DP83822IF: 100BASE-FX troubleshooting

Hi Fiorella,

Thank you for the detailed query.

Is auto negotiation disabled with 100M speed forced on DP83822? (0x0[13:12] = '10')

This is typically required for link up in fiber. If setting this and applying soft reset does not help link up, please share the PHY register dump and schematic for review.

Thank you,

Evan

Hi Evan,

Thanks for the quick reply. I tested again disabling autonegotiation and forcing 100M, but the link is still down. About the signals on the differential lines, does what I described match the expected behavior?

I won't have access to the schematic until Monday due to server issues, but fiber part is basically the same as in the EVM. In the meantime, here is the register dump:

0 : 2100

1 : 7849

2 : 2000

3 : A240

4 : 1E1

5 : 0

6 : 4

7 : 2001

8 : 0

9 : 0

A : 4100

B : 1000

C : 0

D : 0

E : 0

F : 0

10 : 4004

11 : 108

12 : 0

13 : 800

14 : 0

15 : 0

16 : 100

17 : 61

18 : 400

19 : 4001

1A : 0

1B : 7D

1C : 5EE

1D : 0

1E : 2

1F : 0

25 : 240

26 : 0

27 : 0

3E : 0

3F : B4FF

40 : C11D

42 : 0  

101 : 2002

106 : B0BB

107 : 605

10F : 300

111 : 6003

114 : 400A

116 : 14A

121 : 199A

122 : 1027

123 : 51C

126 : 461B

129 : F

130 : 4750

155 : 1

170 : E52

171 : C85C

173 : FF1E

177 : 189B

180 : 0

181 : 0

182 : 0

183 : 0

184 : 0

185 : 0

186 : 0

187 : 0

188 : 0

189 : 0

18A : 0

215 : 1AF

21D : 600

403 : 9FCF

404 : 20

40D : 8

410 : 2000

416 : 870

418 : 0

41F : 0

421 : 5

428 : 0

450 : F41

456 : 8

460 : 551

461 : 410

462 : 301

463 : 0

465 : FF00

467 : B6F

468 : 0

469 : 40

4A0 : 1000

4A1 : 0

4A2 : 0

4A3 : 0

4A4 : 0

4A5 : 0

4A6 : 0

4A7 : 0

4A8 : 0

4A9 : 0

4AA : 0

4AB : 0

4AC : 0

4AD : 0

4AE : 0

4AF : 0

4B0 : 0

4B1 : 0

4B2 : 0

4B3 : 0

4B4 : 0

4B5 : 0

4B6 : 0

4B7 : 0

4B8 : 0

4B9 : 0

4BA : 0

4BB : 0

4BC : 0

4BD : 0

4BE : 0

4BF : 0

4C0 : 0

4C1 : 0

4C2 : 0

4C3 : 0

4C4 : 0

4C5 : 0

4C6 : 0

4C7 : 0

4C8 : 0

4C9 : 0

4CA : 0

4CB : 0

4CC : C

4D0 : 302

4D1 : 18B

4D4 : 7220

4D5 : FBC1

4D6 : 1C1

3000 : 2100

3001 : 7849

3014 : 0

703C : 0

Thank you,

Fiorella

Hi Fiorella,

Thanks for confirming and sharing the register dump. The behavior shared in the waveform is expected, but TD+/- not transmitting is a symptom of link down.

The register configuration appears valid for Fiber 100M. If applying a soft reset to the PHY through registers does not bring link up, the issue is likely with the schematic or layout. I will review when shared.

Thank you,

Evan 

Hi Evan,

Thanks for confirming registers and RD signal are ok. Here are the schematics, what is marked has been taken out (red because of design mistakes, blue for fiber testing). Led GPIO is floating.

Thank you for your assistance

Hi Evan,

A short update on tests: I just realized that somewhere on the code there was an initialization function which forced MDIX by default, with register 0x19 ... so actually it was the TD+- pair that had the active signal, and RD+- is 0V (small noise around 100mV). I've disabled it and I'm using a clean code just to read the registers now, but I still can't get a link up from phy or switch.

