The present invention relates to a measurement arrangement for determining the characteristic line parameters by measuring scattering parameters (S-parameters) as a function of the frequency of an electrical signal line according to the features of claim 1.
Model to hardware correlation measurements on all packaging levels are essential in today's development process of high performance computers. Different measurement techniques in time and frequency domain require different measurement set-ups and test site designs. One demand for the test site is to be equivalent to the product. Therefore, transmission lines on a chip need to be measured in the product line power and ground wiring distributed in all metal layers on chip. In addition it is not only of interest to measure a single transmission line but also with a product like wiring channel utilization. This is essential to image the real signal coupling behavior on the chip and the shielding effect of metal layers between top metal layers and the semi conducting substrate.
A known measurement technique is the so-called S-parameter measurements, see Zinke/Brunswig, “Lehrbuch der Hockfrequenztechnik”, Springer-Verlag, 1989. S-parameters are reflection and transmission coefficients of an n-port network. The equivalent for a single transmission line e.g. is a two port network characterized by a 2×2 S-parameter matrix.
A two-port network is described by the relationship
wherein S11, S22, S12 and S21 are the S-parameters, namely
S11=Input reflection coefficient with the output port terminated by a matched load,
S12=Output reflection coefficient with the input terminated by a matched load,
S12=Reverse transmission (insertion) gain with the input port terminated in a matched load,
S21=Forward transmission (insertion) gain with the output port terminated in a matched load, and
the variables a1, a2 and B1, b2 are complex voltage waves incident on and reflected from the first and second port of the two-port network.
In the present case the S-parameter measurements are an advantageous measurement technique because the S-parameter are easier to measure and work with at high frequencies than other kinds of parameters.
Furthermore different methods are well-known in the state of the art to extract other characteristic frequency dependent line parameters, such as characteristic impedance z(f) or propagation constant ν(f) etc., from the S-parameter measurements, so that these parameters can easily be obtained from the S-parameter measurements. Thomas-Michael Winkel, Lohit Sagar Dutts, Hartmut Grabinski, “An Accurate Determination of the Characteristic Impedance of Lossy Lines on Chips Based on High Frequency S-Parameter Measurements”, IEEE Multi-Chip Module Conference MCWC'96, pp. 190-195, February 1996, Thomas-Michael Winkel, “Untersuchung der Kopplung zwischen Leitungen auf Silizium-Substraten unterschiedlicher Leitfähigkeit unter Verwendung breibandiger Messungen”, Ph D. Thesis, University of Hannover, November 1997.
A special requirement for the high frequency S-parameter measurements in this case is that the transmission lines are not connected to any active device on chip. Due to this, parallel signal lines would be floating if not connected to any driver and receiver. A problem occurs when parallel lines on the test site have to be connected to some point in absence of drivers, receivers and transistors.
One option is to leave both sides of the signal lines open, but floating lines do not correspond to the product and will therefore alter the measurement results.
A second option is to connect both ends of the parallel signal lines to ground. In this case all signal lines act as ground lines which is also not corresponding to the product.
In principle a driver has a low impedance while a receiver has a high impedance. Therefore a third option is to connect one side of the parallel signal line and leave the opposite side open. This option imitates the product but the problem that occurs here is that high frequency measurements are band limited to less than 4 GHz because in the higher frequency range both measurements ports present a different electrical behavior on both ports. While one port just sees open parallel lines the opposite port just sees grounded parallel lines. As a result, for frequencies >4 GHz not only one signal line mode will be excited in the test structure.
As evidenced from the forgoing discussion, it is desirable to provide a measurement system for determining the S-parameters as a function of the frequency of an electrical signal line which does not suffer from the above-note drawbacks and leads to a significant gain of the measurement bandwidth.
The present invention relates to a measurement arrangement for determining the characteristic transmission line parameters by measuring the S-parameters as a function of the frequency of an electrical signal line that achieves an increased measurement bandwidth, namely a measurement bandwidth >4 GHz.
The measurement arrangement according to the invention is characterized by what is specified in the independent claim 1.
Advantageous embodiments of the invention are specified in the dependent claims.
the inventive measurement arrangement comprises a signal line under test—measuring line—and several neighboring signal lines, wherein the measuring line as well as the neighboring signal lines having a first and a second end, representing port 1 (S11) and port 2 (S22) of a two-port network. According to the invention one end of each neighboring signal line is terminated by a low impedance and the other end of each neighboring signal line is terminated by a high impedance, so that the first and second ends of all neighboring signal lines are terminated by a low impedance and a high impedance, respectively, and the number of neighboring signal lines having a low impedance on their first ends or their second ends is equal or nearly equal to the number of neighboring signal lines having a high impedance on their first or second ends.
