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.

ADC32J25: Interfacing with FPGA

Part Number: ADC32J25
Other Parts Discussed in Thread: LMK04828, , ADC32J24

Hello,

Now we have the board, and the board has been verified including the clock generation with LMK04828 and FPGA

We are now trying to get the JESD204B working between ADC32J25 and FPGA.  

We are leveraging a previously working FPGA design with JESD204B.

The 1st step is to see if we can receive a test pattern from ADC32J25.

Attached is the ADC register configuration.

Would you be able to take a look at the configuration?

Regards,

Andrew

7571.adc_config.txt
Fullscreen
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
00400100
00400300
00400400
00400602
00400700
00400800
00400900
00400A03
00400B30
00400C00
00400D00
00400E00
00400F00
00401300
00401500
00402700
00402A56
00402B01
00402F00
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

  • Hello Andrew,

    I'm not sure what exactly you would like for me to check from your configuration. When I loaded it to the GUI, I was able to obtain a capture, meaning the ADC is outputting data (picture below where I have my analog input operating at 25-MHz and 10dBm).

    Also, 5516.ADC3xJxx_128MSPS_Operation_LMK_Setting.cfg is the default config file in case that's what you were looking for. You can load it by clicking on the "Low Level Vi" tab, and then the "Load Config" button.

    If you could also write you set up the GUI to do, that would be very helpful as well.

    Thanks,

    Andrea

  • Hi Andrea,

    Thanks for your help.

    I set up the configuration to generate the test pattern (K.25) as opposed to capturing the ADC data.

    I've noticed in the ADC3000 GUI screencapture, it looks like all the registers are set to zeros.

    So for normal ADC operation, the defaults are all zeros.  Is this correct?

    I can format my config file so you can load it.

    I have a question on the format.

    If you look at the format I used (intended to load into our FPGA)

    For example, 

    00400602

    • the leading two "0"s are fill bits
    • "40" indicates I'm writing to the ADC
    • "006" is the register address
    • "02" is the register value

    Could you see if the format would work?

    Let me also write you on the clock architecture.  

    But I just wanted to see if the data format I used is consistent.

    Regards,

    Andrew

  • Hi Andrea,

    We are following the write format described on pg 55 of the datasheet

    So the "0" corresponds to the fact that A14 should be "1".

    Thank you,

    Andrew

  • config_fpga1.cfg

    I've reformatted the config file so you should be able to load into the GUI.

    I also noticed in ADC3000 GUI 

    • If I select ADC32Jxx, both LMK04828 and ADCJxx show up
    • If I select ADC32xx, only ADC32xx shows up

    I am assuming either one is okay.

    Thanks again,

    Andrew

  • Hello Andrew,

    So for normal ADC operation, the defaults are all zeros.  Is this correct?

    Correct, the default for all registers for the ADC32J24 are 0x00.

    Could you see if the format would work?

    Looking at the last config file you uploaded, that should be able to work IF you change line 1 from "ADC32xx" to "ADC32Jxx" and use the ADC32Jxx GUI page. If not, the data won't be read into the GUI. Also, it looks like in line 12 you did not finish writing that register write. Default_ADC32Jxx.cfg is the file from the default GUI for the ADC32Jxx page.

    I am assuming either one is okay.

    When I loaded the file yesterday to the GUI, I accidentally used the ADC32xx GUI page; however, we recommend to use the ADC32Jxx GUI page instead for this part. All the tests that I ran today to answer your questions where with the ADC32Jxx GUI page.

    I believe I have answered all your questions, if not, please do let me know or if you have any further questions.

    Good Luck,

    Andrea

  •  config_fpga2.cfg

    Hi Andrea,

    We are still troubleshooting.  

    In order to have a better visibility, we would like to see if we can detect a test pattern.

    Could you take a look at the screenshot and see if the test pattern generation is configured correctly?

    Also included is the config file.  Would it be possible for you to load the config file into your setup for verification?