Another thing that noticed is that sometimes after a power up, SD_EN would show a square wave instead of constant 3V, about 20% duty cycle. This has happened only twice almost randomly and I'm still figuring out the conditions to reproduce it. Do you have any idea why this happens? Could it be related to link issues?

Thanks,

Fiorella

Hi Fiorella,

Thank you for sharing the schematic and testing updates.

I will need more time to review, please expect feedback tomorrow.

Thank you,

Evan

Hi Fiorella,

After depopulating LVPECL termination, the schematic looks good for Fiber. Expected Fiber termination is:
50ohm PU on TD/RD +/- -> cap coupling to SFP Module

Signal detect behavior may be related to noise seen on RD+/-, this is unclear without scope capture.

Do you have access to the 822EVM? I'd like to remove SFP module & link partner compatibility as possible causes, if these can be tested with 822EVM.

Thank you,

Evan

Hi Evan,

Thank you for checking the schematic

Unfortunately, I don't have the 822EVM.

Here are some notes about last tests:

• Signal Detect only shows that behavior when connected to a specific SFP transceiver (Flexoptix S.1303.2M.IG). Even if it is just the SFP alone, without fiber cable
• When other transceiver or no transceiver is connected, there is a smaller amplitude noise, and SD stays high

Here are some captures of RD+- noise with Flexoptix SFP

SD signal on yellow and RD- on blue

RD+- noise with other transceivers

I would appreciate your comments and suggestions.

Thanks,

Fiorella

Hi Fiorella,

Are you able to test with different SFP modules or another fiber link partner?

We have validated these SFP modules on 822EVM with the same termination noted above:

- AFBR-5715PZ
- AXFE-1314

Could you also please confirm:

• Is the same SFP module being used on both sides of link?
• In one case you noted SD_EN would show a 20% square wave - have you been able to reproduce conditions for this? If this case is not observed, is SD_EN always low?

Thank you,

Evan

Hi Evan,

Apologies if my previous responses led to confusion. 

Yes, in all tests the SFP modules are the same at both ends of fiber cable.

I did reproduce the conditions to 20% square wave in Signal Detect (Pin 24). The behavior occurred only when connecting the transceiver: Flexoptix S.1303.2M.IG. And the same signal remained with and without connecting a fiber cable or link partner. Just the empty SFP generated the wave as you can see on the images of my previous response. 

RD+- particular noise shown on the first 3 pictures above, also occurred only with Flexoptix SFP

When that is not the case, and 20% wave is not observed:

• SD pin is always high at 3V, due to pullup resistor (R71 in my schematic and R73 on EVM)
• RD+- smaller noise is also shown on last image
• This happens when there is no SFP connected and also when using other SFPs different than Flexoptix.
• Same behavior whether link partner is connected or not.

I'll request the SFPs you mentioned but probably will take some time to arrive and test.

I'd appreciate your comments on this behavior, and if there is anything we can conclude or test to better understand the problem until the new parts arrive.

Thank you,

Fiorella

Hi Fiorella,

Based on the current signal and register results, my conclusion is that the PHY is configured is correctly for fiber, but there is some hardware issue with the board, SFP module, or link partner causing RD to be significantly attenuated.

When probing TD+/- of the link partner, do you see signal with similar amplitude as TD+/- from 822? Is it possible to probe along the path for this signal to isolate where the attenuation is occurring?

Thank you,

Evan

Hi Evan, 

Thanks for your comments and suggestions, that makes sense for me too. I have not been able to probe the signal on the link partner yet. The project involving this PHY has been deprioritized so it will take longer for me to continue testing. I'll keep you updated once I can resume it.

Thanks again for your support,

Fiorella

Hi Fiorella,

Thanks for confirming. I have the fiber setup with 822EVM ready on my side for when you are able to continue on this debug.

Thank you,

Evan