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.

DS90UB953-Q1: Cascaded AWR1243 data serialization for interface to Fusion board and TDA2Px

Part Number: DS90UB953-Q1
Other Parts Discussed in Thread: AWR1243

Hi, we are developing a cascaded system with four AWR1243 devices and four (possibly 2?) DS90UB953 for serialization, which will then connect via coax FPD to a Fusion board attached to a TDA2Px. I am working on the clocking scheme. We want the AWRs to take 256 samples on all four receivers during a 20us ramp. At 12 bits, complex, that is 307.2Mbps per receiver during the ramp. Using four CSI-2 outputs, it’s within the max 450Mbps spec per channel. I would like to have the 25MHz clock on the Fusion board be the master clock for the data flow, run the UB953 serializer in synchronous mode, and from it derive the reference clock for all four of the AWRs. One of the AWRs will be the master for the 20GHz FM_CW clocks fed to itself and the other three AWRs. Let me know if there are any showstoppers in this approach. My questions are:

1. From the UB953 datasheet, table 6 shows that for synchronous mode, with a 25MHz reference clock (f), the CSI bandwidth is f x 128 = 3.2Gbps. Given the AWR output data rate of 307.2Mbps per receiver, can I run two CSI-2 outputs from each AWR at 614.4Mbps each (total 1.2288Gbps) and connect two AWRs to one UB953 (total 2.4576Gbps on four lanes) for serialization?

1a. If the above is NOT true, and I need to use four UB953s, what’s the best way to prevent I2C address conflicts since each UB953 only has a choice of two addresses? Can I somehow utilize the Slave ID Alias feature (UB953 datasheet sections 7.7.52 – 7.7.67). I’m guessing not. It seems that it’s for I2C devices connected to the deserializer. This topic is not detailed in the diagram on the Jan 19th E2E post https://e2e.ti.com/support/sensor/mmwave_sensors/f/1023/t/657535 .

3. What frequency should the reference clocks to the AWRs be? We want a 20us chirp with a 2us idle time for a 22us Total chirp time. Frequency change over the 20us is 200MHz. Perhaps this has no influence on ref clk frequency and I can just generate 50MHz from the UB953 and distribute?

4. For configuring registers on the UB953 via the BCC from the deserializer/Fusion/TDA2Px, is the connection to the I2C bus internal to the UB953? Thus, addressing the UB953s from the controller on the TDA2Px is possible?

5. For the physical FPD coax connections to the Fusion board, do you supply or offer the mating coax cable assemblies? If so, what are the part numbers? Also the PN 59S20X-400L5-Z on the Fusion BOM does not appear to be correct/current. It should be 59S20X-400T5-Z. See http://rosenberger.de/ok/images/documents/db/59S20X-400T5-Y.pdf

6. Since the AWRs will be hardware triggered (using SYNC_IN), should I just use independent GPIO lines from the TDA2Px board? I’d rather not have the time uncertainty of the GPIO over the BCC and through the UB953. The TDA will not be far from the AWR, probably less than 6 inches.

If you want to see my tentative block diagram, let me know.

BTW, there may be an error in Figure 6-1 on the AWR datasheet regarding the 20GHz clock muxes. The FM_CW outputs appear to be connected the MUX input, and there is no output from the mux that the RFSYNTH block is connected to.

Thank you.

  • Hello,
    There are multiple questions here, many of them related to AWR or TDA parts. We can help with the FPD-Link UB953 questions first and forward this to the other forums for the rest, or it might be better if you create separate threads for the different products.
    Can you provide your system block diagram?
  • Thank you for the link to the training videos.  5.3 answered my question about the I2C interface.   I also found information on the frame sync internally generated by the UB960 and how it can be configured to be output on any of the UB953 GPIOs (my question 6 is resolved).  I modified my block diagram accordingly.  I have two AWRs going into a single UB953 in my diagram and wish to receive confirmation from TI that it can be done.  Sorry if the pdf doesn't come through. If so, I'll try a different method. Thanks

    AWR_Bd_DetailedBlockDiag_X2.pdf

  • Hello-

    The 953 can only interface to a single CSI-2 TX (1, 2, or 4 CSI-2 data lanes and a CSI-2 Clock lane).

    Regards,
    Davor
  • Thank you, I will go back to a single 953 per AWR device.  Referring to the link in my question 1a, I have not found an I2C to SPI bridge as shown in the TI block diagram.  Can you recommend a part?  If not, I also came across info in Figure 2 of the 953 EVM user guide (snlu224.pdf) that shows "HS_GPIO (SPI)" on the serializer and deserializer.  Would this be a better approach than the I2C to SPI bridge (if one exists) to interface to the AWR device which is SPI only?  I'm concerned with the vast bus speed difference the bridge would have to accommodate, but on the other hand the 953 only has 4 GPIO and using all 4 of them for SPI doesn't leave any for FrameSync.

  • HEllo Victor,
    1) no this is not advised to mix CSI-2 outputs to one port.
    1b) the deserializer has capability to map each Rx port differently and assign different alias to each serializer. Alternatively you could overwrite each I2C address in the serializer after initialization.
    3) the CSI-2 input clock has no effect on the serializer reference. As long as the CSI-2 rate is below the max throughput this should be fine. There is an N/M divider if you need other than 25 MHz CLK_OUT frequency.
    4) yes, this is typically preferred method.
    5) It is possible there are newer P/N , but they should be all compatible for this series. These are standard Fakra cables here are some reference to AMP and Pasternack, I think you might find on digikey also, I am not sure.

    www.google.com/url

    www.google.com/url

    6)The 954 953 supplies a frame synch output for all 4 FPD-Link over the back channel which can have very good alignment. Please reference 954 datasheet. If you can direct connect a GPIO this should also work.

    Regards,
    Liam
  • Thank you Liam for your responses.  I've not had any luck finding an "I2C/SPI bridge" as shown in Piyali's response in this thread: https://e2e.ti.com/support/sensor/mmwave_sensors/f/1023/t/657535 .  Did he identify one?  I'm not sure there is a way for me to ask him directly through E2E.

    I understand that there is a cascaded reference design in the works.  How is the SPI communications being handled to the AWR devices in that design?  I would like to avoid having a microcontroller or FPGA on my board just to accommodate this.  Would it be possible to use the SPI[2] MOSI, MISO, and SCLK from the TDA2Px EVM board's EXP_P3 connector (page 39 of sprr314) and a GPIO from the 953 for the CS of the AWR?  Let me know if you want to see the block diagram of this proposal.

    Thanks - Victor

  • Hello,
    For follow up on a previous question to a different forum, please use the button for 'ask a related question' .