TN 0035: A Discussion about USP devices 

Author: Chris Albertson
Date: 970818
Revision: #0 979818
Key Words: instrumentation

I have been learning about UPSs. Here is a little info that may help. I am not so good at writting TNs but maybe this an related info will find its way into one.

Along the same lines as the build v. buy for the reset board. It seems some UPSes can perform some of the tests Tom wants the MkIV to do. So this could give some choises when it comes time to do the standard engineering dill of allocating requiremnts to subsystems.

First a couple side isues.

  1. Tom wrote about a lack of a standard connector for re-seting a PC. Every PC I have seen, going back to the first IBM 8086 powered PC can be reset by shorting two pins on a standard 0.1 inch spaced header. The watchdog boards will connect to this header and short the pins either under software command (seems dumb to do) or because an on-board timmer has expired. The board also supplies a DB25 connector so a reset can be triggered by an external device and/or the reset can be anouced to some external device. Hardware that performs this function is one part of a larger system that needs to by planned out.

  2. There is a "HOWTO" about how to assemble a high availabilty Linux system. The normal application is an ISP who has a remote office filled with Linux boxes that interface phone modem banks to a high speed trunk line. Offices are scatterd around a city to provide local number toll free Internet access. These distributed sites feed a central office where the "real" servers are located. The "offices" are many times run "lights out". That is with no human operater physically present. At least not a full time operator. The "office" is many times just sub-leased space. In some cases just a self in someones phone closet. Needless to say the continued operation of these machine is critical to the company's bussiness. Running one of these setups is not a whole lot diferent from running a remote TASS site. The bottom line is that a forced hardware reset or power cycle is a last ditch resort. It's use implies that some other watchdog/safty subsystem as failed. It is akin to a life vest on a cruise ship. You hope 1) the crew sees the iceburg, 2) the water tight compartments work, 3) the life boat floats, if all of the above fails then only do you realy need the life vest.

Now back to UPSs. I find there are three basic types in terms of how they interface to a computer.

  1. brain dead UPS
  2. The smart UPS
  3. The Very smart UPS

The Very smart UPS:

If you want to end any worry about power probllems, UPS failures Spikes, etc. get a type #3 very smart UPS of at least 1KVA size and hook eveything up to it, computer, STAMP, telescope drive, TEC supply, the works. If 1KVA is not big enough then get a bigger one. The only problem is cost. This solution could run to $1000.00 or more.

The very smart UPS addresses all the concerns about UPSes. It will do it's own test periodically by switching over to batery power and measuring the batery voltage under load as a gage of batery condition. It will tell the computer about an AC power fail also tell the computer when a low batery condition occures. It can also inform the computer when the batery performence (ability to accept charge) is getting low. There is a whole set of built in dianostic tests that can be run remotely.

The "brain dead UPS":

No ability to send data to the computer. When the AC fails there is a beep but if there is no operator near by to hear and react to the beep-beep-beep alarm all the UPS does is delay to time when the AC to the computer fails by five mintes or so. Not so usefull.

BUT, the "brain dead UPS" can be helped and made as usfull as the smart UPS for no or little cost or work. See below.

The smart UPS:

Actualy not that smart. This was the only kind of smart UPS before they started butting microcontrollers in UPSes. these UPSes have a serial port but do not send/receive data. all comunication is via the modem control lines. Basically the UPS just drops the "Device Ready" RS232 line when the AC power fails. Some better ones will drop a second line a minute or so before the battery runs out of power. This kind of interface does not require a computer inside the UPS and is the type of interface assumed by most software today as it is the "lowest common denominator" interface and can be expected of any UPS.

Here is what the software does: When the "device ready" pin on the comm port drops the software schedules a shutdown for N minutes in the future. When the pin comes back up (if it does) any scheduled shutdown operation is canceled. So if the power outage is < N minutes we do not actually shutdown anything, just ride it out.

This type of UPS is good enough for un-attended operation if you can live without the builtin dianostics and health monitoring.

How to make a "brain dead" UPS smarter

Now, I happen to own a cheap "brain dead" UPS and need the features of a smart UPS. I though it would not be to hard to use a relay to trigger the device ready line on an unused comm port. Maybe a power cube and 2n2222 transitor an a pull up resistor I thought. Well there is an even simpler method I found in some Linux document: Just connect any old external 300/1200 baud moden from your just pile to a comm port. Power the modem from the _wall_ AC plug. Now when the AC fails the modem's devive ready line also fails. I'm glad I saved that old modem.

It turns out my brain dead UPS is not so dumb. It does a test every seven days where it switches to batery power and tests the batery voltage under load and the recharge rate. It will light up a LED if the batery appears to be near its end of life. They say this will happen in 3 to 5 years of normal use. It uses a 15AH sealed lead acid batery. The "replace batery" alarm will sound weeks before the batery acualy fails so I only need to go near the UPS to look evey few weeks. The CRT does not need an UPS so a small one will do. My $67.00 200VA unit works fine for my needs.

Back to Index of Technical Notes
Back to TASS home page