Power meter basics
Power meters today are almost always digital. They convert an analog signal, referenced to fifty ohms, to a four digit display of power in watts, milliwatts or microwatts, or in decibel milliwatts (dBm). Nine out of ten microwave engineers prefer decibels for power measurements, and you will too once you understand how to think in dB.
By far the best power meter to put into your microwave test setup is the Hewlett Packard 438A dual power meter (or Agilent E4419B dual power meter for those of you with a more modern lab, since the E4419B obsoleted the 438A power meter).
The Hewlett Packard 438A dual power meter allows you to make relative measurements between two power heads (A with respect to B, or B with respect to A), without the need for scrap paper or a calculator. This feature allows you to automatically track the relationship between input and output power of a device over a range of power levels and directly read out the gain compression without plotting Pin/Pout.
Another feature of the HP438A that is quite useful is its ability to take readings relative to a reading on the same power head, that you store in memory. This is done by pushing the "REL" (relative measurement) button. Still another excellent feature is the ability to add offsets to either of the power heads. This is very useful when you have to remove losses outside your DUT from your measurements (which is almost always).
Power heads
Power heads are the sensors that convert RF and microwave signals to analog voltages, which are read and reported by the power meter. There are three types of sensors in use in power heads. The first two, thermistor and thermocouple sensors, convert the "heat" of the incident signal to a voltage proportional to power. Thermistor power heads are regarded as "old-fashioned", if you see any of these in your lab they are hooked to an old analog meter such as the HP 432A. The third type is a diode detector, which rectifies the signal into a DC voltage. Diode detectors provide a lower power range than thermistor power heads (and a lower maximum safe power), so if you are using a power head that measures down to -75 dBm, chances are you are using a diode detector power head. Another hint that you are using a diode detector is that HP (or Agilent) uses the suffix "D" for diode, as in "8487D". The exception to this rule is the now-obsolete 8484A diode detector power head.
If you want to learn more about power heads, Agilent's web site has an excellent application note on this topic and about a zillion others.
For power meter measurements, you will need to choose one or more power heads. Agilent 8480 series power heads that are commonly available are shown in the following table (some of these may be obsolete by now). It is important to stay within the "best" power response range of a power head ("throw away the bottom ten dB and top five dB of its range) . For example, if the power head is specified to for -75 to -20 dBm, don't try to use it below -65 dBm or you will experience slow settling times and flaky results.
Table 1. Examples of HP power head capabilities, 8480 seriesHP model number / Frequency range / Specified power range (dBm) / Best power range (dBm) / Maximum input power (dBm) / RF connector
8481A / 10 MHz
to 18 GHz / -35 to +20 / -25 to +15 / 25 / N
8481H / 10 MHz
to 18 GHz / -15 to +35 / -5 to +25 / 35 / N
8482A / 100 KHz
to 4 GHz / -35 to +20 / -25 to +15 / 25 / N
8484A / 10 MHz
to 18 GHz / -75 to -20 / -65 to -25 / 20 / N
8485A / 50 MHz
to 26.5 GHz / -30 to +20 / -20 to +15 / 25 / 3.5 mm
8487D / 50 MHz to
50 GHz / -75 to -20 / -65 to -25 / 20 / 2.4 mm
R8486A / 26.5 to
40 GHz / -30 to +20 / -20 to +15 / 25 / WR-28
Power handling of power heads
Never exceed the maximum power rating of the power head(s) you are using. The cost to replace a power head that has been blown up by exceeding its maximum power is $1K to $3K, depending on which one you destroy. You can always "increase" the power range of a power head by adding an attenuator to its input, which you can remove mathematically. Be sure that the attenuator you choose can handle the power as well. Note that a "poor-man's power attenuator" can always be constructed from a coupler and a high-power load. For example, a 10 dB coupler terminated with a 10 watt load makes a 10 dB/10 watt attenuator measured through the coupled port.
Interpolating the cal factor at frequencies between data points
Check out this page!
Calibrating the power meter
Before you start calibrating the HP 438A power meter, hit the preset button to put the meter into a known state. This eliminates all offsets and filtering and other weird stuff from leftover from the last user that might mess with your measurement and ruin your day. According to the HP manual, the meter needs one hour to warm up to be at full accuracy. Have another donut.
Calibrating a power meter is a two step process. First, the meter should be zeroed. When you zero a meter, be sure that it has no incident RF power by turning off all sources on the bench. Both power heads should be zeroed approximately every two hours to prevent drifting. Once in a while the meter will politely ask you to "PLEASE ZERO", this is normal. Again, just be sure that there are no stray signals at the power head when you zero it.
