W5100 UDP pings but doesn't send data

Question

Hello! I’m trying to send two bytes of data from w5100 chip to PC using UDP protocol. Seems that I have done everything by the datasheet algorithm: Did software reset Set gateway IP address to 192.168.11.3; Set subnet mask to 255.255.255.0 Set MAC address of w5100; Set IP address to 192.168.11.4; Wrote S0_MR = 0x02 (UDP mode); S0_Port = 0x07D0 (2000); S0_CR = 0x01 (open) Checked S0_SR, it’s SOCK_UDP, so init is OK —Sending: Checked S0_TX_FSR, it’s 0x800; Set S0_DIPR = 192.168.11.198 (my PC), S0_DPORT = 0x0FA0 (4000); Checked TMSR (TX mem size), it’s 0x55; Read S0_TX_WR, it’s zero; Wrote two bytes 0xff, 0xff into start_address, which equals 0x4000 (as gS0_TX_Base) ; Updated S0_TX_WR value, now it’s 2, wrote is back; Set SEND cmd : S0_CR = 0x20; Checked S0_CR, it’s zero, so sending seems to be OK. Chip pings ok, but data’s not received by PC, monitoring by WireShark. Couple of times got the data though, but only two times out of about 30 launches. What am I doing wrong? Help please

Eugeny · Answer

Bob_Black: Set MAC address of w5100 What is MAC address exactly? You shown values for everything except this information. Bob_Black: Couple of times got the data though, but only two times out of about 30 launches Of these “couple of times”, did you have correct data arrived? With only two bytes FF FF in packet data field? What you can do now connect your PC directly to the W5100-based device and try again looking what it shows in Wireshark not setting any filters in there; share your code driving W5100 with us.

Eugeny · Answer

Bob_Black: my code is on VHDL I am not sure I am going to read and trace it. Why not use case statement for state machine? What you currently have is totally non-readable. I bet you simply mess state of the machine during operation. Where’s waiting for MR’s bit 7 to become 0? What is the purpose of cnt?

Bob Black · Answer

I agree that the code is difficult to read, but on time diagrams it’s more obvious, that it writes into registers fine, and reads what is expected. Eugeny: Where’s waiting for MR’s bit 7 to become 0? I check MR right after reset, and it’s zero. No need to wait. Besides, there is expected value in status register after initialization, so the problem at least is not before that. That’s all you can help me with?

Eugeny · Answer

Make a dump of states machine passes through. Thus should be the easiest thing to do on your code. After dump you look into the states, and ensure that the sequence is correct.

Bob Black · Answer

Thank you, problem is solved. Turned out a time delay was needed between chip initialization and sending procedure. For some reason chip couldn’t send data right after initialization