DS90UB638-Q1: Unstable Lock/Pass

Barry W

Part Number: DS90UB638-Q1

Tool/software:

We are testing 5 prototype systems with DS90UB635 / DS90UB638 pairs.

4 of them are performing well. One reports unstable Lock and Pass.

We enabled CMLOUT per the data sheet instructions and the eye diagram looks very good to me. We are connecting in twisted pair configuration, over a short link, in 2Gbps (half rate) mode.

It does not appear that the 638 registers support the necessary functionality for the "Margin Analysis Program" testing.

Can you offer any additional debugging techniques we might use to understand why Lock and Pass fail so frequently on this particular system?

5 months ago

0 Hamzeh Jaradat 5 months ago

TI__Guru** 107810 points

Hi Barry,

Have you tested the failing DES board with a known working SER board?

Have you tested the failing SER board with a known working DES board?

How you tested with a different cable type and length?

Can you provide register dumps from SER and DES, from a working system and a not working system?

0 Barry W 5 months ago in reply to Hamzeh Jaradat

Intellectual 890 points

Thank you for these suggestions.

We will try them and respond with the register dumps.

0 Hamzeh Jaradat 5 months ago in reply to Barry W

TI__Guru** 107810 points

Waiting your feedback.

0 Barry W 5 months ago in reply to Hamzeh Jaradat

Intellectual 890 points

We are adding support to our Linux driver to read all the registers.

Hope to have it working soon.

0 Hamzeh Jaradat 5 months ago in reply to Barry W

TI__Guru** 107810 points

Sounds good. Thank you.

0 Barry W 5 months ago in reply to Hamzeh Jaradat

Intellectual 890 points

Here are register dumps from the 638 and 635 when they are connected.

I dumped the registers twice for each device to see which values might be changing.

