VectorStar™ MS464xB Series Microwave Vector Network Analyzer Calibration and Measurement Guide : Broadband/mmWave Measurements (Option 7, Option 8x) : Measurement Tips
 
Measurement Tips
While calibrations are covered in detail in another section of the measurement guide, there are some peculiarities relative to broadband and mmWave operation that require some comment.
The mmWave bands are often executed in waveguide thus requiring that media type selection (to take care of dispersion correction).
SOLT/SOLR are often not recommended due to the difficulty of fabricating a reasonable open standard. An open waveguide flange radiates quite effectively and, as a result, is both unstable and has a relatively high return loss.
SSLT/SSLR are commonly used, particularly at lower frequencies and do require a good load standard. The offset short lengths must be known with some precision.
SSST/SSSR is quite popular, particularly at higher frequencies since a load standard is not required. Accurate knowledge of the short offset lengths is critical.
LRL is also a popular technique and is quite effective although it can be sensitive to the condition of the waveguide flanges.
BB Mode
In classical broadband mode (3738-based), a hybrid coupler combining the bands into a single W1 coaxial connector is typically used with a default breakpoint of 67 GHz (i.e., at 67 GHz and below, the base VNA is used for both source and receiver; at 67.000000001 GHz and above, the external modules are used for source and receiver). This structure is key to allowing the combined sweep. To calibrate this connector type, typically two calibrations are recommended: SOLT from the low frequency limit to 67 GHz and SSST from 67 GHz to 110 GHz. The two calibrations are performed separately and then combined using calibration merge (see calibration section for details). The only constraints are the two calibrations must be of the same type (e.g., both 1-port S11 or both full 2-port) and the total point count must not exceed the current maximum for the instrument (25000 or 100000). Note that other calibration types (e.g., LRL) can be successfully combined with this connector type.
Note 
It is important that the system be in the appropriate 3739 receiver configuration even when performing the lower frequency calibration. This ensures that the correct receiver mode is used in the merged calibration.
In the 3739-based systems, a similar protocol is used with a 67 GHz breakpoint also typically used for ME7838A/AX and ME7838E/EX (and related) systems (W1 connector-based). For the ME7838D and related systems (using the MA25300A modules) which are 0.8 mm-connector based, a breakpoint for the calibrations of 80 GHz is typically used. In all cases, it is generally a SOLT calibration at the lower frequencies and an SSST calibration at higher frequencies. For the ME7838G and related systems, a coaxial-mode flange-based interface is used for direct connection to wafer probes so the same merged calibrations are used less often (and an on-wafer LRM calibration, for example, is broadband inherently). Adapters from the MA25400A module’s native interface to a 0.8 connector are available and calibrations as discussed above using the 0.8 calibration kit are possible. Adapters to WR-5 and WR-6 waveguides are also available. Traditional waveguide calibrations covering the respective bands are then possible.
The 1mm (W1) and 0.8 mm calibration kits both contain 3 shorts, an open, and a load for both connector genders (along with many adapters, tools and verification components). For the SOLT calibration, the coefficients are setup assuming that SHORT 1 is used for that calibration.
Broadband Cal
In some software versions (including and after V2017.06.01), a calibration method termed ‘broadband cal’ is available for ME7838X systems that allows one to perform the SOLT/R and SSST/R calibrations in one session without using the merge. This broadband calibration saves some connections and a number of keystrokes. Also, since the thru/reciprocal is only connected once, some steps in error coefficients at the breakpoint frequency due to connector repeatability can be reduced. Note that since different algorithms are used above and below the breakpoint, some small steps in data are to be expected (within the uncertainties of each calibration).
As shown in the modify setup dialog below, the concept of the breakpoint frequency is still relevant and one can use the 67 GHz or the 80 GHz (for the 0.8 mm calibration kit and ME7838D) default values or one can enter different numbers. If the current frequency range is either entirely above or below the entered breakpoint, only one of the calibration types will be used but all standards will still need to be connected. Note that the published uncertainty curves only apply if the default breakpoint frequencies are used. Note that options are available on the 3656B/C calibration kits to have .s1p definitions of the standards (in addition to the traditional .ccf files which define the components by means of frequency-dependent circuit models). These .s1p-based calibrations can offer improved uncertainties, particularly for reflection measurements (and all measurements for reciprocal-based calibrations). Lower frequency cal kits (e.g., 3652 and 3654 series) also offer this option.
Edit Cal Params—TWO PORT CAL SETUP Broadband Cal, Merged (SOLT\R-SSST\R) Dialog Box
1. Thru Line selected allows user entries for length, line impedance, line loss and frequency.
S2P Thru selected provides buttons for loading, viewing, and characterization (to generate S2P files).
Note that this calibration method choice will only be available if the system is in an appropriate 3739 mode. The reciprocal choice is available as it is for the base open-short-load and triple-offset-short calibrations but the same reciprocal (or thru) will be used for the entire sweep. If it is desired to use different thrus/reciprocal elements, then the traditional merge method described above should be employed. During the calibration, the open, the load, three shorts and the thru/reciprocal must be measured. Short 1 and the thru/reciprocal data will be used for both upper and lower calibrations. As with the regular versions of these calibrations, losses in the opens and shorts can be defined. See SOLT/SOLR Calibration for more information on these entries, but the loss is entered in terms of dB/mm at a specific reference frequency (fref). Losses at other frequencies are found by scaling as the square root (sqrt(f/fref)). If 0 is entered for the reference frequency, the loss is assumed to be constant with frequency.
As discussed above, the ME7838G and related systems (using the MA25400A module) do not general use this same type of broadband calibration since those systems are almost exclusively used on-wafer with inherently broadband calibrations. Adapters to 0.8 mm can be used with the MA25400A modules and a broadband calibration in 0.8 mm can be performed to 145 GHz.
Modular BB Approach
In the modular BB approach, the multiplexing is instead done behind the high frequency reflectometers and happens at two disjoint frequencies instead of at a common breakpoint. This typically does not affect the calibration break frequency as the latter is more determined by the characteristics of the calibration kit. Thus when using Anritsu’s 3656x W1 calibration kit, the same 67 GHz breakpoint is commonly used for both the classical and modular broadband systems.
Some additional comments related to use of the modular BB system:
Since the thru path of the modules is DC-connected, the internal VNA bias tees can be used (the bias routes from the VNA rear panel through the VNA front panel ports to the modules, and then to the W1 plane). For Kelvin-based measurements, Anritsu Kelvin bias tees can be mounted on the rear of the modules directly (see Figure: Trace Noise and Stability Over About 1 Hour). A variety of bias tees are available with different video bandwidths and current handling capabilities are available; contact Anritsu for more information. The internal DC resistance of the modules is very low so the drop from the bias tee to the DUT can be minimized. Contact Anritsu for a detailed report.
Because the bias tees often have 70 GHz bandwidth and are very well matched, they have little impact on system raw directivity, in part because they are positioned behind the high frequency couplers.
Connection of a Kelvin Bias Tee to a mmWave Module
Isothermal measurements are often of interest and the position of the module very close to the reference plane sometimes raises the question of thermal effects. When normally mounted, the temperature rise at a wafer probe tip is less than 1 degree C. If even low temperature rises are required for on-wafer or other applications, mounting variations are available to reduce this rise to under 0.3 degree C (about the limit of the thermometry being used). Contact Anritsu for more information.
It is sometimes desirable to position the modules remotely (or further than the usual 1 m cables would allow) because of a complex test setup or for antenna measurement and other applications. Up to 5 m cable runs can be used with reasonable performance (power and dynamic range below 54 GHz will be affected by insertion loss changes, but mmWave performance will not be because of saturating mixers and multipliers) and 15 m cable runs can be used in some circumstances (covering 30–125 GHz for the 3743A/AX module). For information on configuring longer cable sets or for special modifications for even more remote applications, contact Anritsu. Trace noise and short term stability from measurements with a 15 m system are shown in Figure: Stability Data for Modular BB System. This measurement was done with conventional horn antennas in a laboratory environment and mainly serves to show reasonable close-in trace performance with long cable runs.
Trace Noise and Stability Over About 1 Hour
Trace noise and stability over about 1 hour are shown here for a 15 m antenna-based measurement.
 
