1

I was shopping for a small UPS for my home networking lab closet; total load under 300w at 120v. But then realized I have a spare AGM car battery, 12v solar panel maintainer and 12-120v sine wave inverter. So I'm DIY home brewing up a project.

I sourced a 15A automatic transfer switch designed for this application - essentially a DPDT relay from what I can tell online. This is for line voltage loads and utility to backup inverter seamless switching. And I'll mount and integrate it to home electrical standards in metal boxes and EMT conduit on its own circuit. The 5v and 12v modem and wifi loads are getting 7812 and 7805 voltage regulators to keep them pure DC off the battery.

While I'm comfortable with the plug and play aspect of this design I am a bit concerned with the reported 80ms cutover time of the relay and power transfer. Goal being to keep a small server and network gear with internal AC/DC supplies from rebooting or struggling during any power flickers and utility to inverter switching.

Is there a capacitor or induction coil design you can help spec that I can place on the load side of this to help smooth this out? A big motor start capacitor or something would be my uninformed thought but don't have the EE skills to know what to do with it. I realize the computer power supplies likely have similar in their circuits but would like to help if I can. I will have TrippLite ISOBARs in line which I believe have induction coils in them.

Am I over thinking this at 80ms delay? Or is there a drop in big coil or cap I should integrate? I realize that commercial UPS designs probably have the inverter in use at all times with a trickle charger, so it's a constant AC > DC > AC flow but I'd like this transfer design to save wear on my basic inverter. Power outages are rare enough here I'm not battling ongoing outages or anything; maybe a few times per year. Any other holistic design aspects I'm missing? TIA!

tldr: Is 80ms too long of a cutover to keep a server and big network switch from rebooting?

Details on

Neil
  • 11
  • 3
  • Is your 12V solar panel maintainer powerful enough to power the modem/wi-fi loads and quiescent inverter current as well as recharge the battery after it's powered those loads during hours of darkness? – Finbarr May 13 '20 at 15:18
  • The solar panel would supply maintenance trickle current only. I'd use a utility powered 12v battery charger to supplement recovery in the case of an extended outage. – Neil May 13 '20 at 16:01
  • Then what powers the modem/wi-fi and inverter in normal operation? – Finbarr May 13 '20 at 16:04
  • Modem / wifi via 12 and 5v regulators would get sufficient power through the solar panel including overnight recovery. I may need to add to the solar array but what I have has a 1.39A rated peak output. The inverter would not be used during normal operation. The transfer switch at the heart of my question will toggle from utility to inverter power upon failure of utility power. – Neil May 13 '20 at 16:19
  • 1
    Your inverter will need to be connected to the battery all the time, so it will take some current all the time even if it's not powering the server. And your panel will only give that peak current in full, direct sunlight. During the winter you'll get a lot less current and for a lot less time and will drain a lot more from the battery with a lot less time to replenish it. Have you calculated exactly how many solar panels you will need? – Finbarr May 13 '20 at 16:26
  • No, this is a bit of an experiment with spare parts from my garage. I'll have to test what the standby draw is off the inverter. In an extended outage I'd likely gracefully turn down the server and utilize the battery for wifi and device charging. I could also support the battery during standby with a 14v 4a car battery charger wall wart should the solar needs be significant. – Neil May 13 '20 at 16:32
  • I don't want to run my 200w actual draw server and 50w of enterprise network gear through my solar charging experiment during normal operation. Primary desired use cases would be keeping these running uninterrupted during power blips shorter than a few minutes, as the utility company does work on things or their automated fail-over fixes the grid. – Neil May 13 '20 at 16:35
  • I have installed generators and electrical for commercial kitchens' walk in coolers, but they have momentum to overcome any sub second power drops during transfer. Looking for anyone with datacenter or UPS electrical engineering design experience. – Neil May 13 '20 at 16:46
  • "I'll have to test what the standby draw is off the inverter". That's not what I'm talking about. The inverter will consume power, even with nothing plugged in, just to maintain line voltage at its output to be ready for when it's needed. Looking at a typical 300W inverter its no-load input power is about 6W, which is half of what your panel generates under optimum conditions. – Finbarr May 13 '20 at 17:01
  • Okay, thanks for the clarification - that 6w draw is what I meant when I said I'd test it, but sounds like it's more significant than I had hoped. Thanks for pointing it out. – Neil May 13 '20 at 17:21
  • Let us [continue this discussion in chat](https://chat.stackexchange.com/rooms/107965/discussion-between-neil-and-finbarr). – Neil May 13 '20 at 17:28