Reg	638 -1	638-2
0x00	0x60	0x60
0x01	0x00	0x00
0x02	0x1E	0x1E
0x03	0x20	0x20
0x04	0xDF	0xDF
0x05	0x01	0x01
0x06	0x00	0x00
0x07	0xFE	0xFE
0x08	0x1C	0x1C
0x09	0x10	0x10
0x0A	0x7A	0x7A
0x0B	0x7A	0x7A
0x0C	0x81	0x81
0x0D	0xB9	0xB9
0x0E	0x08	0x08
0x0F	0x7F	0x7F
0x10	0x00	0x00
0x11	0x00	0x00
0x12	0x00	0x00
0x13	0x00	0x00
0x14	0x00	0x00
0x15	0x00	0x00
0x16	0x00	0x00
0x17	0x00	0x00
0x18	0x00	0x00
0x19	0x00	0x00
0x1A	0x00	0x00
0x1B	0x00	0x00
0x1C	0x00	0x00
0x1D	0x00	0x00
0x1E	0x04	0x04
0x1F	0x02	0x02
0x20	0x30	0x30
0x21	0x01	0x01
0x22	0x00	0x00
0x23	0x00	0x00
0x24	0x01	0x01
0x25	0x00	0x00
0x26	0x00	0x00
0x27	0x00	0x00
0x28	0x00	0x00
0x29	0x00	0x00
0x2A	0x00	0x00
0x2B	0x00	0x00
0x2C	0x00	0x00
0x2D	0x00	0x00
0x2E	0x00	0x00
0x2F	0x00	0x00
0x30	0x00	0x00
0x31	0x00	0x00
0x32	0x00	0x00
0x33	0x20	0x20
0x34	0x40	0x40
0x35	0x00	0x00
0x36	0x00	0x00
0x37	0x10	0x10
0x38	0x00	0x00
0x39	0x00	0x00
0x3A	0x00	0x00
0x4C	0x01	0x01
0x4D	0x10	0x13
0x4E	0x2C	0xED
0x4F	0x32	0x32
0x50	0x00	0x00
0x51	0x00	0x00
0x52	0x00	0x00
0x53	0x03	0x03
0x54	0x00	0x00
0x55	0x00	0x00
0x56	0x01	0x3B
0x57	0x00	0x00
0x58	0x5D	0x5D
0x59	0x00	0x00
0x5A	0x00	0x00
0x5B	0x30	0x30
0x5C	0x88	0x88
0x5D	0x20	0x20
0x5E	0x00	0x00
0x5F	0x00	0x00
0x60	0x00	0x00
0x61	0x00	0x00
0x62	0x00	0x00
0x63	0x00	0x00
0x64	0x00	0x00
0x65	0x94	0x94
0x66	0x00	0x00
0x67	0x00	0x00
0x68	0x00	0x00
0x69	0x00	0x00
0x6A	0x00	0x00
0x6B	0x00	0x00
0x6C	0x00	0x00
0x6D	0x7C	0x7C
0x6E	0x88	0x88
0x6F	0x88	0x88
0x70	0x2B	0x2B
0x71	0x2C	0x2C
0x72	0x00	0x00
0x73	0x04	0x04
0x74	0xB7	0xBB
0x75	0x09	0x09
0x76	0x60	0x60
0x77	0xC5	0xC5
0x78	0x00	0x00
0x79	0x01	0x01
0x7A	0x0F	0x0F
0x7B	0xFF	0xFF
0x7C	0x20	0x20
0x7D	0x00	0x00
0x7E	0x00	0x00
0x7F	0x00	0x00
0xB0	0x04	0x04
0xB1	0x10	0x10
0xB2	0x02	0x02
0xB3	0x08	0x08
0xB4	0x25	0x25
0xB5	0x00	0x00
0xB6	0x18	0x18
0xB7	0x00	0x00
0xB8	0x8C	0x8C
0xB9	0x33	0x33
0xBA	0x83	0x83
0xBB	0x74	0x74
0xBC	0x80	0x80
0xBD	0x00	0x00
0xBE	0x00	0x00
0xBF	0x00	0x00
0xC0	0x00	0x00
0xC1	0x00	0x00
0xC2	0x00	0x00
0xC3	0x00	0x00
0xC4	0x00	0x00
0xC5	0x00	0x00
0xC6	0x00	0x00
0xC7	0x00	0x00
0xC8	0x00	0x00
0xC9	0x00	0x00
0xCA	0x00	0x00
0xCB	0x00	0x00
0xCC	0x00	0x00
0xCD	0x00	0x00
0xCE	0x00	0x00
0xCF	0x00	0x00
0xD0	0x00	0x00
0xD1	0x43	0x43
0xD2	0x94	0x94
0xD3	0x33	0x3F
0xD4	0x60	0x60
0xD5	0xE0	0xE0
0xD6	0x00	0x00
0xD7	0x03	0x02
0xD8	0x07	0x07
0xD9	0x7F	0x7F
0xDA	0x04	0x04
0xDB	0x0B	0x6B
0xDC	0x00	0x00
0xDD	0x00	0x00
0xDE	0x00	0x00
0xDF	0x00	0x00