    Thank you for support.

    Andrew

  • Hello Andrew,

    Could you take a look at the screenshot and see if the test pattern generation is configured correctly?

    By just looking at the screenshot, the GUI should be setup correctly for that page. I set my GUI up the same way for that page and I obtained a capture showing the pattern toggling (from the Codes window; however, from the Real FFT window I saw nothing):

    Here is my config with a working system (both LMK04828 and ADC32J25).

    TestPattern_Toggle_ADC32J25.cfg

    Also included is the config file.  Would it be possible for you to load the config file into your setup for verification?

    When I loaded the config file into my setup, the ADC was not picking up a clock, and hence, I was not able to obtain a capture. This means your configuration for the LMK04828 is incorrect since the part is not able to lock the PLL (when LED D4 is lit, it means the LMK04828 PLL is locked). If you tell me what frequencies you would like to output, if you have a set input frequency, and if there are any other requirements (such as SYSREF, etc) that you need from the LMK04828, I can help create a config with the correct LMK04828 setup.

    Also, note that you don't need to check the "Align Test Pattern" option to output a test pattern. You only need to check that box if you need to align the outputs of the test patterns across the different channels. Hope this helps and let me know if you need anything else from me.

    Good Luck,

    Andrea

  • Hi Andrea,

    I thank you so much for your support.

    I think we are getting close to figuring out the root cause, something related to clocks.

    Just realized the sampling clocks and SYSREF signals were set to LVDS but the circuit design is intended for LVPECL.

    I set the signal type to LVPECL20 now.

    Here are the TICS Pro files

    LMK04828b_v061623a.tcs

    HexRegisterValues_v061623a.txt
    Fullscreen
    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    R0 (INIT) 0x000090
    R0 0x000010
    R2 0x000200
    R3 0x000306
    R4 0x0004D0
    R5 0x00055B
    R6 0x000600
    R12 0x000C51
    R13 0x000D04
    R256 0x01000F
    R257 0x010155
    R258 0x010255
    R259 0x010301
    R260 0x010402
    R261 0x010500
    R262 0x010670
    R263 0x010711
    R264 0x01080F
    R265 0x010955
    XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

    Also in terms of SYNCP/N signals, they are designed as straight CMOS.  So there is no 100 Ohm resistor between the P/N traces.

    3755.fibersense_v1p23.pdf

    • Sheet 2: FPGA pins for ADC32J25
    • Sheet 15: ADC32J25
    • Sheet 18: LMK02828B

    Would it be an issue for SYNCP/N not to have a 100 Ohm resistor in between?  

    Looking at the table in Section 7.12 of the ADC32J25 datasheet, VIH = 1.3 and VIL = 0.5.

    So perhaps the design may still work.  Otherwise some rework may be needed.

    Appreciate your support.

    Thank you,

    Andrew

  • Hello Andrew,

    Here are the TICS Pro files

    I'm working on converting and loading the .tcs file to the ADC GUI, so I'll get back to you by early next week with an answer. In the meantime, are you using PLL1 at all? All your CLKinXs are disabled so no signal will get through to PLL1 even if it's powered up.

    Would it be an issue for SYNCP/N not to have a 100 Ohm resistor in between?  

    It is not an issue as long as FPGA pin for this diff pair has a 100-Ω internal termination.

    Best,

    Andrea

  • Hi Andrea,

    Thanks for the update.

    We have verified the LMK048288 is working again (not certain anything in the FPGA or just some mistake in LMK04828 registers).

    We have also verified that the FPGA can take LVDS (without 100 ohm) from LMK04828.

    Now with the FPGA working again I think if you can share ADC GUI settings I can try to figure out how to translate to TICS Pro settings.

    Regards,

    Andrew

  • Hello Andrew,

    I am glad you got the LMK04828 to work and provide a sampling clock to the ADC.

