W5500 stuck on socket close

Question

Hello, I have a W5500 chip which is stuck after writing Sn_CR_CLOSE to CR, reading back CR always returns 0x10 (=Sn_CR_CLOSE), it never returns 0 as the operation is never completed. This happens on all the 8 sockets. The configuration of the W5500 is the default after a reset. The hardware seems all ok and this same board was working a week ago. The test code is like this: spi_init(); set_w5500_spi_functions(); uint8_t r = getVERSIONR(); assert(r == 0x04); // ok for (uint8_t sn = 0; sn <=7; sn++) { r = getSn_CR(sn); assert(r == 0); // ok } for (uint8_t sn = 0; sn <=7; sn++) { setSn_CR(sn, Sn_CR_CLOSE); for (int32_t i = 0; i <= 1000; i++) { r = getSn_CR(sn); if (r == 0) break; // it never breaks r = getSn_SR(sn); // always returns 0 (=SOCK_CLOSED) } } I have 2 other boards which do not have this problem, after writing CR=0x10 they return CR=0 at the first reading. What happened on this chip?

Eugeny · Answer

Never tried to close socket which is already in closed state. Second suspicion is you writing and reading register data from wrong location (e.g. problems in SPI communication).

Eugeny · Answer

Do other registers read properly? Make a reg dump to see if all the values are correct default values.

gama · Answer

Yes, I’ve checked most of the global and sockets registers (excluding ip addresses, ports and tx/rx pointers). They are at the default value: most of them are zero, but I can correctly read the ones which are not zero (RCR=08h, PHYCFGR=B8h, Sn_TTL=80h, Sn_RX/TXBUF_SIZE=2 …), the only difference is Sn_IMR which is 1Fh instead of FFh, but it is ok the missing bits are reserved.

Eugeny · Answer

Does link go up when you insert network cable connected at the other end?

gama · Answer

The link status is indifferent for this problem: on the working boards it works with or without link, on the stuck board it never works. But on the stuck board the link is never up with the cable connected, also the link led is always off. Hardware-wise everything else seems ok.

Eugeny · Answer

gama: The link status is indifferent for this problem If I would not think there’s a difference, I would not ask. gama: But on the stuck board the link is never up with the cable connected, also the link led is always off. Hardware-wise everything else seems ok. Not ok. Even with unconfigured W5500 (e.g. just after power up) link must be up when live cable is connected. You must check if circuitry is properly powered at the PHY side, and if there’s any connection/PCB issue.

gama · Answer

As I tried to explain in that line, with “indifferent” I meant that this problem doesn’t change with the link on or off, and since without link I have a chip working and one not, I can probably assume that the link detention not working is not a cause but another result of the faulty chip. To recap, a week ago I had 3 boards working without problems, now one board is not working. Without link I can compare the boards pin to pin, the only differences are: EXRES1 pin is 0V instead of the correct 1.2V and on the TX couple I cannot see the “idle frames”, on the working one I can see them repeated every 20ms. I’m quite sure that the problem is the W5500 chip. I’m trying to understand what happened to this faulty chip, its situation is: CR doesn’t reset itself for any command, consumption is 20mA less than normal, but the SPI communication works correctly (and the registers read/written are correct). It seems like the W5500’s SPI peripheral still works and can read/write the chip’s RAM, but the microprocessor in charge of everything else doesn’t work anymore. Does anyone know in what case CR cannot reset itself? Fried microprocessor? Bad cloned chip?

Eugeny · Answer

gama: Does anyone know in what case CR cannot reset itself? Fried microprocessor? Does it matter? If your goal is to solve the problem with non-working module, replace chip, but only after you ensure problem is not external to the chip. gama: I’m trying to understand what happened to this faulty chip It is not proven yet. EXTRES1 relates to analog circuit, 0V level may mean you have problem with 3V3A. Did you check it as I asked in my previous reply? gama: Bad cloned chip? Given the price of the chip I do not think there will be clones or unauthorized copies.

gama · Answer

Eugeny, I don’t know what you trying to do, but nevertheless thanks for trying to help, now it’s ok. Is there anyone from @wizteam who knows in what case CR cannot reset itself? And possibly the cause of this situation, so I can try to avoid it.

