Clifton Laboratories 7236 Clifton Road  Clifton VA 20124 tel: (703) 830 0368 fax: (703) 830 0711

E-mail: Jack.Smith@cliftonlaboratories.com
 

The results of my first measurement series lead me to update the K3's firmware and perform a power calibration. This resulted in significantly better agreement between commanded power and actual output power.

The figure below shows the test setup. The K3 and HP 437B are under control of a program I wrote with EZGPIB. Not shown in the setup is the DC power supply, an Astron 20 ampere unit. I've written about EZGPIB and the Prologix USB-GPIB adapter at EZGPIB and Prologix GPIB Adapter

The 8482A sensor is thermocouple-based, accurate over the range 100 KHz - 4 GHz. The 437B is a digital instrument that reads the sensor output and makes it available over a GPIB interface.

The 8482A's basic accuracy is rated, over the 160 meter - 6 meter frequency range of the K3, is 1.5% or less. This figure is the square root of the sum of the squares of all the individual uncertainties associated with the 8482A. The error associated with the 437B is roughly the same, so the instrument error is around 3%.

This vintage HP power meter/sensor combination are calibrated against a 1 milliwatt 50 MHz power reference oscillator built into the power meter. I've compared the 1 mw reference in the 437B and in an HP435B analog power meter and found them to agree within 1%. I've also found agreement within 1% of the -10 dBm 30 MHz reference output of two Advantest R3463 spectrum analyzers. Finally, I've compared the power meter and sensor readings at a selection of frequencies and power levels against the voltage readings obtained with a Tektronix TDS430A 400 MHz digital oscilloscope and likewise found agreement within 1%.

These measurements suggest that HP's 3% figure is somewhat pessimistic and the 8482A/437B combination is better than that.

Since the 8482A sensor is rated for 100 mw, it's necessary to use a power attenuator ahead of it. I use a Bird 8323 100 watt, 30 dB attenuator, with a Minicircuits HAT-3 3 dB attenuator on the output side. This extends the sensor's range to 200 watts into the attenuator.

Over the frequency range 100 KHz to 54 MHz, the measured attenuation of the Bird 8323 and HAT-3 are within ±0.04 dB of -32.97 dB. The plot below shows the attenuation as measured with an HP87510A vector network analyzer. I used a single -32.97 dB correction factor entered into the HP437B.

The commanded power levels (in watts) are:

The program commands the K3 to the transmit frequency and then steps the output to each of these specific output powers. The output power is set using the K3's "PC" command. The transmitter is activated via the "tune" command, sent over the RS232 command line. Four seconds later, the 437B reads the power. The K3 is then un-keyed for a 6 second cool down period.

Although I set the last power command to 125 watts, this is beyond the K3's rated power, so I've deleted that from the data set. I also suspect that some of the power drop for levels above 100 watts are due to my Astron 20A power supply suffering voltage sag.

The first measurement run I took was with my K3's "as shipped" firmware and factory calibration. (I purchased my K3 assembled from Elecraft.) The firmware revisions in this run are:

1. The K3 shifts from low power mode to high power mode at 12 watts. There's a clear discontinuity in power seen at this shift point.
2. The 50 MHz band shows a major drop at 10 watts. This results from a quirk in the K3's  power setting function. When in the 50 MHz band, my K3 will not accept power settings between 8.1 and 12 watts. When commanded to 10 watts, the K3 firmware sets the power at 8 watts.
3. The K3's wattmeter is calibrated at 14 MHz for two power levels, 5 and 50 watts. My K3 was not accurate at 5 watts, 14 MHz, nor at 50 watts, 14 MHz, being significantly high in both settings.
4. In general, the frequency bands between 1.8 and 24 MHz are grouped reasonably closely together, with 28 and 50 MHz bands showing greater divergence. This suggests errors in the K3's directional coupler.
5. The second type of error is a consistent overage, in that the measured output power exceeds  the commanded power always high with the error decreasing with increasing transmit power. This suggests a problem in the way the detected RF level is transformed to DC and read by the K3's A/D converter and processed into an RF power reading.
6. The drop in power for levels > 100 watts at higher frequency is mostly due to droop in output voltage of the 20 A Astron power supply used in the tests.

These results lead me to update the K3's firmware to the newest beta editions (as of 19 December 2008) and to perform the power calibration process described in the K3's manual. I did the following:

1. Update firmware to:
    

   Parameter	Value
   K3 S/N:	1378
   K3 Main FW Revision:	RVM02.73
   K3 Main DSP Revision:	RVD01.98
   K3 AUX DSP Revision:	RVA01.98
   K3 FRONT PANEL FW Revision:	RVF00.02

    

2. Calibrated the A/D reference voltage. (Was 5.00; measured at 4.972V and  reset to 4.97.
    
3. Calibrated the wattmeter function at 14 MHz, 5 and 50 watts as described in the K3 Operator's Manual.
    
4. Calibrated transmitter gain, using the automatic calibration feature in the K3 Utility.

The figure below shows a slightly different view of the same data - frequency flatness and error at 50 and 100 watts commanded power.

At 50 watts, the wattmeter calibration point, up through 24 MHz, the K3's wattmeter, post calibration, shows very good performance, with the maximum error being less than 2 watts. Even at 28 MHz, the error is only 5%, which is quite good.

At 100 watts, the data is quite consistent with the 50 watt measurements, but with an offset error of around 7 watts. This suggests a linearity or gain error in the K3's power measurement firmware.

We also see the same drop with increasing frequency. In working with directional couplers a few years ago, I've seen this behavior result from stray capacitance and inductance in the coupler. It's challenging to make a flat over the range 1.8 to 54 MHz. In terms of dB, the K3's drop in commanded power at the  50 watt level at 50 MHz is only about 0.35 dB. This is a rather modest error.

1. If you have the appropriate test equipment, recalibrating your K3's wattmeter and gain settings with newer firmware loads is beneficial.
    
2. Measuring RF power accurately is not a simple task, and most power meters available to hams are not all that accurate. Telepostinc's LP-100/LP0100A, when factory calibrated, are an exception and provide much better accuracy.
    
3. Elecraft has room for improvement in wattmeter accuracy through firmware revisions, although kit builders will have the same issue of securing an accurate wattmeter to perform the post-construction calibration.