Ethernet link LED (the green one) is on but no cable is connected!

Hi all,

I'm using LAN8720A ethernet chip in my design with an STM32F4 MCU.

Although I don't connect any cable, the link LED is on.

The chip seems to draw normal current and doesn't heat. Besides, I didn't get any clue if the MCU and ethernet chip communicates or not, so I doubt it's working.

So what does it mean if this LED is always on without a link? Any idea on the debugging steps for ethernet chips generally or specifically?

You followed the designguide in datasheet chapter 3.8.1.1?
Please provide an extract of the Ethernet PHY circuit, showing the LED1/REGOFF pin and the connected parts.

Because the LED pins have different function in different pinstrap defined configurations, you should show the complete schematic.

A likely fault, particularly with prototypes is to have missing solder joints of the small chip scale package.

You can get much information about the PHY state through management interface by dumping all registers.

rho-bot said:

You followed the designguide in datasheet chapter 3.8.1.1?
Please provide an extract of the Ethernet PHY circuit, showing the LED1/REGOFF pin and the connected parts.

Click to expand...

Here is my schematic. The design is actually the same as the STM32F4 Discovery Base Board.

Complete ethernet schematic:
https://obrazki.elektroda.pl/5658664300_1470635423.png

LED part:
https://obrazki.elektroda.pl/3550649100_1470635424.png

- - - Updated - - -

FvM said:

Because the LED pins have different function in different pinstrap defined configurations, you should show the complete schematic.

A likely fault, particularly with prototypes is to have missing solder joints of the small chip scale package.

You can get much information about the PHY state through management interface by dumping all registers.

Click to expand...

It was a total pain to solder the IC. I spent some IC but everything seems to be fine at the end, I checked with needle probes.

I guess the the ethernet chip is not working although it has an oscillation on its crystal, couse I don't receive any response when I trace the MCU.

I believe there are at least two schematic errors:
- LED2 must be equipped with an external pull-down resistor to select the right REFCLK mode and active high LED2 level
- VDDCR is the internal 1.2V regulator node and must never be connected to 3.3V. Supplying more than 1.5V will damage the chip. VDDCR will be connected to external bypass capacitors only in the selected internal regulator mode.

See below a working STM32 interface

FvM said:

I believe there are at least two schematic errors:
- LED2 must be equipped with an external pull-down resistor to select the right REFCLK mode and active high LED2 level
- VDDCR is the internal 1.2V regulator node and must never be connected to 3.3V. Supplying more than 1.5V will damage the chip. VDDCR will be connected to external bypass capacitors only in the selected internal regulator mode.

See below a working STM32 interface

View attachment 131389

Click to expand...

Thanks for your valuable reply. As I checked, these are different with the reference design, so obviously I misconnected that VDDCR pin and forgot to connect the pulldown resistor. I will fix these errors in the 2nd version and see if it works.

I played on the card and managed to change the design by cutting lines and soldering thin jumper wires.

I also changed the chip in case it burned, but still no response.

I measure 1.2V on the internal regulator pin(VDDVR) and there is a little noisy(120mVpp) 666mV dc voltage on the both ends of the 25MHz crystal.

What do these may mean? Any idea on the possible problems?

I would also suggest to add a weak pulldown to LED1 pin, to get sure Int. LDO is activated always.
You might also consider to add additional pullups on MODE0-2 pins (depends on the pin capacitance of the STM32 & traces)
VDDCR pin needs some ext. buffer caps connected to stabilize (1µ||470p, see chater 3.9.2), you -if still connected- have 2x 1n...
Reckeck your oscillator circuit, looks somehow strange with the series resistors and the additional C to GND. Maybe it's not starting to oscillate. The design FvM posted looks more common ;-)
With pulldown on LED2 pin, you should see 50MHz clock on pin14 (I hope the STM32 connected is an input to the signal). If not, you have an oscillator problem....

rho-bot said:

I would also suggest to add a weak pulldown to LED1 pin, to get sure Int. LDO is activated always.
You might also consider to add additional pullups on MODE0-2 pins (depends on the pin capacitance of the STM32 & traces)
VDDCR pin needs some ext. buffer caps connected to stabilize (1µ||470p, see chater 3.9.2), you -if still connected- have 2x 1n...
Reckeck your oscillator circuit, looks somehow strange with the series resistors and the additional C to GND. Maybe it's not starting to oscillate. The design FvM posted looks more common ;-)
With pulldown on LED2 pin, you should see 50MHz clock on pin14 (I hope the STM32 connected is an input to the signal). If not, you have an oscillator problem....

Click to expand...

Do the values of the capacitors really matter much? In fact figured out that I put one capacitor with a wrong value to VDD2A pin.

Also there is a 1Mohm resistor between the two ends of the crytal also in my referance design, but I forgot to put it. Is this a fatal mistake?

