I watched as a quite familiar company began installing a backup power generator for the building. It took several months, but was finally connected and ready for operation.
It was tested several times and the all-knowing IT department experts said everything was good-to-go. BTW, this installation was not a broadcast site, but you may know the company.
About three weeks after the system was declared ready for service, the building suffered a power failure. Let’s see what happened.
First, the generator didn’t start quickly enough. Instead of the few seconds promised, it took several minutes. The reason it took several minutes was because the automatic switchover function didn’t work, and the help desk technicians had to manually start it. Once the generator came online, dozens of large data servers began simultaneously grasping for power like a politician grasps for money.
Can you say "electrical surge?"
With this unexpected high demand overloaded the generator, you guessed it, the breaker tripped. This again dumped servers. Now, even after spending tens of thousands of dollars for an emergency power generator, the company was still totally offline. What went wrong?
I'll start by noting that this is typical of what can happen when people not trained in power calculations are in charge of making critical power-related decisions. It is easy to let the IT people handle it, especially in organizations that don’t have an engineering department. However, as this example shows, power-equipment-related decisions are better left to engineers and companies who understand the intricacies of sizing, generating and distributing electrical power. Most facilities are probably better off hiring outside expertise when it comes to AC power systems.
Let’s review some basics of power.
Power (P) in watts = E x I or I2R or E2/R
The above formulas apply to resistive loads only.
However, most AC devices have a reactive factor, which affects the calculation. This would be the case for motors or switching power supplies, which are ever present in computer systems. Now we need to add a third element to our power calculations, the Power Factor (Z).
Apparent power (S) measured in volt-amps = E x I (resistive load only) or I2 x Z or E2/Z
The power dissipated by the equipment is sometimes called true power and is expressed in watts. It may also be listed on the device in volt-amps (VA). As shown above, in reactive devices, one watt doesn’t necessarily equate to one VA. If you mistakenly assume these two types of power are the same, expect the kind of problem described at the beginning of this column. Such errors can lead to making incorrect UPS, generator and circuit breaker decisions.
The simplest formula for calculating the required power to be provided by a UPS system is:
VA = (I x E)/power factor
Example: 2.5A x 120V = 300/PF
If the PF is 1, then watts do equal VA. However, that’s not typically the case.
In addition to the formulas, it's important to realize that UPS manufacturers have adopted a de facto standard where the wattage rating is only about 60 percent of their VA rating. Why? Spectsmanship. The VA number is typically higher than the watts number. In a sales comparison, bigger numbers are better. Let’s look at two examples where it might be easy to mistake the capability of a desired UPS system.
Example 1: Will a 1000VA UPS power a 900W heater?
We know from the above specification that the power supply is capable of 1000VA. However, we need to factor into that number the true power in watts. Because the UPS industry typically specifies UPS performance in VA rather than in watts, the real available UPS power will be 60 percent of VA.
60% x 1000VA = 600W of true available power
Therefore this 1000VA UPS will not properly power a 900W heater.
Example 2: Will a 1000VA UPS properly power a 900VA video server?
Let’s assume the video server has a power-corrected supply, meaning the PF is approximately one. All large data equipment manufactured after about 1996 use a power factor corrected supply. For this example, it means the server’s real power consumption is about 900W.
However even though the UPS VA output and server VA power requirement appear compatible, there is one more calculation we need to make.
Take the UPS output in VA and multiply it by the power factor of 60 percent. Once we do this, it's clear the UPS is only capable of providing about 600W of power. So, even though upon first glance the UPS appears to be capable of powering the server, closer inspection shows that’s not true.
1000VA x 0.60PF = 600W
The UPS is capable of only 600W output, meaning it cannot power the video server.
The last time I looked, there was no formula for translating bits into watts. That suggests to me most companies ought to leave power system calculations and decisions to people who do more than install “software updates.” Next time someone says, “Don’t worry, the IT department has calculated the numbers,” be cautious.
For more information
Here are some additional resources on UPS technology you may want to consider. (American Power Conversion provided the background and examples for this blog.)
American Power Conversion
Pentadyne Power Corporation
Staco Energy Products
Several good white papers on harmonics and power factor
MGE UPS Systems
Several good white papers on UPS technology