Stratux making more EMI than it used to

[Ergonomicmike]

1. Stratux version: 1.2r1-0360e

2. Stratux config:

   SDR

   • [ ] single
   • [X ] dual

   GPS

   • [X ] yes
   • [ ] no type: VK-162

   AHRS

   • [ ] yes
   • [X] no

   power source: Anker E5

   usb cable: Anker

3. EFB app and version: iFly v9.8.7

   EFB platform: Andrioid KK

   EFB hardware: Asus Zenpad

4. Description of your issue:

This isn't a code issue, per se. But something changed in the Stratux code that is making more RFI now. (I'm gonna guess it's the faster psuedo-AHRS.)

About a month ago, the squelches on both our COMMs (King KX-165 and KX-155) started "breaking." (That is, started letting noise thru.) We noticed it on Tower (118.4 MHz), Ground Freq (121.8 MHz), Guard (121.5 MHz) and Unicom (122.8). Whether it does it in the 130's, I don't know.

Tonight I had the brainstorm to reboot the Stratux in flight. As soon as I hit the reboot command, the noise in the COMM stopped! When the Stratux came back on line, so did the noise.

This started in early February. So that would be coincident with the Feb 3 sh updates.

This is a Pi 3. No copper foil shielding like I had on my old Pi 2. Aside from the software, nothing has changed in my config. The Stratux is where it's always been. The antennas and SDR's have been where they've always been, etc.

I don't think that the developers can really code around EMI. More like the Project might start thinking about shielding the box if others report similar problems with their COMMs.

[cyoung]

I do think this may have to do with CPU load.

https://github.com/cyoung/stratux/blob/master/docs/RPi3/RA-029073-45-00B%20-%20part1.pdf page. 22

A number of users have reported similar issues and it seems to be worse (according to reports) at ~125.0 MHz.

We were never able to figure out the source exactly, but it may be the time to start working on this.

[Ergonomicmike]

Interesting. I quickly skimmed the report and I see the spike at COMM freqs. (Correlates with what I happened on.) But I didn't see what they had the Pi doing for the test. Just powered up? Running Jessie? Calculating Primes?

If it's the CPU radiating, then we ought to be able to shield some of this down. I've got copper foil at the hangar. But will be a least a week until I'm able to get out there again.

Actually, this sounds like something right up @Axtel4's alley.

[Nokomis449]

Probably unrelated but I was getting similar noise on certain comm freqs, and finally narrowed it down to the 12v power cable leading to my iFly 740. A trip to Radio Shack for a small choke fixed the problem. They're cheap enough; might be worth a try.

[Ergonomicmike]

I've got two simultaneous threads going on this issue (one here and one in Slack). In the Slack thread, I mentioned that, since I'm running my Stratux on an Anker, this particular problem is radiated. BUT, that doesn't mean that the Stratux isn't also conducting RF into ship's power for those who are hooked up.

Did you tell iFly about the conducted EMI from the 740? They should have chokes internally on the power.

Years ago I had to put a choke (also from Radio Shack) on an HID power supply in the Glasair.

Since then I thought Radio Shack declared bankruptcy. Are there still stores around?

[Nokomis449]

It was AP that told me to try a choke. Interference occurred both with ship's power and 12v portable battery. Radio Shack filed for bankruptcy and closed a bunch of stores, but a few remain(ed) open. It's been a few months since I've been to one, but the shelves were was pretty sad.

[Ergonomicmike]

Interesting that a choke on ship's power fixed the problem even when you had the problem with battery power. Might point to what Chris found today, that moving his power wires around changed the squelch break. So that might be telling us that the power cable is acting as an antenna.

I've got a clip-on choke from a computer cable I will put in my flight bag to test next time I'm at the hangar. Would be an easy fix that everyone could do. (Easier than soldering on a copper shield.)

[cyoung]

If anyone else is experiencing this, give this config.txt a try (replace /boot/config.txt with the file, reboot). I'll be testing this further this week.

config.txt

[cyoung]

By the way, you can access the "boot" partition on a Windows/Mac/Linux and modify the config.txt file that way.