It's essential to have a certain amount of bypass capacitance for each node. In case of the 3.3V supply, there will be probably more capacitors at other point of your circuit, having e.g. 100 nF at each supply pin should be sufficient then. Proper operation of the internal 1.2V supply depends on sufficient VDDCR bypass capacitance.

Regarding 1 Mohm resistance, I found that it's not populated in the reference design, so it's apparently not required. But apparently doesn't hurt because I have in in a number of designs.

Yes the Caps on VDDCR are mandatory, its the supply for the core logic :!:
You already mentioned the ripple you observed, which is 120mV and at spec limits (1.14V .. 1.26V)...

The 1Meg is not that critical. Needed to support start of oscillation, but depends on the crystal you use.
Please show us your current schematic, as it seems there are diffences to the one you posted.

Have you checked pin14 for 50MHz? It will give us a hint where to focus on.

FvM said:

It's essential to have a certain amount of bypass capacitance for each node. In case of the 3.3V supply, there will be probably more capacitors at other point of your circuit, having e.g. 100 nF at each supply pin should be sufficient then. Proper operation of the internal 1.2V supply depends on sufficient VDDCR bypass capacitance.

Regarding 1 Mohm resistance, I found that it's not populated in the reference design, so it's apparently not required. But apparently doesn't hurt because I have in in a number of designs.

Click to expand...

1Mohm resistance is populated in my reference design (STM32F4 Discovery Base Board) in fact, but I guess it is not that mandatory, howevery I will put it in my 2nd version design just in case.

My VDDCR pin has the right connection now and the internal 1.2V regulator seems to be working properly.

- - - Updated - - -

rho-bot said:

Yes the Caps on VDDCR are mandatory, its the supply for the core logic :!:
You already mentioned the ripple you observed, which is 120mV and at spec limits (1.14V .. 1.26V)...

The 1Meg is not that critical. Needed to support start of oscillation, but depends on the crystal you use.
Please show us your current schematic, as it seems there are diffences to the one you posted.

Have you checked pin14 for 50MHz? It will give us a hint where to focus on.

Click to expand...

Here is the new design:




I also check all the pin voltages of reference discovery board and my design, for the signal pins, this gives the mean voltage.

Here is the comparison between the two:

No Name Disco(V) Design(V)
1 VDD2A 2,90 3,01
2 LED2/nINTSEL 0,00 0,41
3 LED1/REGOFF 2,90 0,00
4 XTAL2 0,65 0,67
5 XTAL1/CLKIN 0,62 0,66
6 VDDCR 1,22 1,21
7 RXD1/MODE1 0,00 2,82
8 RXD0/MODE0 0,00 2,82
9 VDDIO 2,90 3,01
10 RXER/PHYAD0 0,00 0,00
11 CRS_DV/MODE2 0,00 2,82
12 MDIO 2,90 3,01
13 MDC 0,00 0,00
14 NINT/REFCLKO 1,43 0,00
15 nRST 2,90 3,01
16 TXEN 0,00 0,00
17 TXD0 0,00 0,00
18 TXD1 0,00 0,00
19 VDD1A 2,90 3,01
20 TXN 2,90 3,01
21 TXP 2,90 3,01
22 RXN 2,90 3,01
23 RXP 2,90 3,01
24 RBIAS 0,00 1,20

There is no oscillation on the crystal.

Good so far. Question is if the 25 MHz oscillator and refclock output are working as well.

Consider that probing the crystal oscillator nodes with a passive oscilloscope probe might stop oscillations. Most likely at oscin, possibly at oscout node too. If you don't have a high impedance active probe, a 1k - 10k series resistor for the probe might help to see the oscillations. Refout can be probed without problems.

FvM said:

Good so far. Question is if the 25 MHz oscillator and refclock output are working as well.

Consider that probing the crystal oscillator nodes with a passive oscilloscope probe might stop oscillations. Most likely at oscin, possibly at oscout node too. If you don't have a high impedance active probe, a 1k - 10k series resistor for the probe might help to see the oscillations. Refout can be probed without problems.

Click to expand...

It seems refout has no signal. Doesn't that mean the IC isn't able to boot? In discovery boards on the other hand, the crystal has 25MHz oscillation, that I measure with the same oscilloscope.

Here are the new updates:

- connected a 1Mohm resistor between the two ends of the crystal.
- fixed a wrong capacitor connection on VDDCR pin (I had 1nF//1nF instead of 1nF//1uF).
- applied a 25MHz signal from a function generator and see if it works (it didn't work).
- connected only ground and Vcc signal of discovery base board and see that the ethernet chip's crystal has an oscillation without any signal from the MCU.

After all these, there is no oscillation on the crystal and the ethernet chip still doesn't work.

see that the ethernet chip's crystal has an oscillation without any signal from the MCU

Click to expand...

Yes, that's also my observation. LAN8720 crystal is even oscillating with reset input hold low.

There must something basically wrong with your board.