[hpsdr] Level calculation error in Two Tone Test generation routine

[john_eckert at keysight.com]

Hello All,

With all due respect I still believe the levels in HPSDR are incorrect.  I am very familiar with the math, the understanding of peak vs. average and how to measure power.

For the last 30 years I have worked for the world's largest test and measurement company and I have helped dozens and dozens of companies make RF/MW measurements.  I only bring this up because I'm hoping you will hear me out.

Let's take a look at:

1.       History

2.       Industry conventions

3.       A little theory

4.       Signal Generator Example

5.        ARRL Test Method

6.       HPSDR Code

Simply said, IMD or two tone measurements are an indicator of how a device will create signals at its output that are not present on its input.


History:
The most straightforward technique to measure amplifier linearity is to apply a range of known input powers to the amplifier under test and accurately measure its corresponding output powers.  In practice this gives the designer valuable information to tweak his design, but how do we use this information to predict the size of the distortion products?   Nowadays we can apply an FFT using a digital processor and get the answer, but not widely available pre-PC days.  Another technique, passing two tones through the amplifier reveals the non-linear terms as spectral components in the frequency domain.  You can think of the amp as an analog computer computing the output based on its linear and higher level coefficients.  These components are easily measured with a tunable receiver or displayed on a modern spectrum analyzer.


Industry:
Amplifiers used in cell phones, MRI machines, satellites, RADAR systems, ... have a range of specifications, such as frequency range, noise figure, 1dB compression, PAE, gain, and match to name a few.  As you can imagine there are many interpretations on how to measure these parameters.  Let's look at power.  Some manufacturers specify the output power level at the point where the gain is reduced by 1dB.  The 1dB compression point.  Others may give the saturated power.  A system designer on the other hand may request from an amp supplier that he needs a part 'puts out' +30dBm with 3rd order IMD >40 dB at 3.2 GHz.

Let's look at the 1 dB compression measurement.  In this case you measure the gain at a low input level and increase input power until the gain drops by 1dB and report the output power level.  As it usually turns out if you make the same measurement starting at a high power level and reduce it until the gain increases 1db you get a different result!  The so-called amplifier memory effect.   The thing that makes Warren's job so hard.


Theory:
When you apply a single CW(non-pulsed) tone at a given power level to an amplifier and add a second tone at the same power level it will double the output power or put another way it will increase the power by 3dB.   If the amplifier is still operating in its linear power range all is well.  However, if the first tone was already driving the amp to full linear power adding the second tone will drive it into saturation. Clipping and its accompanying distortion will result.  If we reduce the power of the first tone by one half and add the second tone with the same power the result is an output power equal to original single tone/CW case.

The question to be answered is, how much of the of the amplifiers range do you want to use for the measurement?   I contend that you want to measure it at the same level that you operate it at, if your goal is to understand how it performs in that range.  We understand that there is a large amount of non-linearity at the high end of the power range of an amplifier, backing down the power certainly cleans things up.  Running the IMD test at half power does not correctly reflect the real performance under realistic operating conditions.

Spoiler alert:  The ARRL measures it the way I suggest,  I'll get to that in a bit.


Example:
Laboratory grade signal generators are specified by frequency coverage, spectral purity, power output and level accuracy to name a few.  The output level specification is a very important specification if your device is sensitive to power.  Examples are receivers, mixer characterization, power sensor calibration.  A quality signal generator will give you precisely the output power you have set it to.  It will do it for days on end, in spite of internal temperature changes and supply voltage fluctuation.   And, the biggest and, is that it will put out the desired power independent of the modulation type!!!!  A simple diode detector keeps tabs on the peak voltage and maintains it via the ALC loop.

If you happen to have a vector(IQ) signal generator set the output power to -80dBm with a CW signal on your favorite ham band and note that your well calibrated radio spectrum display reports -80dBm.  Now switch on two tone modulation and you will see two tones each drop to -83dBm.  The total power is sum of these.  Math wise convert to watts, add and convert back to dBm = -80dBm.  Or just remember adding two equal amplitude signals doubles the power which is +3dB.

Now if you turn on four tones, the original two tones will drop another 3dB and all four tones are now at -86dBm.  Total power is still -80dBm.

If you connect an oscilloscope to the load you will see that the peak to peak voltage is exactly the same in all these cases even though the waveforms differ considerably.



ARRL Test Method:
From:
http://www.arrl.org/files/file/Technology/Procedure%20Manual%202011%20with%20page%20breaks.pdf

The ARRL procedure applies two tones of audio to the transmitter.  The power is measured on an inline power meter, then attenuated and applied to a spectrum analyzer.   Below are snippets of the test procedure.  In summary, with the two tones applied, the transmitter is adjusted to full output power and the spectrum analyzer is adjusted to minimize its own IMD and the results are the difference in amplitude of the tones and the third order products.  We have all of these components built into our radios.

"4.5.3.3 Tune the DUT per the operator's manual for the test frequency of 3.900 MHz. Turn on the two-tone generator and set both tone switches to ON. With the DUT in the LSB mode, set the generator LEVEL and ATTENUATOR controls for the maximum audio input as specified by the manufacturer. If the manufacturer does not list a specification for this figure,  adjust the 2-tone generator's amplitude for maximum rated RF output of the transmitter with the transmitter's microphone gain near maximum. Observe the transmitter power as shown by the wattmeter. Ensure that the output power of the DUT is not greater than the manufacture's maximum power output rating. Unkey the transmitter and set the step attenuators for approximately -46 dBm input to the spectrum analyzer."


NOTE:  " adjust the 2-tone generator's amplitude for maximum rated RF output of the transmitter".   PowerSDR is not doing this.  It is setting the power to one half the user power level setting.


"4.5.3.5 Adjust the REF LEVEL control (and step attenuators, if necessary) for the peak of the two pips to be at -6 dB. The spectrum analyzer is now calibrated. The amplitude of each IMD distortion product may now be read in dB PEP (dB below the peak envelope power) directly from the display."

NOTE:  The -6dB is an adjustment for the spectrum analyzer not the transmitter.
"The spectrum analyzer is now calibrated."  IMD is a relative measurement.  Changing the attenuators, in front of or inside the spectrum analyzer, or changing the spectrum analyzer reference level does not change the conditions of the transmitter.  The reason the ARRL does this is to insure the measured IMD is not being produced inside the spectrum analyzer.


HPSDR Code:
Here is the code I refer to, 'ttmag1 = ttmag2 = 0.49999 * Math.Pow(10.0, ttmag / 20.0);'  is setting the magnitude of the tones.   Hence, ttmag1 and ttmag2.  They are essentially scaled values referenced to the power control setting.  IQ is figured out later.


Final Comments:
I started this journey with a 5 watt Flex 1000 and have made several contributions to PowerSDR over the years.  I am now running a pseudo homebrew setup using OpenPowerSDR driving a liquid cooled solid state amp.  I stumbled on this two tone issue while trying to understand which stage/stages in my amplifier chain are the major contributors to my total TX IMD.

My hat is off to all those that contribute to making the OPENHPSDR the best platform for amateur radio experimenters.  It's simply wonderful that we can make a radio that outperforms the best of the commercial offerings and that they are running as fast as they can to catch up.   Unfortunately, when they do catch up, the radio will still be their way and not your way!

Power To The People (pure power).

73,
John
K2ox


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openhpsdr.org/pipermail/hpsdr-openhpsdr.org/attachments/20161123/0c91e5f1/attachment.htm>