[Ergonomicmike]

Yeah, I remember. I leave the Stratux in the plane nowadays, so have to SSH in there. I especially don't want to move anything until I've got a better handle on the RFI. (Update: I thought Chris was reminding me that I can SSH from Windows. After talking with Chris via Slack, he's telling me I can modify the sd card directly from Windows. I haven't tried opening the sd card from Windows because I always get a warning from Windows that the card is corrupt. But might try it next time.)

[Ergonomicmike]

I did some testing in the hangar today with Comm 2 (King KX-155) on 125.0 MHz. I made a video.

You can hear the squelch breaks in the video as I do some "hands on" testing.

There is a difference in the tone of the radiated RFI for the two different config.txt's. But the new config.txt does not solve the problem. The problem is radiated RFI, and it seems to be coming mostly from the controller chip. (Followed by the CPU and then something by the SD card.)

Here's the video: https://vid.me/gpZw

Jump to 3:20 for the final solution.

I didn't get a chance to fly it with the shielding today. But given how quite it is now, I'll be surprised if I have any more problems.

Note that this was with an Anker battery. (As opposed to ship's power.) I tried a clip-on choke first on the power lead. No difference that I noticed. (I'm gonna guess that, being battery based, the impedance looking into the Anker is very low.) Someone who uses ship's power should check for conducted EMI after they shield their Stratux.

[cyoung]

In my testing with the new config.txt, the 125 MHz EMI was still present but there were less emissions at other COMM freqs. See the peaks around the left and right of 125 MHz here, these were eliminated on my setup. They appeared as the CPU frequency changed from 600 MHz to 1.2 GHz depending on CPU load. Raspbian is configured by default to use only these two frequency steps.

As noted on Slack, your test raises the question as to how the aluminum heat sinks that some users put on the LAN chip change the EMI situation.

[cyoung]

Here is a summary of the config.txt used over all releases. Most notable was the switch back to a floating CPU frequency with v0.8r2 (Jessie).

I suspect that this is related to #451.