As a result of the special connection pattern both ports look at least nearly identical. Therefore, only one signal mode is excited and the frequency bandwidth is increased significantly.
In accordance with a feature of the invention, the low impedance is formed by a closed-ended line (connection to ground) and the high impedance is formed by an open-ended line.
In accordance with still another feature of the invention, the measuring line is in a plane arrangement and the neighboring signal lines are arranged in-plane to the measuring line in a line pattern matter or in a parallel arrangement.
Preferably, neighboring signal lines arranged directly adjacent to each other have a different impedance on their first ends and their second ends, so that the first ends and second ends of all neighboring signal lines are alternatingly terminated by a low impedance and a high impedance, respectively. This leads to an alternating arrangement on port 1 and port 2, respectively. This means both ports have an identical appearance and as a result the frequency bandwidth is increased to more than 20 GHz.
According to another feature of the invention, the number of neighboring signal lines on both sides of the measuring line is equal.
Further, the neighboring signal lines arranged directly adjacent to the measuring line may have a different or identical impedance on their first ends and their second ends, respectively.
In accordance with still another feature of the invention, the measuring line and the neighboring signal lines are signal lines in a multi-layer chip, wherein the direction of the signal lines between two adjacent layers is rotated by 90°, and the measuring line and its neighboring signal lines are arranged in the same layer—measuring layer—in a parallel arrangement, and the signal lines in the layers adjacent to the measuring layer—neighboring layer lines—are also arranged in a parallel arrangement and having a different impedance on their first ends and their second ends, respectively, so that the first ends and second ends of all neighboring layer lines are terminated by a low impedance and a high impedance, respectively, and the number of neighboring layer lines having a low impedance on their first ends or their second ends is equal or nearly equal to the number of neighboring layer lines having a high impedance on their first ends or their second ends.
Preferably, neighboring layer lines arranged directly adjacent to each other have a different impedance on their first ends and their second ends, so that the first ends and second ends of all neighboring layer lines are alternatingly terminated by a low impedance and a high impedance, respectively. This alternating arrangement of the neighboring layer lines in connection with the alternating arrangement of neighboring signal lines lead to a significant gain of measurement bandwidth. Experiments have shown that due to the inventive arrangement of on chip wiring the bandwidth is increased up to 20 GHz.
According to another feature of the invention the measuring line and the neighboring lines are arranged as a bunch.
In order to achieve a nearly identical appearance of both ports of the bunch the ends of the neighboring signal lines with a low impedance and a high impedance, respectively, are arranged in an equal or nearly equal manner regarding an imaginary cross-sectional area of the bunch.
Additional objects, advantages, and features of the present invention will become apparent from the following description taken in conjunction with the accompanying drawings.
In order to image the real signal coupling behavior on the chip 10 additional signal lines 12, the so-called neighboring signal lines, were added adjacent to a signal line under test 14, the so-called measuring line, in the same layer.
The neighboring signal lines 12 were connected by vias to ground on one side 16, here port 1, to imitate a driver and left open on the opposite side 18, here port 2, to imitate a receiver.
In order to image the shielding effect of metal layers between top metal layers and the semi conducting substrate additional signal lines 20, the so-called neighboring layer lines, were added in the bottom metal layers. All neighboring signal lines 20 were also connected to ground on one side and left open on the opposite side.
As a result the measured reflection parameters S11 and S22 as depicted in
Due to random and systematically measurement errors the measurement uncertainty for the magnitude is usually ˜3%. For frequencies >4 GHz the difference between both reflection parameters exceeds this value for the magnitude as well as for the phase. This means more than just one signal line mode is excited in this test structures. Therefore, all extracted transmission line parameters are just valid up to 4 GHz.
In order to increase the frequency bandwidth of the extracted data, the test site design needs to be modified according to the invention. To ensure that the measured reflection parameters for port 1 (S11) 16 and port 2 (S22) 18 are nearly equal. This goal can be achieved by making both ports 16, 18 symmetrical from an electrical point of view.
As shown in
As a result of this change, the measured reflection parameters for port 1 (S11) 16 and port 2 (S22) 18 are nearly identical as depicted in
10 multi-layer chip
12 neighboring signal lines
14 measuring line
16 port 1
18 port 2
20 neighboring layer lines