The second part of calibrating a power head is the "cal adjust". Here the power meter applies a very stable 50 MHz, 1.000 mW signal to the power head as a reference level and internally adjusts the gain of the meter to match the particular power sensor's response. Start by screwing the power head onto the reference oscillator output of the power meter. Use the minimum possible number of adapters (and no cables) between the power meter and the power head, but don't stress about this, since most stuff contributes negligible loss at 50 MHz. For the highest sensitivity power heads such as the 8484A and the 8487D, you will need to insert the 11708A precision 30 dB/50 MHz attenuator between the power head and the reference oscillator output to create a stable 1.000 microwatt signal. Remember that the 11708A attenuator is a calibration item and should NEVER be used in a test setup. Next push the "CAL ADJ" button. Then enter the 50 MHz calibration factor printed on the side of the power head (usually 100%), push "ENTER" and the meter will do its thing. If you want to check the results of the cal adjust step (highly recommended for neophytes!), leave the power head connected to the reference oscillator, input the calibration factor at 50 MHz and turn the oscillator back on. The meter should read 0.00 dBm or 1.000 mW. The cal adjust procedure should be performed to both power heads of the dual power meter, no more than 4 hours prior to measurement.
Finally, you need to enter the calibration factor at the frequency of operation at which you will be testing. This information is printed on the side of the power head and is a number usually from 90% to 100%. If you use the wrong cal factor by a few percent, the error is usually small. For example, if you leave the cal factor on 100% after cal adjust but it should be set to 98% for your measurements, your data will read 0.09 dB too low (equal to 10xlog[.98]). If you want to be super accurate, you can interpolate the cal factor when you are measuring frequencies that fall between the calibration points, which are typically every gigahertz.
One thing to consider when you are taking power measurements over a wide frequency band... instead of inputting the cal factor(s) each time you change frequencies, you can leave the cal factor set at 100%, and make corrections to the data later in a spreadsheet. You should be using a spreadsheet to keep track of input and output network losses and other stuff, not to mention plotting the data, so it is no big deal to enter the cal factors here as well. Heck, you can even use it to do the pesky linear interpolation between calibrated frequency points in the spreadsheet! The way to correct data that was taken using 100% CF is to divide the measured data (in watts, or milliwatts, or microwatts, but not in dBm) by the proper cal factor at each frequency point.
Power meter measurement errors
It is easy to take power data with a power meter. It is not so easy to take accurate, repeatable data, unless you understand all of the problems that can occur, so listen up!
Errors due to standing waves
A significant source of error with power meter measurements is the standing wave ratio of the power head beating against the DUT. The VSWR of the power head is usually low (less than 1.18:1 to 12.4 GHz for the 8481A power head for example). The bigger problem is usually your DUT. Suppose you are testing an amplifier with 3.0:1 VSWR (r=0.5) with a power head that has 1.18:1 (r=0.082). Because we don't have a handy Greek alphabet on this web site, just pretend that the preceding "r" was a "rho"... The high and low errors can calculated as:
An eight percent error in power is a sizable 0.36 dB error. You can avoid SWR errors if you use attenuators on both sides of your DUT to pad down its VSWRs. Of course, you will have to mathematically remove these losses from your final data.
Errors due to power head non-linearity
As mentioned before and described in Table 1, it is important to operate power heads within their best power range to reduce linearity errors. You can determine the linearity error of your setup by making a "through" connection in place of the DUT, displaying A/B power ratio set relative to the lowest input power of your experiment. Then sweep the input power up to the maximum power while observing the A/B reading. If the setup is done correctly you should see no more than +/- 0.1 dB errors across a power range of 20 dB or more.
A note about the above linearity test... for devices under test with substantial gain or loss, you won't be making a apples/oranges comparison of the linearity of the test setup with and without the DUT unless you put some further thought into this. If you are characterizing an amplifier with 20 dB gain, you could test the amp with a 20 dB attenuator on its output. Then the net gain, with and without the DUT will be approximately the same (zero dB), so that you can use the same power head over the same range of power. Or you could swap power heads between the calibration and measurements steps. Don't lose sleep over the "through" linearity verification, just be sure to keep the power heads within their "happy range".
Errors due to finite coupler directivity
Coming soon!