File	Date	Comment	Config	Config Ident	CPU
stratux-pre-8b4fea41f2-08252015.img.config.txt	8/25/15	Wheezy. RPi2 only	7dfb47e91e8a4b507f3f4921ec9c97a0	A	floating
stratux-pre-a408a8cf13-08152015.img.config.txt	8/15/15	Wheezy. RPi2 only	7dfb47e91e8a4b507f3f4921ec9c97a0	A	floating
stratux-pre-e17c0a5302-08312015.img.config.txt	8/31/15	Wheezy. RPi2 only	8aaf2de4fcf2c7f43bebc7a798663cfd	B	floating
stratux-v.0.1-08072015.img.config.txt	8/7/15	Wheezy. RPi2 only	7dfb47e91e8a4b507f3f4921ec9c97a0	A	floating
stratux-v0.2r1-f47a224c1d-09042015.img.config.txt	9/4/15	Wheezy. RPi2 only	8aaf2de4fcf2c7f43bebc7a798663cfd	B	floating
stratux-v0.3b1-7e61c7556a-09112015.img.config.txt	9/11/15	Wheezy. RPi2 only	8aaf2de4fcf2c7f43bebc7a798663cfd	B	floating
stratux-v0.3b2-5a568e27ce-09162015.img.config.txt	9/16/15	Wheezy. RPi2 only	8aaf2de4fcf2c7f43bebc7a798663cfd	B	floating
stratux-v0.3b3-c3e6d41ebe-09192015.img.config.txt	9/19/15	Wheezy. RPi2 only	8aaf2de4fcf2c7f43bebc7a798663cfd	B	floating
stratux-v0.3b4-75f60562af-09242015.img.config.txt	9/24/15	Wheezy. RPi2 only	773ea70f6708ceeff741ab812092621f	C	floating
stratux-v0.4r1-34f55beb04-10052015.img.config.txt	10/5/15	Wheezy. RPi2 only	773ea70f6708ceeff741ab812092621f	C	floating
stratux-v0.4r2-a2a7b11494-10072015.img.config.txt	10/7/15	Wheezy. RPi2 only	773ea70f6708ceeff741ab812092621f	C	floating
stratux-v0.4r3-6f5f69b-10202015.img.config.txt	10/20/15	Wheezy. RPi2 only	773ea70f6708ceeff741ab812092621f	C	floating
stratux-v0.4r4-293d0e74a8-10222015.img.config.txt	10/22/15	Wheezy. RPi2 only	773ea70f6708ceeff741ab812092621f	C	floating
stratux-v0.5b1-ab889fb116-11212015.img.config.txt	11/21/15	Wheezy. RPi2 only	773ea70f6708ceeff741ab812092621f	C	floating
stratux-v0.5b2-e6ad9aa830-12212015.img.config.txt	12/21/15	Wheezy. RPi2 only	773ea70f6708ceeff741ab812092621f	C	floating
stratux-v0.5b3-680284ee1e.img.config.txt	12/27/15	Wheezy. RPi2 only	773ea70f6708ceeff741ab812092621f	C	floating
stratux-v0.5b4-d682bb5cf0.img.config.txt	1/1/16	Wheezy. RPi2 only	773ea70f6708ceeff741ab812092621f	C	floating
stratux-v0.5b5-cfd1541708.img.config.txt	1/7/16	Wheezy. RPi2 only	773ea70f6708ceeff741ab812092621f	C	floating
stratux-v0.5b6-1be927e672.img.config.txt	1/26/16	Wheezy. RPi2 only	1d76ea8d7359baf024a9a0fd951177b8	D	900MHz
stratux-v0.6r1-bbbcd3bbe1.img.config.txt	1/29/16	Wheezy. RPi2 only	1d76ea8d7359baf024a9a0fd951177b8	D	900MHz
stratux-v0.7b1-02bf2251d8.img.config.txt	2/15/16	Wheezy. RPi2 only	1d76ea8d7359baf024a9a0fd951177b8	D	900MHz
stratux-v0.8r1-796d0a2632.img.config.txt	3/7/16	Wheezy. RPi2 only	1d76ea8d7359baf024a9a0fd951177b8	D	900MHz
stratux-v0.8r2-db130aab76.img.config.txt	3/21/16	Jessie. RPi2+3	7b1933cfa3adf9b9dfe68afebfd97476	E	floating
stratux-v0.9b1-89107818b8.img.config.txt	5/27/16	Jessie. RPi2+4	d26f0c1655c39904674ed33f95975982	F	floating
stratux-v0.9b2-64710fe60d.img.config.txt	6/21/16	Jessie. RPi2+5	d26f0c1655c39904674ed33f95975982	F	floating
stratux-v1.0r1-945c8bf6bc.img.config.txt	7/13/16	Jessie. RPi2+6	d26f0c1655c39904674ed33f95975982	F	floating
stratux-v1.1b1-0107b59010.img.config.txt	12/12/16	Jessie. RPi2+7	794ca6d7ac129bbe22dd7d25075fd4de	G	floating
stratux-v1.2r1-9ee46170ff.img.config.txt	1/29/17	Jessie. RPi2+8	794ca6d7ac129bbe22dd7d25075fd4de	G	floating

[merricke]

My solutions for EMI and power success:

First of all, thanks Chris and other contributors for the Stratux project. I’m one of the happy beneficiaries. I’m a newby, not sure if any of this is new to the group.

Conclusion: EMI was brought under control by grounding the RPi heat sinks and also adding a ‘rat’s nest’ grounded random wire bundle over the Pi board. Power supply is with a dual port usb car charger, Aukey CC-S7, and powered hub, Anker Ultra Slim USB hub w/USB charger. The Pi goes to one port of the charger while the hub gets power from the other port.

Story: My objective is to not use a battery supply. Adequate power for the Pi with the car chargers was a problem with peripherals attached (VK-62 and one SDR). The Pi was running at the low end (red light blinking) using various power sources. I tried:

1. Cheap dual port car usb charger, DC/DC converter but inefficient and put noise on Pi carrier
2. Cheap adjustable DC/DC converter board. Provided good power but very noisy.
3. Made a linear supply. Clean but hot.
4. Anker car charger. Clean, Aukey slightly higher output under load.
5. Aukey car charger. Clean, power adequate but marginal, about 4.7 v on Pi bus.
6. Added Anker powered usb hub for peripherals. 4.9 v on Pi bus.

