RFI from a Raspberry Pi

[Razvan Peteanu]

I thought of sharing this: the amount of interference from a Raspberry Pi (I've been experimenting with receiving SIDs with an SDR and was looking at running GnuRadio on an RPi so that it can be installed at a remote location more easily), particularly at VLF frequencies.

Below is a screenshot from SDR# that was listening to NAA on 24kHz (via an Airspy + Spyverter).

It turned out that the noise generating by the RPi is significant and it overwhelms the signal (this is without a preamp). Interesting that it shifts in frequency; I don't have an explanation for it, maybe difference circuits were enabled/disabled depending on some background services running.

Shutting it down only left the noise from the power supply itself (with a slow frequency drift).

It finally went quiet when I plugged out the power supply from the wall, leaving the signal from NAA (fainter at 25.2kHz is NML). Remarkably, all this was capture on a Thinkpad just nearby, which doesn't generate this huge amount of RFI, even though it uses a switching power supply.

This makes me rethink the use of RPi (and also explains why I couldn't pick up NAA when looking at the input with GnuRadio); I'll build a Faraday cage and try again. This was an RPi3 with a 2.5A power supply.

Razvan

[Marko Cebokli]

To me it looks like it's your power supply. After PI switch on, it has a different load, and shifts frequency, but the spectrum looks qiute similar (the distance of the sidebands from main carrier).

Try another power supply.

 

Marko Cebokli

--
--
You received this message because you are subscribed to the Google
Groups "Society of Amateur Radio Astronomers" group.
To post to this group, send email to sara...@googlegroups.com
To unsubscribe from this group, send email to
sara-list-...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/sara-list?hl=en
---
You received this message because you are subscribed to the Google Groups "Society of Amateur Radio Astronomers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sara-list+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Marko Cebokli]

In fact, the sidebands are +-100Hz, with a weaker 50Hz line, a further very strong hint, that the problem is the PSU.

 

Marko Cebokli

[KD7JYK DM09]

: To me it looks like it's your power supply. After PI switch on, it has a

: different load, and shifts frequency, but the spectrum looks qiute similar
(the
: distance of the sidebands from main carrier).
: Try another power supply.

This is common to SMPSU (switching mode power supply unit). Such a supply
should never be considered around receivers or even sensitive electronics.
They are common and popular only due to cost of manufacture. Step one in
most any RFI situation is to ensure you only use linear (transformer based)
power supplies. A switching mode supply, typically those less than a couple
of hundred dollars are broadband RFI generators with a pair of nice antennas
on them (input/output wires). The expensive supplies also make noise, but
the additional cost is for filtering to minimize what you hear or place it
where it doesn't matter in a particular application. I had one supply that
generated so much noise that despite being underground and covered with a
steel and concrete building it eliminated all HF reception at 135' distance.

Linear supplies are usually in the twenty-five to fifty cent range (US),
cost wise at a second hand store. Get the spec of your noisy supply, for
example, 12VDC @ 2.5 Ampere, go to a "thrift shop", find a heavy iron core
linear supply with the same specification, pay for it with change found in
the parking lot on your way into the store, solve your PSU RFI problems
forever. Most linear supplies aren't well regulated, look for one that is,
or scrap one out of another piece of equipment such as a stereo or other
consumer product or use a hobby radio supply instead for a few dollars more.

For your RPi, see if someone makes an aluminum enclosure for it (save time,
effort and cost), put it on, use ferrite chokes on all wires to cut down RFI
that has a potential to affect equipment at many, many meters away to only a
few centimeters. Properly made cables with adequate grounding and shielding
also help as do ones with adequate "twist" in the wires. Trial and error
for those, swap out cables until noise is no longer a problem.

Kurt

[Mario Cannistrà]

I think Kurt made a very good list. For my project on 15-30 MHz for Jupiter it made a huge difference to place the R-PI3 in a aluminum case. Changing the power supply to a good linear one also proved fundamental.

My small additions to this list:

- I moved the SDR far from the R-PI using a USB cable but I had to replace that cable with one with better shielding. This one is thick, with good shield and does not emit so much RFI like the thinner previous one did.

- placed ferrite chokes also on this usb cable

- added an EMI filter inside the linear power supply since I had noise coming from the mains as well...

mario

[Razvan Peteanu]

Thanks Marko, Kurt and Mario, very good suggestions.

BTW, I've also had bad RFI at VLF from an active USB cable which, ironically, I was using so that the computer running SDR# was far away (15m) from the SDR. A nearby Thinkpad with a plain USB cable turned out to be more quiet.

[Keith Payea]

I’d like to add my two cents here.  Switch mode supplies can be made very quiet, but the ones you buy off ebay from Asia are made as cheaply as possible and often don’t meet any kind of EMC standards.  You get what you pay for.  This is why the SMPS inside the Thinkpad does not cause problems, but the inexpensive one with the RPi does.

 

Also linear supplies are not as quiet as you think they are, especially at ELF/VLF.  The rectifier diodes can generate many harmonics of the line frequency when they switch.

 

Cheers,

 

                Keith

[Bruce & MaryRose Randall]

That is correct about rectifier diodes.  For a switch mode supply that I designed for Mil 461 standard,  some extra EMI filtering was needed over a commercial design because of diode noise among other things.  461 test down below 10KHz because the military uses those frequencies.

 

The noise occurs because the diodes switch off in about 1 to 2 microseconds and current will be flowing in the reverse direction then suddenly stop when the rectifier diodes recover.  The sudden stopping of the current makes a spike with harmonics out to 500KHz or so.  Most of the trash is between 10KHz and 50KHz.  Fast recovery rectifiers that recover in 100nS or so reduce that noise a lot.  Fast recovery rectifiers are also more prone to failure from line surges.

 

In a linear supply, the transformer should help a bit on spikes, but mostly above 50KHz, so FCC 15 testing never sees it, but 30KHz SID receiver does.

 

A note on switchers:  Many change frequency with loading, so noise in the SID band may go down if a load resistor is added to the supply.  Blowing a couple of watts in a resistor may be a good tradeoff.

 

Also some low frequency noise is radiated by a magnetic field, so a closed copper band around the supply may help.  Switcher power transformers always leak some magnetic field.  Connect a copper band to ground as well for electrostatic shielding. 

 

I used some flea market cell phone chargers supplies at work for small 5V supplies.  The noise ran me crazy!  Supply had FCC 15 mark on case.  A quick lab test confirmed these did NOT pass part 15 conducted emissions.  Upon disassembling a supply, I found the EMI filter parts were left off of the circuit board!  Chinese cost reduction??

 

Sorry ramble is a bit long, I may be suffering from PTSD from the EMI wars of the last 35 years.

 

Bruce Randall  NT4RT

[KD7JYK DM09]

: My small additions to this list:

: - I moved the SDR far from the R-PI using a USB cable but I had to replace
: that cable with one with better shielding. This one is thick, with good
: shield and does not emit so much RFI like the thinner previous one did.
: - placed ferrite chokes also on this usb cable
: - added an EMI filter inside the linear power supply since I had noise
: coming from the mains as well...

Those are all very good! I do a lot of weak signal and
long-time-integration mode work. The power supply that eliminated HF radio
at 135', after some work is no longer detectable beyond 3" from the supply
or wires. It's a switching mode supply and it took a few days of playing
around, but it can be done. Unfortunately due to an oddball voltage, no
linear supply is/was readily available.

Kurt