The "Best Setup" for measuring power transfer characteristics
Figure 1 below shows a highly useful Pin/Pout measurement setup, based on many years of experience slaving away out in the lab. This test bench could be automated using LabView if you are so inclined, but it works quite well for taking data manually. The setup exploits the A/B relative power measurement capability of the dual power meter to quickly find compression points of your device under test (DUT) without even plotting any data. The setup can be used to measure P1dB of two-port networks (amplifier, limiter, multiplier) and as well as three-port networks (frequency translators such as mixers).
Figure 1. Power test bench in operation
Referring to the figure, you'll need one RF signal source for amplifier measurements, and two for mixer measurements. Beyond the need for a second source, there are other key differences between measuring an amplifier and measuring a mixer. For example, you will have to keep track of the losses of the input and output networks as well as the power head cal factors at the different input and output frequencies (RF and LO) for mixer measurements.
Be sure that the sources you use are able to handle the RF and LO frequencies that your DUT requires. You should pick synthesized sweep oscillators over old-fashioned sweepers, so that frequency errors are minimized. Although many sources provide a built-in variable attenuator function that allows you to control the RF power level, you should consider using an external, infinitely-variable attenuator to control the signal level as shown, since this way you can quickly adjust the signal to within a few hundredths of a dB. We like to use waveguide rotary-vane attenuators for this purpose.
Signal source 1 must provide the power level needed to put the DUT well into compression, and source 2 needs to provide the proper local oscillator power level for mixer tests. For high power levels you can add power amplifiers to either source, but you may have to consider the effects of power amplifier broadband noise on the measurement, particularly if you use a traveling-wave tube (TWT). Perhaps more importantly, you will have to consider the effect of high-power signals on each and every component in the setup so that you don't roast anything. Remember, if you do barbeque a component, consider sending us a photo of the remains for the Microwave Mortuary!
Within the input network,the coupler following source 1 samples the input signal to the DUT. Attenuator A1 may be used to adjust the input power to power head A to put it into its "best" range (10 dB to 30 dB less than the maximum power). It also serves to reduce SWR errors on the input side of the DUT. The isolator that follows the input coupler helps reduce SWR errors if your DUT has a poor input match. It also prevents directivity errors, by keeping reflected power from the DUT from corrupting the power meter reading at power head A. If you use a high-directivity coupler in the input network this should take care of the problem without the need of an isolator. Be sure the coupler and isolator operate over the required bandwidth. To determine the measurement uncertainty of the input network due to directivity, you can perform the following check before you measure your DUT. Insert a matched 50 ohm load where the DUT would go and apply CW signal from source 1. Observe the power level of power head A. Now remove the matched load and attach a short circuit (if a short is not available an open circuit will be nearly as good). The maximum directivity error of the input network will be the difference in the two power meter readings. You should strive for less than 0.1 dB error here.
Cables A and B may not be necessary, they are used for convenience as well as mechanical strain relief. If you didn't use any cables at all, you could hang the coupler, DUT and both power heads of the sweeper's RF output RF connector, which might amount to 50 inch-pounds of lateral torque on the coax connection to the source. We don't have to tell you why that would be bad, do we?
In the output network, attenuator A2 can be used to adjust the power into power head B to keep it within the "sweet spot" during DUT measurements. For example, if your DUT is known to provide 30 dBm saturated output power, you might want a 20 dB attenuator on the output. Filter FL1 is extremely important for mixer measurements, it is there to reject RF and LO leakage that would corrupt the power head B reading. You need to find a filter that will pass the IF frequency and reject the RF and LO frequencies (by 30, 40, or 50 dB or more!) Don't have such a filter laying around the lab? Build one!
Configuring the power test bench for your DUT
Before you measure the power transfer characteristics (input versus output power) of a nonlinear microwave device, you should have an idea what to expect. Check out our page on nonlinear devices.
How do you know what power heads, couplers, and attenuators to use? The first step is to get to know what you are measuring, and what power heads you have available for the measurement. Go to the manufacturer's web sites, read the data sheets, heck, print them out. The two most important parameters for your DUT are gain (or loss), and maximum output power. Also, decide what frequency band you're interested in. Check out your cables, adapters, power heads, couplers, attenuators, and make sure that everything works well within your frequency band. Not sure about where your connectors crap out? Visit our page on microwave connectors!
The next two paragraphs need some further discussion, check back soon!
Let's start with the output network. What is the maximum output power of your DUT? You need to arrange the output network so that this power is about 10 dB below the specified range of your power head B. Suppose you are testing a two watt amplifier, and you have a 8481A power head (20 dBm is its highest specified power). You should choose a 2 watt, 20 dB attenuator for the output network. Then the highest power your power head will see is 13 dBm.