

Lids in place

Lids in place

Performance

The device indicates presence of the link autonegotiation signals by faint blinking of the LEDs. Only one LED of the pair should be active. Only one pair is active when the switch is MDI-only, both pairs light up alternatingly when MDI/MDI-X is present and the switch is trying both Rx and Tx pairs.

With some switches, there are no apparent signals present. The signals can be coaxed into existence by plugging the cable from the port into another switch or a device, letting the link establish, and then unplugging it and plugging it to the tester. For several seconds the LEDs will blink, then the signals will disappear. It is suspected this is the behavior of the Energy-Efficient Ethernet standard. Some switches that support management/configuration can disable it.

Videos of operation

Canyon 8-port 10/100Mbps powersaving switch

After power-on or disconnection from an active device. In about 10 seconds the switch goes to power-saving mode, sending only occassional brief pulses that are difficult to see.
Netvin switch, dumb and ancient

Old dumb switch, downlink-only port, no MDI/MDI-X
Netvin switch, dumb and ancient

Old dumb switch, uplink-only port, no MDI/MDI-X
blue Zcomax router/switch/wifi AP

No power-saving, autonegotiation attempts on both pairs
TP-Link TL-SG1024, rack gigabit, power save

Power saving, occasional autonegotiation attempts on both pairs
WRT-54GL

No power-saving, autonegotiation attempts on both pairs

TODO

Plans for Mk.2

Next generation should have a battery, an amplifier for the LEDs so the signals are clearer to see even in full daylight, and a microcontroller for sending simple autonegotiation pulses in adjustable duty cycle (few seconds on, few seconds off) to light up the LINK signals on the switches. This will allow, in addition, to track which port in the facility belongs to which switch.


If you have any comments or questions about the topic, please let me know here:
Your name:
Your email:
Spambait
Leave this empty!
Only spambots enter stuff here.
Feedback: