The Most Common Power Meter Mistake

By Jason Kovatch, RF Test Engineer

At least 50% of all power meter users make this mistake. It has happened to you. You measured an RF
power level and later someone said, but I measured the power and got a different number. Who was
right? Measuring RF power appears easy. Modern wattmeters and power meters are simple to use, and can
provide digital measurement data to several decimal places in dBm or watts. But if they are so accurate,
why is it so difficult to make good measurements? If you are making the most common mistake, correcting
for it will dramatically improve your measurement accuracy.
For the most common power meter application, measuring the level of power in a signal, many users tend
to forget that a power meter is not measuring what you want to know- the power level of the signal.
Instead, it is measuring the total power over the entire bandwidth of the sensor, which for measurement
purposes is practically infinite!
Power meters versus Wattmeters
Most people recognize that a calibrated power meter is a superior measurement instrument compared to a
wattmeter. But what is the difference between a wattmeter and a power meter? Generally, a wattmeter is
similar to a power meter in that they both measure broadband power, but unless you are correcting the
power meter for frequency, you are using it as a wattmeter. The perceived plug-and-go measurement
ability of power meters is one of the misconceptions that endear power meters to users. This paper
describes how to account for frequency in your power meter measurements, making them as accurate as
possible.
Correction factors, what are they?
Power meter sensors that do not employ electronic calibration come with a graph or tabular data showing
the calibration factor and correction factors by frequency. Most users will take the reference calibration
factor (CF) in percent and use it to cal the power meter at the cal reference frequency, for instance at 50
MHz. But what are all the other numbers for? They are correction factors, similar to the reference CF, and
indicate the response of the sensor to power as measured at those frequencies across the entire
measurement range of the sensor. The numbers are usually a percent of the full scale response, and can
vary from 80% or so to 100%. Lets say your sensor reference CF is 100% (not uncommon) and the CF of
the frequency at which your signal of interest occupies is 95%. If you, like many users, carefully perform a
cal and then plug the sensor to the signal port and make your measurement, it is still in error by at least
5%, or about 0.2 dB. Since power sensors are available with correction factors as low as 90%, your
measured power can be in error by as much as 10% without taking other known uncertainties into
consideration. Assuming the power to be measured is within the sensor range, the signal-to-noise ratio is
acceptable, and the VSWR of the measurement port is acceptable, then the most important correction
you can make to improve the error is to account for the frequency of measurement.
First, remember that measurement accuracy at higher power levels makes a much bigger difference than
those made at low power. At -20 dBm, a 0.2 dB error is only about 0.5 uW, but at +55 dBm the error is
over 15 Watts!

21222 30th Drive SE, Suite 200, Bothell, WA 425-485-9000 www.ar-worldwide.com

Page 1 of 3
091410

Correcting for the test frequency

Fig.1 Power sensors with Cal Factor (CF) data.

CF as low as 90% is shown. For frequencies not enumerated, use straight-line interpolation of adjacent
data. If your measurement frequency is not enumerated or between data provided, you are using the wrong
sensor.
Measurement system slope and offsets
Applying corrections by frequency to power measurements goes beyond simply correcting the sensor. If
the RF power level to be measured is not connected directly to the power meter head, whether the RF
path is just an attenuator or an entire test bench setup, you must account for the slope of the RF path to
correct your measurement. A common method uses a measurement Offset to add the attenuation of the
path loss the power measurement displayed. Unfortunately, a single offset is frequently used, as the meter
may only retain a single value. The Offset value must be changed for each measurement frequency that has a
Page 2 of 3
091410

different loss. Some e-cal power meters allow you to input a table of offsets that will be interpolated by the
meter. This is a great feature, but will occur only when you input the measurement frequency. High quality
attenuators will come with calibration data, or measure the slope with a network analyzer for reference.
Bandwidth and Noise
Due to the high bandwidth of a power sensor, when using a power meter to measure power in a signal, the
signal-to-noise ratio is important. Remember that Noise Power = KTB, and for a sensor BW of 26 GHz, B
> 100 dB! Noise power can dominate the total power measured when signals fall as low -50 dBm or so.
Adding a post-amplifier will not help the signal-to-noise, as the noise will be amplified as well as the
signal. In this situation a filter is wanted, ideally a band-pass filter at the frequency of interest. Lacking a
band-pass filter, a high- or low-pass filter (or both) can help. Check the ON/OFF power level of the
system noise without the signal to see if the noise power is a factor in your measurement.
When making measurements of signals at higher power levels it is important to be aware of strong
harmonics or other spurious signals that may contribute too much to the total power measured. Check the
signal with a spectrum analyzer to make sure. In this case, again, filters need to be used to subtract that
power before the sensor can see it.
At lower power levels a good spectrum analyzer (SA) may make a better measurement than a broadband
power meter, as the SA will employ detection and filters that exclude the noise power.
VSWR can make a difference
Another practical power meter measurement correction that you can easily make is to assure yourself that
the measurement uncertainty created by VSWR is minimized. When measuring with a sensor on an
unknown port, you can make a quick check with a 6 dB pad to see if the measurement improves. Adding
6dB of loss increases the return loss by 12 dB, and will improve poor VSWR by a substantial amount,
allowing a better measurement. The easiest way to do that is to use the power meter Relative
measurement feature to zero the displayed power level, then add the pad and see if the level changes by
6 dB. As long as the signal level is still in the sensor range, if it changes more than the pad value the
VSWR is adversely affecting the measurement, and you should use the pad. It might need more
attenuation than 6 dB. In any case, you must enter the percent correction as shown on the attenuator for the
frequency being measured or you must input the measurement frequency.
To reduce the measurement uncertainty of a relative measurement, try to keep the power applied to the
sensor at the low end of the power range. The relative uncertainty can be as high as 6% at the high end,
and as low as 1% at the low end.
Many factors contribute to making a successful power measurement- using the correct type of detector for
your signal of interest, making allowances for gating and various forms of modulation, and more. But if you
can avoid the biggest mistake and remember when you use a power meter that you are making a
broadband measurement, not just measuring a discrete signal, you are on the way to getting better and
more consistent results.

Page 3 of 3
091410