Noise at roughly 60 KHz intervals
I'm seeing a BUNCH of noise signals from about 11MHz to 22 Mhz (?) at roughly 60 Khz intervals.
Is this a known electronic design issue? Is it possibly a signal from one of the computers in the vicinity? If so, is there a fix?
Thanks!
Bill
Is this a known electronic design issue? Is it possibly a signal from one of the computers in the vicinity? If so, is there a fix?
Thanks!
Bill
Comments
These seem to have appeared suddenly, well, I don't think I had them a couple of weeks ago and here they are.
Got to move it out to the shop!
Best,
Bill
More
30/60KHz spurious signals.
I
have a KiwiSDR with BBG that was purchased as a set. It is supplied from a 5V
3.5A linear psu.
After sorting out the interfacing to the network, it worked.
I have a choice of two antennae, a 12m longwire, or a PA0DRT copy active
mini-whip. There were a few unwanted signals to clear up with ferrite rings but
as I go down towards the noise floor there are spurious signals that DON'T go
away with application of ferrites or screening.
Also
I can see similar unwanted signals on a selection of already wired and active KiwiSDR
web receivers.
I
chose one of the stronger spurii showing on my RX, at 23127KHz +/-3-500Hz, as a
working point, and I
tried and tried to reduce its level in the receiver, the spurs are about 30dB
above the noise floor.
Very similar spurii can be seen on several other KiwiSDR RXs accessed via the SDR.hu
list and I can post the screen dumps when Ive worked how to do that.
For
the moment I have given up as I can't get rid of these signals.
They are even still there, at reduced levels, when I have removed the antenna
and all my peripheral circuitry, leaving just the power cable (with large
ferrite ring), and the CAT6 cable (with FT240-31 rings at either end) with the
antenna input shorted out.
In fact I surmise that the unwanted signals come directly from the IP network
communication.
Has
anyone else seen this?
And better still has anyone found a way around the situation/problem??
The
network switch is a Netgear Prosafe GB switch, model GS105 that is supplied
from a linear psu; and the connecting CAT5e cable is about 5m long.
I am trying to find someone to lend me an
alternative switch.
Peter
The latest version installed is 1v14.
The only solution (partial) I found is to keep the Ethernet cables as short as possible. That can be seen
on the welcome photo of my SDR.
Also no one I know has a spare switch for me to borrow, will I have to buy one??
And I see that someone else has also commented on a very similar situation.
See -Does someone have an idea about these birdies ? by F1JEK/P
here is an illustration showing that it is just my installation that has this phenomenon.
Attachments:
https://forum.kiwisdr.com/uploads/Uploader/43/db768d1376ffbdbb53955d457238c8.png
https://forum.kiwisdr.com/uploads/Uploader/0d/d9be363404b5dbcf4805ff7ea58882.png
Attachments:
https://forum.kiwisdr.com/uploads/Uploader/97/88b9d269f30e7d02c874d88baa7ef3.png
https://forum.kiwisdr.com/uploads/Uploader/18/c26473a3d6461713b3408fdc271f2a.png
https://forum.kiwisdr.com/uploads/Uploader/e0/22fb9c6b8b059eb84a9757da876b56.png
https://forum.kiwisdr.com/uploads/Uploader/8f/2e40d393d5147587270b21c8b7cba9.png
https://forum.kiwisdr.com/uploads/Uploader/94/b15d7ef56761ea36d7880592ddf2da.png
Thanks for the info and advice Martin, I'm slowly working through cabling permutations and lots of ferrite rings.
I've gotten another KiwiSDR with ser. no. 2050 for second installation and this is attached to a "Beagle Bone Green" with a label BBG217021961 stuck to it. So I can do a direct comparison after the tests with different ferrite toroids attached in various places are completed. So far no real progress has been made and I haven't found or borrowed a different switch either, new ones cost money, plus I'm a little fed up with getting nowhere fast.
Is there any data not yet published relating to the older BBGs obtained separately to a KiwiSDR so that we can check against the ser. no.s of the BBGs supplied in the sets.
WA2ZKD wanted to know if i'd tried a direct connection between the KiwiSDR and a PC.
