Spectrum analyzer self-calibration

Hewlett-Packard Journal, June, 1991 by Timothy L. Hillstrom, Joseph F. Tarantino

The HP 3588A contains extensive self-calibration capability that allows it to achieve very good amplitude accuracy and LO feedthrough specifications compared to those previously achievable. On power-up and periodically thereafter, the HP 3588A performs self-calibrations of the frequency flatness for all input ranges, the IF level and shape, LO feedthrough, and the source amplitude.

Internal Hardware Calibrator

The internal hardware calibrator provides a precision amplitude reference level of -20 dBm [ or -]0.02 dB from 200 kHz to 150 MHz. This signal can be switched internally to the receiver input for automatic range flatness and IF level calibration. Fig. 1 shows the calibrator block diagram.

The limiter conditions the input signal and desensitizes the output level to input level, allowing up to 20 dB of variation at the input with negligible output variation. The dc servo around the differential switcher provides precise control of the sum of the collector currents, which indirectly controls the average (dc value) of the individual collector currents. If the input to the switcher is reasonably symmetrical (i.e., low even-order harmonic content), then the fundamental of the switcher is exactly related to the dc value by

fundamental rms amplitude = (dc amplitude) [square root of 8] / [pi].

Thus the servo loop can, in theory, exactly control the level of the fundamental of the output. Practically, second-order effects such as finite switching times and base-to-collector feedforward exist. However these effects can be well-controlled with a simple frequency compensation network. The impact of employing this calibrator in the HP 3588A is reflected in the typical absolute amplitude accuracy specification of 0.2 dB from 30 k?Hz to 150 MHz.

First LO Feedthrough Nulling

It is important to minimize LO feedthrough to minimize low-frequency residuals and to reduce the problem of multiple tones in the IF at low frequencies. The HP 3588A incorporates a circuit to reduce the feedthrough level to at least 20 dB below range under all conditions.

LO feedthrough nulling is accomplished by a circuit which feeds a small amount of the main (swept) LO signal around the first mixer. The in-phase and quadrature components of this signal are adjusted by the instrument's main processor while monitoring the response when tuned to zero Hz. A very simple gradient search algorithm is used to find the minimum feedthrough level. The tuning is done by means of to "electronic potentiometers" which adjust the resistive and reactive impedance components of the feedforward circuit. The resistive component is adjusted by varying the current through a pin diode, while the reactive component is adjusted by a half-bridge composed of two WC (voltage variable capacitor) diodes (see Fig. 2).

IF Filter Shape Calibration

The overall cascaded IF shape must be measured so that the contribution of the IF response away from its center frequency does not degrade the performance for narrowband zoom measurements. This is accomplished by sweeping past an internal fixed-frequency signal so that the IF frequency response is traced out in reverse. This response is then reversed and interpolated as needed to correct any subsequent narrowband zoom measurements.

Source Calibration

The HP 3588A source output amplitude is calibrated using the previously calibrated receiver. The output amplitude is measured during self-calibration at two amplitude settings using a direct internal path between the source output and receiver input. From figure in its low-distortion mode, the lower autorange threshold must also be reduced by 10 dB in this mode. Detection of an underrange condition at this level is difficult, so the autorange algorithm is defeated in this case. However, a single autorange can still be performed, allowing a one-time optimization of the signal level.

The main signal path includes a 150-MHz low-pass filter between the input amplifier and the first conversion to prevent out-to-band signals from appearing at the mixer input. This filter is a ninth-order Chebyshev design with an integral amplitude equalizer. The filter is implemented with printed circuit inductors and surface mount capacitors and consistently demonstrates passband frequency response flatness of 0.4 dB. Special care was taken to shield this filter so that section-to-section crosstalk does not compromise stopband performance. Because this filter is also used in the source section of the instrument, development time was reduced. We were also able to reduce manufacturing cost since all of the parts are placed by machine, and we were able to reduce the time for final test since no adjustments are required.

The first conversion follows the low-pass filter and translates the input signal to a fixed first intermediate frequency of 310.1875 MHz. The first local oscillator is synthesized and covers a frequency range from 310.1875 MHz to 460.1875 MHz. The first LO will be discussed in detail later. As in any receiver that must accept a broadband input, spurious products, harmonic and intermodulation distortion products, and frequency flatness all require careful attention. It is primarily the first mixer that sets the harmonic distortion and residual spurious performance for the instrument. Since the design criteria to satisfy these goals were in conflict to some extent, achieving the desired first mixer performance represented a significant design effort. To achieve good distortion performance it is necessary to operate the mixer at high LO drive levels while maintaining symmetry of the drive waveform. However, any mixer has finite isolation between its ports, so higher LO levels result in large amounts of LO signal at the output of the mixer. If these signals are not prevented from reaching subsequent conversions they can produce residual (false) responses. Careful filtering and shielding were required to minimize these residual responses. In addition, a small amount of the first LO signal is fed around the first mixer to cancel the mixer's intrinsic feedthrough (see "Spectrum Analyzer Self-Calibration," page 47).