The present invention relates to a signal separation circuit for a radio frequency (RF) network analyzer in accordance, respectively, with the preambles to claim 1, 2, and 4. An example of such a signal separation circuit is the Hewlett-Packard Model 8503A S-Parameter Test Set. This instrument is an accessory for radio frequency network analyzers which facilitates measurement of the S-parameters (scattering parameters) of a device under test (DUT) over a range of frequencies.

In general, an RF network analysis measurement system contains several separate modules. First is an RF source to provide a stimulus to a device under test (DUT). Second is a signal separation circuit. or test set, for routing the stimulus to the DUT and for sampling the energy that is reflected from, or transmitted through, the DUT. Also, energy is sampled from the signal that is incident upon the DUT in order to provide a reference for all relative measurements. Third is a tuned receiver to convert the resulting signals to an intermediate frequency (IF) for further processing. In vector network analyzers both the magnitude and phase relationships of the original signals must be maintained through the frequency conversion to IF to provide usable measurements. Fourth is a detector to detect the magnitude and phase characteristics of the IF signals, and fifth is a display on which to present the measurement results.

In the known signal separation circuit, a radio frequency stimulus received at an RF input port is routed to two test ports for connection to the DUT by a reverting (double-pole double-throw) switch. With the reverting switch set to its first position, the DUT receives the stimulus from test port 1 and the response of the DUT (the transmitted signal) is incident on test port 2; with the reverting switch set to its second position, the DUT receives the stimulus from test port 2 and the transmitted signal is incident on test port 1. The circuit includes two directional bridges which are coupled to the test ports and configured to sample HF signals that are incident on the test ports from the DUT. Thus both the reflected and the transmitted part of the RF stimulus is sampled and may be routed to the network analyzer for evaluation with either setting of the reverting switch. This enables a measurement of the S-parameters of the DUT without a need for manual reconnections.

Measurement accuracy may be improved in a computer controlled system by a process known as vector error correction where a set of calibration devices with known characteristics is measured and a set of equations is solved to determine the errors associated with the network analyzer itself. This calibration model is then stored in the computer for later correction of data measured on an unknown device.

With the known signal separation circuit, however, the electrical characteristics of the path between the reference port and the device under test may vary in a non-repeatable fashion on each reversal of the signal flow to the device under test because of non-repeatability of the reverting switch, which may cause errors not accounted for in a previous calibration of the system.

Relative to this prior art the present invention solves the problem of improving the known signal separation circuit in such a way that a complete measurement of the S-parameters of a device under test may be performed without a loss of calibration.

A fully error corrected measurement of the four vector S-parameters, and of the vector transmission and reflection parameters, is accomplished in "real time". Vector testing using a single set up of a DUT can be performed across a broadband frequency range from RF to millimeter bands. Dynamic accuracies of 0.05 dB in magnitude and 0.3 degrees in phase can be accomplished for a device with 50 dB of insertion loss. An overall dynamic range of 100 dB, resolutions of 0.001 dB in magnitude, 0.01 degrees in phase, and 10 picoseconds in group delay, and corresponding measurement stabilities are attained depending on the particular frequency range-and signal separation circuit used. The signal separation circuits, or test sets according to the present invention incorporate broadband signal separation devices, balanced broadband power splitters, and high conversion efficiency samplers with flat frequency response and low crosstalk, and provide optimized performance for different frequency ranges and connector types. A dedicated interface provides control from a main network analyzer.

An embodiment of the invention will now be described in detail with reference to the accompanying drawings.

• Figure 1 shows a simplified block diagram of an RF vector network analyzer;
• Figures 2a and 2b show a schematic and a cross sectional view respectively of a wideband RF directional bridge for use in an embodiment of the present invention;
• Figures 3 through 6 show detailed block diagrams of four test sets as shown in figure 1;
• Figures 7.1 through 7.24 show detailed schematics for selected parts of the block diagrams shown in figures 3 through 6;
• Figures 8.1 through 8.6 show a detailed block diagram of the IF/ detector section of the vector network analyzer shown in figure 1;
• Figures 9.1 through 9.6 show a block diagram and the related equations used for adjusting offset and gain errors in the IF/detector section of the vector network analyzer shown in figure 1; and
• Figures 10.1 through 10.34 show detailed schematics for selected parts of the block diagram shown in figures 8.1 through 8.6.

