Kiwi Noise Floor

In trying to isolate and reduce noise I disconnected the antenna at the SMA connector on the Kiwi. The radio signals drop on the spectrum of course, but only to a certain point, visually flat from about 105 dBm at higher frequencies to a minimum of about 110dBm at about 4Mhz. I have seen other Kiwi's with lower readings on their spectrums (even with antenna connected), so it got me wondering if my slightly higher noise floor was power supply or network related or a calibration issue. I do have a pretty clean SMPS power supply (Meanwell brand) and Ethernet is connected via Apple Express. Both connecting cables are wound multiple times through large mix 31 toroids. I will try a DC power supply again this weekend but I'm curious for anyone's thoughts. P.S. Disconnecting the GPS cable from the Kiwi made no difference either.



  • Mark,
    You did terminate the antenna input with a 50 ohm resistor? Then changing power supplies might show variances of power noise. You can connect direct to the KiwiSDR, computer to KiwiSDR, except when using Windows 10... ( ). This would reduce the sources of network noise and might help in the noise evaluation. G8JNJ posted this:

    Good luck,
  • edited March 2020
    It's interesting that you are having issues with an ostensibly clean switch-mode power supply (SMPS) with everything else disconnected. Here's a bit if info on trying to resolve such issues.

    What you should expect:

    A properly-functioning KiwiSDR with nothing connected but power and Ethernet has a noise floor of about -155dBm/Hz. If one scales this for a 2400 Hz (SSB) bandwidth one should expect a noise floor of about (-155 + 34 = ) -121dBm - which is about S1 or S2 - at all frequencies.

    If you see much higher than about S1-S2 with nothing but power and Ethernet connected - and you have NOT recalibrated the S-meter via the Admin interface - then you need to start doing some sleuthing.

    Not having included a screen shot, it's difficult to tell the nature of the noise. Typically, switching supply (SMPS) noise is "clumped" at lower frequencies (below 2-3 MHz) with apparent bands coinciding with the unit's harmonics - typically at 15-60 kHz intervals, 30-ish kHz being most common. At higher frequencies these clumps become less discrete, often tending to run together, the nature of the noise floor often being betrayed by AC mains frequency modulation if one listens using AM. It should go without saying that one should find a 5 volt linear supply - able to handle a peak of about 1.3 amps - for testing to establish a baseline before going through too much more effort.

    A few comments:

    - With no antenna (HF or GPS) connected, the Ethernet cable should not contribute significantly to the overal noise floor of the KiwiSDR. (This isn't necessarily the case when an antenna is connected, but that's for a later discussion.)

    - As a sanity check, I would use an inexpensive 5-8 port consumer-grade switch (one that has zero chance Power-over-Ethernet capability) for testing. I've run across some switches that have POE connections that are not galvanically-isolated that have the potential of bringing noise to the KiwiSDR - even when the POE is "off" - causing havoc in terms of conducted noise and susceptibility issues when transmitting on HF.

    - Make sure that the voltage reaching the KiwiSDR is, in fact, 5.0-5.3 volts. If it drops too much below this (misadjusted power supply, cable resistance) all bets are off when it comes to performance of the KiwiSDR. I doubt that a bit of undervolt will cause it to be noised up so much, but it's worth checking.

    - Assuming that you have eliminated any "through" paths for common-mode RF (e.g. disconnecting any antennas, proper galvanic isolation of the Ethernet connection) and you *still* see the noise it's worth noting that winding the power or Ethernet cable on a toroid will do absolutely no good as that offers only common-mode noise suppression.

    - If the DC output of the power supply is very noisy (e.g. 10s-100s of millivolts of ripple) a common-mode choke will, by definition, do nothing at all - meaning that the expense and effort of a toroid on the DC line will do nothing for that. One would, at the very least, need an L/C Pi-type network to suppress differential noise from the power supply - a simple circuit that will require rather large-value (hundreds of uF) low-ESR switching-type electrolytics, a large-ish choke (10s of uH capable of handling the DC current with acceptable voltage drop) and a few smaller capacitors. Again, this aspect can be easily checked with the use of a known-quiet linear power supply.

    If you do determine that your power supply is the culprit, you have exactly two choices:

    - Find another power supply that is known to be quiet. Linear supplies are typically (but not always) very "quiet" in terms of RF, but a typical, cheap off-the-shelf switcher is likely not to be and is going to be a gamble.

    - Take a known-reliable switching supply and make it "completely" quiet: This simply cannot be done to a noisy supply solely with the addition of ferrite devices on the input and output leads (toroids, snap-on chokes). To do this properly, it will be necessary to put this supply in its own metal enclosure and filter the daylights out of the in/out leads - and if done well, noise on the input/output leads can be reduced below the point of detection. I have done so on several occasions, although it takes a bit of effort, but if you want to entirely eliminate the supply as a potential noise source, it is what is required. I have documented doing this here:

    * * *

    After eliminating the power supply as a noise source there are other things that may be done in the Kiwi configuration, such as:

    - Setting the Ethernet to 10 Mbps. This can reduce the noise from the Ethernet PHY - most effective at the higher frequencies and to reduce QRM to other receivers (particularly VHF/UHF) that may coexist at the receive site.

    - Setting the SPI clock to the lowest rate. Particularly useful for reducing harmonic noise - especially on 2 meters - from the KiwiSDR itself. Unfortunately this tends to bottleneck data on this line, the most obvious effect being slower waterfalls when multiple users are connected.

    Finally, at most receive sites getting rid of the noise of the KiwiSDR itself is the easy part - it's the reduction of noise from other, nearby devices - most of them in the same household - that is the greater challenge and it is because of this that a large percentage of publically-available KiwiSDRs are just not very good at their intended purpose.

    Good luck and 73,

  • Just for the record: The Meanwell PSUs I bought for testing are not at all quiet. It could be that the ones I got didn't perform well. At any rate, I quickly brought the Bel power into use again.
  • "Just for the record: The Meanwell PSUs I bought for testing are not at all quiet. It could be that the ones I got didn't perform well. At any rate, I quickly brought the Bel power into use again."

    In any case, supplying the kiwi with a DC battery directly should allow for best case scenario I would think (as far as the power supply goes)?
  • Batteries often use buck converters to hold the voltage at 5v. So a battery pack is not necessarily a better option than other power supply units.
  • @kramcd not necessarily. Good linear PSUs are mostly noise-free. Moreover, many can be adjusted to the roughly 5.2-5.3 VDC ideal voltage. And while not "cheap", a Bel 3A power is only around 50-60 USD from Mouser, excellent value for money in my opinion. Tweakable voltage, and fires up the Kiwi every time without hiccups.
Sign In or Register to comment.