2 Answers2

1

That is a bad choice of battery. It's a starting battery, not a deep cycle battery. You'll destroy it lickety split. You may want to hang out with the off-grid homesteader / solar / wind crowd for awhile and learn how to set up systems like this.

Also since you're willing to spend this kind of scratch on a battery, maybe evaluate some of those Tesla modules that are out there. They respond much better to deep cycling than even a deep cycle battery, which this is not.

80ms is typical of a mechanical transfer switch. But further, you will have trouble because of it throwing automatically, and possibly having some false starts, as power cuts out and returns. To be clear, 80ms is five full AC cycles - high low high low high low high low high low. So there is just no chance of doing this with DC-style inductors or capacitors. There's no such thing as a DC battery. And even then, how would it be cut in? The transfer switch's contacts are in transit.

Really, you should be reinventing the whole kit-n-kaboodle to run on low voltage DC. The server's power supply should input LVDC. The router, switch, modem whatever should input low voltage DC (probably already are). Essentially it will be hung off the battery at all times with makeup power coming from a battery charger.

Now, you don't need a transfer switch.

The kind of money you're spending on that gear, you could get a LV power supply for a server, and also a nice lithium Tesla style battery with some range.

Harper - Reinstate Monica
  • 21,639
  • 27
  • 68
0

Unfortunately your 80ms switchover time is far from "seamless" and is not likely to be sufficient in this instance, although without a specification for hold-up time for your server and switch power supplies it's impossible to say. However, for comparison, the ATX specification only says that a power supply should hold the PWR OK output to the motherboard up for at least 18ms from the removal of mains power; this corresponds to just over a single cycle of the mains AC supply at 60Hz.

There's no easy fix by adding a capacitor or inductor to hold an AC supply up the same way you can with DC, but let's assume for the sake of argument that this is possible and work out how big a capacitor would be needed if the supply were 120V DC. Your 300W load will draw 2.5A and let's assume that the voltage mustn't drop by more than 10% (i.e. 12V) in 100ms to be on the safe side and be sure that the equipment doesn't get reset. That's a rate of change of 120V/s, which using the formula $$I=C\frac{dV}{dt}$$ gives $$2.5=C*120$$ hence $$C=0.0208F$$ So you're looking for something like a 22,000uF capacitor with a working voltage of at least 150V. A quick look at my normal component supplier shows only one part listed that will do the job, and it measures 219mm in length, 76mm in diameter and costs £100 plus VAT. About the same volume as your entire PC power supply and probably a lot more expensive.

And that's only one of the two major issues with your plan. The solar panel you have is designed to keep a battery topped up while it's not being used but is not capable of very much more. With a modem typically taking something like 8W, a wi-fi access point taking maybe 10W, the media converter mentioned in the comments taking 6W and the no-load demand of the inverter at around 6W you're looking at a continuous load of 30W from a panel that can only deliver 16W with maximum sunlight hitting it straight on. Take into account the reduction in output between dawn/dusk and peak sunlight and the reduced sunlight hours in the winter and you're probably going to need twelve such panels to cope with the load.

Best advice: Carry on shopping for a small UPS. I'd get one that has a USB or serial port that connects to your server and can shut it down as the battery starts to run out and bring it back online when power is restored. It's a far more practical solution.

Finbarr
  • 5,501
  • 3
  • 19
  • 33