Probably still several layers to peel away here.  A quick look brings more questions.

As mentioned, terminated coax noise floor doesn't tell us much.

But Dec 12 v 13 heat map is very suspicious. 

        Noise actually drops on 60 & 630 bands from Dec 12 to Dec 13.  !!  uhhh.. Really?  It would seem to imply two mechanisms with some cancellation on the 13th. Yccch.

Sharp steps from the 13th on at least 12-20 are a Bad Thing©. Propagated noise doesn't do this. More fodder for the local ingress theory. Compare your shape over a day with a known quiet site/system, perhaps KPH.

Is the T1-1 the very closest thing to the Kiwi?  It should be or you risk providing a return path through a chassis/common of the filters to something. You don't want this, you want to float that end of the Kiwi to CM currents.

Amplifiers in the right place can provide relative suppression of ingress, in the wrong place they don't help. Until you have peeled away the, apparently, multiple mechanisms it is hard to even say where to put one the closer to what you consider to be the antenna is generally better.

I don't know what ou mean by 'only used as a probe'. Is that just saying this is not a main stream rx for you?  I didn't get the relevance nor the inclusion of selected bands (at least two of which are suspicious as described above).

More peeling required.

Glenn, Clint,
At https://wspr.live/gui/d/ereVvgn7z/station-noise-stats?orgId=1&refresh=1m&var-site=TI4JWC&var-receiver=All&var-band=All&var-type=rms_level
you will find my noise raspberryIKWI recordings. Dec-11 is 50-ohm termination at the KIWI. Dec-12 levels with a 50-Ohm termination at the cable at the base of the antenna.
Dec-13 and on are when running normally. The antenna is only used as a probe at 60M, 180, 630.
The noise levels Radio-only .vs. radio-feed-line range from a low of 2 db for 17, 20 M  to 23 db for 80 M
From the antenna to KIWI, in the shack I have in series T1-T transformer, AM filter, FM-filter.
I have recommended 14 db amplifier on the bench but I need a "limited Attenuation High pass filter"
Unfortunately, with only one good eye, my construction days are passed. 

John

Not what you really want to know. A monopole uses the 'ground system' as an image plane. Even with a good radial system and the best baluns it is probably not possible to suitably distinguish CM from differential+CM noise.

A termination only tells you that current on the coax with no 'sink' at the antenna is low enough. It usuallly is but if you can't terminate and see the Kiwi noise floor then you do have a problem to fix somewhere, possibly CM through the groundplane of the kiwi in/out the LAN/SMA or PS/SMA. Anything that keeps you from a 50-ohm Kiwi noise floor needs to to be fixed first so go ahead and try it just don't think that you are nearly done...

Glenn,
The antenna is a Hy-Gain AV-680, 9-BAND HF VERT ANT, 80/40/30/20/17/15/12/10/6M. I am guessing that the"shorting method" does not apply. 
Can I learn anything by disconnecting at the antenna feed point and applying a short or 50-ohn load?

John

John
That's helpful. It probably means that the conversion from CM to differential on your feedline is occuring on the antenna side of the T1-1.   If you have a symmetric antenna, one on which you can short a desired differential signal to examine an undesired CM component, you can verify this.  If OTOH you have an asymmetric antenna like a monopole this won't be possible since shorting the connections shorts both modes.

Earth currents and actual near-field sources are still possibilities too. There may be others.

Glenn,

Rob provided me with a T1-1 assembly months ago. Other than the first test with a 50-ohn load at the KIWI, I always use the T1-1.

John

John,
If you can't find that PCB you (or anyone reading) can easily make something  from an inexpensive TV/FM 4:1 transformer that will serve purpose:



The picture shows examples of a Mini-Circuits T1-1 along with  a homebrew transformer with 2turns:2turns wound on a core from one of the TV baluns at the top.
With typical core materials it won't have quite as low insertion loss in HF as the T1-1 but to it will have pretty low inter-winding-capacitance and good common-mode suppression for testing.
Here's  a quick example I just wound and measured.  It has about 2.5 dB insertion loss at 15 MHz compared to the T1-1 which is well under 1 dB but is plenty adequate to identify the large amount of CM that we're discussing. This is at 5 dB/division, green is T1-1 Yellow is TV balun wound with wire size small enough to get through the holes in the core:



If anyone wants the SMA-connectored PCB shown in the top photo, drop the attached xfrmr.brd file onto Osh Park's web page, pay them a small fee and receive it postpaid in the US within about 10 days.  Connectors are normal SMA or (preferrably PCB mount SMAs available on eBay. 

Don't be tempted to use a Mini-Circuits T6-1, perhaps because the specifications show a lower low end. In fact that part is not a conventional flux-coupled transformer, it's a transmission line and has too much coupling between input and output for our purpose.

If inserting one of these at the Kiwi's SMA connector doesn't make any difference then you've just proven that this isn't your problem and need to look elsewhere, but many times it is.

Glenn n6gn

--

Do you have a low interwinding capacitance isolation transformer located at the KiWi SMA or can you at least temporarily try one? Wind a couple of turns on primary and secondary of  a high-mu core  if you don't. Even if it doesn't have sufficient low-end for your needs it will test the hypothesis. and is probably an easier way than first changing PS and LAN connections.

