Some terms in this article:

ED(e00): Directional error

ES(e11): Source matching error

ER(e10e01): Reflection tracking error

e22: Load matching error

T1/T2(e10e32): Transmission tracking error

SOLT: Short-Open-Load-Through

SOLR: Short-Open-Load-Reciprocal

TOSM: Through-Open-Short-Match

As far as the author knows, many network sub-users have some misunderstandings in calibration, thinking that as long as the calibration piece is selected correctly, everything will be fine, but the choice of the calibration piece is only the first step in the correct calibration, and the final test result is accurate. The guarantee is far from enough. The following will introduce to you a place where many errors occur in the use of network sub-calibration-where unknown through calibration should be used but SOLT calibration is used incorrectly. So what is the difference between Unknown Direct (SOLR) and SOLT? This probably starts from the origin of the two calibration algorithms.

The calibration of the early vector network is a very long and complicated iterative calculation process [1]. It was not until 1971 that Kruppa and Sodomsky [2] used an 8-term error model (not the well-known 8-term error model) to clearly describe the calibration method of the two-port vector network for the first time. Modified direct calculation method. The method of Kruppa and Sodomsky uses the test results of open, short, load and thru standard parts to calibrate the three reflection error items of ED, ES, and ER of each port, and the two transmission error items of T1 and T2. Hewlett-Packard (now Keysight) made further modifications to this error model in 1978 [3], and became the more familiar 10-term error model. The corresponding calibration method was also named SOLT or TOSM calibration algorithm and has been used all the time. to date.

Figure 1 10-term error model [5]

With the TRL calibration method proposed by Engen and Hoer [3] in 1974, the two-port vector network 8-term error model (7-term error model) was introduced into the vector network calibration method for the first time. It has a great impact. The unknown through calibration (SOLR) we are going to introduce today is derived from this error model. The 8-term error model is different from the 10-term error model in that it introduces a fourth receiver (the vector net of the early 10-term error model only needs 3 receivers), which can take advantage of the extra The test results are used to determine part of the unknown parameters of the calibration standard; that is, the vector network can be calibrated by using some standard parts with known parameters, without the need to know all the parameters of the standard parts like the 10-item error model.

Figure 2 8-term error model [5]

The 8 error terms in Figure 2 can be normalized and combined into 7 terms in actual calculations, so the 8-term error model can also be called the 7-term error model.

Therefore, the difference between unknown through calibration (SOLR) and SOLT is reflected in the difference in the requirements of different error models for the known parameters of the calibration standard. SOLT is derived from the 10-term error model. It is necessary to establish 10 equations to solve the 10 error terms. 10 equations represent the known parameters that require 10 standard parts; while the unknown through is derived from the 8-term (7-term) error model , So only 7 equations need to be established. So in addition to the 6 single-port standard reflection parameters, SOLT also needs to know the complete 4 S parameters of the through standard to calibrate the vector network; while the unknown through calibration does not require the through standard parameters to be fully known, only the through standard. It only needs to satisfy the relationship of S21=S12, which is why the unknown through calibration is called this name.