DP83869HM: Strap setting is not recognized, registers readback fail

adding pictures:

Schematics:
8228.sch.pdf

Hi Jurag,

I have started looking into this. At first look it looks like that your register reads are not working fine. For example, I see that register 0001 = 616D after power up but in datasheet I see that value of bits [15:12] can not be different from 0111. Lets clean up the register read/write operation first so that we can rely on it for checking status of phy. Let me know the following for this :

1. Are you using 25MHz clock from crystal or some other source? I see both in the schematic.

2. If not crystal, then is this clock active before the power of 869 is up?

3. What is the MDC clock frequency? Is if free running or does it stop after each register read?

Also on LED pins I see two options of LEDs. Which one was is finally populated?

Usb2mdio link : http://www.ti.com/tool/USB-2-MDIO

--

Regards,

Vikram

Thank you for your reply Vikram,

The register 0x01 reports 0x6169 after power up and in a second it changes to 0x616d - probably making a link when optical cable is inserted.

1. Yes, I am using 25MHz crystal directly connected to PHY IC.

2. Crystal is operating normally; I can measure the 25MHz waveform.

3. MDC signal shows clock bursts about every second, or more often if I am configuring anything through CLI or in windows. MDC frequency is 2.5MHz. You can find scope screenshots below. This is probably some periodic link-check.

I exported better schematics with variant assembly. You can see exactly what is populated at my current sample. I removed LEDs just in case they would interfere with STRAP setting.
I am still not able to download USB2MDIO tool. It reports "Sorry, the page you’re looking for can’t be found." after I sign in and try to download by "request" button.

Scope screenshots: 

detail:

Do you have any ideas what else I could try? 
Your support is appreciated.

sch-variant-assembly-conpc008.pdf

Hi Juraj,

Here is what I can conclude so far :

1. Your straps are in correct mode by straps : Rgmii to Fiber.

a. Because Rgmii lines are toggling : both clock (125MHz) and data are toggling. Without Rgmii mode entry they will not toggle.

b. Fiber link is up with link-partner.

2. Register read writes has some issue :

a. 0x0001[15:11] should always be 0111.

b. When you are trying to write register to program delay , readback is showing incorrect value.

3. Rgmii data transfer to MAC is not successful because most likely you are not able to program the delay correctly (register write issue).

I guess you tried the link I sent over for downloading usb2mdio. Can you raise query for the same? I am not sure how to resolve the download issue but related team will be able to help you faster.

Also is there any other device connected on the same mdc-mdio bus? If yes, can we power down those devices and see if then controller can talk better with phy? (To rule out any clash)

--

Regards,

Vikram

Hello Vikram,

you are correct in all the points. That is the scenario I see. 

The register access for some registers does not operate properly. However some registers are acting normally - I  can turn on/off 25MHz clock pin output and measure - it is reacting to changes well.

I will create a new post related to USB2MDIO.

There is only LAN7801 IC interconnected with DP83869HM IC using mdc-mdio bus. They are interconnected directly through 0R resistors. Additionally MDIO has 1.5k pullup to 1V8 level (I tried 2k2 and 1k5).

I am also testing this on 2 boards and they are acting the same.

Are there any other things I could check?
Thanks,

Juraj

Hi Jurag,

Let me know if you were able to figure out usb2mdio for debug.

--

Regards,

Vikram

Hello Vikram,

I tried using USB2MDIO tool to read registers:
With extended register setting turned on:

Register 0000 is: 1140 
Register 0001 is: 616D 
Register 006E is: 0204 
Register 01DF is: 0041 
Register 0032 is: 00D0 
Register 0086 is: 0077 
Register 0C00 is: 1140 
Register 0C01 is: 616D 
Register 0000 is: 1140

and with the extended register setting off:

Register 0000 is: 1140 
Register 0001 is: 616D 
Register 006E is: 1140 
Register 01DF is: F000 
Register 0032 is: 2000 
Register 0086 is: 0007 
Register 0C00 is: 1140 
Register 0C01 is: 616D 
Register 0000 is: 1140

Some register values are different. I guess registers with address higher than 0xE need to be read with extended register setting on. 

Moreover, when I disconnected MDC MDIO connection between by MAC and PHY and set up RGMII delay (PHY enabled on TX and RX as default and turned off on MAC), it started working. I got data connection and working ping.

One more thing is bothering me in the design. The MAC and operating system cannot tell the link status of PHY because the MDC MDIO is disconnected and always assumes that cable is connected. 

If MDC and MDIO is connected, I can see cable connection status, but I have no data - original problem.

I connected a scope to MDC and MDIO and decoded the communication between MAC and PHY:

1. after power up MAC scans all the PHY addresses and tries to red 0x01 register until it finds PHY

2. reads reg 0x01 val 0x616D

3. reads reg 0x01 val 0x616D

4. reads 0x02  - 0x2000

5. reads 0x03 - 0xA0F1

6. reads 0x04 - 0x0020

7. writes 0x04 - 0x0AE0

8. reads 0x9 - 0x1300 

9. writes 0x9 - 0x0200, 1G FD, auto master/slave

10. writes:
0xD - 0x0003

0xE - 0x003C

0xD - 0x4003

0xE - 0x0060

11. writes to 0x0 - 0x1200 - autonegotiaion restart

I do understand some of these commands, but I do not know what changes after point 10, and value of point 6.

Thank you for checking this.
Juraj

Hi Juraj,

I am not sure whether I understood the present debug status correctly. Were you able to figure out :

1. Why MAC turning-off/on + MDC/MDIO disconnected is correcting the problem? Is it a MAC issue?

2. Do you need help to how to read/write extended register (with address greater than 1F) ? Here is an example to read lets say write reg<0x0123> of mmd 1F = 0xA0A0 :

000D 001F

000E <reg_address>

000D 401F

000E <value>

To read :

000D 001F

000E <reg_address>

000D 401F

000E

Let me know where you need my help.

--

Regards,

Vikram

Thank you for your reply Vikram,

I assume that driver in PC may configure the MAC certain way and then MAC will issue some commands to PHY that will configure the PHY a wrong way. In case the MAC and PHY MDIO interface is disconnected, the MAC does not have a chance to re-configure PHY. 

Would this be a plausible explanation?

Juraj