Wednesday, November 14, 2018

Tesla Powerwall RF sensitivity to RF transmissions - and how to deal with it.

This article is about my experience in causing RF interference to a powerwall, but I have written on two other articles about amateur radio operation along with having a Tesla Powerwall - these articles may be found here:
Figure 1:
Left to right:  Back-up Gateway (BUG), disconnect, two Powerwalls.
The cable raceway is just out of view, below the BUG.
 The BUG controls everything and was likely being bothered by RF
energy conducted to it via the Ethernet cable.  While it is possible to
connect a Powerwall via wireless, the reported experiences of other
Powerwall owners is that this method is less reliable in
the long term than a hard-wired cable.
Click on the image for a larger version.
In the first article above ("Reducing RFI from the Tesla Powerwall 2) I mentioned that I may have had an experience with RF interference to the Powerwall - but that the circumstances were inconclusive:  There was another problem that may have caused the symptoms (e.g. one of the mains connections' clamps had not been tightened at the time of installation).

This time, I do have a tale to tell about radio frequency interference (RFI) to my Powerwall 2 system.

The history:

Years ago, because it was convenient, I placed my DSL modem in my ham shack. Other than low-level spurs from the modem's plug-in power supply (wall wart) which were easily solved, it never caused any problems - nor would my HF operation seem to bother my DSL modem - except on 160 meters when I ran more than about 50 watts.  I found this remarkable because the wire from the DSLAM (the distant interface from the phone company) came through the window only about a foot away from the windowed transmission line that carried the transmit RF power to the antenna.  At some point I dropped the POTS (Plain Old Telephone Service) dial-up in lieu of VOIP and at the network interface (the box outside the house) I disconnected the internal house wiring as it was no longer needed, back-feeding this internal wiring from the VOIP box to allow the continued use of the phones.

Several years ago I got a linear amplifier capable of full-legal power (1500 watts in the U.S.) to use on the HF bands to assist when conditions were poor and on some frequencies this high-power operation would cause intermittent drop-outs in Internet connectivity when I transmitted.  It seemed that the longer I operated, the fewer these drop-outs were, possibly due to the modem "re-training" itself to deal with the (apparently) degraded connections.

When I got the Powerwall its Ethernet connection came through the outside wall and into the house near the DSL modem as it was a convenient place to make the connection - and a UPS was nearby.  I recently decided to relocate the DSL modem to another room, one farther away from the ham shack and closer to where the underground wire from the telephone company came into the house and this involved a bit of additional wiring of Ethernet cable, and since my entire house is effectively on a UPS (via the Powerwall) it didn't matter where I plugged it in now.

Without the DSL modem in the shack I installed another Ethernet switch to manage the multiple connections that were made at that point:  One to the shack computer, another to the garage's Ethernet (to provide Internet connectivity of the solar inverters), another to a KiwiSDR and yet another to another switch where even more things were connected.  Also on this switch is the Ethernet connection to the Powerwall.

On the blink:

I'd done the relocating of the networking gear earlier in the afternoon about a week and a half ago and it wasn't until that evening when I got around to tidying things up slightly and putting a transmitter on the air - in this case, my 630 meter (472-479 kHz) station - to make a few contacts.  When I keyed the transmitter - which produces about 75 watts of RF - the lights in the house dimmed, a UPS beeped and the lights when bright again - so I quickly un-keyed.

After swearing to myself and hoping that this was a coincidence I waited for a minute or two and tried again - with the same result:  The power flickered, the UPS beeped and the lights went back to normal.  Bringing up the Tesla app I looked at the Powerwall's back-up history and saw that I now had two outages, each less than 30 seconds:  That in itself was unusual because the Powerwall typically stays off the grid for a couple minutes even after utility power returns to make sure that the it is stable.

"$#!+", I though to myself again!

I then powered up the HF station.  Things were OK on 75 meters at 100 and 1500 watts and things were also OK on 40 meters at 100 watts - but I triggered the same "blink" response at any power level above about 600 watts.

Now began the methodical investigation.  The first thing that I did was to disconnect the Ethernet connection to the Powerwall from the switch that I had just installed:  No problems at all on any frequency or power level.

This was getting interesting:  Why would connecting the Ethernet cause a problem?

Ethernet connections are supposed to have galvanic isolation via a transformer!  To be sure, there is a small amount of capacitive coupling between the two windings, but this was on the order of a few 10s of picofarads - not nearly enough to cause a problem at 630 meters - or so one would think!

