GENERAL FOUR-PORT ON-WAFER HIGH FREQUENCY DE-EMBEDDING METHOD

Abstract

The present invention provides a general four-port on-wafer high frequency de-embedding method. The method comprises: for each on-wafer de-embedding dummy, building a model considering the distributive nature of high frequency characteristics of the on-wafer de-embedding dummy; obtaining the intrinsic Y-parameter admittance matrix of said N on-wafer de-embedding dummies by calculation or simulation by using said models; and solving the equation set which the corresponding measurement and calculation or simulation data of said on-wafer de-embedding dummies satisfy for the elements of the related admittance matrices of the parasitic four-port network to be stripped in de-embedding and model parameters of models on which said calculation or simulation is based.

Description

TECHNICAL FIELD

The present invention relates to a high frequency de-embedding method, particularly an on-wafer high frequency de-embedding method, and more particularly a general four-port on-wafer high frequency de-embedding method.

BACKGROUND OF THE PRESENT INVENTION

The high frequency characterization measurements, performed on various active and passive electronic devices, semiconductor devices and integrated circuits thereof prepared on a wafer made of silicon and other semiconductor materials, are the basis for device modeling and model parameter extraction and circuit design and evaluation. This high frequency characterization measurement, particularly the high frequency characterization measurement for model parameter extraction of devices, must correspond to specific ports of a device under test (DUT), the locations of which are precisely defined. For this reason, it is necessary to remove all the parasitics, added by instruments, cables, probes and on-wafer testing set-up required to perform the high frequency measurement, from the raw data of the measurement. This is known as the so-called calibration procedure which is a key technique for ensuring accurate extraction of model parameters of high frequency devices.

With regard to the on-wafer high frequency measurement, calibration in a broad sense consists of two steps: off-wafer calibration and on-wafer de-embedding. In the first step, by off-wafer calibration, a test reference plane is moved from vector network analyzer (VNA) to the tip of on-wafer high frequency testing probes by using a set of impendence standard substrate (ISS) calibration dummies (generally made of alumina ceramic substrate). Techniques usually adopted here include short-open-load-thru (SOLT), line-reflect-reflect-match (LRRM), thru-reflect-line (TRL), four-port method (16-term error model), and the like. In the second step, by on-wafer de-embedding, the test reference plane is further shifted to ports of the DUT by using a suite of on-wafer de-embedding dummies. In recent years, a general four-port on-wafer de-embedding method has been proposed to strip the related parasitics of on-wafer measurements. The physical basis of the so-called general four-port method is to describe all the parasitics between the tips of the probes and the ports of the DUT by including them in a general four-port network, the four ports of which are defined at the tips of the probes and the ports of the DUT, respectively. FIG. 1 is a schematic view of an on-wafer high frequency characterization measurement set-up, wherein input and output testing ports of the VNA configured to test S parameters of the DUT are connected through high frequency cables to input and output on-wafer high frequency testing probes, respectively. The tips of the said probes are made contact with the surfaces of input and output on-wafer probing pads, respectively, and the sad probing pads are connected through on-wafer input and output access lines to input and output terminals of the DUT, respectively. FIG. 2 is a schematic view of the corresponding general parasitic four-port network. Here, the tip of the input testing probe, the input terminal of the DUT, the output terminal of the DUT, and the tip of the output testing probe are defined as port 0, port 1, port 2 and port 3 of the said parasitic four-port network, respectively, and those ports are denoted by P0, P1, P2 and P3, respectively. The following equation (1) can be obtained through proper theoretical derivation:

Y _M =Y _ee −Y _ei(Y_A+Y_ii)⁻¹Y_ie  (1)

where, Y_A denotes the Y-parameter 2×2 admittance matrix from P1 to P2, i.e., the intrinsic two-port Y-parameter 2×2 admittance matrix of the DUT; Y_M denotes the Y-parameter 2×2 admittance matrix from P0 to P3, i.e., the measured whole two-port Y-parameter 2×2 admittance matrix of the DUT including all the on-wafer parasitics; and Y_ee, Y_ii, Y_ei and Y_ie are four Y-parameter 2×2 admittance sub-matrices of the Y-parameter 4×4 admittance matrix for describing those parasitic four-port network characteristics. Equation (1) is transformed to Equation (2):

Y _A =−Y _ii −Y _ie(Y_M−Y_ee)⁻¹Y_ei  (2)

As can be seen, Y_A as the left side of the Equation(2) can be obtained as long as that the right side of Equation (2) is known after the measured value of Y_M is substituted. That is, the intrinsic high frequency characterization measurement result Y_A of the DUT is obtained by stripping all the on-wafer parasitics from the raw data of the whole high frequency characterization measurement Y_M of the DUT, achieving the purpose of the high frequency characterization measurement de-embedding. Therefore, total 16 elements of the four Y-parameter 2×2 admittance matrices Y_ee, Y_ii, Y_ei and Y_ie for describing said parasitic four-port network need to be determined. A proper equation set, which is composed of equations obtained by applying Equation(1) to a sufficient number of on-wafer de-embedding dummies with both measured Y_M and theoretical Y_A known, can be solved for the said 16 matrix element. Accordingly, for the general four-port method, five two-port on-wafer de-embedding dummies are usually adopted, i.e., Open O, Short S, Left L, Right R, and Thru T, the equivalent circuits of which are shown in FIG. 3. where their corresponding intrinsic two-port Y-parameter admittance matrices are respectively:

$\begin{array}{cc}{Y}_{\mathrm{AO}}=\left[\begin{array}{cc}0& 0\\ 0& 0\end{array}\right]& \left(3\right)\\ Y_{\mathrm{AS}}=\left[\begin{array}{cc}\infty & 0\\ 0& \infty \end{array}\right]& \left(4\right)\\ Y_{\mathrm{AL}}=\left[\begin{array}{cc}{Y}_L& 0\\ 0& 0\end{array}\right]=\left[\begin{array}{cc}{G}_L+j\omega C_L& 0\\ 0& 0\end{array}\right]& \left(5\right)\\ Y_{\mathrm{AR}}=\left[\begin{array}{cc}0& 0\\ 0& Y_R\end{array}\right]=\left[\begin{array}{cc}0& 0\\ 0& G_R+j\omega C_R\end{array}\right]& \left(6\right)\\ Y_{\mathrm{AT}}=\left[\begin{array}{cc}s& -t\\ -t& s\end{array}\right]& \left(7\right)\end{array}$

Respectively substituting Equations (3)-(7) into Equation (1) obtains an equation set formed of an enough number of equations. After solving said equation set, by using Equation (2), the intrinsic Y-parameter admittance matrix Y_A of the DUT can be calculated from the measured whole Y-parameter admittance matrix Y_M of the DUT, i.e., the high frequency de-embedding for stripping said on-wafer parasitics is completed. Specifically, if said parasitic four-port network is passive and contains no any anisotropic material, and further more specifically, if said parasitic four-port network is symmetrical, the number of such on-wafer de-embedding dummies may be decreased to four, even three, as needed. In this way, the complexity and workload of de-embedding related testing structure designs, wafer fabrication, testing and data processing can be effectively decreased.

As described above, said general four-port high frequency de-embedding method is to further shift the test reference planes to input and output ports of the DUT, respectively, on the basis of off-wafer calibration in which the test reference planes are moved from the VNA to the tips of the on-wafer testing probes by using a suite of ISS calibration dummies. Actually, said general four-port high frequency de-embedding method is also suitable for directly moving the test reference planes from VNA to input and output ports of the DUT, without requiring off-wafer calibration based on the ISS calibration dummies, and this is called one-step calibration method. However, in doing this, the premise is that it is necessary to change, in the general four-port network as shown in FIG. 2, the definition of location of P0 from the tip of the input testing probe to the input testing port of the VNA and the definition of location of P3 from the tip of the output testing probe to the output testing port of the VNA, while maintaining the definition of locations of P1 and P2 as the input terminal of the DUT and the output terminal of the DUT. It is to be declared here that the general four-port on-wafer high frequency de-embedding method involved in the present invention may be applicable to on-wafer de-embedding after the completion of off-wafer calibration based on the ISS calibration dummies, and also may be used, as the one-step calibration method, in on-wafer de-embedding which directly moves the high frequency test reference planes directly from VNA to the input and output ports of the DUT without requiring off-wafer calibration based on the ISS calibration dummies.

The general four-port high frequency de-embedding method has the following advantages: the starting points and end points, from and to which, respectively, the test input and output reference planes need to move, are defined as ports of a general four-port network, respectively, and in this way, all the on-wafer parasitics to be stripped in de-embedding are contained in said general four-port network, without making any assumptions on the specific form of the interior structure of said general four-port network containing all the parasitics, and hence, the universality of said de-embedding technique is ensured. However, a problem with the prior art of said general four-port high frequency de-embedding method is that idealized assumption is made on the network characteristics of the required two-port on-wafer de-embedding dummies, and those idealized on-wafer de-embedding dummies are actually unachievable in reality, and consequently, errors are inevitably introduced. Specifically, this is first manifested in Open and Short dummies. Referring to Equations (3) and (4), ideal Open requires y₁₁=y₂₂=0 for Y_AO and ideal Short requires y₁₁=y₂₂=∞(infinite) for Y_AS. Meanwhile, ideal Open and ideal Short further require that there is no any coupling between the input port and the output port, and hence, for both Y_AO and Y_AS, y₁₂=y₂₁=0. However, actually, on-wafer Open and Short dummies are non-ideal no matter how they are designed and fabricated. Due to the presence of parasitics (for example, parasitic resistances, parasitic capacitances, parasitic inductances or the like), the admittances (y₁₁ and y₂₂ of Y_AO) of the actual Open are not equal to 0 and the admittances (y₁₁ and y₂₂ of Y_AS) of the actual Short are not infinite, either, and also coupling inevitably exists between the input port and the output port of the actual Open and Short dummies, and hence, for both Y_AO and Y_AS, y₁₂ and y₂₁ are not strictly equal to 0, either. Secondly, with regard to the Left and Right dummies, the prior art of the general four-port high frequency de-embedding is implemented by connecting conductors G_L and G_R to the input port and the output port, respectively, on the basis of the Open dummy. Although, as shown in FIGS. 3 (c)-(d) and Equations (5) and (6), the prior art also takes into account parasitic capacitances C_L and C_R connected in parallel with G_L and G_R to form access admittances Y_L and Y_R of the Left and Right dummies, respectively, determined by the distributed nature of high frequency operation, particularly as the operating frequency is increased from radio frequency and microwave band to millimeter wave band, artificially-designed conductors (i.e., resistors) cannot, once fabricated on a wafer, be strictly and accurately described only by a simple lumped equivalent circuit having conductors (resistors) and capacitors connected in parallel. That is, forcibly setting y₁₁ and y₂₂ of Y_AC and Y_AR to be respectively equal to G_L+jωC_L and G_R+jωC_R does not accord with the fact. AgOn, since the Left and Right dummies are built based on the Open dummy, the non-ideal characteristic problem with the actual Open dummy as described above still exists here. As a result, simply zeroing y₁₂, y₂₁ and y₂₂ of Y_AL and y₁₁, y₁₂ and y₂₁ of Y_AR in Equations (5) and (6), respectively, inevitably causes corresponding errors.

SUMMARY OF THE PRESENT INVENTION

In order to overcome this problem, the present invention provides an improved general four-port on-wafer high frequency de-embedding method which is still based on the general parasitic four-port theory as described above. However, no idealized lumped assumption is made on the necessary on-wafer de-embedding dummies. Instead, by using the distributive theoretical calculation or simulation results in combination with the corresponding high frequency characterization measurement data, the de-embedding of on-wafer DUT high frequency characterization measurement data is finally completed by optimizing and calibrating corresponding calculation or simulation results. On the basis of inheriting the advantage of the universality the prior art of the general four-port high frequency de-embedding has that all the on-wafer parasitics to be stripped are contained in said general four-port network without any assumption on the specific form of the interior structure of said network, the present invention gives full consideration to the non-ideal nature of the practically fabricated on-wafer de-embedding dummies. No idealized lumped assumption is made on the necessary on-wafer de-embedding dummies as in the prior art, and the universality of the prior art of general four-port high frequency de-embedding is inherited and further developed.

In order to achieve this purpose, a general four-port on-wafer high frequency de-embedding method is provided, including the following steps:

1.1: fabricating, together with a device under test (DUT) to be de-embedded, N on-wafer de-embedding dummies;

1.2: measuring to obtain the whole Y-parameter admittance matrix Y_M of said DUT and the whole Y-parameter admittance matrix Y_Mj (j=1,2, . . . , N) of each of said on-wafer de-embedding dummies;

1.3: for each on-wafer de-embedding dummy, building a model considering the distributed nature of high frequency characteristics of the on-wafer de-embedding dummy;

1.4: obtaining the intrinsic Y-parameter admittance matrix Y_Aj(p_i, p₂, . . . p_M) (j=1,2, . . . , N) of said N on-wafer de-embedding dummies by calculation or simulation by using said models,

where, p₁, p₂, . . . , p_M are M model parameters of said models on which said calculation or simulation is based, and 4N−16≧M;

1.5: solving an equation set Y_Mj=Y_ee−Y_ei (Y_Aj(p₁, p₂, . . . p_M)+Y_ii)⁻¹Y_ie(j=1,2, . . . , N) for elements of four sub-matrices Y_ee Y_ii, Y_ei and Y_ie of the admittance matrix Y of said parasitic four-port network to be stripped in de-embedding and said model parameters p₁, p₂, . . . , p_M as unknowns,

wherein, as shown in the following equation, Y_ee, Y_ii, Y_ei and Y_ie as four sub-matrices, form the admittance matrix Y of said parasitic four-port network:

$Y=\left[\begin{array}{cc}{Y}_{\mathrm{ee}}& Y_{\mathrm{ei}}\\ Y_{\mathrm{ie}}& Y_{\mathrm{ii}}\end{array}\right]$

1.6: substituting Y_ee, Y_ii, Y_ei and Y_ie obtained in Step 1.5 and the whole Y-parameter admittance matrix Y_M of said DUT obtained by measurement in Step 1.2 into Equation Y_A=Y_ii−Y_ie(Y_M−Y_ee)⁻¹Y_ei, to obtain by calculation the intrinsic Y-parameter admittance matrix Y_A of said DUT.