Description of the Block Diagram

Figure 1 shows a block diagram of an RF vector network analyzer system. The measurement system consists first of a main network analyzer 101 with a second IF/detector section 103 and a data processor/display section 105. The main network analyzer 101 is fed by one of four configured test sets 107 which provide the signal separation circuitry 108 and first IF frequency conversion circuitry 113 for reflection/ transmission (one incident signal) or S-parameter (two incident signals) measurements up to either 18 or 26.5 GHz. The frequency converter 113 alone is also available to permit the addition of user supplied signal separation devices 108 for specially configured test needs. The third main component of the measurement system is a compatible RF source 109 such as an HP 8340A synthesized sweeper, available from the HewlettPackard Co., Palo Alto, California, which can be used in either a stepped frequency mode, in which synthesizer class frequency accuracy and repeatability can be obtained by phase locking the source 109 at each of the over 400 frequency steps over the frequency range selected by the main analyzer 101 or the swept frequency mode for applications where extreme frequency range, high stability, and spectral purity are important such as in narrow band measurements over sweeps of less than 5 MHz. An HP 8350B sweeper with HP 83500 series RF plug-ins covering the entire desired frequency range or with lesser spans can also be used in applications where a more economical source is sufficient. Both the HP 8340A and the HP 8350B include the necessary analog interface signals as well as full digital handshake compatibility with the main analyzer 101. This digital handshake compatibility allows the main analyzer 101 to act as the controller for the entire system by directly managing the source 109 to provide all of the inputs such as start frequency, stop frequency, centering, span, and modulation, as well as constraints that the source 109 normally places on itself internally. For example, if a user by means of the main analyzer 101 requests the source 109 to sweep to an incompatible frequency such as 50 GHz, the source 109 will respond to the main analyzer 101 that such a frequency cannot be accommodated and the main analyzer 101 in turn informs the user of the situation. Therefore, the user need only be concerned with his interface to the main analyzer 101 and can use any source 109 that has implemented the required handshake protocols. Because the main analyzer 101 is in control of the source 109, it is also possible to automatically select a different frequency range or mode (stepped or swept) to be applied to each of the ports 1 and 2.

Several concepts have been incorporated in the IF/detector section 103 of the main analyzer 101 to increase the precision of IF processing and signal detection. Most of the phase lock hardware 125 in the phase lock loop resides in this section 103. Harmonic mixing number and local oscillator pretuning are controlled digitally via lines 127 and 129 and offer phase lock and tracking performance that is precisely repeatable from sweep to sweep. Before the first IF signals proportional to al, a2, bl, and b2 are sent to the synchronous detectors 131 and 133, they are down converted to a second IF at 100 KHz by mixers 138 and go through a pair of multiplexers 136 and variable gain amplifiers 134 in the second IF section 135. Amplifier gain is controlled and calibrated digitally and is varied by autoranging to optimize the second IF signal levels 130 and 132 available to the synchronous detectors 131 and 133 resulting in an order of magnitude improvement in signal to noise performance and dynamic accuracy for the detector output signals xl, yl, x2, and y2. Likewise, the synchronous detectors 131 and 133 employ a digital architecture that allows for precise control of their 90 degree phase shift function which resul is in improved accuracy as well as common mode rejection of local oscillator phase noise effects. Finally, the detected signals xl, yl, x2, and y2 are multiplexed with a sample-and-hold/multiplexer (MUX) 137 and then digitized by an analog-to-digital converter (ADC) 139 with 19 bits of resolution. Each ADC conversion takes approximately 40 microseconds and four readings are made for each RF frequency data point _':₀ provide the real and imaginary data for both the reference signal 13: and test signal 132.

