Nipron PSU with "built-in" UPS

Rather than having a UPS supplying AC power to the mains input of a PSU I thought it would be more efficient if a battery were able to supply DC power direct to the PSU

After a bit of googling I came across this

The Nipron Non-Stop Power Supply "NSP"

Does anyone have any views / experience of such a device?
The battery goes in a spare 5.25inch drive bay
Good for a few mins of uptime; more than sufficient to cover against brown-outs, blown fuses, turning off the wrong plug socket!!
Remote shutdown for anything longer

Use is a Windows Home Server box

The concept is perfectly sound. It may come down to how good the Nipron implementation is, and what Nipron's quality is like.

The battery would be way too small to be effective.

Originally Posted by drescherjm

The battery would be way too small to be effective.

surely the effectiveness would depend on the load and the required up time??

but i agree, one isn't going to get a lot of up time

On top of not having uptime these single battery UPSs that fully discharge in less than 3 minutes tend to not last more than a year or 2 making the user have to replace the battery more often.

It would however be fine to eliminate a few second outage.

Originally Posted by drescherjm

On top of not having uptime these single battery UPSs that fully discharge in less than 3 minutes tend to not last more than a year or 2 making the user have to replace the battery more often.

It would however be fine to eliminate a few second outage.

The SERVICE life of the battery is more a function of the charger used than it's capacity assuming the proper type of battery chemistry is selected for the job at hand.

Something like these units may be a handy way of handling brownouts, cord kicks and short interruptions due to maintenance or grid switching. If I had one, I'd probably tell the system to start shutdown after 60 seconds. There's no need to run the battery down to 0% and if the power isn't back on by 60 seconds there's a diminishing probability that it will be back on by 120. (In the grids I've lived on anyways)

I mentioned that because at work batteries tend to last a lot shorter on APC BackupUPS 350 (and similar) than the same exact battery in a 1000 VA APC unit with two of the same batteries. I usually get 3 to 5 years out of the UPSs with two batteries and 1 to 2 with one battery. This may be do to the fact that 1 battery is pretty much does not have any spare capacity while 2 batteries of a BackupUPS 1000 at 75% would be just fine for a 150W to 300W load.

Google's servers have used this same type of implementation for a few years now with no problem. The idea is certainly sound, but the uptime is short on them. Battery life is entirely dependent on quality of the batteries used.

Originally Posted by Paul_Johnson

Google's servers have used this same type of implementation for a few years now with no problem. The idea is certainly sound, but the uptime is short on them. Battery life is entirely dependent on quality of the batteries used.

yeah, i remember reading about that some time ago link

large image of google server
PSU seems to be a custom part by a company called Magnetek

Originally Posted by drescherjm

I mentioned that because at work batteries tend to last a lot shorter on APC BackupUPS 350 (and similar) than the same exact battery in a 1000 VA APC unit with two of the same batteries. I usually get 3 to 5 years out of the UPSs with two batteries and 1 to 2 with one battery. This may be do to the fact that 1 battery is pretty much does not have any spare capacity while 2 batteries of a BackupUPS 1000 at 75% would be just fine for a 150W to 300W load.

I think the difference here is going to be the charger logic/quality in a APCs higher line UPSs. Using a 24V DC side in a UPS will halve the current draw required for a given output power over a 12V DC side. However the UPS systems these batteries are installed in also usually double and triple in capacity. If you load them up you're back to the same current demands.

Demanding high current OUTPUT will usually only restrict runtime, but not service life, since the chemical process can't "keep up" with the need for power and all of the battery's chemical energy cannot be extracted before the terminal voltage drops too far. This is why you'll often see 100% runtimes of 5min, but 50% runtimes of 20min. More chemical energy can be extracted from the battery at a slower rate.

What kills the service life (ie. measured in years) of a battery prematurely is usually the charger, or using the wrong battery design and chemistry for the intended use. CHARGING a battery too quickly can kill it. Floating a battery at too high a voltage can kill it. Keeping the battery too warm will kill it. The battery itself is a combination of design factors for float batteries (since 99.99% of the time it's just sitting there on a charger) and deep cycle (since 0.01% of the time you run it down to dead). Each have different design criteria and UPSs need both.

So, smarter chargers with more logic to respect the battery and extended it's service life will always give longer battery service life. The more expensive/higher end the UPS, the smarter the charger on average.

Originally Posted by Paul_Johnson

Google's servers have used this same type of implementation for a few years now with no problem. The idea is certainly sound, but the uptime is short on them. Battery life is entirely dependent on quality of the batteries used.

In google's case, they probably only need 7-10s of power before gensets are online. This is actually pretty smart of google because:

- they don't need to built infrastructure for a data center UPS or -48VDC system just to run servers for up to a minute until the gensets start

- if the gensets don't start, there's a good chance they're screwed anyways

- they are probably constantly upgrading servers anyways and the service life of the whole system is probably equivalent to the service life of the battery (3 yrs maybe)

- there isn't a huge centralized UPS system with transfer switches etc.. requiring 100% uptime (as opposed to 99.999%) which runs the risk of taking the entire datacenter down.

- they can perform all kinds of maintenance in or near the racks without worrying about the servers inside, since their "couple of minutes" runtime is satisfied by built-in power.

- they could build a smaller -48VDC plant to run all of their core equipment which can be sourced from Cisco, Juniper, Brocade, whoever with -native -48VDC power supplies. In that case, again, you only need batteries and rectifiers - no fancy switching - much more reliable.

Also keep in mind that PC's use DC power, the same power that batteries provide. So depending on the implementation, the built-in batteries might be powering the components directly (using some sort of voltage regulation for the different rails) versus being stepped up to ~120vac by the UPS then back down to DC by the power supply. There's quite a bit of "lost" energy during all the conversions. They can also opt to use lithium ion or lithium polymer batteries which provide a MUCH better power/weight ratio than your typical sealed lead acid batteries.