Solving the equation set Y_Mj=Y_ee−Y_ei (Y_Aj(p₁, p₂, p_M)+Y_ii)⁻¹ Y_ie (j=1,2, . . . , N) for the elements of four sub-matrixes Y_ee, Y_ii, Y_ei and Y_ie of the admittance matrix Y of said parasitic four-port network to be stripped in de-embedding and said model parameters p₁, p₂, . . . , p_M as unknowns in Step 1.5 comprises the following steps:

2.1: assigning initial values to said model parameters p₁, p₂, . . . p_M, respectively;

2.2: obtaining the value of Y_Aj(p₁, p₂, . . . p_M)(j=1,2, . . . , N) by calculation or simulation by using the assigned model parameters p₁, p₂, . . . p_M;

2.3: after solving an equation set Y_Mj=Y_ee−Y_ei (Y_Aj(p₁, p₂, . . . p_M)+Y_ii)⁻¹ Y_ie(j=1,2,3,4) by using the known measured values Y_Mj(j=1,2,3,4) of the first four on-wafer de-embedding dummies and said calculated or simulated values Y_Aj(p₁, p₂, . . . p_M)(j=1,2,3,4) to obtain the values of Y_ee, Y_ii, Y_ei and Y_ie, substituting the known measured values Y_Mj(j=5,6, . . . , N) of the remaining on-wafer de-embedding dummies and said solved values of Y_ee, Y_ii, Y_ei and Y_ie, into Y_Dj=−Y_ii−Y_ie (Y_Mj−Y_ee)⁻¹Y_ei (j=5,6, . . . , N) to obtain by calculation the de-embedded Y-parameter admittance matrices Y_Dj(j=5,6, . . . , N) of said remaining on-wafer de-embedding dummies;

2.4: comparing the calculated Y_Dj(j=5,6, . . . , N) with the calculated or simulated values Y_Aj(p₁, p₂, . . . p_M)(j=5,6, . . . , N) of the corresponding remaining on-wafer de-embedding dummies already obtained in Step 2.2,

determining final values for said undetermined model parameters p₁, p₂, . . . p_M which are necessary for the calculation or simulation of the on-wafer de-embedding dummies, if a difference between the two meets the set error standard, and

correcting the values of the said model parameters p₁, p₂, . . . p_M and reassigning them, respectively, and then turning back to Step 2.2, if the difference between the two does not meet the set error standard.

In order to achieve this purpose, a general four-port on-wafer high frequency de-embedding method for the four-port network of passivity, reciprocity and symmetry is provided, including the following steps:

3.1: fabricating, together with a device under test (DUT) to be de-embedded, N on-wafer de-embedding dummies of passivity, reciprocity and symmetry;

3.2: measuring to obtain the whole Y-parameter admittance matrix Y_M of said DUT and the whole Y-parameter admittance matrix Y_Mj (j=1,2, . . . , N) of each of said on-wafer de-embedding dummies;

3.3: for each on-wafer de-embedding dummy, building a model considering the distributed nature of high frequency characteristics of the on-wafer de-embedding dummy;

3.4: obtaining the intrinsic Y-parameter admittance matrices Y_Aj(p₁, p₂, . . . p_M) (j=1,2, . . . , N) of said N on-wafer de-embedding dummies by calculation or simulation by using said models,

where, p₁, p₂, . . . , p_M are M model parameters of models on which said calculation or simulation is based, and 2N−6≧M;

3.5: solving an equation set Y_Mj=Y_ee−Y_ei(Y_Aj(p₁, p₂, . . . p_M)+Y_ii)⁻¹Y_ei(j=1,2, . . . , N) for the elements of the three sub-matrices Y_ee, Y_ii and Y_ei of the admittance matrix Y of said parasitic four-port network to be stripped in de-embedding and said model parameters p₁, p₂, . . . , p_M as unknowns,

wherein, as shown in the following equation, Y_ee, Y_ii, and Y_ei, as three sub-matrices, form the admittance matrix Y of said parasitic four-port network:

$Y=\left[\begin{array}{cc}{Y}_{\mathrm{ee}}& Y_{\mathrm{ei}}\\ Y_{\mathrm{ie}}& Y_{\mathrm{ii}}\end{array}\right];$

and

3.6: substituting Y_ee, Y_ii and Y_ei obtained in Step 3.5 and the whole Y-parameter admittance matrix Y_M of said DUT obtained by measurement in Step 3.2 into Equation Y_A=−Y_ii−Y_ei (Y_M−Y_ee)⁻¹Y_ei, to obtain by calculation the intrinsic Y-parameter admittance matrix Y_A of said DUT.

Solving the equation set Y_Mj=Y_ee−Y_ei(Y_Aj(p₁, p₂, . . . p_M)+Y_ii)⁻¹ Y_ei (j=1,2, . . . , N) for the elements of the three sub-matrices Y_ee, Y_ii and Y_ei of the admittance matrix Y of said parasitic four-port network to be stripped in de-embedding and said model parameters p₁, p₂, . . . , p_M as unknowns in Step 3.5 comprises the following steps:

4.1: assigning initial values to said model parameters p₁, p₂, . . . p_M, respectively;

4.2: obtaining the values of Y_Aj(p₁, p₂, . . . p_M)(j=1,2, . . . , N) by calculation or simulation by using the assigned model parameters p₁, p₂, . . . p_M;

4.3: after solving an equation set Y_Mj=Y_ee−Y_ei(Y_Aj(p₁, p₂, . . . p_M)+Y_ii)⁻¹ Y_ei (j=1,2,3) by using the known test values Y_Mj(j=1,2,3) of the first three on-wafer de-embedding dummies and said calculated or simulated values Y_Aj(p₁, p₂, . . . p_M) (j=1,2,3) to obtain the values of Y_ee, Y_ii and Y_ei (it is unnecessary to completely solve Y_ei, referring to claim 5.4), substituting the known test values Y_Mj(j=4,5, . . . , N) of the remaining on-wafer de-embedding dummies and said obtained values of Y_ee, Y_ii and Y_ei into Y_Dj=−Y_ii−Y_ei(Y_Mj−Y_ee)⁻¹Y_ei (j=4,5, . . . , N) to obtain by calculation the de-embedded Y-parameter admittance matrices Y_Dj(j=4,5, . . . , N) of said remaining on-wafer de-embedding dummies;

4.4: comparing the calculated Y_Dj(j=4,5, . . . , N) with the calculated or simulated values Y_Aj(p₁, p₂, . . . p_M)(j=4,5, . . . , N) of the corresponding remaining on-wafer de-embedding dummies already obtained in Step 4.2, determining final values for undetermined model parameters p₁, p₂, . . . p_M which are necessary for the calculation or simulation of the on-wafer de-embedding dummies, if a difference between the two meets the set error standard, and

correcting the values of the model parameters p₁, p₂, . . . p_M and reassigning them, and then turning back to Step 4.2, if the difference between the two does not meet the set error standard.

Solving the equation set Y_Mj=Y_ee−Y_ei(Y_Aj(p₁, p₂, . . . p_M)+Y_ii)⁻¹Y_ie (j=1,2,3) to further obtain by calculation the de-embedded Y-parameter admittance matrices Y_Dj(j=4,5, . . . , N) of said remaining on-wafer de-embedding dummies in Step 4.3 comprises the following steps:

5.1: obtaining by calculation matrices Z_2A=(Y_A2−Y_A1)⁻¹, Z_2M=(Y_M2−Y_M1)⁻¹, Z_3A=(Y_A3−Y_A1)⁻¹ and Z_3M=(Y_M3−Y_M1)⁻¹, where exponent −1 represents matrix inversion;

