1
\$\begingroup\$

I have recently produced a number of prototypes using the Microchip ENC424J600 Ethernet IC.

I can't a full circuit diagram as this is a commercial product but I have reproduced the relevant part of the circuit below. What I am seeing is that as soon as I apply power the green LED lights, indicating a good link, even without an Ethernet cable plugged in.

If I am to believe the data sheet this LED should not require me to have good communications with the micro that is driving it and this LED should work provided it is powered, the oscillator is running and the Ethernet cable is plugged in and connected to a suitable device such as a hub.

Does anyone have experience with this IC and can suggest what I may be doing wrong.

If I have understood the data sheet correctly this is a typical application circuit enter image description here

As as been pointed out in comments the data sheet does not show the TPIN transformer with the centre tap pulled up to 3V3. My original circuit does not have this connection. I added this because it is shown on the circuit diagram of Microchips demo board for this IC. The LED performs in the same way in either case.

I have also scoped the OSC1 pin to confirm it has a clean signal.

\$\endgroup\$
6
  • \$\begingroup\$ have you probed the crystal with an oscope to make sure it's oscillating? Although the datasheet says that the LEDs will light up as long as the "Ethernet magnetics are present and wired correctly," it is possible that they will not turn on until the auto-negotiation process is completed. If you did not lay out your board with proper impedance matching, then the auto-negotiation process might fail...... Maybe try configuring it as 10Mbps (to make auto-negotiation easier) and then see if the LEDs light? \$\endgroup\$
    – John M
    Apr 14, 2016 at 17:13
  • \$\begingroup\$ Section 2.5 says that the LEDs can be used as a check for the hardware connections. If you look at figure 2-5, I see that the TPIN+/- center tap is connected via 10nF to the termination network, but not pulled up to 3.3V. In your schematic it is. I am not sure if the different bias voltages (e.g. GND and VCC) are a problem for the TPIN/TPOUT pairs, but you may wish to correct the circuit according to datasheet. \$\endgroup\$
    – Hans
    Apr 14, 2016 at 17:17
  • \$\begingroup\$ @johnny_boy thanks for your input, I have a clean 25MHz coming in generated by a CMOS oscillator. The light is coming on even without an Ethernet cable plugged in. \$\endgroup\$ Apr 14, 2016 at 17:18
  • \$\begingroup\$ @WarrenHill The light coming on without the ethernet cable plugged in is expected, as the datasheet says "[the led will turn on as long as] Ethernet magnetics are present and wired correctly." Check out the text directly under figure 2.7. Perhaps this means that the LEDs will indicate "are the magnetics correct?"... just speculation. \$\endgroup\$
    – John M
    Apr 14, 2016 at 17:52
  • \$\begingroup\$ As @Hans has mentioned, you have your TPIN transformer & biasing resistors center-tap pulled up to 3.3V but the application schematic does not. Do you have a reason for doing this ... ? \$\endgroup\$
    – brhans
    Apr 14, 2016 at 18:43

1 Answer 1

1
\$\begingroup\$

From the datasheet:

Both LEDs automatically begin operation whenever power is applied, a 25 MHz clock is present and the Ethernet magnetics are present and wired correctly. A connection to the host microcontroller via the SPI or PSP interface is not required. LEDA and LEDB can, therefore, be used as a quick indicator of successful assembly during initial prototype development.

Then it talks about another "bit" in a configuration register, I'd imagine that this would change it to something you'd expect, but you have to connect to it with a micro to write to this register.

STRCH: LED Stretching Enable bit 1 = Stretch transmit, receive and collision events on LEDA and LEDB to 50 ms 0 = LEDA and LEDB outputs show real-time status without stretching

There is also a register that lets you monitor the status and led state. REGISTER 8-2: EIDLED: ETHERNET ID STATUS/LED CONTROL REGISTER

The best way to verify systems that are being prototyped is to test all the pieces one by one. Start with the spi, see if you can write and then read configuration registers. Then test the ethernet, send a small packet and see if it shows up in the buffer of the phy chip.

If you can't read a register on the spi side then you have a problem with the chip or the setup, and you'll need to do some schematic and hardware troubleshooting. With any design, first check power, check both DC and AC levels (ie, switch the meter to ac, if there is too much ripple you might have problem. Also with power, check the startup conditions. I've had cheap power supplies that had an overvolt condition on startup that was messing with my design. Then check the clock signal.

If your pulling this design directly from the manufacturers example, it might be a good idea to compare this design to the manufactures. If you assembled it yourself examine the pins and check for solder bridges. Also, if there is some way for you to measure the chips temperature or current, make sure it's reasonable. If that doesn't work, replace the chip and follow stringent ESD and package handling guidelines (humidity can kill chips, esp when soldering).

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.