Occasionally I suffered SD card corruption during shutdown with the marginal PS’s, so I had to re-image the card a few times. Haven’t had that problem lately.

EMI problems were from two sources. One is the 125.0 carrier from the Pi. Other is information being impressed on the Pi’s carrier by noisy power supplies. With enough modulation and a strong carrier, all communications were being impacted by Stratux. Using a clean PS eliminated the transmitted noise information, and grounding the Pi’s heat sinks brought the Pi’s carrier power down enough to avoid spurious breaks of com squelch.

My testing on the bench was crude, just used a handheld radio. The results provided enough contrast that I didn’t bother putting any components on a scope.

Discovered by accident that a 3’ piece of loose signal wire attached to the ethernet can and the cpu heat sink also attenuated the Pi carrier. Tried wrapping board in metal screen and using longer wire but no gain was noticed over just jumbling the 3’ wire on top of the Pi.

I also observed noise from the SDR on 120.0, however, it wasn’t strong enough to trip my squelch. I don’t know if that would interfere with a 120.0 com channel.

My setup is old, analog King KY-197 com’s, RPi3, VK-62 GPS, and thumb drive size SDR (from Vilros, summer of 2016). My screen is a Windows laptop running the Ifly app.

[Ergonomicmike]

3 foot or 3 inch? Got any photos?

[merricke]

3 feet. I’ll try to remember to get some pix Wed. Can you put pix on Github or do you put them on another site?

[Ergonomicmike]

I've left images before. I think you just click on "Attach files" below.

[merricke]

Pix of the grounded heat sinks & jumbled wire (emi shield). I used solid hookup wire which is 'haywire twisted' to one heat sink and deformed to make sufficient mechanical contact with the other heat sink. The gray wire is 28ga stranded wire.

[Ergonomicmike]

Interesting. I expect that the grounded shields are coupling the EMI from the chips to ground. (Earlier, in a video somewhere, I had put my finger over the chips (without heat sinks) and found that my finger radiated away energy, making more noise.) If so, then it might be that short braid would also do the trick too.

[merricke]

Depending on where you're doing your tests, your body may be adding 60 Hz noise to the system. Try using clip leads to ground the heat shields to the USB or Ethernet cans to see if you get similar results to mine. I think it's a 2 step process: -reduce the Pi's 125.0 carrier strength with EMI shielding (grounding the heat sinks & the 'wire cage') -reduce modulated information (below 8 KHz for my KY197) on the carrier that opens the squelch circuit by using a clean power supply.

I'm assuming battery supplies also use DC/DC converters so they may not necessarily be noise free. I didn't test with any battery supplies. And adding your own filters (capacitors or chokes) my alter the output characteristics of the converter instead of smoothing the current.

[GitUser275]

As a new Stratux owner I've been browsing some of the old discussions. I found this one interesting so I ran an EMI current probe over my unit. I found a signal at 125.01 MHz with the probe centered over the micro SD card (marker 1). There were also signals at 115.25 MHz and 120.08 MHz when the probe was over the NooElec RTL receivers (markers 2 and 3). I don't know yet if these signals will cause any problems with my aircraft radios. I didn't find any other signals as strong as these.

In the images the yellow trace is with Stratux powered down. The magenta trace is with the current probe over the micro SD card. The cyan trace is with the current probe above the NooElec RTLs.

[merricke]

That's good info, and it concurs with my crude tests just using a handheld radio. I'm wondering, at 120 I was getting a high pitch squeal, maybe from the 2 carriers (or is the SDR active during the RPi test?)?

[Ergonomicmike]

Very cool @GitUser275 ! The 115.25 MHz explains the noise I'm getting on 116.0 MHz. Next time I do a video (below) I'll go down to 115.25.

Per your findings, I also get noise on 121.8 MHz, not to mention 125.0.

What frequency was your SDR set to? Per the video below (and others I've done), for me it's the 1090 SDR that makes most of the non-125 MHz noise.