5.2: calculating quantities

$r_{2p}=\frac{z_{2A11}+z_{2A12}}{z_{2M11}+z_{2M12}},r_{3p}=\frac{z_{3A11}+z_{3A12}}{z_{3M11}+z_{3M12}},r_{2M}=\frac{z_{2A11}-z_{2A12}}{z_{2M11}-z_{2M12}},r_{3m}=\frac{z_{3A11}-z_{3A12}}{z_{3M11}-z_{3M12}},y_p=\frac{r_{3p}\left(y_{A311}+y_{A312}\right)-r_{2p}\left(y_{A211}+y_{A212}\right)}{r_{2p}-r_{3p}},y_m=\frac{r_{3m}\left(y_{A311}-y_{A312}\right)-r_{2m}\left(y_{A211}-y_{A212}\right)}{r_{2m}-r_{3m}},$

x_p=r_2p(y_p+y_A111+y_A112)(y_p+y_A211+y_A212) and x_m=r_2m(y_m+y_A111−y_A112)(y_m+y_A211−y_A212), where z_2A11 and z_2A12 are respectively z₁₁ and z₁₂ of Z_2A; z_2M11 and z_2M12 are respectively z₁₁ and z₁₂ of Z_2M; z_3A11 and z_3A12 are respectively z₁₁ and z₁₂ of Z_3A; z_3M11 and z_3M12 are respectively z₁₁ and z₁₂ of Z_3M; y_A111 and y_A112 are respectively y₁₁ and y₁₂ of Y_A1; y_A211 and y_A212 are respectively y₁₁ and y₁₂ of Y_A2; y_A311 and y_A312 are respectively y₁₁ and y₁₂ of Y_A3;

5.3: calculating quantities

$y_{\mathrm{ii}11}=\frac{y_p+y_m}{2}\mathrm{and}y_{\mathrm{ii}12}=\frac{y_p-y_m}{2}$

to obtain the matrix

$Y_{\mathrm{ii}}=\left[\begin{array}{cc}{y}_{\mathrm{ii}11}& y_{\mathrm{ii}12}\\ y_{\mathrm{ii}12}& y_{\mathrm{ii}11}\end{array}\right];$

5.4: obtaining by calculation the square of elements y₁₁ and y₁₂ of the matrix Y_ei:

$y_{\mathrm{ei}11}^2=\frac{x_p+x_m\pm 2\sqrt{x_px_m}}{4}\mathrm{and}y_{\mathrm{ei}12}^2=\frac{x_p+x_m\mp 2\sqrt{x_px_m}}{4},$

where the plus-minus sign is selected such that, at the low frequency limit, y_ei11² tends to infinity and y_ei12² tends to zero, and both y_ei11² and y_ei12² continuously vary with frequency;

5.5: obtaining by calculation a matrix Z_Ai=(Y_A1+Y_ii)⁻¹;

5.6: calculating quantities

$y_{\mathrm{Ai}11}=z_{\mathrm{Ai}11}\frac{x_p+x_m}{2}+z_{\mathrm{Ai}12}\frac{x_p-x_m}{2}\mathrm{and}$ $y_{\mathrm{Ai}12}=z_{\mathrm{Ai}12}\frac{x_p+x_m}{2}+z_{\mathrm{Ai}11}\frac{x_p-x_m}{2}$

to obtain a matrix

$Y_{\mathrm{Ai}}=\left[\begin{array}{cc}{y}_{\mathrm{Ai}11}& y_{\mathrm{Ai}12}\\ y_{\mathrm{Ai}12}& y_{\mathrm{Ai}11}\end{array}\right],$

where z_Ai11 and z_Ai12 are respectively z₁₁ and z₁₂ of Z_Ai;

5.7: obtaining by calculation the matrix Y_ee=Y_M1+Y_Ai;

5.8: obtaining by calculation a matrix Z_Mej=(Y_ee−Y_Mj)⁻¹(j=4,5, . . . , N);

5.9: calculating quantities

$y_{\mathrm{Mej}11}=z_{\mathrm{Mej}11}y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Mej}12}+z_{\mathrm{Mej}21}\right)\frac{x_{p}-x_m}{4}+z_{\mathrm{Mej}22}Y_{\mathrm{ei}12}^2\left(j=4,5,\dots ,N\right),y_{\mathrm{Mej}12}=z_{\mathrm{Mej}12}y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Mej}11}+z_{\mathrm{Mej}22}\right)\frac{x_p-x_m}{4}+z_{\mathrm{Mej}21}y_{\mathrm{ei}12}^2\left(j=4,5,\dots ,N\right),y_{\mathrm{Mej}21}=z_{\mathrm{Mej}21}y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Mej}11}+z_{\mathrm{Mej}22}\right)\frac{x_p-x_m}{4}+z_{\mathrm{Mej}12}y_{\mathrm{ei}12}^2\left(j=4,5,\dots ,N\right)\mathrm{and}$ $y_{\mathrm{Mej}22}=z_{\mathrm{Mej}22}y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Mej}12}+z_{\mathrm{Mej}21}\right)\frac{x_p-x_m}{4}+z_{\mathrm{Mej}11}y_{\mathrm{ei}12}^2\left(j=4,5,\dots ,N\right)$

to obtain a matrix

$Y_{\mathrm{Mej}}=\left[\begin{array}{cc}{y}_{\mathrm{Mej}11}& y_{\mathrm{Mej}12}\\ y_{\mathrm{Mej}21}& y_{\mathrm{Mej}22}\end{array}\right]\left(j=4,5,\dots ,N\right),$

where z_Mej11, z_Mej12, z_Mej21 and z_Mej22 are respectively z₁₁, z₁₂, z₂₁ and z₂₂ of Z_Mej; and

5.10: obtaining by calculation the de-embedded Y-parameter matrix Y_Dj=Y_Mej−Y_ii(j=4,5, . . . , N) of said remaining on-wafer de-embedding dummies.

Calculating the intrinsic Y-parameter admittance matrix Y_A of said DUT by using the whole Y-parameter admittance matrix Y_M of said DUT obtained by measurement in Step 3.6 comprises the following steps:

6.1: obtaining by calculation a matrix Z_Me=(Y_ee−Y_M)⁻¹;

6.2: calculating quantities

$y_{\mathrm{Me}11}=z_{\mathrm{Me}11}y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Me}12}+z_{\mathrm{Me}21}\right)\frac{x_p-x_m}{4}+z_{\mathrm{Me}22}y_{\mathrm{ei}12}^2,y_{\mathrm{Me}12}=z_{\mathrm{Me}12}y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Me}11}+z_{\mathrm{Me}22}\right)\frac{x_p-x_m}{4}+z_{\mathrm{Me}21}y_{\mathrm{ei}12}^2,y_{\mathrm{Me}21}=z_{\mathrm{Me}21}y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Me}11}+z_{\mathrm{Me}22}\right)\frac{x_p-x_m}{4}+z_{\mathrm{Me}12}y_{\mathrm{ei}12}^2\mathrm{and}$ $y_{\mathrm{Me}22}=z_{\mathrm{Me}22}y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Me}12}+z_{\mathrm{Me}21}\right)\frac{x_p-x_m}{4}+z_{\mathrm{Me}11}y_{\mathrm{ei}12}^2$

to obtain a matrix

$Y_{\mathrm{Me}}=\left[\begin{array}{cc}{y}_{\mathrm{Me}11}& y_{\mathrm{Me}12}\\ y_{\mathrm{Me}21}& y_{\mathrm{Me}22}\end{array}\right],$

where z_Me11, z_Me12, z_Me21 and z_Me22 are respectively z₁₁, z₁₂, z₂₁ and z₂₂ of Z_Me; and

6.3: obtaining by calculation the intrinsic Y-parameter admittance matrix Y_A=Y_Me−Y_ii of said DUT to be de-embedded.