I then grabbed another Ethernet switch - a small, in expensive Linksys 5-port switch and connected everything to it:  No problems.
Figure 2:
Inside the raceway where several cables - including the Class-2 Ethernet
RS-485 cables for the Neurio run.  This picture shows the bonding of the
CAT-5's shields to the raceway (yellow tape and the blue wire that connects
to a mounting screw of the raceway using a spade lug and star washer). 
Also visible are some snap-on RF chokes (round gray, square black).
As mentioned, simply grounding the shield did nothing to solve the RF
ingress issue since the energy was being conducted on the cable's internal
wires already:  It took the addition of the chokes here and elsewhere (see
the other photos) to reduce the level being conducted on the Ethernet cable.
There are no exposed conductors in this raceway so it is "safe" to work in.
Click on the image for a larger version.

At this point I may have been able to get away with using that small, cheap switch, but it was very old and it was "only" a 100 Mbps-capable switch which meant that it would be a bottleneck for traffic between computers.  I was also determined to make it such that I it would not matter to what I connected the Powerwall's Ethernet cable as I wished to avoid a future problems should I forget why it was there!

When shielded cable doesn't help:

A bit of investigation revealed that the CAT-5e cable to the Powerwall was shielded.  One might first think that this should have solved the problem - but you'd be wrong:  Shielding is useful for containing energy within the cable and prevent it from radiating, but if the cable in question is, itself, longitudinally conducting RF energy from one place to another, this shielding has absolutely no useful effect on its own!  In other words, if a piece of equipment at one end is somehow allowing RF to be induced onto the Ethernet cable, shielding will do nothing at all to prevent the conduction of RF to the far end - and it may well make it worse!

Still, the shield might prove useful to allow shunting some of the RF current on the shield so out of due diligence I went outside to the electrical raceway below the Powerwall where the conduit - which is all metallic and bonded to everything else, including the Powerwall - that conveyed the Ethernet cable from the house.  There, I carefully opened the outer jacket of the cables and made a connection to the shields which I then bonded to the metal raceway as depicted in figure 2.

Figure 3:
Inside, showing both the cable that goes between the Powerwall and the
Neurio in the garage (the left-hand cable with the two white snap-on chokes
in a loop) and the Ethernet cable that connects between my switch and
the Powerwall.  The Ethernet not only has two of the white snap-on chokes,
but it several turns are also wound through a Mix 75 toroidal core, seen
near the upper-right corner of the picture.
While this was enough to stop RF susceptibility issues on the HF bands,
these devices did not offer enough choking inductance on the Ethernet
cable to prevent problems at 630 meters.  Another Ethernet choke was
wound on another Mix 75 toroidal core - See Figure 4.
Click on the image for a larger version.
Because I had previously installed some Ferrites on the Ethernet cable and the other (shielded) CAT-5 cable that made the RS-485 connection to the Neurio in my garage I was hoping that this bonding - and the self-capacitance of these cable, both of which were now bonded - would shunt the RF energy to it and solve the problem.

It didn't.  No surprise there.

I decided to get serious about the problem and started checking the ferrite devices that I'd previously installed.  Winding a few turns on each I measured the inductance and found that they did provide a reasonable amount of reactance at 40 meters - typically 5-10 microHenries which provides between 200 and 400 ohms of impedance at 7 MHz - but clearly, this was not enough as I was still having problems at 7 MHz.  This would be especially true at 630 meters where the reactance would be only a few 10s of ohms - hardly enough to significantly impede RF energy at that frequency.  I then began to rummage about through my collection of ferrites, running a few turns of wire through each.

It became clear that the devices that I had previously used were typically of "Mix 43" or similar, best-used for the higher HF frequencies (and into VHF):  A few turns through one of these will give 10-ish microHenries of inductance, but I needed far more than this so I switched my attention to finding "Mix 75" and "Mix 77" devices - ferrite material that had an order of magnitude or so more permeability and would also yield much higher impedance:  A few turns on these would yield the hundreds of microHenries and offer several kilo-ohms of reactance to better-block RF - especially at 630 and 2200 meters.