lawrence · Answer

Hi @gama Thank you for the test with W5500. It is so interesting ethernet function test. Actually, In Sn_CR register, 0x10 means SOCK_CLOSED. This is defined in ioLibrary, as #define Sn_CR_CLOSE 0x10 It is so good way to use W5500 with own made driver. But if there is some problem, it could be helpful to refer the official ioLibrary: github.com Wiznet/ioLibrary_Driver/blob/master/Ethernet/W5500/w5500.h //***************************************************************************** // //! \file w5500.h //! \brief W5500 HAL Header File. //! \version 1.0.0 //! \date 2013/10/21 //! \par Revision history //! <2015/02/05> Notice //! The version history is not updated after this point. //! Download the latest version directly from GitHub. Please visit the our GitHub repository for ioLibrary. //! >> https://github.com/Wiznet/ioLibrary_Driver //! <2013/10/21> 1st Release //! \author MidnightCow //! \copyright //! //! Copyright (c) 2013, WIZnet Co., LTD. //! All rights reserved. //! //! Redistribution and use in source and binary forms, with or without //! modification, are permitted provided that the following conditions This file has been truncated. show original Thank you

gama · Answer

Hello @lawrence, thank you for your reply. Yes, I’m using the same file ‘w5500.h’. This is my problem: I write the CR registry with Sn_CR_CLOSE = 0x10: setSn_CR(sn, Sn_CR_CLOSE); but when I read back the CR registry: uint8_t r = getSn_CR(sn); the value read is always 0x10 and never 0x00 (CR should reset itself to 0x00) no matter how much I wait. I’m really sure that I have a faulty/damaged W5500, also: it consumes 20mA less than a working W5500, on the TX couple lines I cannot see the “presence/idle frames”, the EXRES1 pin is at 0V instead of the correct 1.2V (but VBG is at 1.2V). Other signals/powers/clocks of the W5500 chip are all correct. In what situation the CR register cannot reset itself to 0x00 once written? And possibly, do you know what situation could have damaged the chip? Thank you.

lawrence · Answer

@gama Sorry for the late, Could you share all register’s reset value in 2 cases? one is after spi init and second is after socket open. We can figure out damaged chip or not with the log. I will also try to check some cases. Thank you,

lawrence · Answer

@gama Could we know the init value of Sn_CR register before write any command? If it is not 0x00 in init, as you said we consider any damaged case.

gama · Answer

Lawrence, unfortunately I was forced to replace that W5500 chip with a new one, with the new one the board worked correclty. Now I’m trying to manually mount the problematic W5500 on a old board to give you a full register dump, please be patient. However, I had already written down a partial dump of the registers of that W5500, it was done just after the reset and without the ethernet link: Common registers: getVERSIONR = 0x04 getPHYCFGR = 0xB8 getMR = 0 getIR = 0 getIMR = 0 getSIR = 0 getSIMR = 0 getRCR = 0x08 getPMAGIC = 0 Socket register (n = 0…7): getSn_MR = 0 getSn_CR = 0 getSn_IR = 0 getSn_SR = 0 getSn_TOS = 0 getSn_TTL = 0x80 getSn_RXBUF_SIZE = 0x02 getSn_TXBUF_SIZE = 0x02 getSn_IMR = 0x1F getSn_KPALVTR = 0 So to reply your question, getSn_CR was zero before writing any command. After this, if I write CR: setSn_CR(sn, Sn_CR_CLOSE) I get: getSn_CR(sn) = 0x10 getSn_SR(sn) = 0 this is true for any socket (sn = 0…7). Thank you for your help.

gama · Answer

@lawrence, I mounted the W5500 on an a different board, this board was working with a different W5500. Now the board doesn’t work anymore, the W5500 has the same behavior, it cannot reset CR. This makes me think this W5500 is faulty. For at least 2 days before the problem, the chip was working, it was connected to the ethernet cable and it was powered at 3.3V. The 3.3V rail is from a step-down converter which is a bit noisy, it has a ripple of 200-300mV, but on the same 3.3V rail there is also the microprocessor, and after the problem the microprocessor still worked without problems. Attached the full dump before and after writing CR to 0x10.

