Sam Wetterlin



There are two parts to the initial calibration of the Modular Spectrum Analyzer. First, the response over signal level must be calibrated for each final IF filter. This is called Path calibration. Its primary purpose is to establish the relationship between raw readings of the analog-to-digital converter and the true input power level. This requires applying a broad range of input levels of known strength, at a single frequency. For an MSA with the VNA feature, this calibration will also establish the extent to which the internal phase shift changes with signal level.


The second part is calibration of frequency response over signal level. This calibrates the response of the input mixer (combined with variations in the LO to the mixer). It requires applying known input levels at various frequencies spanning the full range. It is actually not critical that the absolute input levels be known, as long as we know how the input varies over frequency. If Path calibration is performed at 1 MHz, then for frequency calibration we don’t need to know the exact level at 1 MHz, but we need to know how the input at other frequencies compares to that at 1 MHz.


A builder with access to a signal generator for the 0-1GHz range, capable of producing calibrated outputs over the range 0 dbm down to -120 dbm, does not need any further equipment to perform these calibrations. But for others there are several accessories that will be helpful in calibration. I have combined the schematics for these into an ExpressSchem file, and the PCB layouts into an ExpressPCB file.


The Switched Attenuator can be set to attenuations from 0-57.5 db in 2.5 db steps. It consists of stages of 20-20-10-5-2.5 db.  For calibration it is intended to be used at a frequency no higher than 2 MHz, though it is functional to 500 MHz or so. The objective was to make the attenuation values as precise as possible, to make the input and output impedances as close as possible to 50 ohms, and to make the phase delay fairly equal at all settings. The error in each stage of the Switched Attenuator is intended to be no more than 0.02 db. When multiple stages are engaged, the total error can be the sum of the individual errors. This error level is difficult to achieve with commercial stepped attenuators, or with combinations of commercial fixed attenuators.


The Switched Attenuator can provide many of the various signal levels needed for Path calibration. But to go beyond 57.5 db of attenuation, it is necessary to add one or more fixed attenuators. The files referred to above include schematics and PCB layouts for such fixed attenuators. One 20 db attenuator and one 40 db attenuator will generally be sufficient, but many more values are shown in the schematics. These attenuators are intended to have very precise values at low frequencies, and tests show that they hold their values well to 1 GHz. They are not as precise as some commercial attenuators when you exceed 1 GHz, but at low frequencies they are much more precise than any commercial attenuators I have tested.


The attenuators provide the ability to create precisely varied signal levels, but the user still needs to start with a signal of known strength. The Calibration Source is intended to fill this need. It generates square waves with a very good 50% duty cycle, minimal overshoot/undershoot, and a very good 50-ohm output impedance. The key to its use is that the exact level of the fundamental can be determined by a DC measurement with an ordinary voltmeter.


This provides the perfect signal source to feed the Switched Attenuator for use in Path Calibration. It can be built with an on-board oscillator to provide a frequency in the 1-2 MHz range. (Actually even 10 MHz is feasible, but for calibration the lower frequencies are better.)


The Switched Attenuator, fixed attenuators and Calibration Source are all intended for use in Path Calibration, and provide everything that is needed for that calibration. But for Frequency Calibration, we need the ability to generate a broad range of frequencies, all at known signal levels. To generate the various frequencies, we can use a signal generator, or we can use the MSA’s tracking generator (TG) if it has one. In either case, though, the output level is not likely to be known precisely.


That brings us to the Leveler. The Leveler uses an AD8367, which is a variable gain amplifier with both an RF output, and a DC output that is proportional to input power level. The power detection capability is used to control the gain, such that the output comes out at a fixed level. Hence, no matter what the input level (within reason), the output is leveled at a precise value. Up to 500 MHz, the input can vary from 0 dbm down to -30 dbm with excellent leveling. By 1 GHz, the input must be in the range 0 dbm down to -15 dbm. The ideal input is anything within a few db of -10dbm, which is the nominal level of the TG output.


Over a considerable range of input levels, then, the Leveler output will be approximately -7.7 dbm.  Each AD8367 will fix the output at a slightly different level. For our purposes, that is fine, because our only requirement is that (1) the output not vary with modest changes in input level and (2) whatever the output is at 1-2 MHz (or whatever our reference frequency is to be) the output at other frequencies must either be the same, or must have a predictable relationship. In fact the AD8367 output stays constant within about +/- 0.1 db up to 500 MHz, and then gradually drops, falling about 1 db by 1 GHz. However, tests of three separate builds show that the profile of each AD8367 follows a predictable path. Therefore, good results can be achieved by using the following assumed output levels:


                  1 MHz       -7.70 dbm

                40 MHz       -7.80 dbm

              175 MHz       -7.60 dbm

              400 MHz       -7.80 dbm

              900 MHz       -8.60 dbm

            1000 MHz       -8.85 dbm

            1050 MHz       -8.85 dbm

The levels shown in the table appear to be accurate within 0.05 db to 500 MHz, with the error increasing to perhaps 0.5 db by 1 GHz.


The Leveler can be used with any signal source that provides a modestly decent sine wave output. With a signal generator, the output can be fed to the Leveler, and the leveler output can be used for Frequency Calibration, after passing through an attenuator of approximately 30 dbm. The exact attenuation is not critical, so long as it remains constant over the frequency range. Various frequencies are set, the MSA is set to a zero-width scan at that frequency, and the Calibration Manager is then told to retrieve the frequency and measured power.


A more automated procedure is possible if the TG is used, because the MSA can control the TG frequency. If the TG is fed to the Leveler, followed by the attenuator and then into the MSA, the user can set the MSA to scan the range 1-1000 MHz in perhaps 100 steps, and then tell the Calibration Manager to retrieve the entire set of readings at once. This provides calibration in approximately 10 MHz steps with very little effort. But it is necessary to take into account the fact that the Leveler output is not identical at all frequencies. There is a simple trick to do this. Frequency Calibration is normally done with a default frequency calibration file in place, which provides for zero adjustment at all frequencies. All we have to do is replace that file with one that makes adjustments for the known profile of the Leveler. Based on the profile above, the following frequency correction factors will do the job:


                  1    0

                40    0.10

              175    -0.10

              400    0.10

              900    0.90

            1000    1.15

            1050    1.15


The left column is frequency in MHZ; the right column is the correction factor, which is positive if the Leveler output at a frequency is less than that at 1 MHz. The above table can be used to produce a frequency calibration file as follows: (1) select the frequency file in the Calibration Manager; (2) click Display Defaults to get a default table with two entries; (3) replace the two default entries with the above table by using copy/paste; and (4) click Save File. 


After establishing the initial frequency calibration file, you can proceed with automated Frequency Calibration by setting up a broad frequency scan and re-entering the Calibration Manager to record the readings.


This initial calibration file can also be used for the “manual” Frequency Calibration method, with a signal generator providing input frequencies through the Leveler one frequency at a time. With this file in place, whether you do manual or automatic Frequency Calibration, you can do a preliminary measurement of the leveler (plus attenuator) with the MSA at the reference frequency (i.e. the one at which you did Path Calibration). Whatever that measured level is, you treat that as the true power level for all frequencies being calibrated.