DS90UB953-Q1: noise/corruption pattern

Michael Koschutnig

Part Number: DS90UB953-Q1

Dear TI support team,
I am currently working with the following hardware setup:
An image sensor with MIPI CSI-2 interface, 4 lanes, 800Mbps/lane
connected to DS90UB953TRHBTQ1
connected to DS90UB954TRGZRQ1 via differential connection / STP cable
connected to a NXP i.MX6Q MIPI CSI-2, 4 lanes, 800Mbps/lane

The image sensor streams monochrome images at ~24fps, 640x480px in MIPI RAW12 format. The image shows a noise/corruption pattern in area px585/0 to 630/480. A screenshot is available here: http://upload.bluetechnix.at/file/0r96rah009seb3x/mipibridge.jpg
The hardware is based on the DS90UB953-Q1EVM reference schematic. The serializer is running in synchronous mode.
What could be the source of this pattern? Error registers show no errors and pattern generator image seems ok.
Thank you,
MK

over 7 years ago

0 Palaniappan Manickam over 7 years ago

TI__Genius 10475 points

Hello,
Do you see any CSI-2 errors or do you have a CSI analyzer to look at the packets? Have you tried BIST to check the forward and back channels between 953 and 954 are working ok?

0 Michael Koschutnig over 7 years ago in reply to Palaniappan Manickam

Prodigy 30 points

Hi,

I can see no CSI-2 errors nor BIST errors. Unfortunatly I have no access to a MIPI CSI analyzer.

0 Hamzeh Jaradat over 7 years ago in reply to Michael Koschutnig

TI__Guru** 107380 points

Hello MK,

if error registers on the UB954 do not show any error, then the FPD-Link communication is okay.
please check the noise on the power supply pins of the image sensor. This could create noise.
also you can measure the noise on the powersupply pins of the Deserializer. the measured noise (ripple) values should not exceed the maximum numbers determined in the Datasheet.

0 Michael Koschutnig over 7 years ago in reply to Hamzeh Jaradat

Prodigy 30 points

Hello Hamzeh,
we checked the power supply and it seems as if power supply vpp was too high for the deserializer. We added additional capacitors and now voltage is stable and the corruption pattern is gone.
The serializer had the same power supply, though the noise did not show any impact in the image (but we added the additional capacitor there, too).
Strange that the serializer's test pattern did not show any corruption with this bug.
Thank you for your help!

All the best,
Michael

Interface

Interface forum

DS90UB953-Q1: noise/corruption pattern