The output of the ADC 139 is then passed on a 16 bit bus 141 to a hign speed central processor (CPU) 143 which includes a microprocessor such as a Motorola 68000 as well as the associated microprocesor system interrupt and 1/0 control circuitry. Because the CPU 143 is integrated into the main network analyzer 101 it is possible to utilize a multi-tasking architecture to make more efficient use of time than has previously been possible. This architectural integration also permits substantial increases in data processing flexibility and system control performance. Via a dedicated system interface and bus 145, the Integrated within each test set 107 is the first IF frequency converter 113 with three channels 113a, 113b, and 113c for reflection/ transmission measurements and four channels 113a, 113b, 113c, and 113d for S-parameter measurements. RF to IF conversion is achieved through a sampling technique equivalent to harmonic mixing. An harmonic of a tunable local oscillator 115 is produced by an harmonic generator 116 to mix with the incoming RF signal to provide the first IF signal at 20 MHz for the incident signal al on the input port 1, the incident signal a2 on the output port 2, the reflected or transmitted signal bl on the input port 1, and the reflected or transmitted signal b2 on the output port 2. Frequency tuning for the local oscillator 115 is controlled by a phase lock loop 125 that compares the signal al or a2 in the reference channel first IF to an IF reference oscillator 119 in the IF/detector section 103. Any difference between the frequency of the signal al or a2 in the reference channel first IF and the IF reference oscillator 119 results in an error voltage on the error voltage signal line 121 via switch 123 that tunes the local oscillator 115 to the frequency that produces the desired first IF. Switch 123 is toggled to select the most appropriate signal al or a2 to lock on to based either on internal criteria within the system or as defined by the user. When using the internal criteria, if the incident signal port is port 1, al is selected by switch 123, and if the incident signal port is port 2, a2 is selected by switch 123. This scheme allows the local oscilator 115 to track the incoming RF when the RF frequency is changing with time as in the swept mode. The integrated test set 107 permits high RF to first IF conversion efficiency even at 26.5 GHz, making possible both high sensitivity and wide dynamic range measurements. The test set architecture eliminates the extensive RF switching needed in previous test sets, removing the significant uncertainties caused by the lack of repeatability of mechanical switches. The reflection/transmission test sets 107 require no internal switching since the fourth channel 113d is not required, and the S-parameter test sets 107 use only one electronic PIN diode switch located inside of the test set 108 such that it cannot contribute to uncertainties as it is switched prior to the ratio node of the power splitter. CPU 143 controls the RF source 109, the test set 107, and, along with the sample selection and timing circuitry 146, all of the IF processing functions including the phase lock hardware 125, autoranging in the IF amplifiers 134, detection by the synchronous detectors 131 and 133, and digitization by the ADC 139. The CPU 143 periodically initiates a self calibration sequence for the IF amplifiers 134, synchronous detectors 131 and 133, and the ADC 139 and the resulting gain, offset, and circularity changes are stored in memory 147, so that the changes in the IF amplifiers 134 can be subtracted from measured results. The CPU 143 also performs all data processing functions for the system. The signals in the IF section 103 are detected as linear real and imaginary components of a vector quantity and the CPU 143 processes the detected data into a variety of formats for presentation on the CRT display 149. By digitally computing the various measurement formats, improvements in dynamic range and meaningful resolution are gained over traditional analog circuit processing techniques.

With the known network analyzer systems, an external computer was required in order to characterize and remove systematic errors. With the present RF vector network analyzer, this capability exists internally with enough storage capacity (i.e., 256K bytes of random access memory (RAM) and 256K bytes of bubble memory) in the memory 147 to retain up to two 401 point 12-term error corrected traces of data. (Note: each byte of memory consists of eight bits of data storage.) In addition, the measured data can be converted to show the response of the DUT 111 as a function of time (time domain) using an internal Fourier transform process. All data processing takes place virtually in real time by means of parallel data processing in the CPU 143 aided by the incorporation of a dedicated, floating point, complex number, vector math processor 151 designed specifically for fast vector computations. The multiplication of two complex numbers by the vector math processor 151 requires only one operation with the product available within 20 microseconds, so that error corrected measurement results are avail- _"abre 1000 times faster than in the prior art. By means of an internal vector graphics generator 153, the real time processed data is then immediately presented on the CRT 149, on a digital printer/plotter 155, or via an IEEE-488 (HP-IB) interface and bus 157 to external devices. Present as well as past states of front panel controls 159, past and present traces of data, and entire system calibrations can also be stored in and recalled from the memory 147 or loaded and read from a built-in tape drive 161 by means of the system interface and bus 157 under control of the CPU 143.