gama · Answer

Seems I cannot upload files, here is the dump: Before writing CR: — Common registers — getMR: 00 getGAR: 00 00 00 00 getSUBR: 00 00 00 00 getSHAR: 00 00 00 00 00 00 getSIPR: 00 00 00 00 getINTLEVEL: 00 00 getIR: 00 getIMR: 00 getSIR: 00 getSIMR: 00 getRTR: d0 07 getRCR: 08 getPTIMER: 28 getPMAGIC: 00 getPHAR: 00 00 00 00 00 00 getPSID: 00 00 getPMRU: ff ff getUIPR: 00 00 00 00 getUPORTR: 00 00 getPHYCFGR: b8 getVERSIONR: 04 — Socket 0 registers — getSn_MR: 00 getSn_CR: 00 getSn_IR: 00 getSn_SR: 00 getSn_PORT: 00 00 getSn_DHAR: ff ff ff ff ff ff getSn_DIPR: 00 00 00 00 getSn_DPORT: 00 00 getSn_MSSR: 00 00 getSn_TOS: 00 getSn_TTL: 80 getSn_RXBUF_SIZE: 02 getSn_TXBUF_SIZE: 02 getSn_TX_FSR: 00 00 getSn_TX_RD: 00 00 getSn_TX_WR: 00 00 getSn_RX_RSR: 00 00 getSn_RX_RD: 00 00 getSn_RX_WR: 00 00 getSn_IMR: 1f getSn_FRAG: 00 40 getSn_KPALVTR: 00 — Socket 1 registers — getSn_MR: 00 getSn_CR: 00 getSn_IR: 00 getSn_SR: 00 getSn_PORT: 00 00 getSn_DHAR: ff ff ff ff ff ff getSn_DIPR: 00 00 00 00 getSn_DPORT: 00 00 getSn_MSSR: 00 00 getSn_TOS: 00 getSn_TTL: 80 getSn_RXBUF_SIZE: 02 getSn_TXBUF_SIZE: 02 getSn_TX_FSR: 00 00 getSn_TX_RD: 00 00 getSn_TX_WR: 00 00 getSn_RX_RSR: 00 00 getSn_RX_RD: 00 00 getSn_RX_WR: 00 00 getSn_IMR: 1f getSn_FRAG: 00 40 getSn_KPALVTR: 00 — Socket 2 registers — getSn_MR: 00 getSn_CR: 00 getSn_IR: 00 getSn_SR: 00 getSn_PORT: 00 00 getSn_DHAR: ff ff ff ff ff ff getSn_DIPR: 00 00 00 00 getSn_DPORT: 00 00 getSn_MSSR: 00 00 getSn_TOS: 00 getSn_TTL: 80 getSn_RXBUF_SIZE: 02 getSn_TXBUF_SIZE: 02 getSn_TX_FSR: 00 00 getSn_TX_RD: 00 00 getSn_TX_WR: 00 00 getSn_RX_RSR: 00 00 getSn_RX_RD: 00 00 getSn_RX_WR: 00 00 getSn_IMR: 1f getSn_FRAG: 00 40 getSn_KPALVTR: 00 — Socket 3 registers — getSn_MR: 00 getSn_CR: 00 getSn_IR: 00 getSn_SR: 00 getSn_PORT: 00 00 getSn_DHAR: ff ff ff ff ff ff getSn_DIPR: 00 00 00 00 getSn_DPORT: 00 00 getSn_MSSR: 00 00 getSn_TOS: 00 getSn_TTL: 80 getSn_RXBUF_SIZE: 02 getSn_TXBUF_SIZE: 02 getSn_TX_FSR: 00 00 getSn_TX_RD: 00 00 getSn_TX_WR: 00…

irina · Answer

Hi~ @gama The clocks used to perform register write operations and register erase operations are different. When w5500 operates register read/write, it uses the SPI clock connected to the master. the w5500 uses the clock of phy to operate the register clear operation. Therefore, if the PHY does not operate normally as @Eugeny said, the register clear operation is not performed.

