Do It Yourself UPS for low wattage / high runtime.

soylentgreen

I have a server rack which uses low power servers and networking equipment. Total draw is about 100W. Everything is on a UPS but the runtime is about 2 hours. I'd like to have more like 24-48 hours.

I know that APC, TrippLite, and CyberPower all sell UPS models with external battery packs, but I've priced out a 1 - 2kWh system and it's $1000+ for all brands.

I'm thinking about a project to do this myself - appreciate your comments:

Poor-Mans Double-Conversion Setup:

• Buy a 30A / 12V power supply (about $50) something like this http://www.jameco.com/webapp/wcs/stores/servlet/Product_10001_10001_2101497_-1
• 100Amp-Hour or larger 12V AGM battery (about $250) and also connect to a...
• MorningStar 300W sine waver inverter. (about $250)
• Total cost under $600.

Hack a commercial UPS

• Buy a sine wave UPS that runs on 12V such as http://www.cyberpowersystems.com/products/ups-systems/pfc-sinewave-series/CP1000PFCLCD.html (about $150)
• Hook this to a 100Amp-Hour or larger 12V AGM battery (about $250)
• Total cost about $400.

I can see pros & cons to both - the double-conversion setup will give me really clean power, but at a small efficiency loss and higher price. The Hacked UPS is cheaper, but I worry that the circuitry is not really made to run for 24 hours straight (although one would expect that if it's a "1000W" UPS and I'm only loading it at 100W, that it probably won't melt down), and also I wonder if it would get confused being connected to such a large amp-hour battery.

Ideas?

MarkC

UPSs are being used for continuous emergency power generation successfully. You might contact techntrek of this forum for actual data from his Prius/APC SURT UPS being used for days of emergency backup power. I have a similar system, but have only tested for hours, not days - have not had a major outage yet. These units are so robust and over designed, it seems unlikely that they cannot operate as backup power for a couple of days - especially under partial loads.

BB.

I would tend towards a "real battery charger" vs a DC power supply. The Iota is designed for charging a battery and an IQ4 module can be used with on Utility Power.

The problem with DC power supplies is that they may have some modes that are not really great to have with batteries. A frequent power supply protection mode is to use a "crowbar" to short circuit the output if there is an over voltage--Not a good idea with a battery bank.

The power supply you linked to has "hiccup mode" if the output is over current... So, if you have a power failure and the battery is significantly discharged, it is possible for the power supply to go into over current limit and reduce its average output current a lot--And actually not recharge the battery at all (especially if you still have your 100 Watt DC load).

Lastly--This is the part that has always bugged me... Eventually stuff fails. your AC to DC Power Supply, the Battery Bank after 3-7 years, the AC Inverter will run ~5-10+ years between failures, etc... Compared to how reliable is your Utility Power... A UPS can (in theory) switch over to AC mains power or Inverter power if there is a failure in either. With a pure AC inverter solution, if the inverter fails (and it will eventually), you are out of luck unless you do some other fail over trick that is fast enough to avoid rebooting the computers.

One solution that I liked was the dual AC input computer power supply.

http://www.directron.com/redundant.html

You can connect the AC1 and AC2 to two completely independent AC power systems (AC mains and AC inverter for example). And remember that with redundant components/systems, each redundant unit must be able to indicate it has failed so it can be replaced before its mate dies too.

The whole issue of redundant and reliable computer power and systems vs that single point of failure--I always worry that too much complexity actuality reduces reliability overall.

If your AC mains are not reliable, then using a MorningStar inverter certainly would be a good solution.

And running backups or even parallel server racks (redundant and verified?) is important here too. Eventually everything fails.

-Bill

soylentgreen

BB, good points about failure: using the double-conversion method is adding many new points of failure. I like the dual power supply as well.

soylentgreen

I went ahead and built it.

Hacking a commercial UPS

• Bought a sine wave UPS that runs on 12V such as http://www.cyberpowersystems.com/pro...000PFCLCD.html (about $120)
• Found a great deal on two used 145 amp hour AGM telecom batteries (search the web for "Alcatel Lucent 145" or "12AVR145ET" and you'll see some used ones for about $125 each ($250 total)
• Automotive amplifier wiring kit from Frys (or equivalent) which includes 8 Gauge cables, connectors, and nice 60amp fuses. ($15 each, qty 2): $30
• Total cost : $400

I removed the small 9 Amp*Hour AGM from the UPS, wired in the two monster 12AVR145ET in parallel with proper fusing, and ran a test: My 100 Watt server rack ran happily for about 10 hours, and using the voltage/capacity curve here : http://i.stack.imgur.com/EheHZ.png I figure they are performing about as advertised, and my 290 amp hour rig would run for about 30 hours. Pretty cool.

After restoring power to the UPS, the built-in battery charger is VERY SLOWLY bringing them back up to voltage at about 0.01 volts per hour, so it'll take 5 days or so to hit the float voltage (which should be around 13.5V per specs, but I'm curious to see what the UPS itself is set for).

Here's a question: if you are recharging AGM batteries with a dramatically under-sized charger (e.g. C/100 give or take) will they eventually get to 100% sitting at float voltage after a few months?

