Squarewave noise

#1

My RS has been running fine since February 2019 but recently may have developed a serious problem or so it appears. Swarm would show occasional short bursts of the problem then the problem would kick-in for maybe an hour or so. Right now it sees the problem for over 4 hours continuously. Actual detections are there but heavily masked by this interference.

The problem or noise seemed like it could be the power supply but I swapped in a new supply today and there was no change.

In then next week or so I plan to try another RPI and also temporarily relocate it to another part of the house so I can work on it. Also I am using a passive POE cable so when I move it :to the bench", I will not use that.

Maybe the RPI be the cause? (Trying another RPI should eliminate that possibility.)

Maybe the A/B board has an issue?

Could it be some other external device in the house? (Moving it temporarily to another part of the house may help eliminate that possibility.)

  • Mike

RSH.R150F.2019-08-13T20_11_55.logs.tar (3.7 MB)

#2

Hi Mike, welcome and thanks for your message. This definitely looks like a problem but it’s not one we’ve ever seen before. Your logs look fine, nothing stands out to me from them, however the wave makes me think that there is either a large external RF source nearby, or the data stream is somehow being corrupted prior to leaving the Shake.

I took a look at your data in SWARM and found it interesting to me that the problem is not constant, i.e. it happens every couple of minutes sometimes, and other times it happens every 2-3 seconds. It’s kind of a strange pattern.

We want to identify this problem because as I said we’ve never seen anything exactly like it before. That will require being methodical about eliminating possibilities. Let’s start with the simplest ones.

  1. RF noise — after looking at the data this would be my guess. Can you send us a photo of the installation? Try moving the Shake to elsewhere in your house to see if it changes the amplitude of the wave at all.
  2. Power — it does look like the station has power issues at certain points in time. Did you replace with a 2.5 A or greater supply? If so, is the device on a UPS or plugged directly into the wall?
  3. OS corruption — try burning a new image to the SD card to see if that changes anything. If the card itself is corrupted you may need to replace the card entirely.
  4. Pi issue (we think unlikely) — Replace the Pi as you suggested.

If 1-4 all fail we will reevaulate next steps.

Cheers and keep us up to date on how it goes.
Ian

#3

Thanks Ian. I’ve got other unrelated work going on here keeping me busy so I can only get to this as time permits.

  1. RF noise
  • Attached are the install photos.
  • I did tap off of the A/D board to get 5VDC for a temperature & humidity sensor (DHT22) as the pin on the header was no longer accessible. I can disconnect that to see if it affects the problem.
  • The idea of the box and POE was to make it easier to deploy at outdoor locations. I did get a super good deal on the box.
  1. Power
  • I had a Mean Well adjustable 5VDC/5A PS near the Dell 6224P (48V POE) switch. But the other day I swapped it out for another Mean Well PS (3A model) and immediately saw the problem return leading me to say the PS is not the issue.
  1. OS corruption
  • Yes, I thought something could be an issue there especially since I somewhat regretfully have done some semi-successful standard RPI updates which I believe is not recommended for RS builds. I can do the re-image when I get a chance.
  1. Pi issue - I agree it is unlikely to be an issue. It will be a pain to swap Pis but I can do this as the last item on this list.

The photos are posted below.

  • Mike

#4

I just made some changes:

  • Reloaded the SD as suggested and added no other SW.
  • Removed 3-wires from DH22 sensor
  • Connected CanaKit 2.5A power supply in place of passive POE
  • Ethernet cable is of course now connected directly between Dell 6226 switch and the RPI with no POE cables in between. (I indicated 6224P in my previous post but that was in error.) The metal box cover as well as weight are not currently in use.

So far (27 minutes) the helicorder looks good.

  • Mike
#5

take a closer look - it is still there just lower in intensity

Ken

2 Likes
#6

Mike, first of all very nice installation overall. However as Ken says the signal is still weakly there. How far did you move it in order to test? Can you try it somewhere other than at your house?

