TDoA doesn't work with a mix of 4ch and 8ch Kiwis [fixed in v1.214]

G4DYAG4DYA
edited August 2018 in Problems Now Fixed
Hi all. I can get sane TDoA results if I use all 4ch Kiwis, or all 8ch Kiwis, but not with a mixture of the two types.

73,

Richard G4DYA

Comments

  • jksjks
    edited July 2018
    Thanks -- that is very valuable information. One difference between RX4WF4 and RX8WF2 modes is the size of the audio buffer in the FPGA. I didn't think this would make a difference based on how the timestamps were applied. But obviously there is an offset. I'll try and get this fixed today.
  • jksjks
    v1.214 is out with a fix for this problem. The list of TDoA Kiwis will be limited to those running v1.214 or later so this problem won't continue to be seen.
  • ChristophChristoph
    This sounds like there is an offset in the absolute gnss timestamps with the new FPGA firmware.

    @G8JNJ or @jks : could you make an IQ recording with timestamps from a 1PPS (better 5PPS or more) input signal? Then I or someone else can measure the new offset.
  • G4DYAG4DYA
    v1.214 seems to solve the problem. Thank you very much for the quick fix.
  • G8JNJG8JNJ
    @Christoph

    OK will do.

    Can you remind me of the kiwirecorder.py syntax for the recording command string. It's a while since I last used it.

    I'm out today but I've left a KiWi in 8ch mode with the PPS marker running. GPS 1PPS is fed via 20m of coax.

    http://g8jnj.proxy.kiwisdr.com:8073

    Regards,

    Martin - G8JNJ
  • jksjks
    In the new firmware the audio buffer size is unchanged for RX4 mode. For RX8 it is doubles in size since I now have spare FPGA memory and a larger buffer is needed to have the same level of protection for the 4 new channels. What has changed for both modes is that the SPI buffer has doubled in size. So SPI transfers are more efficient. The SPI buffer size figures in the calculation of the NRX_CHANS constant which is now twice as large for RX4 mode and the same for RX8.

    The fix I made was to simply divide NRX_CHANS by 2 for RX4 mode so that it's equal to the value used for RX8 in the timestamp code. I don't understand this code fully, but it seems to have resolved the problem. Now there is also a tuned constant "gps_delay" which I think was measured empirically using the 1PPS setup if I remember correctly. This might have to be adjusted.
Sign In or Register to comment.