In broadband and mmWave measurements, stability is often of interest because of the possibility of long measurement sessions or longer calibrations. It would be desirable if a calibration could last for a longer period of time to reduce the chance of invalid measurements and to reduce the overall time spent calibrating. The modular BB approach offer some advantages in having the high frequency couplers (covering 30–145 GHz) being right at the 1 mm (or 0.8 mm) connector to minimize raw directivity degradation, a small integrated package to avoid thermal gradients, and a tightly integrated control system. The result is good stability in both reflection and transmission over time as suggested in Figure: Stability Data for Modular BB System. In order to optimize stability, minimizing environmental temperature variations certainly helps as does protecting the cable runs from overstress or unnecessary exposure to thermal variation.
 
Stability Data for Modular BB System
Waveguide-based measurements are sometimes needed. The classical approach can simply demount the mux coupler and use the exposed waveguide flange (WR-10, 75–110 GHz in that case). A variety of adapters are available for the modular BB approach in WR-15, WR-12, and WR-10 sizes (50–75, 60–90, and 75–110 GHz nominal, respectively). A bracket comes with these adapters to prevent stress from being applied to the 1 mm connector and to improve measurement repeatability. Waveguide calibration kits (3655 series) are available to support SSLT and LRL calibrations referenced to these waveguide planes. The uncertainties do not differ markedly between coaxial and waveguide-based calibrations in these bands as suggested in Figure: Example Coaxial and Waveguide Uncertainty Curves, but the waveguide measurements can be more subject to repeatability issues depending on the physical setup and flange quality on the DUT or on any extensions (some high quality extensions are included in the calibration kits to help with the latter).
 
Example Coaxial and Waveguide Uncertainty Curves
The frequency plan for the modular BB and mmWave modes was selected to optimize noise floor, spur and trace noise performance. In the True Mode Stimulus mode (available in 4-port, dual source systems; see Dual Source and DifferentialView™ (Option 31 and Option 43) of this guide), a different frequency plan can provide improved trace noise performance at higher frequencies (at the expense of increase spurs which normally impact only low-level measurements which are less common in true mode stimulus applications). This frequency plan is available with the TMS Coherence switch (available via System | Setup | Misc. Setup menu) and should be turned ON for both the 4-port calibration phase and for the TMS measurement. If one later decides to turn this off, the 4-port calibration must be redone.
The ME7838G system is designed primarily for on-wafer use and has a 0.6 mm coaxial flange-based interconnect for attaching probes. Power calibrations can be done with adapters (to 1 mm, 0.8mm, and WR-05 waveguide) and the power reference plane moved to the probe tip using embedding/de-embedding tools and provided generic .s2p adapter files. Other on-wafer power calibration approaches can also be used. RF user calibrations (TRL, etc.) are generally performed on-wafer using substrate-based standards.