This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

DS90UB953A-Q1: Video formats supported

Part Number: DS90UB953A-Q1
Other Parts Discussed in Thread: DS90UB954-Q1

For our industrial imaging application, we are considering using the DS90UB953A-Q1 serializer and DS90UB954-Q1 deserializer to support multiple image sensors. The full frame output of the sensors will exceed 1920Hx1080V pixels, output over a CSI-2 interface in various RAW formats at up to 60fps. The application will configure the sensors for various non-standard partial frame outputs that output only a sub-rectangle from the full sensor.

My questions, without having fully read the 953 data sheet, concern the supported video formats.

The data sheet states that the 953/954 support 1920Hx1080B at 60fps in RAW formats. Does this represent the upper limit of the channel capacity?

Do the serializer and deserializer require different configurations for different video formats?

Do the serializer and deserializer support only standard video formats, or will they support unusual (non-standard) horizontal and vertical resolutions in RAW modes that output only a sub-rectangle from the full sensor?

If so, then is the frame rate limited simply by the SerDes channel capacity, or are there further limitations that depend on the video format? In other words, will selecting a sub-rectangle from the full sensor allow us to achieve a greater frame rate than achievable will the full frame?

Is the 954 the best choice for a companion deserializer to the 953 serializer?

Thank you, M.Reich

 

  • Hello Matthew,

    Here are my answers:

    The data sheet states that the 953/954 support 1920Hx1080B at 60fps in RAW formats. Does this represent the upper limit of the channel capacity?

    [Casey] No this is not the upper capability. It is just a common use case example. The upper capability is defined by the maximum allowed CSI-2 bandwidth which is 832Mbps/lane = 3.328Gbps total for 4 lanes 

    Do the serializer and deserializer require different configurations for different video formats?

    [Casey] No they do not 

    Do the serializer and deserializer support only standard video formats, or will they support unusual (non-standard) horizontal and vertical resolutions in RAW modes that output only a sub-rectangle from the full sensor?

    [Casey] There is no restriction on the video formats whatsoever. In fact, the 953/954 could be used for other data besides video data (user defined CSI-2 data). They just act like a pass through of the packets that they are receiving from the source 

    If so, then is the frame rate limited simply by the SerDes channel capacity, or are there further limitations that depend on the video format? In other words, will selecting a sub-rectangle from the full sensor allow us to achieve a greater frame rate than achievable will the full frame?

    [Casey] See above - as long as the total CSI-2 bandwidth is within spec and the per lane speeds are observed then our two devices just act like a pass through

    Is the 954 the best choice for a companion deserializer to the 953 serializer?

    [Casey] Yes! 954 is a fine choice. Also 960 could be used if you want > 2 RX inputs (hub device) 

    Best Regards,

    Casey 

  • Hello Casey,

    Your answer stating that the SerDes acts ‘like a pass through of the packets’ is just what I was hoping to hear for my application.

    Following up on your answers, and after examining the DS90UB953A data sheet and Application Report SNLA267A, here are a few additional questions:

    Data sheet section 7.3.1.3 “CSI-2 Protocol Layer” describes short and long packets. In order to check the CRC the 953 needs to interact with the data packets at least minimally the CSI-2 level, above the D-PHY level. To do so, it needs to distinguish short packets from long packets, and it probably needs to determine the length of the long packets.

    Does the 953, or deserializers such as the 954, also check the long packet checksum?

    Does either the 953 or the 954 (or 960) have any other interaction with the packets, that is, at the CSI-2 level?

    Your reply make quite clear that the SerDes is transparent to the data formatting. But, just to clarify, why then do all serializer and deserializer product descriptions list specific supported video formats? I’m asking again since our application will use some non-standard formats.

    I have not examined any deserializer data sheets in detail yet. But, the factor that will influence the choice between the 954 and the 960 is maximizing aggregate throughput. Since the maximum data rate of most of our sensors exceeds the 3.32Gbps capability of the 953, a design goal will be to avoid additional limitations. If necessary we could use one deserializer per sensor.

     

    Does multiplexing more than one sensor onto a single CSI-2 output from the 954 or 960 add overhead (neglecting any demultiplexing in the image processor)?

    Do these considerations point to either the 954 or the 960?

    Can you point to a reference that explains the protocol for multiplexing and demultiplexing data streams to and from CSI-2?

    Thank you,

    M.Reich

  • Hello Matthew,

    More answers:

    Does the 953, or deserializers such as the 954, also check the long packet checksum?

    [Casey] The deserializer performs checksum on both long and short packets. Errors are flagged if checksum doesn't match and those packets can be forwarded or not forwarded based on desired configuration. Both sides perform ECC and can correct 1 bit error or detect 2 bit error 

    Does either the 953 or the 954 (or 960) have any other interaction with the packets, that is, at the CSI-2 level?

    [Casey] It is possible to remap the virtual channels in the packet header but aside from that there is no interaction 

    Your reply make quite clear that the SerDes is transparent to the data formatting. But, just to clarify, why then do all serializer and deserializer product descriptions list specific supported video formats? I’m asking again since our application will use some non-standard formats.

    [Casey] These devices are primarily marketed towards automotive applications for CMOS imagers, so these are the most common use cases but we can of course support a variety of other applications like Lidar, Radar, DLP, etc. 

    I have not examined any deserializer data sheets in detail yet. But, the factor that will influence the choice between the 954 and the 960 is maximizing aggregate throughput. Since the maximum data rate of most of our sensors exceeds the 3.32Gbps capability of the 953, a design goal will be to avoid additional limitations. If necessary we could use one deserializer per sensor.

     [Casey] 954 and 960 can both output as much information as they get from inputs so there really shouldn't be any limitation here. I'm not sure what you mean about the aggregate data exceeding 953 though. Please ensure the 953 requirements are met for total bandwidth input

    Does multiplexing more than one sensor onto a single CSI-2 output from the 954 or 960 add overhead (neglecting any demultiplexing in the image processor)?

    [Casey] The overhead actually depends slightly on the forwarding mode. There is some information on this in the 960 datasheet in section 7.4.20. In round robin mode the overhead is slightly higher because there are more SOT/EOT/etc.. If you can use line concatenate mode for example then some of these portions are omitted so the overhead is less 

    Do these considerations point to either the 954 or the 960?

    [Casey] Either one would be acceptable - if you want more sensors into one hub then I would suggest 960

    Can you point to a reference that explains the protocol for multiplexing and demultiplexing data streams to and from CSI-2?

    [Casey] Please see 960 datasheet section 7.4.17 for information on virtual channels and multiple streams per CSI-2 feed

    Best Regards,

    Casey 

  • Hello Casey,

    Thank you for your informative answers.

    This finishes the first stage of examining the TI SerDes line for our application.  They appear suitable.  Now, a more detailed examination of the data sheets.

    Thank you,

    M.Reich