The Pi4 can be tough to make quiet. Pi 4’s don’t have grounded standoff holes so the best way I’ve found is to ground it to a shielded metal enclosure via the USB and ethernet ports.
But today I have tried with Pi Zero and RTL-SDR. I am sorry to say that this setup has not worked for me. The noise is about 3 dB worse now.
I have put copper tape all over the Pi zero plastic box, and connected it with the shield of the USB and HDMI ports.
Before and after that I used rtl-tcp tuned to 433.920 MHz, with the same RF gain. I could see about -63 dB noise floor before the copper tape, and -60 dB after it.
I have used the same rtl-sdr.com v3 dongle, rtl-sdr.com FM notch filter and the same antenna in the same place.
I only noticed that a strong peak in the panadapter seems to have gone away, if I remember it all right.
After the copper tape I have used a ham radio talkie all over the room to see where the noise comes from. It seems not to come from the Rpi zero.
With the outdoor antenna and the PC, however, I cannot receive the noise. It will be better when I will be able to know exactly when there is noise, and when there is not.
I suppose that only part of the noise came from the pi. It could explain why the strong peak disappeared. I am not sure now, but I guess that the problem is complex and need further investigation.
Yes, of course. Thank you again for your interest. I will give you a better answer as soon as possible.
First I want to plan the test a little, so I do not give you a false clue. I should have taken some photos of the waterfall before and after putting the copper tape all over the pi box, but I have not. Sorry about it.
Perhaps I was wrong and the sample rate has no effect.
Now I will connect a HackRF to the laptop, to see any unintentional radiation not directly propagated just by the OTG USB wire. Please note that at a certain time I will close one of the two instances of GQRX, the one that reads from the Pi Zero. You will see how the waterfall gets cleaner when I disconnect from rtl_tcp.
The cause could be the wifi activity, or more probably the USB activity from the Pi and the RTL.
I believe that the frequency band, the noise strength and the unusual spreaded noise peaks are the same I had observed with Pi 4 connected to Kerberos SDR, so I suspect that the problem is the same and the treatment should also be the same. Please let us suppose that for now, since it is easier to test with the Pi Zero, the RTL and a conventional operating system. I also will be able to test with Kerberos and Pi4, but I suppose it would be more difficult.
I believe that ferrites would not be of any help, because the frequency band involved (UHF).
Today I have seen that the noise floor of the RTL is about -80 dB. I have also tested the MagicSDR android app. In the same situation, the noise floor raises to -60 dB with the app, and the noise peaks also raise as well. I suppose this phenomenon is related to some bug in the android app. In this moment it is not important for me. I can study it later.
Tire pressure sensors in some model cars pass psi info to the dashboard at 433MHz, this could be a source for this noise. it’s just my guess thinking about it.
I have to say same thing as Carl. the antenna should not be so close. All electronic devices gives away some kind of RF energy. In Europe its ETSI standards that put the limits of RF. In my work as EMC engineer, i messure in Anechoic Chamber usually on 3 meter distance. And when i see your pictures, have you tried to move antennas away from the computer and RSP PI ?
I am sorry to say that I disagree with all of you. As I said in my previous post, the waterfall gets much cleaner when I disconnect from rtl_tcp. Please refer to the previous post and the attached photographs.
You can see that in some photos there are two instances of gqrx: one for rtl_tcp (from the pi) and another one from the hackrf.
In some other photo there is only one instance: the one for the hackrf. You can check it in the window titles to know which is each one.
In that photo (the one with the hackrf only) you can see how the waterfall has cleaned just a few seconds before, in the middle of the waterfall. This is the moment when I closed the rtl_tcp connection.
So most of the noise is somehow related with the rtl_tcp connection, and the RTL activity. When the RTL is idle there is no noise, or much less noise. When the RTL is busy the noise starts again.
Please note that the laptop screen is still running by the hackrf antenna, at the same distance, when the waterfall gets cleaner.
It is also impossible to explain why the supposed tire pressure monitoring system disappears just when I disconnect from rtl_tcp.
It is true, however, that not all the noise disappears when I disconnect from rtl_tcp. It looks like there is several noises from several different sources. Some noise could certainly be caused by the laptop screen. This noise would not be important, because it will disappear when I will go pedestrian mobile with the KerberosSDR.
However, it could be interesting to further investigate if my laptop is noisy. I will try ASAP.
This is the photo again. Unfortunately, the antenna from the hackrf cannot be seen just by the laptop, but the hackrf cable is short, and certainly the hackrf could not go too far
I still have not fully diagnosed the problem. I suppose there are several noises that add one another. But I have found and replaced a bad USB data cable and it got a few dB of improvement.
The bad cable worked as an antenna and radiated RF, which was recieved by the antenna array, if I properly understood.
I suppose I need a good couple of ferrites for my new cable to achieve a further improvement, probably.
I also use my KerberosSDR with full RF gain (49.6 dB). I can see in the demo videos that usually people set it at a lower level, about only 30 dB. With RTL blog v. 3 dongle, 49.6 dB gain seems to work very well for me, the noise can grow up a little but SNR is better that way. I use the blog’s FM notch filter to avoid saturation of the dongle. I suppose it is the same with Kerberos, isn’t it?
Besides, I have read somewhere that the noise figure of Kerberos SDR is not so good as the noise figure of the v. 3 dongle, which seems to be quite good. Perhaps I am expecting too much from Kerberos? Is there any knowledge about the usual kerberos/kraken noise floor? How can I check if my own noise is near the usual level?
it’s just my guess thinking about it.