On the basis of inheriting such a universality that the existing technology of universal four-port high frequency de-embedding contains all parasitic parameters to be stripped in said universal four-port network without any assumption on the specific form of the interior structure of said network, the present invention gives full consideration to the non-ideal essence of the practically needed de-embedding co-testing structures. No lumped and idealized assumption is made on the necessary de-embedding co-testing structures as in the prior art, and the universality of the existing technology of universal four-port high frequency de-embedding is inherited and further developed.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic view of an on-wafer high frequency characterization measurement set-up;

FIG. 2 is a schematic view of a general parasitic four-port network for an on-wafer high frequency characterization measurement;

FIG. 3 shows the equivalent circuits of five on-wafer de-embedding two-port dummies generally used in the prior art of the general four-port on-wafer high frequency de-embedding method: Open O (a), Short S (b), Left L (c), Right R (d) and Thru T (e);

FIG. 4 is a schematic view of a strip-line Thru (T) on-wafer de-embedding dummy in Implementation 1;

FIG. 5 is a schematic view of an Open (O) on-wafer de-embedding dummy in Implementation 1;

FIG. 6 is a schematic view of a Short (S) on-wafer de-embedding dummy in Implementation 1;

FIG. 7 is a schematic view of a Left (L) on-wafer de-embedding dummy in Implementation 1;

FIG. 8 is a schematic view of a Right (R) on-wafer de-embedding dummy in Implementation 1;

FIG. 9 is a schematic view of a Thru 1 (T1) on-wafer de-embedding dummy in Implementation 2;

FIG. 10 is a schematic view of a Thru 2 (T2) on-wafer de-embedding dummy in Implementation 2;

FIG. 11 is a schematic view of a Thru 3 (T3) on-wafer de-embedding dummy in Implementation 2; and

FIG. 12 is a schematic view of a Thru 4 (T4) on-wafer de-embedding dummy in Implementation 2.

DETAILED DESCRIPTION OF THE PRESENT INVENTION

Implementation 1: one implementation of a general four-port on-wafer high frequency de-embedding method for a case in which parasitics to be stripped form a general parasitic four-port network

(1) Five on-wafer de-embedding dummies are designed and fabricated together with a device under test (DUT) to be de-embedded, i.e., five on-wafer de-embedding two-port dummies generally adopted in the prior art of the general four-port on-wafer high frequency de-embedding method, including: Open (O), Short (S), Left (L), Right (R) and strip-line Thru (T). However, no assumption is made on the form of their equivalent circuits as in FIG. 3. Instead, they are all considered as a general two-port network. Their schematic structural diagrams are respectively as shown in FIGS. 4, 5, 6, 7 and 8, wherein a lower grounded metal, an intermediate signal wire metal, and an upper grounded metal of the strip-line Thru T are respectively made of three layers of metal in the multi-layer metal interconnection process of a semiconductor integrated circuit, and a dielectric layer is filled between the lower grounded metal and the upper grounded metal and surrounds said signal wire metal; the signal wire is removed from the strip-line Thru T dummy to obtain an Open dummy O; the input and output ports of the Open dummy O are connected to the lower grounded metal and the upper grounded metal, respectively, by a via/metal stack, to obtain a Short dummy S; the input port of the Open dummy O is connected to the lower grounded metal and the upper grounded metal by a via/metal stack to obtain a Left dummy L; the output port of the Open structure O is connected to the lower grounded metal and the upper grounded metal by a via/metal stack to obtain a Right dummy L.

(2) By an on-wafer high frequency measurement instruments and techniques, the whole Y-parameter admittance matrix Y_M of said DUT and the whole Y-parameter admittance matrices, which are respectively denoted by Y_MO, Y_MS, Y_ML, Y_MR and Y_MT, of said five on-wafer de-embedding dummies are obtained by measurement.

(3) In the premise of giving consideration to the distributive nature of high frequency characteristics of said on-wafer de-embedding dummies, models for said Open (O), Short (S), Left (L), Right (R) and Thru (T) are built in a passive electromagnetic field simulation software environment, respectively, according to structures as shown in FIGS. 4, 5, 6, 7 and 8, and electromagnetic field simulation is performed by said passive electromagnetic field simulation software on the basis of the built models to obtain intrinsic Y-parameter admittance matrices of said five on-wafer de-embedding dummies, which are respectively denoted by Y_AO (σ, ∈_r), Y_AS (σ, ∈_r), Y_AL (σ, ∈_r), Y_AR(σ, ∈_r) and Y_AT (σ, ∈_r), where, σ and ∈_r, as model parameters of models on which said simulation is based, are respectively the conductivity of metals in said high frequency on-wafer de-embedding dummies and the relative dielectric constant of the dielectric layer material filled among the metals.

(4) An equation set Y_Mj=Y_ee−Y_ei (Y_Aj(σ, ∈_r)+Y_ii)⁻¹Y_ie (j=O,S,L,R,T) is solved for the elements of the related admittance matrixes Y_ee, Y_ii, Y_ei and Y_ie of the parasitic four-port network to be stripped in de-embedding and said model parameters σ and ∈_r as unknowns. Specifically, the following steps are included:

(4-1) on the basis that initial default values are taken for the model parameters σ and ∈_r respectively, obtaining the values of Y_Aj(σ, ∈_r)(j=O, S, L, R, T) by passive electromagnetic field simulation;

(4-2) after solving an equation set Y_Mj=Y_ee−Y_ei(Y_Aj(σ, ∈_r)+Y_ii)⁻¹Y_ie (j=O, S, L, R) by using the known measurement values Y_Mj (j=O, S, L, R) of the first four on-wafer de-embedding dummies and said simulated values Y_Aj(σ,∈_r)(j=O,S, L, R) to obtain the values of Y_ee, Y_ii, Y_ei and Y_ie, substituting the known measurement values Y_MT of the strip-line Thru T and said obtained values of Y_ee, Y_ii, Y_ei and Y_ie into Y_DT=−Y_ii−Y_ie(Y_MT−Y_ee)⁻¹Y_ei to obtain by calculation the de-embedded Y-parameter admittance matrix Y_DT of the strip-line Thru T;

(4-3) comparing the calculated Y_DT with the simulated value Y_AT(σ, ∈_r) of the strip-line Thru T already obtained in Step (4-1), properly correcting the values of the model parameters σ and ∈_r and then turning back to Step (4-1) if the difference between the two does not meet the set error standard, and obtaining the values of Y_Aj(σ,∈_r)(j=O, S, L, R, T) again by passive electromagnetic field simulation by using the corrected model parameters; and

(4-4) once turning back to Step (4-1), starting a cyclic iterative fitting process from Step (4-1) to Step (4-3), performing iterative optimization fitting by optimization algorithms such as inverse modeling, and continuously correcting the values of the model parameters σ and ∈_r used for simulation of intrinsic Y-parameter admittance matrices of the on-wafer de-embedding dummies until a difference between the de-embedded Y-parameter admittance matrix Y_DT of the strip-line Thru T and the corresponding simulated value Y_AT (σ, ∈_r) meets a set error standard, that is, determining the final values for the undetermined mode parameters σ and ∈_r which are necessary for the simulation of the on-wafer de-embedding dummies by this iterative fitting between Y_DT and Y_AT (σ, ∈_r).

(5) The solved Y_ee, Y_ii, Y_ei and Y_ie and the whole Y-parameter admittance matrix Y_M of said DUT obtained by measurement are substituted into the right side of Y_A=−Y_ii−Y_ie(Y_M−Y_ee)⁻¹Y_ei to complete said general four-port on-wafer high frequency de-embedding. That is, the intrinsic Y-parameter admittance matrix Y_A of said DUT is calculated by using the whole Y-parameter admittance matrix Y_M of said DUT obtained by measurement.

Implementation 2: one implementation of a general four-port on-wafer high frequency de-embedding method for a case in which all the on-wafer parasitics to be stripped form a parasitic four-port network of passivity, reciprocity and symmetry.