When you do reconnect the DHT22, you may want to use Pin 01 (3.3V) instead of the 5V connector on the Shake board. We are looking into whether hosting the DHT22 from the board would have unintended consequences. Read more about our GPIO usage here: https://manual.raspberryshake.org/specifications.html#gpio-pins

#7

Per Ken’s suggestion (in email) I put another ethernet switch (8 port gigabit TrendNet) between the RS and the big switch (Dell 6224) using a 2-3 foot cable at the RS end. Here are the before and after waveforms:

To answer your question about how far I moved the RS, I did not move it. I only made those several changes indicated in my post then set it back down on the basement floor. I can fairly easily bring it up the the 1st floor which is on the opposite side of the house.

  • Mike
1 Like
#8

Oh, on the DHT22 comment, my thinking at the time was to use 5 volts and not risk 3.3 per comments by the adafruit.com folks (“Sometime 3.3V power isn’t enough in which case try 5V power.”).

https://learn.adafruit.com/dht?view=all

  • Mike
#9

This looks much much better but I’m baffled as to why. Can you isolate the problem to the router, or perhaps an unshielded PoE connector? You said all PoE hardware between the router and the Pi was removed, or is the injector cable still the main internal case connector?

Absolutely, but with such short leads I can almost guarantee you’ll be okay! From my own experience with DHT22s, 5V is typically used to compensate for significant power drops over long leads.

#10

Good point on the short leads for the DHT22. Thanks!

I just moved the RS to the room on the 1st floor as a test. It’s noisy but the results may be better.

I will leave it there for the time being as a test but it will be picking up much more noise as I have a kitchen project and more that I am working on.

Mike

1 Like
#11

Are you using a Cat5 or Cat6 ethernet cable? I can’t understand why this would affect things but perhaps there’s some crosstalk occurring somewhere.

#12

In the basement it was using CAT 5e shielded but I think only 1 end had the shield terminated. There is a Dell server as well as two 24-port switches down there and maybe one is starting to generate noise. Maybe this is a good time for me to consider another ‘permanent’ location, even outside as I can get ethernet outside easily.

1 Like
#13

I had this sort of problem big-time when I connected my RS to a Ubiquiti F-POE-G2. It has a powerful Ethernet port that can drive gigabit speeds out to 100 meters. I used a one meter cable to the RS. I think the problem is that the RPI Ethernet transformers have a capacitor-coupled connection to ground (DC minus) from the ethernet transformers on the board. If these transformers saturate with signal, it will drive spikes into the local DC minus bus. If the geophone has any coupling to ground it will now generate an unwanted signal. I guess having a plexiglas box is a plus in that situation :wink:

This problem is part of the reason I put an ethernet isolator in line with my RB installation. A less costly “cure” might be to insert 25 or 50 feet of really cheap ethernet cable in line with the RPI (tightly coiled). That should knock down the signal level enough to avoid transformer saturation.

Edit: I should clarify that my noise looked different than Mike’s. I was very clearly related to times when the RS was being queried for data. No noises were generated when the PI responded with its rather puny output level …

2 Likes
#14

Okay, I quickly made that possible ‘cure’. I guess it was about 18 feet coiled on a 1-1/4" thin-wall stainless steel pipe. I assume it was a 25’ length and I had to leave enough to reach the passive POE injector near the switch. So except for this choke and stock OS, everything (HW, location) is back to original. Yes, it’s back in the basement at the original location.

  • Mike
1 Like
#15

Wow. That’s really interesting but it totally makes sense since the Pi has no true ground and has to discharge somewhere if the capacitors are overloaded. This will be going right into the manual. Thank you very much for sharing and clarifying this!

Data looks way better now @bronsaxitide. Thanks for your patience through this process. Let us know if anything else comes up.

#16

The root cause of this kind of interference saturation of the tiny Ethernet transformers from too much signal. This kind of thing does not matter much unless you have piggy-backed some analog circuitry onto the PI. And it’s more of an issue if you have a metal enclosure in proximity to the geophone. The cure is just to lower the Ethernet signal level and/or provide common mode (galvanic) isolation of the Ethernet signal using a device sometimes called a “medical Ethernet isolator”. Using a wireless LAN connection would also achieve this result, but may cause some other issues due to the variable latency of WLAN communications.

Ken

2 Likes
#17

Thanks Ian and Ken. It has been running 16 hours now and looking all good!

  • Mike
2 Likes