The vector math processor 151 is constructed from a series of commercially available medium scale integrated circuits.

Description of the Test Sets

The wideband test sets 107 to 26.5 GHz include a high performance RF triaxial directional bridge 1901 as described in European Patent Application entitled "Broadband Radio Frequency Directional Bridge And Reference Load", Serial No. 84114631.9, filed December 1, 1984, and shown in figures 2a and 2b, coupled to each of the DUT ports 1 and 2 as shown in figures 3 and 4. The directional bridge 1901 is a balanced Wheatstone bridge 1903 that extracts a floating vector signal for measurement in a single-ended detector system without disturbing the balanced configuration. Included in this high performance RF directional bridge 1901 is a combination reference load and balun 1905 which provides signal separation over the entire frequency range from 45 MHz to 26.5 GHz, and also permits the application of a DC bias as part the RF input Vi_n to the DUT 111 via a conventional RF bias tee 2105. In contrast, the narrower band test sets 107 as shown in figures 5 and 6 utilize a conventional directional coupler 2201 for each port to cover the frequency range of 0.5 to 18 GHz. By incorporating the signal separation devices 108 in the test sets 107, broadband vector measurements are made possible with just one connection of the DUT 111 between port 1 and port 2.

Each of the test sets 107 contains its own built in power supplies 2001, to simplify various system configurations and each of the test sets 107 has its own HP-IB interface 2003, coupled to the system bus 145 in order to provide control and identification to the main analyzer 101. Each of the test sets 107 is connected respectively to section 103 via a first IF multiplexer 2002 or 2102 to provide daisy chaining of several test sets. The first IF multiplexer 2002 and 2102 are in turn connected respectively to the al, bl, and b2 connections for the reflection/transmission test sets in figures 3 and 5, and the al, a2, bl, and b2 connections for the S-parameter test sets in figures 4 and 6. The S-parameter test sets also include: front panel indicators 2104 to signal the active test port, a conventional bias tee 2105 on each of the test channels to provide voltage bias 2107 needed in the testing of active devices, PIN diode transfer switches 2109 under control of the main analyzer 101 via the system bus 145 and a switch interface 2110 for switching the RF input between the ports 1 and 2, and variable attenuators 2111 under control of the main analyzer 101 via the system bus 145 and an attenuator interface 2113. Various RF pads 2015 and test and reference extentions 2117 are provided to adjust and balance the RF power levels.

Each of the test sets has a frequency converter 113 to provide the first IF conversion of the RF signals in immediate proximity to the RF input and the test ports. Within the frequency converters 113 are the VTOs 115, the first IF samplers 2019, pulse generators 2021 to drive the first IF samplers 2019, and first IF amplifiers 2023 and 2123. The first IF amplifiers 2123 also include an input band pass filter 2131, a filter amplifier 2133, and an output low pass filter 2135 to provide additional signal shaping. Each of the VTOs 115 is driven by a sample/ hold circuit 2025, a summing node 2027, and a buffer amplifier 2029 coupled to the phase lock circuitry 125 in section 103.

Figures 7.1 through 7.24 show detailed schematics for selected parts of the circuitry associated with the test sets 107 as shown in figures 3 through 6. Figures 7.1 through 7.4 show the first IF multiplexers 2002 and 2102, figures 7.5 through 7.10 show the VTO 115 and related drivers, figures 7.11 through 7.16 show the HPIB interface 2003, figures 7.17 through 7.21 show the attenuators 2111 and the PIN diode switch 2109, and figures 7.22 through 7.24 show the first IF samplers 2019 and the first IF amplifiers 2023 and 2123.