(1) Four on-wafer strip-line Thru de-embedding dummies of passivity, reciprocity and symmetry and different in width are designed and fabricated together with a device under test (DUT) to be de-embedded, i.e., Thru 1, Thru 2, Thru 3 and Thru 4 which are respectively denoted by T1, T2, T3 and T4. Their schematic structure diagrams are respectively as shown in FIGS. 9, 10, 11 and 12, wherein said strip-line Thru signal wires have widths of W₁, W₂, W₃ and W₄, respectively, but they all have the same length of L, and they all have the same thickness of t. The dielectric layers between the signal wires and the upper and the lower grounded metals have the same thickness of b/2.

(2) By using on-wafer high frequency measurement instruments and techniques, the whole Y-parameter admittance matrix Y_M of said DUT and the whole Y-parameter admittance matrices Y_Mj (j=1,2,3,4) of said four on-wafer strip-line Thru de-embedding dummies are obtained by measurement.

(3) In the premise of giving consideration to the distributive nature of high frequency characteristics of said on-wafer de-embedding dummies, analytical models for said four on-wafer strip-line Thru de-embedding dummies are built, respectively, and following intrinsic Y-parameter admittance matrices of said four Thru de-embedding dummies are calculated by the built models:

$Y_{\mathrm{Aj}}\left(\sigma ,{\varepsilon }_r\right)=\left[\begin{array}{cc}{y}_{\mathrm{Aj}11}\left(\sigma ,{\varepsilon }_r\right)& y_{\mathrm{Aj}12}\left(\sigma ,{\varepsilon }_r\right)\\ y_{\mathrm{Aj}12}\left(\sigma ,{\varepsilon }_r\right)& y_{\mathrm{Aj}11}\left(\sigma ,{\varepsilon }_r\right)\end{array}\right],y_{\mathrm{Aj}11}\left(\sigma ,{\varepsilon }_r\right)=\frac{\left(W_j+0.441b\right)\sqrt{{\epsilon }_r}}{30\pi b\tanh \left\{\begin{array}{c}\frac{0.0027\mathrm{bL}}{\left(b-t\right)\left(W_j+0.441b\right)}\left[1+\frac{2W_j}{b-t}+\frac{1b+t}{\pi b-t}\ln \left(\frac{2b-t}{t}\right)\right]\\ \sqrt{\frac{\pi {\epsilon }_r{\mu }_0f}{\sigma }}+j\frac{2\pi \mathrm{fL}\sqrt{{\epsilon }_r}}{c}\end{array}\right\}},y_{\mathrm{Aj}12}\left(\sigma ,{\varepsilon }_r\right)=-\frac{\left(W_j+0.441b\right)\sqrt{{\epsilon }_r}}{30\pi b\sinh \left\{\begin{array}{c}\frac{0.0027\mathrm{bL}}{\left(b-t\right)\left(W_j+0.441b\right)}\left[1+\frac{2W_j}{b-t}+\frac{1b+t}{\pi b-t}\ln \left(\frac{2b-t}{t}\right)\right]\\ \sqrt{\frac{\pi {\epsilon }_r{\mu }_0f}{\sigma }}+j\frac{2\pi \mathrm{fL}\sqrt{{\epsilon }_r}}{c}\end{array}\right\}}$ $\left(j=1,2,3,4\right),$

where, σ and ∈_r, as model parameters of models on which said calculation is based, are respectively the conductivity of metals in the on-wafer Thru de-embedding dummies and the relative dielectric constant of the dielectric material, μ₀=4π×10⁻⁷H/m is the permeability of vacuum, f is the frequency of test and simulation, and c=3×10⁸ m/s. is the speed of light in vacuum.

(4) An equation set Y_Mj=Y_ee−Y_ei(Y_Aj(σ, ∈_r)+Y_ii)⁻¹Y_ei(j=1,2,3,4) is solved for the elements of the related admittance matrixes Y_ee, Y_ii and Y_ei of the parasitic four-port network to be stripped in de-embedding and said model parameters σ and ∈_r as unknowns. Specifically, the following steps are included:

(4-1) on the basis that the initial default values are taken for the model parameters σ and ∈_r respectively, obtaining the values of Y_Aj(σ, ∈_r)(j=1,2,3,4) by calculation;

(4-2) after solving an equation set Y_Mj=Y_ee−Y_ei(Y_Aj(σ,∈_r)+Y_ii)⁻¹Y_ei (j=1,2,3) by using the known measurement values Y_Mj (j=1,2,3) of the three de-embedding dummies T1, T2 and T3 and said calculated values Y_Aj(σ,∈_r)(j=1,2,3) to obtain the values of Y_ee, Y_ii and Y_ei (it is unnecessary to completely solve Y_ei, as long as the square of its elements is obtained), substituting the known measurement value Y_M4 of the T4 de-embedding dummy and said obtained values of Y_ee, Y_ii and Y_ei into the right side of Y_D4=−Y_ii−Y_ei(Y_M4−Y_ee)⁻¹Y_ei to obtain by calculation the de-embedded Y-parameter admittance matrix Y_D4 of the T4 de-embedding dummy, specifically including the following steps:

(4-2-1) obtaining by calculation matrices Z_2A=(Y_A2−Y_A1)⁻¹, Z_2M=(Y_M2−Y_M1)⁻¹, Z_3A=(Y_A3−Y_Aj)⁻¹ and Z_3M=(Y_M3−Y_M1)⁻¹, where exponent −1 represents matrix inversion;

(4-2-2) calculating quantities

$r_{2p}=\frac{z_{2A11}+z_{2A12}}{z_{2M11}+z_{2M12}},r_{3p}=\frac{z_{3A11}+z_{3A_{12}}}{z_{3M11}+z_{3M12}},r_{2m}=\frac{z_{2A11}-z_{2A12}}{z_{2M11}-z_{2M12}},r_{3m}=\frac{z_{3A11}-z_{3A12}}{z_{3M_{11}-z_{3M12}}},y_p=\frac{r_{3p}\left(y_{A311}+y_{A312}\right)-r_{2p}\left(y_{A211}+y_{A212}\right)}{r_{2p}-r_{3p}},y_m=\frac{r_{3m}\left(y_{A311}-y_{A312}\right)-r_{2m}\left(y_{A211}-y_{A212}\right)}{r_{2m}-r_{3m}},x_p=r_{2p}\left(y_p+y_{A111}+y_{A112}\right)\left(y_p+y_{A211}+y_{A212}\right)\mathrm{and}$ $x_m=r_{2m}\left(y_m+y_{A111}-y_{A112}\right)\left(y_m+y_{A211}-y_{A212}\right),$

where z_2A11 and z_2A12 are respectively z₁₁ and z₁₂ of Z_2A; z_2M11 and z_2M12 are respectively and z₁₁ and z₁₂ of Z_2M; z_3A11 and z_3A12 are respectively z₁₁ and z₁₂ of Z_3A; z_3M11 and z_3M12 are respectively z₁₁ and z₁₂ of Z_3M; y_A111 and y_A112 are respectively y₁₁ and y₁₂ of Y_A1; y_A211 and y_A212 are respectively y₁₁ and y₁₂ of Y_A2, y_A311 and y_A312 are respectively y₁₁ and y₁₂ of Y_A3;

(4-2-3) calculating quantities

$y_{\mathrm{ii}11}=\frac{y_p+y_m}{2}\mathrm{and}y_{\mathrm{ii}12}=\frac{y_p-y_m}{2}$

to obtain the matrix

$Y_{\mathrm{ii}}=\left[\begin{array}{cc}{y}_{\mathrm{ii}11}& y_{\mathrm{ii}12}\\ y_{\mathrm{ii}12}& y_{\mathrm{ii}11}\end{array}\right];$

