Grounded modem and DC-powered server; causing internet connectivity issues

[harrisonpatm]

Posting here on the offchance that someone might know what's going on.

I have a number of devices in my house powered directly off my 48VDC battery bank solar system, via converters. This is a good setup because I only have a 2000w inverter currently for those devices that can only be powered off 120v AC, and I don't want to overload it. The more devices that the inverter doesn't need to power, the less it will trip.

My modem and router are powered via a 48v to 12v buck converter. Two years ago when I was setting up my system I realized that my home's coaxial cable was grounded, as it should be. I realized it when I was working on the 48v positive bus outside, in bare feet, and got a small jolt. I traced it to the modem: since the coaxial cable is grounded, and the buck converters are non-isolated, there must be a shared ground in the modem's wiring that's creating a path for the negative DC current back through the modem, and through the grounding wire. Interesting, but ultimately I didn't do anything to change that. I was more careful when working on the bus bars, and I didn't have any other issues.

More recently, the past few months, I've been using various small PCs in homelabbing for fun and learning. Because I want to put anything on the battery that I am able, I've been trying to power them directly off my battery/solar, since they're on 24/7 and I want to reduce power. And I keep running into an issue that I don't understand enough to troubleshoot, that I wanted to post about on the offchance that the knowledgeable people on the forum here might know how to fix it. When powering my server off my battery, and plugging in the ethernet cable, I get little or no connectivity. Consistently. I've tried a number of combinations. If the device is being powered by my battery, and using a wired connection, I get no internet. If I power it with mains AC via a standard power brick or PSU, it connects fine. If I power the device off my home battery, but use WIFI to connect, also no issue. Even if I'm using a 48vdc-to-120vac inverter, I still have issues. I've used multiple different kinds of PCs to test this theory: my first server was an old tower-style PC, I've used some reporpused thin clients, and my newest media server build is a 1L Thinkcenter with external drives. All of them are having the same issue of no connectivity when trying to power them off the battery. And lately, with this newest build, it's even causing my modem to crash and have to be restarted every few hours.

So, does anybody have any ideas how to fix this? The past couple of days I've managed to get my latest build to be DC-powered, by simply plugging my modem back into the wall, but it's still having connectivity issues every couple hours that are instantly solved by unplugging/plugging the modem. I thought about disconnecting the grounding wire from the coaxial cable coming into the house, but I'm pretty sure that could lead to an issue if we get a bad lightning storm (though I may do it anyway, temporarily, just to further confirm that this is the source of the issue). I thought about finding a proper isolated DC-DC converter just for the modem, but even 2A rated devices are $60-100. Next I'm going to try to power this latest sever from a standalone inverter, but I purpose-built it to run directly off DC, so that feels like a bit of a letdown.

Any other suggestion? I don't know enough about electrical engineering to know what's the best path here. Thanks!

 

[harrisonpatm]

As I'm thinking more about it this morning, I thought of some new search terms that I hadn't before. If I install a ground loop isolator between the modem and the coax cable going into it, would that help?

 

[neptronix]

Wish i could advise here but i've never ventured this far to have these problems.

The easy answer is to run it off a known clean source of AC power and just accept the loss of 1-2 watts of inefficiency.
I run all computer equipment on a UPS unit because it effectively cleans up AC power and prevents damage to digital equipment.
UPS units typically have a super clean output.
Consumer grade DC to DC and inverters typically don't have the best output. too choppy.. i would measure the quality of the output before rawdogging it.

 

[fechter]

I'd guess it's some kind of ground loop problem. You might try looking for an isolated dc-dc converter 48v to 12v.

 

[harrisonpatm]

Consumer grade DC to DC and inverters typically don't have the best output. too choppy.. i would measure the quality of the output before rawdogging it.

I would typically agree with you, except in this case, my server is a mini 1L pc that would be ordinarilly run off a standard 20v wall-wart-style power brick. I don't have an oscilloscope, but I would guess that a 48v-to-20v converter is no worse quality than a 20v PSU. But in any case, I agree with you:

The easy answer is to run it off a known clean source of AC power and just accept the loss of 1-2 watts of inefficiency.

