When we moved into our current house, I had a Matrix 5000 UPS that I used to power my server room. This 5kVA unit was a really impressive device that ran on 220V and could have its “brain” replaced without powering down the load. I had four external battery packs for it, which provided over an hour of runtime at full load. Last summer it stopped working as a UPS for some reason, and then the isolation transformer failed catastrophically six months later:
Unfortunately, this unit is no longer made by APC, and there isn’t much information about it online.
I really liked this setup because the UPS was in the garage, fed by a dedicated 220V 30A circuit, and the UPS directly fed a small breaker panel which took two circuits up to the server room. That way, everything was protected by the UPS from surges and small power outages. Since the failure, I’ve been running on small UPSes directly at the critical machines, and using surge protectors for the non-critical stuff.
Recently, I came into a couple of SmartUPS 3000RMXL units for a good price. Two of these have good batteries in them, but the batteries in the other two are on the way out. Since they are “XL” units, they have external battery connections, allowing you to add additional runtime by daisy-chaining 48V battery packs. I decided to construct a large battery bank for this unit and build the connectors required to hook it up.
The batteries I used are C&D Technologies UPS12-370FR, which are 100A/Hr commercial gel cell batteries designed for custom UPS systems. Why did I choose these? Well, because they are designed exactly for this sort of thing. Oh, and I also got them for free!
I put one of the SmartUPS units on the platform I originally had for the MatrixUPS and changed my previous L6-30 (220V) plug on the wall to an L5-30 (120V) for the new unit. Then I created two 5-15 pigtails to bring two of the 15A circuits off the back of the UPS in to the two legs of the breaker panel. With this configuration, the UPS carries the entire server room, for about 2200W continuous load. With the existing weak batteries, the UPS claims about 10 minutes of runtime (although I don’t believe it).
The internal battery on the UPS connects to a port on the front near the battery bay using an Anderson SB120 plug. This is actually routed to the rear of the unit where two SB120 plugs are connected by a continuity module in normal operation. The continuity module simply connects the UPS to the internal batteries.
In order to get the external batteries into the mix (and keep the internal ones there for a buffer during maintenance) I created a cable that I expect is quite similar to what APC provides for use with its approved external packs. This cable is two 8AWG conductors for each terminal on the UPS side of the circuit, which goes to the external battery pack. An additional short 10AWG jumper also exits this connector and goes to the internal battery side. The finished cable looks like this:
A close up of the UPS-side connector:
Note that, when looking at the back of the UPS, the SB120 plug closest to the line input and circuit breaker is the one connected to the UPS’ charging and inverter circuits. The one furthest from this is the one that loops to the internal battery. I made sure that the heavy gauge wire was connected to the UPS side, with the jumper to the battery so that you don’t end up pulling all your current through the smaller jumper wire. The SB120 plugs are designed to be stacked in this manner, and have holes designed for bolts as pictured.
Next, I stacked all four of the external batteries on top of the UPS, taking care to locate them near the edges of the unit where there is most likely to be internal support for the weight. With more room and care, this could have been done better, and I may change it in the future. Once the batteries were physically placed, I started to link them up in series with short 2AWG jumper wires:
In one of the links, I added a 200A fuse to make sure that any short across the terminals wouldn’t result in a huge amount of current flow and/or an exploded battery. I also then attached the battery cable to the 48V side, leaving the cable still disconnected from the UPS itself:
Just in case you are wondering, 48V DC is plenty to give you a noticeable shock across a sweaty arm (trust me), so care should be taken once you get to this point.
After checking my connections one more time, as well as the voltage and polarity of the output at the connector, I plugged it into the UPS. Immediately the UPS’ fans kicked on to cool the charging circuit as the batteries were a few tenths of a volt lower than the UPS keeps them at normally. This subsided after a minute or two. At that point, I reconnected the internal battery on the front. The finished connection looks like this:
Obviously I still need to do some cleanup.
Next, I logged into the UPS’ web interface and told it that an additional four battery packs were connected. The internal pack contains two sets of four 7AHr batteries in parallel to give about 14AHr of capacity at 48V. Four of these would be 56AHr, which I figure is probably close to the limit of what APC would recommend plugging into a single UPS, and may be closer to the actual capacity of the new batteries at high load. All lead-acid batteries are rated at one or two capacities, given a specific load. The faster you pull current out of them, the lower total capacity they have.
After performing a calibration test, the UPS now reports (and actually provides) 1hr45m of runtime with the additional pack. This is pretty good for a 2200W load, especially considering that the non-critical stuff gets shut down almost immediately after a power failure, lowering the load and increasing the runtime.
I should also point out that I have a very plain 3500W camping generator that doesn’t use an inverter and produces very mediocre output. It’s fine for most things (including electronics) but most UPSes are unable to handle and correct the input when running on the generator. These models have the ability to set the input tolerance very low, which allows them to run happily when plugged into the generator. The MatrixUPS was never able to do this, even on the lowest sensitivity setting.
I have a few more pictures, and high-resolution versions of the above in my gallery.