(4-2-4) obtaining by calculation the square of elements y₁₁ and y₁₂ of the matrix

$Y_{\mathrm{ei}}\text{:}y_{\mathrm{ei}11}^2=\frac{x_p+x_m\pm 2\sqrt{x_px_m}}{4}\mathrm{and}$ $y_{\mathrm{ei}12}^2=\frac{x_p+x_m\mp 2\sqrt{x_px_m}}{4},$

where the plus-minus sign is selected such that, at the low frequency limit, y_ei11² tends to infinity and y_ei12² tends to zero, and both y_ei11² and y_ei12² continuously vary with frequency;

(4-2-5) obtaining by calculation a matrix Z_Aj=(Y_A1+Y_ii)⁻¹;

(4-2-6) calculating quantities

$y_{\mathrm{Ai}11}=z_{\mathrm{Ai}11}\frac{x_p+x_m}{2}+z_{\mathrm{Ai}12}\frac{x_p-x_m}{2}\mathrm{and}$ $y_{\mathrm{Ai}12}=z_{\mathrm{Ai}12}\frac{x_p+x_m}{2}+z_{\mathrm{Ai}11}\frac{x_p-x_m}{2}$

to obtain a matrix

$Y_{\mathrm{Ai}}=\left[\begin{array}{cc}{y}_{\mathrm{Ai}11}& y_{\mathrm{Ai}12}\\ y_{\mathrm{Ai}12}& y_{\mathrm{Ai}11}\end{array}\right],$

where z_Ai11 and z_Ai12 are respectively z₁₁ and z₁₂ of Z_Ai;

(4-2-7) obtaining by calculation the matrix Y_ee=Y_M1+Y_Ai;

(4-2-8) obtaining by calculation a matrix Z_Me4=(Y_ee−Y_M4)⁻¹;

(4-2-9) calculating quantities

$y_{\mathrm{Me}411}=z_{\mathrm{Me}411}y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Me}412}+z_{\mathrm{Me}421}\right)\frac{x_p-x_m}{4}+z_{\mathrm{Me}422}y_{\mathrm{ei}12}^2,y_{\mathrm{Me}412}=z_{\mathrm{Me}412}=y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Me}411}+z_{\mathrm{Me}422}\right)\frac{x_p-x_m}{4}+z_{\mathrm{Me}421}y_{\mathrm{ei}12}^2,y_{\mathrm{Me}421}=z_{\mathrm{Me}421}Y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Me}411}+z_{\mathrm{Me}422}\right)\frac{x_p-x_m}{4}+z_{\mathrm{Me}412}y_{\mathrm{ei}12}^2\mathrm{and}$ $y_{\mathrm{Me}422}=z_{\mathrm{Me}422}^2Y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Me}412}+z_{\mathrm{Me}421}\right)\frac{x_p-x_m}{4}+z_{\mathrm{Me}411}y_{\mathrm{ei}12}^2$

to obtain a matrix

$Y_{\mathrm{Me}4}=\left[\begin{array}{cc}{y}_{\mathrm{Me}411}& y_{\mathrm{Me}412}\\ y_{\mathrm{Me}421}& y_{\mathrm{Me}422}\end{array}\right],$

where z_Me411, z_Me412, z_Me421 and z_Me422 are respectively z₁₁, z₁₂, z₂₁ and z₂₂ of Z_Me4;

(4-2-10) obtaining by calculation the de-embedded Y-parameter matrix Y_D4=Y_Me4 Y_ii of the T4 de-embedding dummy;

(4-3) comparing the calculated Y_D4 with the simulated value Y_A4 (σ, ∈_r) of the T4 de-embedding dummy already obtained in Step (4-1), properly correcting the values of the model parameters σ and ∈_r and then turning back to Step (4-1) if the difference between the two does not meet the set error standard, and obtaining the values of Y_Aj(σ,∈_r)(j=1,2,3,4) again by calculation by using the corrected model parameters; and

(4-4) once turning back to Step (4-1), starting a cyclic iterative fitting process from Step (4-1) to Step (4-3), performing iterative optimization fitting by a trial and error method, and continuously correcting the values of the model parameters σ and ∈_r used for calculation of intrinsic Y-parameter admittance matrices of the de-embedding dummies until a difference between the de-embedded Y-parameter admittance matrix Y_D4 of the T4 de-embedding dummy and a corresponding calculated value Y_A4(σ, ∈_r) meets the set error standard, that is, determining the final values for the undetermined mode parameters σ and ∈_r which are necessary for the calculation of the de-embedding dummies by this iterative fitting between Y_D4 and Y_A4(σ,∈_r).

(5) The Y_ee, Y_ii and Y_ei (actually, it is unnecessary to completely determine a specific value for all elements in Y_ei, as long as the square of the elements is determined) solved in Step (4) and the whole Y-parameter admittance matrix Y_M of said DUT obtained by measurement in Step (2) are substituted into the right side of Y_A=−Y_ii−Y_ei (Y_M−Y_ee)⁻¹Y_ei to complete said general four-port on-wafer high frequency de-embedding of the parasitic four-port network, to be stripped, of passivity, reciprocity and symmetry. That is, the intrinsic Y-parameter admittance matrix Y_A of said DUT is calculated by using the whole Y-parameter admittance matrix Y_M of said DUT obtained by measurement. Specifically, the following steps are included:

(5-1) obtaining by calculation a matrix Z_Me=(Y_ee−Y_M)⁻¹;

(5-2) calculating quantities

$y_{\mathrm{Me}11}=z_{\mathrm{Me}11}y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Me}12}+z_{\mathrm{Me}21}\right)\frac{x_p-x_m}{4}+z_{\mathrm{Me}22}y_{\mathrm{ei}12}^2,y_{\mathrm{Me}12}=z_{\mathrm{Me}12}y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Me}11}+z_{\mathrm{Me}22}\right)\frac{x_p-x_m}{4}+z_{\mathrm{Me}21}y_{\mathrm{ei}12}^2,y_{\mathrm{Me}21}=z_{\mathrm{Me}21}y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Me}11}+z_{\mathrm{Me}22}\right)\frac{x_p-x_m}{4}+z_{\mathrm{Me}12}Y_{\mathrm{ei}12}^2\mathrm{and}$ $y_{\mathrm{Me}22}=z_{\mathrm{Me}22}y_{\mathrm{ei}11}^2+\left(z_{\mathrm{Me}12}+z_{\mathrm{Me}21}\right)\frac{x_p-x_m}{4}+z_{\mathrm{Me}11}y_{\mathrm{ei}12}^2$

to obtain a matrix

$Y_{\mathrm{Me}}=\left[\begin{array}{cc}{y}_{\mathrm{Me}11}& y_{\mathrm{Me}12}\\ y_{\mathrm{Me}21}& y_{\mathrm{Me}22}\end{array}\right],$

where z_Me11, z_Me12, z_Me21 and Z_Me22 are respectively z₁₁, z₁₂, z₂₁ and z₂₂ of Z_Me; and

(5-3) obtaining by calculation the intrinsic Y-parameter admittance matrix Y_A=Y_Me−Y_ii of said DUT to be de-embedded.

The foregoing descriptions are merely preferred embodiments of the present invention and the protection scope of the present invention is not limited thereto. Any changes or replacements that readily occur to those skilled in the art within the technical scope disclosed in the present invention should be included within the protection scope of the present invention. Hence, the protection scope of the present invention should be subject to the protection scope defined in the claims.

