Network Analyzer (electrical) - Calibration or Error Correction

Calibration or Error Correction

A network analyzer, like most electronic instruments requires periodic calibration - typically this is performed once per year and is performed by the manufacturer or by a 3rd party in a calibration laboratory. When the instrument is calibrated, it will usually have a sticker fixed to the outside, stating the date it was calibrated and when the next calibration is due. A calibration certificate will be issued.

But a vector network analyzer achieves highly accurate measurements by correcting for the systematic errors in the instrument, the characteristics of cables, adapters and test fixtures. The process of error correction, although commonly just called calibration, is an entirely different process, and may be performed by an engineer several times in an hour. Sometimes it is called user-calibration, to indicate the difference from periodic calibration by a manufacturer.

A network analyzer has connectors on its front panel, but the measurements are seldom made at the front panel. Usually some test cables will connect from the front panel to the device under test (DUT). The length of those cables will introduce a time delay and corresponding phase shift (affecting VNA measurements); the cables will also introduce some attenuation (affecting SNA and VNA measurements). The same is true for cables and couplers inside the network analyzer. All these factors will change with temperature. Calibration usually involves measuring known standards and using those measurements to compensate for systematic errors, but there are methods which do not require known standards. It should be noted that only systematic errors can be corrected. Random errors, such as connector repeatability can not be corrected by the user calibration. However, some portable vector network analyzers, designed for lower accuracy measurement outside using batteries, do attempt some correction for temperature by measuring the internal temperature of the network analyzer.

The first steps, prior to actually starting the user calibration are:

• Visually inspect the connectors for any problems such as bent pins or parts which are obviously off-centre. These should be thrown away, as mating damaged connectors wth good connectors will often result in damaging the good connector.
• Clean the connectors with compressed air at less than 60 psi.
• If necessary clean the connectors with isopropyl alcohol.
• Gage the connectors to determine there are not any gross mechanical problems. Connector gauges with resolutions of 0.001" to 0.0001" will usually be included in the better quality calibration kits.
• Torque the connectors to the specified torque. A torque wrench will be supplied with all but the cheapest calibration kits.

There are several different methods of calibration.

• SOLT - which is an acronym for Short, Open, Load, Thru, is the simplest method. As the name suggests, this requires access to known standards with a short circuit, open circuit, a precision load (usually 50 Ohms) and a through connection. It is best if the test ports have the same type of connector (N, 3,5 mm etc), but of a different sex, so the through just requires the test ports are connected together. SOLT is suitable for coaxial measurements, where it is possible to obtain the short, open, load and through. The SOLT calibration method is less suitable for waveguide measurements, where it is difficult to obtain an open circuit or a load, or for measurements on non-coaxial test fixtures, where the same problems with finding suitable standards exist.
• TRL
• TRL*
• LRL
• Unknown through

The simplest calibration that can be performed on a network analyzer is a transmission measurement. This gives no phase information, and so gives similar data to a scalar network analyzer. The simplest calibration that can be performed on a network analyzer, whilst providing phase information is a 1-port calibration (S11 or S22, but not both). This accounts for the three systematic errors which appear in 1-port reflectivity measurements:

• Directivity -- error resulting from the portion of the source signal that never reaches the DUT.
• Source match -- errors resulting from multiple internal reflections between the source and the DUT.
• Reflection tracking -- error resulting from all frequency dependence of test leads, connections, etc.

In a typical 1-port reflection calibration, the user measures three known standards, usually an open, a short and a known load. From these three measurements the network analyzer can account for the three errors above.

A more complex calibration is a full 2-port reflectivity and transmission calibration. For two ports there are 12 possible systematic errors analogous to the three above. The most common method for correcting for these involves measuring a short, load and open standard on each of the two ports, as well as transmission between the two ports.

It is impossible to make a perfect short circuit, as there will always be some inductance in the short. It is impossible to make a perfect open circuit, as there will always be some fringing capacitance. A modern network analyzer will have data stored about the devices in a calibration kit. (Agilent 2006) For the open-circuit, this will be some electrical delay (typically tens of picoseconds), and fringing capacitance which will be frequency dependent. The capacitance is normally specified in terms of a polynomial, with the coefficients specific to each standard. A short will have some delay, and a frequency dependent inductance, although the inductance is normally considered insignificant below about 6 GHz. The definitions for a number of standards used in Agilent calibration kits can be found at http://na.tm.agilent.com/pna/caldefs/stddefs.html The definitions of the standards for a particular calibration kit will often change depending on the frequency range of the network analyzer. If a calibration kit works to 9 GHz, but a particular network analyzer has a maximum frequency of operation of 3 GHz, then the capacitance of the open standard can approximated more closely up to 3 GHz, using a different set of coefficients than are necessary to work up to 9 GHz.

In some calibration kits, the data on the males is different to the females, so the user needs to specify the sex of the connector. In other calibration kits (e.g. Agilent 85033E 9 GHz 3.5 mm), the male and female have identical characteristics, so there is no need for the user to specify the sex. For sexless connectors, like APC-7, this issues does not arise.

Most network analyzers will have the ability to have a user defined calibration kit. So if a user has a particular calibration kit, details of which are not in the firmware of the network analyzer, the data about the kit can be loaded into the network analyzer and so the kit used. Typically the calibration data can be entered on the instrument front panel, as well as loaded from a medium such as floppy disk or USB stick, or down a bus such as USB or GPIB.

The more expensive calibration kits will usually include a torque wrench to tighten connectors properly and a connector gauge to ensure there are no gross errors in the connectors.