    What do you mean by ADC GUI settings? Above I attached the test pattern you requested to ensure you were able to communicate with the ADC (here is it attached again: TestPattern_Toggle_ADC32J25 (1).cfg). However, for me to create a config, I need to know at what frequency you're sampling (which I believe is 160-MHz from your schematic) and what exactly you need from the ADC. Then, I can generate a file for you and send it over. Once I get more clarification I can better support you.

    Best,

    Andrea

  • Hi Andrea,

    Sorry for delay.  I was able to spend some time and here is where things stand.

    1. I have gotten familiar with the ADC registers
      1. First I tried various combinations of 2Ah (SERDES Test)
      2. I believe this control the patterns at the JESD level so any ADC data will be ignored
      3. I see bcbc... which I believe is K28.5 characters
      4. They are fairly consistent.  Does it mean that most of the clocks are working?
      5. However, I try to send test ADC pattern as you also suggested in the ADC config file, 010101...
      6. But I'm still seeing bcbcbc...
    2. The ADC sampling clock is 160 MHz and the SYSREF is 8 MHz
      1. My LMK04828 configuration is a bit different because I use it to generate various clocks.
      2. However, I have verified I can control the LMK04828.  

    I'm also attaching the schematic.  A couple of things I would like to ask

    1. From the ADC datasheet, the ADC outputs (DAP/M, DBP/M) are CML.
    2. In our schematic they are treated as LVDS with the internal 100 Ohm termination within the FPGA
    3. Hope this is equivalent
    4. As for SYNCP/M~, the FPGA provide 3.3 CMOS signals.  In fact, I am not certain it is possible to generate LVDS from the regular IOs on the FPGA (Xilinx Kintex7)

    At this point, I think everything should be correctly set up:

    1. LMFS = 2221 and K=8
    2. Btw I think there is a typo in the ADC datasheet.  In Table 29. Register 2Bh Field Descriptions, 0 = Default is 9 (I think it should be 8)
    3. Also in 30h and 31h, they use a zero offset, meaning if K=8, Register 31h should be set to 0x07 (not 0x08).

    Attached below are

    1. screen captures of the Vivado ILA
    2. Schematic (ADC is on Sheet 15, JESD on FPGA is on Sheet 5, ADC config control is on Sheet 2
    3. ADC config used

    I sincerely appreciate your support.

    Regards,

    Andrew

    fpga_captures_070123.pptx

    3808.fibersense_v1p23.pdf

    adc_config_070123.txt
    Fullscreen
    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    00C00600
    00400601
    00C00600
    00400100
    00400300
    00400400
    00400602
    00400700
    00400800
    00400902
    00400A03
    00400B20
    00400C00
    00400D00
    00400E00
    00400F00
    00401300
    00401500
    00402700
    XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

  • I just tried the same ADC settings you provided, and just triggered on the rx_tdata (received data in FPGA via JESD).

    The data looks like: 54aa54aa5455a01c

    Is this something close what was expected?

    I understand the test pattern was: 1010...0101...

    Perhaps we may be missing some bits but hopefully it means we are getting closer.

    Thanks again,

    Regards,

    Andrew

    fpga_captures_070123a.pptx

  • Hello Andrew,

    I will get back to you at the end of the week once I get back to the office after the holiday weekend in the US.

    Best,

    Andrea

  • Andrew,

    0xBCBC indeed refers to K28.5 characters and shows that the ADC is in CGS stage of bring-up sequence. See below image. 

    Can you add SYNC to the ILA IP block and also trigger your ILA on SYNC de-assertion (rising edge of SYNC) and verify that data changes from 0xBCBC to K28.0 character and Dx.y data symbols? 4 multi-frames after the SYNC de-assertion the actual ADC data will be present at the FPGA if the ADC has exited ILAS (initial lane alignment sequence).

    What is the rx_tvalid signal in your ILA?  

    Thanks, Chase

  • HI Chase,

    Let me work on this (again sorry for delay due to my traveling).

    Thank you,

    Andrew