Can you also please scan into the 300 to 400 MHz? We're beginning to wonder about GS frequencies. And since there's no way to listen to what's being received on the GS, it would be good to know what's there at fundamental frequencies. Also, please scan from 130 up to 300 - might be something in the 160's who's harmonic could interfere with GS.

In the meantime, I found that some of this noise is being conducted into the power cord and then radiated out. See my video at https://vid.me/ADfa2

I'm in the process of putting braid over the power cord and also the USB cable for the 1090 SDR.

Would appreciate any thoughts on ways to suppress this noise after seeing my tests in the video.

(Updated for clarity.)

[GitUser275]

For the spectrum traces shown above both SDR were active (RX 978 and RX 1090). I tried unplugging one to see which one was causing which frequency but I don't have a capture. I'll run some more tests and report back.

Snap on ferrite beads can be quite effective at suppressing spurious VHF signals on power and signal cables. Put them as close to the source as possible.

Are you at DVT? I recognized the frequencies in the first post of this thread.

[Ergonomicmike]

@GitUser275 Yes, DVT. Per the video, tried a snap on bead. I couldn't tell a difference.(Maybe I don't have the right ones for the freqs we're talking about. @cyoung reports success with his.

[Ergonomicmike]

Also, doing testing in a Glasair. Not having a metal shell for any shielding, ours is the Worst Case airplane for EMI problems.

[GitUser275]

DVT, Mike, and a Glassair? Maybe I know you. J3 at Sunburst and flight review 6/19/2007??

Andy

[Ergonomicmike]

That's me. (Pretty scary, huh?) I keep track of time nowadays by the Glasair. So, we finished it 1995. I thought that Sunburst had been forced to shut down some time back then. (As opposed to only 10 years ago in 2007.) I let my CFII expire after we started flying the Glasair. (But I remember the Cub and how scary it was checking people out in it with those heel brakes.)

[GitUser275]

it's a long time ago but you checked me out to solo in Jim's J3 at Sunburst and much later in 2007, long after Sunburst had closed, I gave you a flight review in your bird. If you think heel brakes were bad - I think I went grey instructing in the right seat of an Aeronca Chief. No brakes on that side.

It's a small world. Good to hook up again in a new area of interest.

[Ergonomicmike]

Ahhh. So Andy D? (of Glider fame?)

[GitUser275]

Yes, that's me. Moving back on topic I made the requested sweeps using the EMI current probe.

For 300-400 MHz I only found very weak signals near the micro SD card slot:

For 130-300 MHz the significant signals were in the vicinity of the two RTL SDR:

In both cases the yellow trace is a baseline with Stratux unpowered. These traces use power average sampling with 100 sample average. Note that the amplitude scale is not the same as for the earlier traces.

[Ergonomicmike]

Are you doing this at work? Or do you have your own analyzer? If the latter, and if you're still around DVT, maybe we can sniff out my unit in the plane?

[GitUser275]

My retirement hobby was to renew my interest in amateur radio and the spectrum analyzer was bought for that. Sure, we could meet some time at your hangar. Is your aol email still valid?

[Ergonomicmike]

Yep.

I sold my old Heathkit HF equipment at a yard sale for $100 a few years ago. (Obsolete I guess, since they didn't have the new HF bands that came in later.)

[merricke]

GU275, can you run the scans with some test leads grounding the heat sinks to the board ground? I'm wondering if you'll get more definitive results than I did at 125.0.

[GitUser275]

As I said earlier - " I found a signal at 125.01 MHz with the probe centered over the micro SD card (marker 1). " If I move my 1.5 inch diameter probe away from the card holder the 125.01 MHz signal is very quickly attenuated. I detected no significant signals with the probe centered over either heat sink.

Later I'll see if the 125.01 peak is influenced by grounding the heat sinks but it may be tricky to get good contact. They appear to be anodized aluminium which is not conductive unless the anodizing is removed.

The probe I'm using is a home brew 1.5 inch diameter shielded magnetic field probe.

[GitUser275]

I looked again with the probe close to each heat sink and I find no strong signals there. There is a strong 250 MHz signal with the probe centered over the RAM chip which is other side of the board. Here are a couple more screen captures.