Reg	635-1	635-2
0x00	0x30	0x30
0x01	0x00	0x00
0x02	0x12	0x12
0x03	0x48	0x48
0x04	0x00	0x00
0x05	0x03	0x03
0x06	0x40	0x40
0x07	0x10	0x10
0x08	0xFE	0xFE
0x09	0x1E	0x1E
0x0A	0x10	0x10
0x0B	0x13	0x13
0x0C	0x26	0x26
0x0D	0x00	0x00
0x0E	0x87	0x87
0x0F	0x00	0x00
0x10	0x00	0x00
0x11	0x00	0x00
0x12	0x00	0x00
0x13	0x00	0x00
0x14	0x00	0x00
0x15	0x20	0x20
0x16	0x18	0x18
0x17	0x3F	0x3F
0x18	0x80	0x80
0x19	0x52	0x52
0x1A	0x52	0x52
0x1B	0x42	0x42
0x1C	0x00	0x00
0x1D	0x00	0x00
0x1E	0x00	0x00
0x1F	0x00	0x00
0x20	0x00	0x00
0x21	0x00	0x00
0x22	0x00	0x00
0x23	0x00	0x00
0x24	0x00	0x00
0x25	0x02	0x02
0x26	0x00	0x00
0x27	0x00	0x00
0x28	0x67	0x67
0x29	0x33	0x33
0x2A	0x01	0x01
0x2B	0x00	0x00
0x2C	0x00	0x00
0x2D	0x00	0x00
0x2E	0x00	0x00
0x2F	0x00	0x00
0x30	0x00	0x00
0x31	0x20	0x20
0x32	0x09	0x09
0x33	0x04	0x04
0x34	0x00	0x00
0x35	0x10	0x10
0x36	0x00	0x00
0x37	0x60	0x60
0x38	0x00	0x00
0x39	0x00	0x00
0x3A	0x00	0x00
0x3B	0x00	0x00
0x3C	0x00	0x00
0x3D	0x00	0x00
0x3E	0x00	0x00
0x3F	0x00	0x00
0x40	0x00	0x00
0x41	0x00	0x00
0x42	0x00	0x00
0x43	0x00	0x00
0x44	0x00	0x00
0x45	0x00	0x00
0x46	0x00	0x00
0x47	0x00	0x00
0x48	0x00	0x00
0x49	0x00	0x00
0x4A	0x00	0x00
0x4B	0x00	0x00
0x4C	0x00	0x00
0x4D	0x00	0x00
0x4E	0x00	0x00
0x4F	0x00	0x00
0x50	0x20	0x20
0x51	0xC0	0xC0
0x52	0x47	0x47
0x53	0x00	0x00
0x54	0x00	0x00
0x55	0xD6	0x1A
0x56	0x00	0x01
0x57	0x05	0x05
0x58	0x00	0x00
0x59	0x00	0x00
0x5A	0x07	0x07
0x5B	0x00	0x00
0x5C	0x00	0x00
0x5D	0x00	0x00
0x5E	0x00	0x00
0x5F	0x00	0x00
0x60	0x00	0x00
0x61	0x00	0x00
0x62	0x00	0x00
0x63	0x00	0x00
0x64	0x00	0x00

+1 Hamzeh Jaradat 5 months ago in reply to Barry W

TI__Guru** 107810 points

Hello Barry,

Based on the provided register dumps, it looks like you have poor link quality. Please check you cable type/length and test using another cable type and length.

This probably indicates a poor channel parameters for the link. Can you measure Return-loss and Insertion-loss and provide the test results to see if they fulfill our channel specifications.

Here is my registers analysis:

DES:
reg 0x20 = 0x30 -- no RX port is being forwarded to CSI TX
reg 0x33 = 0x20 -- CSI TX output is not enabled
reg 0x4D = 0x10/0x13 -- Lock status change, No LOCK or PASS
reg 0x4E = 0x2C/0xED -- multiple error types detected
reg 0x56 = 0x01/0x3B -- number of RX parity errors
reg 0x74 = 0xB7/0xBB -- Line count is not stable
reg 0x7A = 0x0F -- All types of CSI errors
reg 0x7B = 0xFF -- max amount of CSI errors which means received data is corrupted.
reg 0xD3 = 0x33/0x3F -- AEQ values are very high indicating poor link quality.
reg 0xDB = 0x6B -- multiple error interrupts have triggered.

SER:
reg 0x52 = 0x47 -- CRC error detected
reg 0x55 = 0xD6/0x1A -- CRC errors count

0 Barry W 5 months ago in reply to Hamzeh Jaradat

Intellectual 890 points

Thanks for the analysis. I went through the whole list and came to the same conclusions.

The curious thing is the link is short and the CMLOUT looks clean.

I think we will try manually setting the EQ in Register 0xD4 to test lower values.

Does that offer any chance of working?

0 Hamzeh Jaradat 5 months ago in reply to Barry W

TI__Guru** 107810 points

Barry W said:
I think we will try manually setting the EQ in Register 0xD4 to test lower values.

Does that offer any chance of working?

Not sure if that will make any improvement, but you can give it a try.

Can you also check DES reg 0x41 value should be 0xA9.

0 Barry W 5 months ago in reply to Barry W

Intellectual 890 points

Thank again for the help. I really appreciate it. Your comments helped us put together a test to diagnose the problem.

We found 2 things:

1. Running with CMLOUT enabled destroyed our incoming reliability. The register logs I sent were taken with CMLOUT enabled. Once we disabled CMLOUT, the issues were 10x better, but not fixed.

2. The root cause turned out to be EMI from a noisy high voltage switching supply. After applying some filtering to that supply, we saw the errors disappear. Auto-equalization now initializes at a low setting and the parity and CSI errors are gone.

Interface

Interface forum

DS90UB638-Q1: Unstable Lock/Pass