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.

DS90UB954-Q1: UB954 deserializer Issue with CSI clock signal in CSI-2 synchronous mode with non-continuous clock

Part Number: DS90UB954-Q1

Hi,

We are using UB953-954 Ser-Deser pair for connecting to Image sensor IMX390 in our product.
Here we are able to configure the Ser-Deser and Sensor to the desired configuration with I2C operations.
However, the CSI-clock output of deserializer does not match the csi-clock input of Serializer.

Scope readings attached.

The CSI clock as we understand should produce
a) High frequency (~300MHz) signal with ~200mV amplitude duting data transfer period
b) LP11 i.e both P and N pins should be ~1V in frame blanking period. (and thus at the rate of ~60Hz for 50 FPS video.)

As we can see in the attached images, the serializer input CSI clock follows this exect pattern.
However, the Deserializer output CSI clock seems to be deviating in 2 aspects.
a) Amplitude of high frequency signal during data transfer period. (getting only 20~30 mV instead of 200mV)
b) Frequency of LP11 Occurence (observing LP11 State at a rate of 60KHz instead of ~60 Hz)

Note on Images Attached:
csi_clk_voltage_levels.png : csi clk cam -> ser input , csi clk main -> deser output

ser_in_csi_clk.png                : Serializer output clock becoming LP11 at ~52 Hz for 60 FPS.

deser_out_csi_clk.jpg          : Deserializer output clock becoming LP11 at ~62 KHz for 60 FPS.



We are running our ser-deser in CSI-2 Synchronous mode, with non-continuous Clock, configuring the sensor to produce 1920x1080 frames at 60FPS

Also, to isolate the issue we tried the pattern generator code as given in UB954 specification, and did not configure serializer.
The same CSI clock behaviour could be seen with pattern generator configuration as well.

Ser-Deser Configuration is as follows:
Deser : {Reg Address, Reg Value}
        {0x4c, 0x01},
        {0x33, 0x01},  // 4Lane, non-continuous Clock
        {0x20, 0x20},
        {0x42, 0x71},   
        {0x6e, 0x88},
        {0x58, 0x5E},
        {0x6d, 0x40},  // STP cable for FPD
        {0xd5, 0xf0}
        // Serializer and sensor I2C id and Alias programming


Ser :   {Reg Address, Reg Value}
        {0x02, 0x33},
        {0x0E, 0x1E},
        {0x0D, 0xE1}


Does the deser require some other specific configuration to output the CSI-clock in the expected manner.

Please guide us for getting the correct signal in CSI-clock


Thanks in advance

Best Regards,
Ranjith Bhide

  • Hello Ranjith

          Is the Deserializer DS90UB954 locked? LOCK signal should be high if it is locked. 

    Thanks

    Vijay

  • HI Vijay,

    Yes, we have probed the lock signal and pass signal both show High(~3.3V).

    Best Regards,

    Ranjith Bhide.

  • Hi,

    Attaching the scope reading of data and clock lane upon running Pattern Generator:

    Data Lane :

    Clock Lane :

    Clock Lane Scaled up for Time and Amplitude :

    This seems to have same issues as mentioned in the original query.

    The clock seems to be going to LP11 after every ~33us => 30KHz frequency of LP11.

    and the signal amplitude is only ~20 mV.

    Please let us know if this indicates, issue in any particular aspect that we can further look into.

    Best Regards,

    Ranjith Bhide.

  • Ranjith,

    This is normal. The 954 CSI-2 output operates at discrete DPHY clock rates. Most likely what you are seeing here is that the input CSI-2 at the sensor side is running at a much slower rate than the 954 output is configured for and the clock rate is set such that the output doesn't spend any excess time in LP11. On the 954 side the default output speed is 800Mbps/lane and since it is running higher than the input, this just means that the data packets are time compressed and in the excess time where no data is being sent, the CSI-2 output will go into LP11. 

    I can see you are also configuring the 954 clock for continuous mode. This means that the clock will not toggle between LP11 and HS mode like your source is doing. You can enable/disable this as desired in the 954 registers. Finally the scope being used for measuring the 954 output I can see is only 100MHz bandwidth. This is not nearly enough for properly measuring the 954 output since the clock rate is 400MHz by default (and I don't see you doing any adjustment of register 0x1F in your above code). So what you are measuring on the scope is heavily attenuated (but seems overall normal to me - no issue). 

    Best Regards,

    Casey 

  • Hi Casey,

    Sorry for the delayed response, We were having annual shut down during the year end, and we are back today.

    W.R.T the attenuation in 400 MHz clock signal in previous comment of ours, we will check with a CRO of higher bandwidth and get back (This was the scope reading for Pattern Generator configuration in 954 Deser).

    However, the first image attached in the original query  ( https://e2e.ti.com/cfs-file/__key/communityserver-discussions-components-files/138/csi_5F00_clk_5F00_voltage_5F00_levels.png ) is taken from a high bandwidth Oscilloscope, with differential probe. here the green channel is Serializer input mipi clock, orange channel is Deserializer output mipi clock. Here we can clearly see that the voltage level of the clock signal seems to be much smaller the expected 200mV. What could be causing this?

    Regarding the frequency of LP11 occurrence in the clock signal. We came across the following statement in UB954 deser spec under topic 7.4.15 CSI-2 Protocol Layer:

    "Special packets for frame start, frame end, line start and line end information"

    Does this mean that there would be a LP11 state after each line transmission (i.e during line blanking period), which could be resulting in the observed behaviour of high occurence of LP11 state in Deser Output as compared to Sensor output (where LP11 is occurring only after frame end)

    Best Regards,

    Ranjith Bhide.

  • Hello Ranjith,

    The output of the HS clock from the 954 doesn't look unreasonable but it hard to tell the differential swing from the scope shot since the scale is large. The HS transmit differential voltage spec is 140-270mV. Is it possible that your scope is actually providing termination to the measured signal? I see the coupling is 50 ohm but I believe you would need to measure with a high impedance probe in order to ensure that the signal is not double terminated. 

    LP11 happens during all inactive time between MIPI packets. Depending on the frequency of the CSI-2 transmitter, the time it takes to transmit each packet is variable so the space in-between will be LP11. 

    Best Regards,
    Casey