Well, I didn't get networks comms to start at all so I wonder if this is only possible with a USB connection?
Or maybe if I had a 'crossed over' network cable, this connection might work. Does someone have any thoughts on this?
Initially just try to turn things off see if the source is in your control, if it is, good otherwise welcom to the modern world, I can tell what PC my neighbour is using from the QRM signature.
I find some cheap CCTV cameras are just terrible (looking at you ESCAM), broad frequency comb transmitters peaking around 20MHz, you'll need to see where the signal is getting in.
I reduced the direct local pickup problem by using a wireless bridge, ferrite rings on everything and a metal case, grounded.
The CCTV camera will still cause issues as that is a geniune RF signal being transmitted, reduced that mainly by CAT7 cables fibre bridge and grounding.
My PA0RDT benefited from grounding at the base of the feed (with a low pass filter) network cable as feed (CCTV baluns) and another ground just before the RX.
If your Kiwi is online? I'll see if it ties up with what I'm used to as local QRM here.
Stu
In the end I went back to using un-screened CAT 5 with 12 turns through a FT240-31 core (Mouser part number 623-2631803802) which was just as effective on the HF as it was when used with screened CAT 5 at frequencies around 20MHz, but didn't introduce the additional noise on VLF. I think the screen adds an additional ground loop, which the transformers in the Ethernet interface and external choke baluns can't fully isolate / filter because of the very low RF frequencies involved (just my opinion).
I also found that bonding the RF in and GPS connectors directly to the metal case at the point of entry reduced the HF noise floor by a further approx 3-6dB depending upon the exact cable configurations, but your mileage may vary.
Regards,
Martin - G8JNJ
Image below while fighting today's new raft of QRM.
PA0RDT on the bottom, LZIAQ with bike rims top, without the ability to subtract one loop from the other the anything below 150kHz is just about blown away. I was using bendy coax loops for a week or so but QRM so bad today needed the symmetrical metal loops to properly subract.
I see what you are saying about ground loops and shielding, I really only use the shielding on the CCTV networking as that sings, ferrite as well on a longer run but those rings are not cheap (to be honest I've not gone as far as 12 turns, normally six here).
Fibre or radio for the Kiwi as it's easier than working with ground loops.
I personally would wifi link the Kiwi first, so much easier to deal with injected noise if you can isolate the kiwi and antennas from the rest of the copper wiring.
The new FTP cable is a shielded, direct buried, CAT 6 type cable. The shield is not grounded on either end, so only capacitance to the ground it lays on to drain the shield. I should layout a conduit for it to keep that pesky cat and other varmints from chewing holes in it.
Ron - KA7U
Has made a very nice difference though, well done.
Ron - KA7U
Ben - SWLOI33
Ron
KA7U
There is probably a bit of imperfect balance on your loops making them somewhat susceptible to E-field conducted. The other possibility is that connecting the shield causes noise currents to flow between the receive and antenna ends: If this happens, all bets are off as these currents can/do get into everything! I'm somewhat surprised that the VCC and DC return connections don't cause a similar problem - unless you have significant choking inductance (say, a bifilar on a ferrite core) on these lines to decouple them at RF.
On something like that I would wind a couple dozen turns of that flat CAT5/6 cable (I use 6'/2meter jumpers) on a large-ish 77 or 75 mix toroid and have one of these at each end to insert several millihenries of inductuctance longitudinally down the line to quash such currents. In that case, a "DC" ground on the shield at each end ("past" the choke) should have little effect at those frequencies where the added ferrites' reactance is in effect.
73,
Clint KA7OEI
I agree with Clint.
The screen can sometimes be a curse as it introduces low frequency ground loops that may negate the balanced lines. You need to use type 77 or 75 in order to be effective at the lower frequency end of the spectrum, although type 31 would be OK if you are not interested in VLF, and type 43 would be best if you are only interested in say 3-30MHz.
In your circuit I'd also suggest adding a choking balun to the power and ground return and ideally a suitable decoupling capacitor between the two wires on each side of the balun in order to tie them to the same RF potential.
I'd also add a choke to the DC supply to minimise any noise from the supply on the non-grounded +ve line. Something like 12 turns through a BN73-202 core is quite good in that respect.
