DP83867IR: DP83867 eye pattern test fail

Hi Shu-Cheng,

These minor failures are usually caused by board layout. Please follow this app note for layout guidelines: http://www.ti.com/lit/an/snla079d/snla079d.pdf

Regards,

Hung Nguyen

Hi Sirs,

Thanks for your reply.

Could you help to check our layout and schematic as below.

Schematic:

Layout:

Have any suggestion can share?

Because we have the same problem with two models

By the way, the place of MDI looks good!

Hi Shu-Cheng,

1. Are MDI pairs laid out as 100-Ohm differential?

2. Do you have compliance result for 100Base-TX on both pair A and B?

3. Can you remove the ESD protection device and redo the measurement for 1000Base-T?

4. For MDI connection, each center tap of the magnetics should be independently de-coupled to ground via a 0.1uF capacitor. See this app note http://www.ti.com/lit/an/snla246a/snla246a.pdf for more details.

Regards,

Hung Nguyen

Hi Sirs,

 Update as below

1. Are MDI pairs laid out as 100-Ohm differential?   YES  100-Ohm differential

2. Do you have compliance result for 100Base-TX on both pair A and B?   YES  , Please refer as below

100_Base-T.zip

3. Can you remove the ESD protection device and redo the measurement for 1000Base-T?　Yes, Still fail

4. For MDI connection, each center tap of the magnetics should be independently de-coupled to ground via a 0.1uF capacitor. See this app note http://www.ti.com/lit/an/snla246a/snla246a.pdf for more details. Yes,

Hi,

- Your schematic shows 0.25pF cap on each MDI pin. Make sure those caps are not populated.

- The center taps of 4 channels are connected together inside the MagJack. It is not recommended for this device. See app note http://www.ti.com/lit/an/snla246a/snla246a.pdf

Regards,

Hung Nguyen

Hi Sirs,,

Thanks for your reply.

1. The 0.25pF cap is reserved, not populated.
2. We'll discuss with MagJack vender to check this question.

Actually we would like to use EVM to verification, but we not sure EVM will pass this report.
So, does TI have this test report can share?

Thanks!!

Hi Sirs,

Update our test result.

1. The 0.25pF cap is reserved, not populated.

2. We used another MB to test and change the RJ45 Connector that was not connected inside the center tap.

The phenomenon is still same as before, the report as below

change conn_ESE0273082.zip

Any suggestion are welcome. Thanks!!

Hi Sirs,

Our comment as below, could you help check is correct or not?

mii write 0x05 0x001f 0x8000

mii write 0x05 0x0000 0x0140

mii write 0x05 0x0010 0x5008

mii write 0x05 0x0009 0x3b00

mii write 0x05 0x000d 0x001f

mii write 0x05 0x000e 0x0025

mii write 0x05 0x000d 0x401f

mii write 0x05 0x000e 0x0480

mii write 0x05 0x000d 0x001f

mii write 0x05 0x000e 0x01d5

mii write 0x05 0x000d 0x401f

mii write 0x05 0x000e 0xF508

Our fail items on report as below 

Hi Shu-Cheng,

It's interesting that when you change the MagJack, the failing channels changed from A,B to B and D. I'm wondering whether other 1000Base-T tests passed or failed.

1. Can you provide other 1000Base-T test results (including return loss) for all channels?

2. Can you remove the last 4 register writes to see if there is any difference?

mii write 0x05 0x001f 0x8000

mii write 0x05 0x0000 0x0140

mii write 0x05 0x0010 0x5008

mii write 0x05 0x0009 0x3b00

mii write 0x05 0x000d 0x001f

mii write 0x05 0x000e 0x0025

mii write 0x05 0x000d 0x401f

mii write 0x05 0x000e 0x0480

mii write 0x05 0x000d 0x001f     ====> REMOVE

mii write 0x05 0x000e 0x01d5     ====> REMOVE

mii write 0x05 0x000d 0x401f     ====> REMOVE

mii write 0x05 0x000e 0xF508     ====> REMOVE

3. If number 2. does not help, please try turning on the test mode on each channel separately:

mii write 0x05 0x001f 0x8000

mii write 0x05 0x0000 0x0140

mii write 0x05 0x0010 0x5008

mii write 0x05 0x0009 0x3b00

mii write 0x05 0x000d 0x001f

mii write 0x05 0x000e 0x0025

mii write 0x05 0x000d 0x401f

mii write 0x05 0x000e 0x0400     ====> 0x0400 = Channel A; 0x0420 = Channel B; 0x0440 = Channel C; 0x0460 = Channel D

mii write 0x05 0x000d 0x001f     ====> REMOVE

mii write 0x05 0x000e 0x01d5     ====> REMOVE

mii write 0x05 0x000d 0x401f     ====> REMOVE

mii write 0x05 0x000e 0xF508     ====> REMOVE

Regards,

Hung Nguyen

Hi Sirs,

Thanks for your reply.

First, does TI have test report on  DP83867IRRGZT ? because actually we didn't use TI Giga Phy before.

We not sure will it is IC limitation or not?

About your question, the reply as red word 

1. Can you provide other 1000Base-T test results (including return loss) for all channels?

We changed 2~3 pieces of test, it's still fail.

Fail doesn't necessarily have AB or BD. Each piece is different, but all of them will not pass.

2. Can you remove the last 4 register writes to see if there is any difference?

mii write 0x05 0x001f 0x8000

mii write 0x05 0x0000 0x0140

mii write 0x05 0x0010 0x5008

mii write 0x05 0x0009 0x3b00

mii write 0x05 0x000d 0x001f

mii write 0x05 0x000e 0x0025

mii write 0x05 0x000d 0x401f

mii write 0x05 0x000e 0x0480

mii write 0x05 0x000d 0x001f     ====> REMOVE

mii write 0x05 0x000e 0x01d5     ====> REMOVE

mii write 0x05 0x000d 0x401f     ====> REMOVE

mii write 0x05 0x000e 0xF508     ====> REMOVE

Removing these four lines will be abnormal. 

Also confirmed that sending four pairs separately does not affect the test results.

Hi Sirs,

By the way, 

Do you know what is 01D5=0XF508 function?
If REMOVE, the waveform will be smaller.

We would like to adjust the value to make it a little bigger.