Second IF

A detailed block diagram of the second IF/detector section 103 as shown in figure 1 is illustrated in figures 8.1 through 8.6. After the signals a₁, a₂, b₁, and b₂ have been converted to the second IF frequency by the second IF mixers 138, the resulting signals a₁', a₂', bi', and b₂' are sent to the second IF MUXs 136 as shown in figure 8.2. A 100 KHz calibration frequency 2501 produced by clock 119 and a ground input 2502 are also sent to the second IF MUXs 136 so that the second IF channels can be automatically calibrated for both gain and offset errors. This automatic calibration is performed by individually measuring the vector gains of the four cascaded 12 dB amplifiers that make up the amplifiers 2503 each to within 0.001 dB with the help of the ADC 139. Offset errors are removed by applying the ground input 2502 to the MUXs 136, turning off all gain in the amplifiers 2503, and measuring the resulting signal with the ADC 139 for each of four phase offsets (i.e., 0, 90, 180, and 270 degrees) of the synchronous detectors 131 and 133, thus rotating the measument plane used in the synchronous detectors 131 and 133. This change in phase offset and rotation of the measurement plane in the synchronous detectors 131 and 133 is accomplished by adjusting the phase angle of the demodulating signal used for synchronous detection by means of the adjustable phase shifters 2505 as shown in figure 8.3.

As shown in figures 9.1 through 9.6 it can be seen that the true values of X and Y can be determined from the measured values of X_m and Y_m from the equation shown in figure 9.2. First, the offsets Xo and Yo are determined by grounding the input of MUX 136 as shown in figure 9.1, turning off all gains G₁, G₂, G₃, and G₄, and measuring X_m and Y_m for each of four phase offsets 0, 90, 180, and 270 degrees. Xo and Yo are then calculated by the relationship shown in figure 9.3. H is determined by selecting the calibration signal 2501 and turning on the gain G₄. X_m and Y_m are then measured for each of the four phase offsets and the offsets X₀ and Y₀ are subtracted. H can then be calculated using the four quadrature relationships shown in figure 9.4 and performing a least squares error fit to each of the four measured data points as shown in figure 9.5, where A is the level of the calibration signal 2501, the X and Y terms correspond to X_m - Xo and Y_m - Yo, and Sigma is the summation of the four quadrature measurements. Determining the gain and phase of the four amplifiers G₁ through G₄ requires that each be independent of one another since H = G₁ ^* G₂ ^* G₃ ^* G₄. First with only G₁ on, Xm and Y_m are measured for each of the four phase offsets and a corrected X' and Y' are calculated using the correction coefficients previously determined during the offset correction. Using the equations shown in figure 9.6 the complex gain (a + jb) can be calculated that will best translate the four X' and Y' data points into the quadrature relationships shown in figure 9.4. The measurement of X_m and Y_m and calculation of a_l and b₁ as above is repeated sequentially with each of the amplifiers G₂, G₃, and G₄ on one at a time.

Figures 10.1 through 10.34 show detailed schematics for selected parts of the block diagrams shown in figures 8.1 through 8.6. Figures 10.1 through 10.7 show the clock 119, figures 10.8 through 10.12 show the 19.9 MHz local oscillator 2511, figures 10.13 through 10.16 show the second IF mixer 138, figures 10.17 through 10.22 show the IF counter 2513, figures 10.23 through 10.28 show the VTO pretune circuitry 2515, and figures 10.29 through 10.34 show the main phase lock circuitry 2517.

Software Signal Processing

Signal processing in the present RF vector network analyzer begins at the output of the synchronous detector pair 131 and 133 which provide the real (X) and imaginary (Y) parts of the test and reference signals. As explained previously, offset, gain, and quadrature errors are corrected for both of the IF/detector chains via software. The resulting test and reference data is then ratioed to produce the appropriate S-parameters and stored.

Using a one term model (vector frequency response normalization), a three term model (one port model), and up to a twelve term error correction model (comprehensive two port) of the microwave measurement hardware, vector error correction software in conjunction with the vector math processor 151 provides corrected data.

The vector data may be formatted into magnitude, phase, or other formats as desired and stored for providing convenient access for scale and offset changes. Scaled data is stored in the display RAM 217 from which the display generator 153 hardware repetitively creates a plot on the CRT 149 for a flicker-free display. Input and output access is provided via the HP-IB interface 157 and via tape 161. Direct printer and plotter output for the printer, or plotter, 155 may be made.

The software is structured as a multi-tasking system to provide a rapid data update rate by allowing data processing to take place when the data acquisition software is not busy. Overlying command and control tasks interleave data processing with data acquisition cycles.