Your Kiwi must be running the v1.690 (or later) release to use the proxy service and be in the public Kiwi list (,

Please visit (documentation) and (online store)

Compact BB-AI

Testing shows 43 deg C with one core at 100% More testing to follow.


  • I still want to have a go at reversing the stack, seems illogical to have the most thermally active sources in the least thermally accessible locations.
    If only they sold BeagleBones and Capes without soldered headers.
  • If only..... :) I'd do a lot of things different if that was the case. My approach seems to work and make no mods to either the KiwiSDR cape of the BB-AI.
  • Jim,
    What does your broadband spectrum/noise floor look like with a terminated RF input?
  • I'll be looking at noise etc. over the weekend. Since the last pix, I added these at each end

  • When I see devices dependent on fans for safe operation, I get a bit nervous - but know that it's sometimes a necessary evil, particularly when space is confined: It goes without saying that one use a known-good brand-name fan.

    I say this considering the fact that out of four KiwiSDRs that I take care of, all four of the original, generic Chinese case fans have failed by the time a year had elapsed - most had started to get noisy long before this - three of them in areas with large temperature swings, and one that just sits on my desk.

    A direct replacement (slightly thinner, but ball bearings instead of sleeve) for the case fan that is of a known brand is here:

  • Thanks Clint.... I used the available fans for POC and once I know that works, I'll upgrade the fan(s)
  • Glenn, How BB is BB? full "DC" to 30 MHz. Or large range like 5-10. The SPEC function gives a lot of different answers depending on that.
  • I'm interested in seeing a spectral display of 0-30 MHz with perhaps a 10 kHz AM BW parked in the middle (or somewhere) measuring the noise spectral density through the communications channel and S-meter. The spectral display is probably 1024 bins so the same answer should be obtainable either way but might as well have a check.

    I'm interested in seeing any evidence of coherent or BB related stuff. I think that stock kiwi's generally have a 'bulge' coming out of the broad noise floor up above 15 MHz - maybe 2-3 MHz wide - that is probably something on the cpu board. I'm interested to see if things are any different with the AI and if they seem to be acceptable. It's a different CPU, probably with different coupling mechanisms to the kiwi board and I'd like to see if anything new/bad is obvious that would hurt sensitivity anywhere in the entire range.
  • edited December 2019
    Glenn.... you can try it here.
  • Thanks for letting me look. Generally it looks pretty good, maybe even a little better than a stock BB/kiwi. I don't see much coherent stuff except the weak line at 25.2 MHz which I think is network related and some of the usual very low end stuff. Indicated noise floor is around -156 dBm/1-Hz which seems normal and no 'bump' at mid-upper HF such as is visible on some kiwis - maybe only visible when the overall noise floor is lower in the distribution. I've always suspected that was coming from the stock BB.

    I do notice that it seems to operate just a little differently. Specifically, changing zoom level takes quite a while to catch up. While one is waiting, maybe 3-5 seconds, a wider spectrum selection looks black. It does eventually update but seems different. I wonder if this is what F6BIR was mentioning. Try it and see what you think.

    GA and up an antenna on it and lets see how it does with CM. No guarantee that the CM performance will be the same as before. Not only may the actual currents be different, e.g. network goo may be bigger but it's also possible that it is couples differently if the AI has shuffled things around. Probably the kiwi-specific stuff won't change but one can't say much about all this without measurement and study.

    Thanks again.

    Glenn n6gn
  • after adding the antenna, did see some very low level stuff up at 20-ish. Now have 13 ch. of wsprdaemon feed running and temp is OK. That brought my total wsprdaemon load on that box (Odroid XU4) up to 27 channels so had to tweak a few things there.
Sign In or Register to comment.