BB.

Someone here posted a few days ago that Lifeline Concorde AGM batteries recommended a 0.2C rate of charge...

A 0.01C rate of charge at float voltage (Float, which is not supposed to really be charging anything), I would be concerned that the battery would not properly/fully recharge. And if the batteries are older (and hot), their self discharge rate may get higher too (and extend charging times). For flooded cell batteries, their self discharge can approach 1% (and still be a "good battery"). And I would suggest that if you see >2% self discharge rate--The battery should be replaced (that much float current, greater than 2%, can overheat the batteries).

Practically speaking, a 0.05C rate of charge for off grid systems works out pretty well for our rules of thumbs (seasonal / weekend home, etc.). And over 10% rate of charge for daily/full time off grid home.

Practically speaking, if you have a 1% rate of charge--would it be a good emergency system if you have an outage flollowed by another outage a few hours/day later--Would that be acceptable to you (first outage, your battery bank carries the loads--Second outage, battery bank cannot sustain loads without backup genset and/or damage to battery bank).

-Bill

Cski

Instead of hacking a ups, there are battery chargers that are designed to charge batteries and supply power. Here is one exMple http://www.iotaengineering.com/dls55.htm
It may not be exactly what you need but something to look into. I had a radio bank that operated off of 12 and 6 VDC. I used a spare 12 volt blue top 55ah battery and a dc to dc converter to get the 6 volts. Used a similar charger to supply the 12 volts and keep the battery charged. If AC power failed the battery would supply the power. It was able to recharge the battery safely and efficiently. The only difference is you would need an inverter on the output of the battery. If you only bed 100 watts you should be able to get one at a reasonable cost.

soylentgreen

BB. wrote: »

Practically speaking, if you have a 1% rate of charge--would it be a good emergency system if you have an outage flollowed by another outage a few hours/day later--Would that be acceptable to you (first outage, your battery bank carries the loads--Second outage, battery bank cannot sustain loads without backup genset and/or damage to battery bank).
-Bill

That's a real issue. At this location we've had about 4 power outages in 4 years, the longest lasting about 18 hours, which was why I tried for a > 24 hour runtime. Naive statistics would say "one outage per year" = 1/365 chance per day. The chance of two outages on consecutive days is thus 1/(365*365) or less than 1 in 100,000.

Of course, that calculation is bogus, since power outages are not statistically independent!

However, I have enough monitoring software so that if one outage happens, I'd be paying much closer attention and could throw a big battery charger on it or be otherwise prepared for the 2nd outage.

Mostly, my goal here is to avoid nuisance shutdowns of the server rack: if we had a major disaster and lost power for 2+ days, it's unlikely that the cable / DSL companies will be providing connections anyway, so the point would be moot.

lkruper

soylentgreen wrote: »

Mostly, my goal here is to avoid nuisance shutdowns of the server rack: if we had a major disaster and lost power for 2+ days, it's unlikely that the cable / DSL companies will be providing connections anyway, so the point would be moot.

This is something I have wondered about. I have a UPS on the DSL modem at my cabin and a small inverter generator. The phone company keeps my POTS going in a power outage. What about DSL? I am assuming they have generator backup at their locations. Also, I presume they service businesses from the same location who might have better Service Level Agreements than residential. Does anyone have any experience with a long power outage and whether or not DSL was up and running? I have Verizon for my provider.

zoneblue

BB. wrote: »

A 0.01C rate of charge at float voltage (Float, which is not supposed to really be charging anything), I would be concerned that the battery would not properly/fully recharge.

"Float" kinda muddled me a bit when i first tried out midnite's days between bulk feature. Instead of going bulk, absorb, then float, the battery goes straight to float. This has very little effect on the charge current curve throughout the day. It does take a bit longer to reach the point where charge acceptance reaches my "float" threshold, but charging at float doesnt necessarily mean that the current is low. Charge current relates to the combined plate charge and the difference between that and charging voltage. If you want i can dig our some curves to show what i mean.

But from a more practical point of view, your bank will be on true "float" service, except the discharge events. In order to get the capacity restored as quickly as you can to prepare for the next discharge event, you might want to consider an addtional charger. AGMs are said to like a reasonable chage rate to "stir things up". (Not exactly sure what this means. Some posters here feel that some stratification does occurs in agms, and this may help.)

KenZ71

lkruper wrote: »

This is something I have wondered about. I have a UPS on the DSL modem at my cabin and a small inverter generator. The phone company keeps my POTS going in a power outage. What about DSL? I am assuming they have generator backup at their locations. Also, I presume they service businesses from the same location who might have better Service Level Agreements than residential. Does anyone have any experience with a long power outage and whether or not DSL was up and running? I have Verizon for my provider.

Will dsl stay up during a major power outage? All depends on your local government rules.

When Hurricane Irene and her sister Sandy walloped the New England area ATT had generators powering the uverse boxes at the street so I had dsl as long as I could power my router & Modem. Comcast was out for a few days even after the power came back.

ATT only did that because the government required the voip phone service they sold to run even if the power was out.

Too bad their speeds stink.