gama · Answer

Hello @irina_kim, thanks, finally I’ve an answer. So the PHY is not working, can I measure something or do something to investigate the cause of this problem? I need to prevent this kind of problem, and of course, I can have inadvertently damaged the chip myself, but I’m trying to understand how. Sorry if I’m too persistent.

lawrence · Answer

Sorry for the late answer. So do you mean it is not working on Ethernet Linking status? Unfortunately, there is no way to check the chip condition without PHY Link on. Could I know why you need to check chip status before PHY link? And if you share your h/w schematic, we can also check if there is any issues or not. Thank you

gama · Answer

Hello @lawrence, I’ve a faulty W5500 which cannot establish the Ethernet link, as Irina pointed out probably the PHY is not working correctly, but the chip is not entirely fried because the SPI works, the register RAM works and the 25 MHz clock is oscillating correctly. I’m trying to understand what happened to this chip to prevent this kind of problem. Schematic attached. schematic1261×678 16.4 KB Thank you.

Eugeny · Answer

gama: Eugeny, I don’t know what you trying to do, but nevertheless thanks for trying to help, now it’s ok. I have one question and an opinion, by not sure if it is appropriate to say anything.

gama · Answer

Eugeny, seriously, no hard feelings, ask away. I was just that me repeating the same things and you dismissing my questions was a waste of your and my time, and I had little time. Water under the bridge, I hope you understand.

Eugeny · Answer

@gama sure, also no problem from my side. I was waiting for answers for questions I was asking, you were asking more questions without answers. So here’re questions (I broke one big question up into several): does this failure occur after you power off device, and then power on, or may appear spontaneously during operation? As I see you initially said that you close idle sockets just after power up/during initialization, you noticed the problem after power on, right? Or you identified problem after power on, but issue occur during normal operation? what is powering the W5500 - is circuit available? Is it battery? what is the rise time for the power during power on? What is the slope of the power voltage rise? would be great if you have scope and sample the waveform for the voltage on the VDDA rail several times to see the actual power rise cycle (and at best power fall cycle during power off). But you must measure these on the normal W5500, not defective one (as defective may have different active and reactive resistance and impact the readings). And my opinion is that problem should not relate to network cable, it must be somewhere on the board, or outside the board, and the first candidate is power. While you can see filtering circuit on the reference schematic and you have just copied it (however seems ferrite bead is 181, not 121), the VDDA power rise profile heavily depends on the power rise profile at the entry to the filter. If you happen to have battery, closing contacts (powering the filter) will cause very fast voltage rise, and it may cause heavy overshoot after the filter. Digital circuits may have different sensitivity to these overshoots and survive them. If you are interested in theory and simulation, then use LTSpice and see how different power rise profiles affect the filter output voltage shape. Edit: the first question which actually must be asked - seems you have your own board, how do you solder W5500 chips?

gama · Answer

@eugeny: the problem happened only on one chip and I don’t know when it happened, I only realized there was a problem after a power up, but before the board was powered without ethernet cable for 2 o 3 days straight, the problem could have happened then and I didn’t realize the 5500 is powered by a step-down converter from 12-24 VDC, the converter is a MC34063 with a standard circuit, it is a bit noisy but nothing alarming, the ripple is 200mVpp max at 24V input the max slope is about 1V/10ms on VDD and AVDD the slope is quite consistent on multiple power-ups, it is really soft without any overshoot, I measured it on a working W5500 the chip was originally soldered in a oven, after the problem I removed and re-soldered it manually I agree that the power could be the cause, probably I’ve fried it myself somehow. However I find strange that something on the chip still works, I would expect a short-circuit on one of the power rails, also on the same 3.3V there is a STM32F103 which was not damaged, and I’ve found the STM32F to be quite weak. Anyway, now I have 10 boards working for two weeks without problems, so I tend to think it was a sporadic incident.

Eugeny · Answer

So this research did not give us definite answer to what have happened… The WIZnet chips are very sensitive to soldering temperature, when soldering using reflow please ensure you fully comply with the profile in the chapter 6 of the datasheet, and when soldering with iron ensure you set temperature below 325 C (I solder at 315). Soldering at 350 may damage the chip and soldering at 375 will kill the chip’s functionality in weird way.