Skip to content

Latest commit

 

History

History
55 lines (37 loc) · 2.66 KB

Design.md

File metadata and controls

55 lines (37 loc) · 2.66 KB
Raw

Design

Maybe you should first have a look at AD9363 + ZYNQ schematics of PlutoSDR, ANTSDR, and kangyuzhe666's SDR.

A series of "hacking" is required to get things work.

  • AD936X run on 1.8 V or 2.5 V, while most FPGA boards have VCCIO 3.3 V
    • Use level converters like SN74AX8T245. Changing VCCIO is not plausible because USB and ETH PHY need 3.3 V
  • Wiring is both poor and long on FPGA boards
    • So have to use LVDS mode. Openwifi needs LVDS anyways
    • Have to stick to 1RX 1TX. But that's enough for most applications
  • But ZYNQ with 3.3 V HP VCCIO can't run LVDS
    • For receiving, HDMI can run TMDS_33. So re-drive using SN65CML100 -- NO change on PYNQ-Z1
      • Can also change 50 R pullup to 100 R series termination and directly receiving LVDS
      • Can also remove 50 R pullup and use DS90LVDS048A receiver
    • For sending, I use single-ended output on FPGA, and convert these to LVDS using DS90LVDS047
  • PMOD can't run high speed
    • Yet it can run at 80 MHz (40 M sampling rate 1rx-1tx), enough for openwifi, dump1090, and everything else I tried
    • Optionally remove TVS diodes and 200 ohm series resistors, can run at 120 MHz (~61.44 M sampling rate 1rx-1tx)
  • Board shape is strange...
    • Small connectors on PMOD/HDMI side is required
    • I'm worried about mechanical strains caused by HDMI bridges, so using short HDMI cables is being considered.
  • Hand-solder BGA?
    • Heatgun, soldering iron, solder paste, flux are enough, it's really not that hard
    • warning: Boards after SMT assembly will suffer oxidation when re-flowing, so ordering SMT assembly w/o the BGA chip then manually solder the BGA may not work well

Future plan

V1.2 seems good and will be for sale soon. Another batch(v1.2.1) with minor modification will come soon.

...

Another 5 fabricated and proved to be fine. Another batch(v1.2) coming soon.

...

Boards version 1.1 have been made and manually assembled using 1.0's material. It works fine, so another batch of 5 will be made in late Feb. 2022.

...

Boards version 1.0 suffered BGA soldering problems because they have been re-flowed during STM assembly.

...

Since I'm confident unmodified PYNQ-Z1 can make it, board version 1.0 will be designed and sent to fabrication and SMT assembly before Dec 31, 2021. 5 boards are to be made.

If the 5 boards turn to be OK, more will be coming.

Design "explained"(initial testing version, deprecated):