--

At 2031z I removed GPS.
Waiting to see.

John

Today at 1900z I changed the 5 v power supply. From a 20 amp 5.25 v AC powered to an inline 12/5.2 buck converter.
10M to 40 m noise dropped 2--4 dbm, 60m to 630 noise went up. 160 up by 10 dbm.

John

I am back to my normal RX configuration. T1-1, AM filter, FM filterr.
I will do the GPS disconnect test and touch Antenna ground while KIWI is terminated to a dummy load test, Monday, Sunday is our Thanksgiving dinner with the locals.
John
TI4JWC

a quick disconnect of the gps antenna will check if that is a source and the unit will remain operational for that short period.

Agreed and I hope the draft article shows why.   The MW stuff you are seeing is almost certainly coming in conducted.

My only arguments with W8JI is one of scale and bandwidth. He's correct about CM changing radically with that test but  his  implication  (maybe I'm unfairly reading that into what he wrote )  that mitigation over a broad (say 3-4 decade) bandwidth may be done  effectively compared to the need.  

I think most of us underestimate the distance between the level of the propagated noise floor, even in a matched half-wave antenna where it might be down near or below the Kiwi's floor at a quiet site, and the level of CM that can easily/typically be introduced by CM lines in and out of a Kiwi, through the ground and out the other side. 

At HF I measure on the order of 80 dB rejection by the Kiwi, which might sound like a lot but compared to, say, -154 dBm (20 dB over KTB and maybe a reasonable upper HF number) that means the CM better be causing less than -74 dBm equivalent injection.  If one considers the CM impedance pushing those currents to be a bit higher than 50 ohms, then this means they better be smaller than about 1 uA ! I bet for most of our (wired) installations this is a fairly tall order.  For antenna systems with lower levels these numbers only get worse.

I think what you have is a big CM source coming from LAN or PS (probably not GPS since it is shorter and smaller self-capacitance to the world) and going out through the SMA(s) through some path.

What might help you a lot is a low inter-winding  capacitance 1:1 transformer,maybe just a Mini-Circuits T1-1, isolating the grounds of the SMAs for common mode. It's not perfect, still 1 pF but I bet that and care that there's no return path through the GPS SMA return will likely change things. 

As W8JI suggests, you can possibly see the effect by terminating the Kiwi with 50 ohms and then just touching the antenna system ground to the SMA ground at the termination to get a more representative comparison. You may see a huge jump in noise when you do that which would be confirmatory.

Glenn, I just read this from an article by W8JI.
"

Measuring Common-Mode Noise

We sometimes hear we can test or evaluate a system for unwanted noise or signal ingress by disconnecting and replacing the antenna with a dummy load. This idea actually has no theoretical foundation at all. Replacing an antenna with a small load significantly alters common mode impedance of the system, and removes the ingress point (the antenna's feedpoint) entirely. Dummy load substitution significantly changes system common-mode impedance. 

The only real test would come from a dummy load with the same connections and impedances (both differential and common mode) as the actual antenna. In other words the test load has to be the actual antenna to keep feed line common mode ingress the same. Obviously, casual dummy load substitution is a useless test!

The best approach is to use preventative measures in initial system design and installation. Quite often the cost of being safe is less than a few percent of the initial system expense.

New KIWI  jpg attached.
John

Glenn, you are correct. the Low reading of -155 on 11-26th was a 50-ohm termination at the KIWI connector, no cable, no antenna.
At this moment the noise recordings are to a 50-ohm load at the arrestor at the grounded base of the antenna, antenna disconnect, cable only.
Looks like I have significant noise ingress in the low bands. I will do a new KIWI screen-shot after I finish my Saturday chores; wash the dog and dog beds.

John
TI4JWC

It's a v0 draft I'll try to clean it up as I have time and ideas

https://groups.io/g/wsprdaemon/message/562

I still don't know for sure about absolute levels,  a repeat Kiwi plot showing console would let me be more confident. Let me just mention

• The spectrum display is always 1024 bins so in full band this is about a 30 kHz bin size. The 16 equivalent 16 bit channel depth means things go bad at low levels around large signals. What looks like a phase noise pedestal will show up due to the DSP.  This display is really best for an overview, not for quick spectral density measurements. Also without a detector telling you where the RMS floor is you are at the mercy of your eye.  As mentioned, the waterfall is not as valuable though can sometimes show clues as to the nature of the source. Helps if you have the console on to see the speed and time scale.
• Better, for a single frequency is to tune in AM or IQ with 10 kHz bandwidth to an identified no-signal spot and use the S-meter. You want to be sure that you are always getting >-127 reading so the wide BW helps this. You can be narrower if you adjust accordingly and perhaps miss signals that will skew the reading.  If you tune this way, you'll need to subtract 40 dB for the 10 kHz from whatever you read. 
  

Between these two you can get both an overview and a good single frequency measurement.

Grafana is showing a Kiwi system noise floor of around -157 which means you aren't actually observing noise characteristics due to everything else. There may be more steps and they may have larger delta.  Something doesn't add up here for a Kiwi hooked to almost any antenna in a typical environment. At MW and low HF the external noise from a quiet site to a matched antenna should almost always be above the Kiwi floor.  If yours isn't either you have bad match or loss of some kind.  

Glenn, what is the title of your post with the draft article?

John