I currently have the modem, and just the modem (not router/switch) running off a 12v power brick plugged into mains, not my inverter. I'll run that for a bit to see if there's any difference. Though, it's not about the loss of efficiency. When I have everything running off the inverter, it's one piece of equipment that will take everything out if/when it trips or fails completely. Wheras having everything distributed across multiple isolated pieces of equipment takes load off the inverter and reduces server downtime. Because I do get occasional power outages where I live, and it's nice to still have internet during those times.

Also, it's hobby. I can absolutely plug everything into the mains and be done with the problem, but its nice to learn and get experience with these kinds of things.

@fechter , do you think that the coaxial ground loop isolator I linked above would accomplish the same job as an isolated 48v-to-12v converter? Because a good quality, properly isolated converter in the 1-2A range I'm looking for runs $50-100, whereas the coax isolator is $15.

 

[fechter]

A coaxial ground loop isolator would break any ground loop involving the coax, but you could possibly have a loop in the ethernet cable as well. It might be worth trying the isolator.
Some "wall wart" power bricks can run OK on 48vdc. I've had some that worked on 24v. Just for fun you might try running the existing brick off 48v and see if it works.

 

[harrisonpatm]

A coaxial ground loop isolator would break any ground loop involving the coax, but you could possibly have a loop in the ethernet cable as well. It might be worth trying the isolator.
Some "wall wart" power bricks can run OK on 48vdc. I've had some that worked on 24v. Just for fun you might try running the existing brick off 48v and see if it works.

You mean, the 120vac-to-12vdc power adapter? You're saying there's a chance i could give it 48vdc instead of 120vac, and still get 12vdc out?

I have a dozen 12v adapters laying around. This might be interesting.

 

[amberwolf]

You mean, the 120vac-to-12vdc power adapter? You're saying there's a chance i could give it 48vdc instead of 120vac, and still get 12vdc out?

Yes.

I use an old AC-powered 15v battery charger from some powertool for my "DC-DC" on SB Cruiser, which uses a "52v" 14s pack. Even at completely empty on the traction pack it's still enough voltage to start and run the charger and get me 15v out.

I did the same thing on CrazyBike2 and actualy ran all the lights but the car headlight and car horn off of it. SBC only powers the automotive relays that switch everything else on with it, and the 4s battery actually powers the lights and stuff since I have so many it's too much current for the charger to handle.


Keep in mind that the diodes on the AC input side mean some voltage drop from the source, so if it doesn't work you could wire past the diodes (just be sure to get the polarity correct as it's no longer "reversible" ).

Also, without bypassing them, the power rating may be halved since you'll only be going thru one pair of diodes, so if they are specd for minimum capability, they might fail at higher power draws. (haven't ever had that happen, but it could).


I also ran some screw-in CFL bulbs off a 14s pack to experiment with for CB2's headlight at one point...they ran off the battery all the way to the end, but htey would only start when it was near full.


In my general experience, most of the adapters that say they will run on 100VAC-220VAC and 50/60hz work with DC input down to something like 50vdc or a bit less.

Most of the ones that only say 110-115vac and 60hz don't, they have some design bits requiring the actual AC part of the power, or need more voltage to run than the packs I've used supply, etc.

Anything tha'ts a heavy transformer-based unit won't work--those require AC. But anything that's a light SMPS design could.

 

[fechter]

Right. Like amberwolf says, it has to be a switching mode power supply, not an old school iron transformer and you probably won't get the full rated current out of it before it shuts down, so just use one that's big enough. I wouldn't worry about the loss in the input diodes. They are nice to keep so you don't have to worry about polarity on the input.

 

[amberwolf]

Inside the SBC "triangle" (which could've been made a lot neater and easier to work on ); the big black box is the 15v wallwart AC adapter; the automotive relay/fuse units below it, turnsignal unit got spraypainted red on that side (to the right of the relays).

 

[harrisonpatm]

I upgraded my 15-year-old modem, and i also installed a ground loop isolater on the coax cable going into the modem. Couple days now, and no connectivity issues. Now I can actually run my DC-powered home server as planned!