Claims

1. A general four-port on-wafer high frequency de-embedding method, comprising the following steps: 1.1: fabricating, together with a device under test (DUT) to be de-embedded, N on-wafer de-embedding dummies;1.2: measuring to obtain the whole Y-parameter admittance matrix YM of said DUT and the whole Y-parameter admittance matrix YMj (j=1,2, . . . , N) of each of said de-embedding dummies;1.3: for each de-embedding dummy, building a model considering the distributive nature of high frequency characteristics of the de-embedding dummy;1.4: obtaining the intrinsic Y-parameter admittance matrix YAj(p1, p2, . . . pM) (j=1,2, . . . , N) of said N de-embedding dummies by calculation or simulation by using said models,where, p1, p2, . . . , pM are M model parameters of models on which said calculation or simulation is based, and 4N−16≧M;1.5: solving an equation set YMj=Yee−Yei(YAj(p1, p2, . . . pM)+Yii)−1Yie (j=1,2, . . . , N) for the elements of four sub-matrices Yee, Yii, Yei and Yie of the admittance matrix Y of the parasitic four-port network to be stripped in de-embedding and said model parameters p1, p2, . . . , pM as unknowns,wherein, as shown in the following equation, Yee, Yii, Yei and Yie as four sub-matrices, form an admittance matrix Y of said parasitic four-port network:

2. The general four-port on-wafer high frequency de-embedding method according to claim 1, characterized in that solving the equation set YMj=Yee−Yei(YAj(p1, p2, . . . pM)+Yii)−1Yie(j=1,2, . . . , N) for the elements of four sub-matrixes Yee, Yii, Yei and Yie of the admittance matrix Y of the parasitic four-port network to be stripped in de-embedding and said model parameters p1, p2, . . . , pM as unknowns in Step 1.5 comprises the following steps: 2.1: assigning initial values to said model parameters p1, p2, . . . pM, respectively;2.2: obtaining the values of YAj(p1, p2, . . . pM)(j=1,2, . . . , N) by calculation or simulation by using the assigned model parameters p1, p2, . . . pM;2.3: after solving an equation set YMj=Yee−Yei(YAj(p1, p2, . . . pM)+Yii)−1 Yie (j=1,2,3,4) by using the known measurement values YMj(j=1,2,3,4) of the first four de-embedding dummies and said calculated or simulated values YAj(p1, p2, . . . pM)(j=1,2,3,4) to obtain the values of Yee, Yii, Yei and Yie, substituting the known measurement values YMj(j=5,6, . . . , N) of the remaining de-embedding dummies and said obtained values of Yee, Yii, Yei and Yie into YDj=−Yii−Yie (YMj−Yee)−1Yei (j=5,6, . . . , N) to obtain by calculation the de-embedded Y-parameter admittance matrix YDj(j=5,6, . . . , N) of said remaining de-embedding dummies;2.4: comparing the calculated YDj(j=5,6, . . . , N) with the calculated or simulated values YAj(p1, p2, . . . pM)(j=5,6, . . . , N) of the corresponding remaining de-embedding dummies already obtained in Step 2.1,determining final values for undetermined model parameters p1, p2, . . . pM which are necessary for the calculation or simulation of the de-embedding dummies, if a difference between the two meets the set error standard, andcorrecting the values of the model parameters p1, p2, . . . pM and reassigning them, and then turning back to Step 2.2, if the difference between the two does not meet the set error standard.

3. A general four-port on-wafer high frequency de-embedding method of passivity, reciprocity and symmetry, comprising the following steps: 3.1: fabricating, together with a device under test (DUT) to be de-embedded, N de-embedding dummies of passivity, reciprocity and symmetry;3.2: measuring to obtain the whole Y-parameter admittance matrix YM of said DUT and the whole Y-parameter admittance matrix YMj (j=1,2, . . . , N) of each of said de-embedding dummies;3.3: for each de-embedding dummy, building a model considering the distributive nature of high frequency characteristics of the de-embedding dummy;3.4: obtaining the intrinsic Y-parameter admittance matrices YAj(p1, p2, . . . pM)(j=1,2, . . . , N) of said N de-embedding dummies by calculation or simulation by using said models,where, p1, p2, . . . , pM are M model parameters of models on which said calculation or simulation is based, and 2N−6≧M;3.5: solving an equation set YMj=Yee−Yei (YAj(p1, p2, . . . pM)+Yii)−1 Yei(j=1,2, . . . , N) for the elements of three sub-matrices Yee, Yii and Yei of the admittance matrix Y of the parasitic four-port network to be stripped in de-embedding and said model parameters p1, p2, . . . , pM as unknowns,wherein, as shown in the following equation, Yee, Yii and Yei, as three sub-matrices, form the admittance matrix Y of said parasitic four-port network:

4. The special general four-port on-wafer high frequency de-embedding method according to claim 3, characterized in that solving the equation set YMj=Yee−Yei(YAj(p1, p2, . . . pM)+Yii)−1Yei(j=1,2, . . . , N for the elements of three sub-matrices Yee, Yii and Yei of the admittance matrix Y of the parasitic four-port network to be stripped in de-embedding and said model parameters p1, p2, . . . , pM as unknowns in Step 3.5 comprises the following steps: 4.1: assigning initial values to said model parameters p1, p2, . . . pM, respectively;4.2: obtaining the values of YAj(p1, p2, . . . pM)(j=1,2, . . . , N) by calculation or simulation by using the assigned model parameters p1, p2, . . . pM;4.3: after solving an equation set YMj=Yee−Yei(YAj(p1, p2, . . . pM)+Yii)−1 Yei(j=1,2,3) by using the known measurement values YMj(j=1,2,3) of first three de-embedding dummies and said calculated or simulated values YAj(p1, p2, . . . pM) (j=1,2,3) to obtain the values of Yee, Yii and Yei (it is unnecessary to completely solve Yei, referring to claim 5.4), substituting the known measurement values YMj (j=4,5, . . . , N) of the remaining de-embedding dummies and said obtained values of Yee, Yii and Yei into YDj=−Yii−Yei(YMj−Yee)−1Yei (j=4,5, . . . , N) to obtain by calculation the de-embedded Y-parameter admittance matrices YDj(j=4,5, . . . , N) of said remaining de-embedding dummies;4.4: comparing the calculated YDj(j=4,5, . . . , N) with the calculated or simulated values YAj(p1, p2, . . . pM)(j=4,5, . . . , N) of the corresponding remaining de-embedding dummies already obtained in Step 4.1,determining final values for undetermined model parameters p1, p2, . . . pM which are necessary for the calculation or simulation of the de-embedding dummies, if a difference between the two meets the set error standard, andcorrecting the values of the model parameters p1, p2, . . . pM and reassigning them, and then turning back to Step 4.2, if the difference between the two does not meet the set error standard.

5. The special general four-port on-wafer high frequency de-embedding method according to claim 4, characterized in that solving the equation set YMj=Yee−Yei(YAj(p1, p2, . . . pM)+Yii)−1Yei (j=1,2,3) to further obtain by calculation the de-embedded Y-parameter admittance matrices YDj(j=4,5, . . . , N) of said remaining de-embedding dummies in Step 4.3 comprises the following steps: 5.1: obtaining by calculation matrices Z2A=(YA2−YA1)−1, Z2M=(YM2−YM1)−1, Z3A=(YA3−YA1)−1 and Z3M=(YM3−YM1)−1, where exponent −1 represents matrix inversion;5.2: calculating quantities

6. The special general four-port on-wafer high frequency de-embedding method according to claim 3, characterized in that calculating the intrinsic Y-parameter admittance matrix YA of said DUT by using the whole Y-parameter admittance matrix YM of said DUT obtained by measurement in Step 3.6 comprises the following steps: 6.1: obtaining by calculation a matrix ZMe=(Yee−YM)−1;6.2: calculating quantities