What finally worked:
Figure 4:
A high-inductance choke wound on a Mix 75 toroidal core using flat "Cat 6"
Ethernet cable.  This device offers a common-mode choking inductance of
several millihenries at LF, MF and lower HF frequencies and was required
to solve the problem of conducted RF on the Ethernet cable causing issues
with the Powerwall's back-up gateway.  This cable was inserted into the
Ethernet cable between the indoor switch and the Powerwall using a CAT-6
rated double-female joiner.  A combination of a very high
inductance choke like this and several snap-on chokes over several turns
like those seen in Figure 2 and 3 are recommended for broad-band attenuation
of conducted currents.
Flat Ethernet cable is preferred for this as it will better-retain its internal
geometry when wound as tightly as this while similarly-winding round
Ethernet cable in such a manner may impact its signal integrity and
reduce its maximum speed.
Click on the image for a larger version.

What I settled on was a combination of several things:
  • Several turns of the Ethernet cable through some Mix 75 snap-on chokes.  The use of several chokes and several turns maximizes the added inductance.  (see figures 2 and 3.)
  • Some more of the same snap-on chokes were placed over the RS-485 connection to the Neurio in the garage.  There wasn't evidence that RF energy was affecting this line, but I didn't want to take the chance.
  • As depicted in Figure 4 I wound a 6 foot length of flat CAT-6 Ethernet cable over a toroidal (ring) core of Mix 75 ferrite to provide a choking inductance of several milliHenries - several orders of magnitude higher inductance than the other ferrite devices.  This "bulk" inductance would be responsible for blocking RF energy at the 630 and 2200 meter bands at which I often operate.
  • Inside the raceway I placed some additional Mix 75 snap-on devices over both the Ethernet and RS-485 cables, each passing through the cores several times for maximum inductance.

By adding all of this ferrite I increased the impedance (at RF) of the common-mode current to hundreds (if not thousands) of ohms at all of the frequencies on which I am likely to operate.  It is to this effort that one must go to minimize conducted RF energy on such conductors.

Why did it matter which Ethernet switch I was using?  I'm not sure, but I suspect that the Gig-E switch is lacking complete galvanic isolation on at least some of its Ethernet ports.  This issue could occur not only if there is a shielded Ethernet cable with attached shielded connectors on each end - which would be a liability in this particular case - but also if there is some sort of balanced connection to the pairs themselves - such as would be present if the switch had POE (Power Over Ethernet) capability.  The manual for the switch indicates no POE capability, but there is definitely something different about it and the way it connects the cables!  Perhaps a similar model of this switch does have POE and there is, in fact some connection inside - but unless/until I tear it down to find out, I won't really know.

Afterward:
Figure 5:
The "pin" on the BUG (Back-Up Gateway) that I was
requested by Tesla to photograph and forward to them.
This is pin is normally hidden by a sticker, hence the
"void" residue on the metal. This pin moves left/right
to go on/off grid, respectively.
Click on the image for a larger version.

Several days after this event I got a telephone call from Tesla Powerwall Support asking me to remove the sticker from my back-up gateway and see if the "pin" was visible.  When queried, the representative noted that they had recorded several "incidents" with my system - and time and dates of these correlated exactly with my RF interference issues:  I told her that as far as I was concerned, the issue was resolved - but they still insisted that I take a picture of the pin and forward it to them - See Figure 5.

Update:  I was contacted again by Tesla - and this time they asked "if I was able to move the pin".  Not having been asked to do this before, I did so when I was at home again and it did move:  To the right, the house was isolated from the grid and to the left, the house is tied to the grid.

Having forwarded this information, I have yet to hear back.


Parts sources for ferrite devices: 

There are several sources of snap-on ferrite devices described on this page, including:
  • KF7P Metalwerx - link - Supplier of a variety of ferrite devices and many other things.  At the present time he stocks the "Mix 31" devices, but does not stock "Mix 75" snap-on cores at the time of posting, but he does have ferrite rings of both ferrite mixes.
  • Mouser Electronics - link - The "Mix 31" snap-on cores - P/N:  623-0444164181  (Fair-Rite P/N:  0444164181);  "Mix 75" snap-on cores - Mouser P/N:  623-0475164181  (Fair-Rite P/N: 0475164181).  Mouser Electronics has other sizes and mixes of these various devices, including toroids (rings).

Links to other articles about power supply noise reduction found at ka7oei.blogspot.com:

Other solar power related posts at ka7oei.blogspot.com:

This page stolen from blogspot.ka7oei.com

[End]