I've just been building some coax bias Tees and they have three inductors. The Bias Tee inductor itself, an inductor on the +ve DC supply rial and a choke on the incoming DC supply lines. It also has 2 lots of 2,000uF LF decoupling in addition to the normal 0.1uF & 10uF RF values. Because the DC input is carried on the coax you have to be very careful about both common mode and differential mode noise being injected directly into the RF path from the power supply.
If you use CAT5 or a balanced feed you need to be extra careful that you don't inadvertently introduce unbalanced paths that make it end up being the same as using unbalanced coax.
Regards,
Martin - G8JNJ
OK, I have been experimenting to find the best configuration. I terminated the receiver end of the CAT6 FPA cable into a shielded RJ45 plug and used a balun board with a shielded RJ45 jack. The interference level dropped to a low level as compared to a non shielded RJ45 plug plugged into a shielded RJ45 jack, but there remained a significant group of spurs around 18250KHz, peaking at about -98dBm on the kiwiSDR scale. I then used a balun board with a plastic RJ45 Jack and using he shielded RJ45 plug, the level of the spurs remained the same. Then I reterminated the CAT6 FPA cable into a plastic non-shielded RJ45 plug and connected that to the non-shielded RJ45 jack on the balun board. The level of the spurs around 18250KHz dropped to -108dBm, an improvement of -10dBm on the KiwiSDR. This is actually quite good as compared to the average KiwiSDR I've listened to. At the antenna end, the CAT6 FPA cable is terminated into a plastic non-shielded RJ45 plug and the amplifier connects with a non-shielded RJ45 jack. I read on the LZ1AQ web site that grounding the shield at the receiver and leaving the shield open at the antenna would give the best common mode rejection, but I have not tried it with a shielded connection on both ends here. My new amplifier boards are configured to accept the horizontally mounted shielded jacks and I'll try terminating both ends in a shielded jack with the next amplifier build. Until then, the non-shielded jacks and plugs on both ends of the feedline is providing the best results here. I do not understand why.
As to the DC ground configuration, you can check the schematic to see the amplifier power input circuits.
Ron - KA7U
The other thing to consider is that the capacitance of the shield to the inner conductors - probably in the area of 10-15pF/foot. What this means is that along a 100 foot length of this stuff, noise on the shield and/or longitudinally on the twisted conductors will have 1000-1500pF of coupling which, at 1.8 MHz, has a reactance of 88-59 ohms (respectively) which can do a pretty good job of coupling garbage on the line: If the balance of the transformers is anything other than perfect, you will hear the noise.
Many years ago I constructed a system that had galvanic isolation that used twinax and I needed to have extremely good longitudinal isolation. Constructing a Faraday-shielded transformer helped tremendously, having 60dB or so isolation, but the last little bit of isolation was aided by having a fairly high-value potentiometer (5k, I believe) across the input balanced line with the wiper to ground and adjust it for the best null - well over 80dB. Had I needed to go higher than just a few MHz I would have probably needed to include some nulling capacitors as well.
* * *
I see in the loop amplifier diagram that the V+ and ground leads are AC-decoupled with 470uH inductors (ideally around 5.3k at 1.8 MHz, or about 2.6k for the pair). At LF frequencies, these are probably a bit marginal - around 177 (88 for the pair) ohms at 60 kHz.
As far as grounding the shield goes, one typically tries all of the combinations and uses the one that works best. If grounding either end of the shield (but not both) makes a difference, I would kind of suspect that one or both of the transformers lacks sufficient longitudinal isolation, typically caused by mutual capacitive coupling between windings.
All of this indicates that the "shack" end of the cable is extremely noisy from switching power supplies, and connecting the shield at either end simply couples the noise on it, effectively bypassing part of the transformers' isolation, so whatever you can do to decouple the entire cable at RF (especially the "shack" end) will help (e.g. dozens of turns of this CAT-5 cable through a large, mix 75 or 77 toroid) . I would bet that if a large decoupling choke were put at each end of the cable, connecting the shield at either/both ends would make very little difference at RF.
73,
Clint
KA7OEI