LPC2388 & KSZ8721 10Mbit problem

[hasiptuna]

Hello All,

I use LPC2388 and KSZ8721 ethernet with Olimex schematic. 100Mbit mode is working fine. But 10 Mbit mode  is not working. I use uip 1.0.    Missing the last two or three-pack is coming.


Thank you for your interest

[mark]

Hi

If you are using uIP you may find support at http://www.sics.se/~adam/uip/index.php/Main_Page

Regards

Mark

[hasiptuna]

Hello Mark,

I know that site.But This problem is not from software. MAC or KSZ8721 ,It has a problem one of them. If the  software is problematic, 100Mbit Full Duplex mode is not working fine.  Does it work with 10MBit mode at Olimex kit?

Regards

[mhoneywill]

Hi

You could always try this mailing list, http://www.sics.se/~adam/uip/index.php/Mailing_list

The uTasker TCPIP stack is NOT UIP it is much more powerful, Also remember UIP is free where as uTasker requires a license to be purchased.

Are you asking if uTasker works on the Olimex board at 10M?

Regards

Martin

[hasiptuna]

I am so sorry.I didnt know. is Utasker work 10 Mbit at Olimex?
Dribble Nibble status bit is set at 10Mbit mode.Is it problem?

Regards

[mark]

Hi

I haven't heard of problems with 10M on the Olimex board.

As comparison, there is a binary which runs om this board here: http://www.utasker.com/SW_Demos.html
The LAN setting can be set to 10M (rather than auto-neg) to see whether it performs correctly. (Guide here: http://www.utasker.com/docs/LPC23XX/uTaskerV1.3_LPC23XX.PDF)

Note also that, if you are using a 100M hub (probably very rare however nowadays), 10M setting would not be able to work.

Regards

Mark

[hasiptuna]

Hello All,

I found problem.Problem is at Olimex kit schematic. 22 number pin of KS8721 is wrong connection. This pin is E_CRS.This pin have to connect 9 number of KS8721.

Thank you Mark for your interest.

[mark]

Hi

On my Olimex LPC-2378-STK board pin 22 has been broken off the KSZ8721. I wondered why that was so...
But I don't see any connection to pin 9.

In case of any difficulties this may be an important tip.

Regards

Mark

[david_f]

Hi Mark,

I wondered if you have few minutes to check if theses 2 pins are not connected together on your Olimex board. On mine, there is a small wire below the chip that I found out when I soldered the chip out. I don't understand why we need to connect these 2 pins together, the datasheet provides no explanation.

Many thanks in advance,

David

[mark]

Hi David

Unfortunately I am on the road for another weeks and didn't pack in the Olimex board (I just have my IAR one).
I am surprised that the Olimex diagram doesn't seem to correspond (http://www.olimex.com/dev/images/LPC2378-STK-sch.gif). This still shows the connection to pin 22.
I wonder whether the trick was to connect pads 22 and 9 underneath the chip and remove pin 22?

Can anyone else with the board take a very close look at the pin/pad to see whether there is a secret wire underneath there? No idea why though...

David, I wonder whether asking Olimex might result in an explanation?

Regards

Mark

[mark]

Hi David

I have (finally) checked my board using a magnifying glass but can't see any extra wires connected.
This is not 100% because I did it without removing the device, but I looked hard and really didn't get the impression that there was something additionally soldered under the chip.

Regards

Mark

[david_f]

Hi Mark,

I've seen your message in the Sparkfun forum today, thank you very much your help !

Yes, I saw nothing as well on my board until I decided to remove the part. The best way would be to check with a simple multimeter (when I removed the part, I didn't know where to search). I've received no reply from Olimex unfortunately. I contacted a Micrel FAE (here in France) last week, I'll transmit the schematics. He told me that this part (KS8721) is very mature and he should send me some designs quickly.

I'll ket you know about this strange issue,

Thanks again for your help.

Best Regards
David

[mark]

Hi David

I have measured the resistance between the pad 22 and the pin 9. It is indeed  0 Ohms!

This suggests that there is also a connection underneath the chip on my board as well.

Regards

Mark

[david_f]

Hi Mark !

Ah ah, it is interesting ! This means the board and the schematics are differents. I've some problems to receive packets correctly on my design (new PCB) which is based on the schematics (LPC2378-STK). The Micrel PHY is a KSZ8721BL set up in RMII mode as well. To get some boards working, I have to connect a strap between pin 9 & 22. If I remove it, I reveive no packet at all. However, some boards doesn't work even if I connect these pins together : an error from the status descriptors for each packet received (bad length or alignement error or CRC error). When I read the buffer, I have 12/13 first-bytes of a packet correctly (example for an ARP packet : 6 bytes broadcast (0xFF) followed by 6 bytes of my source MAC address (my computer) and then... dummy bytes...). No problem to send packets though, checked with ethereal.

I received an email from Micrel yesterday :

We think the problem may be due to the strap between pins 9 & 22.

For RMII mode (see datasheet) the E_CRS signal (on customer schematic)
should connect to pin 9 RXDV/CRSDV and not pin 22 CRS. This needs to be
corrected as we believe this is the cause of the receive errors
(misalignment).
 
We suspect in the modification that you strapped pin 9 and pin 22
together but also cut pin 22. Pins 9 & 22 are both outputs so you do not
want them connected together (you may be lucky in RMII but certainly we
would not advise connecting the two pins).

Well OK, but it doesn't work... why ? The schematics provided by Micrel attached to the email are all designed for MII (pin 22 and pin 9 are used separetly), not RMII...

To conclude : there is something wrong either in the datasheet or with this chip in RMII mode, or with my PCB design... no reply from Olimex too... I made a small test code to check EMAC. Does someone write another one for LPC2388/78 + Micrel PHY ?

Thanks again Mark,

Best Degards,
David

[david_f]

Hello !

Ok, I finally found something : I forced the pins TXD2, TXD3 and TXEN to the ground, put 2 pull-down 1K resistors (pin 11 = RX_ER and pin 9 = E_CRSDV) and disconnected pins 22 = E_CRS and pin 9 = E_CRSDV, leaving pin 22 floating. I also use a GPIO pin from the LPC to make a hard-reset (pin 48 = PHY_RST). In LPC side, I setup EMAC pins in the PINSEL register first and then apply a hard reset. I think the pull-down resistors helps to setup the straping options properly, I don't know...

Finally, my boards seems to work with these modifications 

However, I've no explanation about the Olimex modifications... Mark, did you have any issue in Half duplex mode with your Olimex board ? The misalignement can be explained by this.

Thanks again !

Best Regards,
David