Centered on RAM: (marker table not captured)

Centered on micro SD card holder:

[cyoung]

@GitUser275 - regarding the 115.25 MHz and 120.08 MHz signals, do these vary with input voltage (to the Pi)?

[merricke]

@GitUser275 , I'm wondering

• the area of noise around 103 MHz on both sides of the board
• the periodic spikes about every 42 MHz
• would the signal at 125 over the SD card drop off if the heat sinks are grounded (mine aren't alodined and are easy to connect to with test leads). My way of checking signal strength was to check squelch requirements at differing distances with my handheld radio (done at home but improvement there matched improvement in the plane).

[Ergonomicmike]

While flying yesterday, ran thru the NAV channels.

This is not an exhaustive list, but to my surprise, found noise on:

108.00 (Some kind of Intermod. Could hear talking. Might not be Stratux.) 108.55 109.25 109.40 110.00 110.15 110.70 111.15 111.95 112.00 112.05 112.85 112.90 112.95 113.75 114.60 114.65 114.70 115.20 (Solid, like a carrier. None of that repetitive noise associated with the SDR.) 115.60 116.40 116.45 117.30 117.35

[GitUser275]

I made a much smaller probe. This one has a loop diameter about 0.5 inch. Using it I was able to explore the processor side of the board. I found the strongest signals in the area equidistant between each heat sink and the header, an area with no large devices. I'll do more more later in the week.

[Nokomis449]

It'd be nice if someone could line their case with copper tape and see if it reduces the noise. https://www.amazon.com/dp/B07663WM7D/ref=sspa_dk_detail_3?psc=1&pd_rd_i=B07663WM7D&pd_rd_wg=EpeiX&pd_rd_r=BX062CXSD8E3WZ0NK9M6&pd_rd_w=az1aW

[GitUser275]

Re - voltage influence

What range of voltage would be of interest? I'm running off a bench supply but via a small adjustable regulator board that has been carefully tweaked to 5.25 V. I did that so I can't overvolt the Pi.

[cyoung]

4.65V - 5.25V or thereabouts.

[merricke]

My frequency notes, I'm only using 978 SDR, not 1090 108.35- static, localizer 115.20- carrier, VOR, from SDR, matches with GU275 115.52- static 120.00- carrier w/high pitch noise, ATC, from SDR, matches w/GU275 125.00- carrier, ATC, RPi, matches w/GU275 129.60- carrier w/static, from SDR 129.87 to 129.95- static 131.12 to 131.15- static 131.60 to 131.70- static

[GitUser275]

I set out to see if the levels of the 115 and 120 MHz signal were voltage dependent but was distracted by something else.

The 115 and 120 Mhz signal pair is localized to the RTL SDR dongles. This pair is present when either RTL is plugged in and when both are plugged in. The pair first appears after power application just before the fan test runs. These levels are shown in the amber trace. 1 minute and 42 seconds after power application the level of the 115 MHz signal rises as shown in the magenta trace.

What causes this change in signal level?

[Axtel4]

There is a 90 sec delay for the R-Pi start up to the activation of the SDRs to give the GPS a chance to lock. Does the SRD levels return to the lower value when they are turned off in the WebUI?

[Axtel4]

@merricke Is that data with or without your "spaghetti" fix?

[GitUser275]

"Does the SRD levels return to the lower value when they are turned off in the WebUI?"

No, the levels stay constant at the high level when both 1090 and 978 are selected to off. However, if Stratux is re-booted with both selected off, the 115.2 MHz signal stays at the low level. The level rises at each SDR when that SDR is later selected to on.

[Axtel4]

@GitUser275 Interesting that it stays the same level when the SDR is turned off in the IU. I suspect the 115.2 MHz signal is the forth harmonic of the 28.8 MHz SDR master clock. The 125 MHz signal is the clock frequency of the USB/Ethernet chip on the Raspberry Pi.

[rhole]

It is possible that turning the SDR off disables only part of the device and the master oscillator may remain on. You could physically remove the SDR to find out. Also know that a good DC ground is not necessarily a good RF ground. I would provide a ground wire (grounding braid is best) 14g or better if there is no metal airframe to use. I’ve seen many noise issues with, for example, handi-talkies greatly improved with a good ground.

Rick Hole