next up previous contents
Next: 5.4 SPICE Modeling Interface Up: 5. Implementation Previous: 5.2 Process Flow Interface

Subsections


5.3 Electrical Test Interface

5.3.1 Introduction

A typical test-chip architecture demands numerous repetitions of the same test device with only slightly varied parameters. Traditionally the development engineer specifies the devices needed and documents their parameters. Based on this documentation the layouter draws the test structures and a test engineer creates test programs for an automated parameter tester. The large amount of test structures as, e.g., seen in a typical CMOS based high voltage process demands the need for an automated one-source template based system to create layout, documentation, and test programs. This system itself has to be flexible enough to allow easy incorporation of new test structures and measurement methods.
The development engineer creates one master file defining a set of test structures and parameter variations (such as e.g. well distance variations). In this file one can also define the measurements to be performed, using a simple meta-syntax. A script that uses a predefined cell library to create the layout of test lines with a standardized pad layout reads this file. The test lines are arranged and placed in the test chip automatically, and the exact position of these lines is provided to the automatic test system. A second script uses measurement templates to convert the master file in test programs suitable for the specific test system. The resulting layout files and test programs are hyper-linked into the documentation that can be used as a reference by the development engineer fur further analysis. An overview about this scheme is shown in Figure 5.3.
Figure 5.3: Electrical test program conversion work-flow
\includegraphics[angle=0,origin=c,width=0.5\textwidth,clip=true]{figures/automated_test_creation.ps}

5.3.2 Electrical Parameter Test

The modules involved in creating the test programs for the electrical parameter extraction use several layers of abstraction.

5.3.2.1 Measurement Routines

On the lowest level there are the test routines for parameter extraction itself, e.g., a threshold voltage extraction or a breakdown voltage extraction routine. These routines contain all the instrument specific commands and calculations that are necessary to obtain device parameters. The routines can make use of a fairly large number of parameters specifying everything up to delay times, integration times, and compliance values (see Table 5.1).


Table 5.1: Parameters for threshold measurement routine
Parameter Comment Parameter Comment
Gate Gate pins Idcpl Compliance for drain current
Drain Drain pins Vsubsforce Optional substrate bias
Source Source pins Isubcpl Substrate current compliance
Substrate Substrate pins Dtime Settling time before
  measurement is taken
Type Specifies NMOS or PMOS Strategy Measurement strategy.
  Supported strategies are
  measure, integrate and average
Idstart Minimum drain current to Stepno For strategy average:
start drain voltage sweep number of points to average
Vdsstep Step width of Steptime For strategy average:
drain voltage sweep time interval between
  measurements
Vdspoints No. of points for Vtsmeas Return value:
drain voltage sweep measured threshold voltage
Vdscpl Compliance for drain voltage  


However, in most cases the test program designer does not want to care about all these details, but wants to use a simple routine where one specifies the pads to use and some basic parameters specific to the DUT. Therefore a set of templates was introduced to provide a translation from an easy-to-use syntax to the more powerful but also more complicated measurement routine syntax (see Listing 5.4).
\begin{lstlisting}[float=tbp, frame=shadowbox,caption={Template definition},
la...
...
</Design_Rule>
<Result>
ILEAK[pA]
BV[V]
S[um]
DR[um]
</Result>
\end{lstlisting}
These templates also make it possible to have several measurement commands call the same routine with different parameters. Thereby one may specify measurements with different current compliance values or settling times for different types of transistors.
The development engineer does not need to deal with the specific parameter set for every basic routine. If the actual templates are not sufficient for the task needed, the specialist responsible for electrical test can provide a new template.

5.3.2.2 Test Program Definition

A test program in the development engineers' view consists of the following parts (see Listing 5.5).
\begin{lstlisting}[float=tbp,frame=shadowbox,caption={Electrical test-program ex...
..., 3, 0.01, 3, RESULT )
</VTSLIN>
<VTSLIN_Var>
</VTSLIN_Var>
...
\end{lstlisting}
This configuration offers the flexibility to define programs for design rule test-chips and similar test-chips consisting of test-lines with only similar devices. However, the test program has to be defined manually for each test line.

5.3.2.3 Automated Creation of Test Program Definitions

For a recent high voltage process development project the evaluation of geometric variations of high voltage transistors was needed. Because of the fact that these transistors have a rather complex layout and therefore a lot of layout parameters that modify the transistor behavior, the testchip turned out to include up to 1536 similar devices to be layouted in up to 129 test lines. In total there have been 466 test lines with only 7 different types of transistors. In order to rule out human errors and make life easier for designers and test program developers, it was crucial to automate as much work as possible.

As input only some basic device layout and tables were needed, to specify which parameters to vary in what sequence. From these input files the layout and arrangement of the devices on the test lines was performed automatically. The resulting arrangement list was used as an input for a script that used a general template measurement program similar to Listing 5.4 and filled up the <NAME> and <PIN_TABLE> section with the correct information. This configuration only requires the seven template files to be created and guarantees that all test lines are measured in the same way.

The resulting program definition files are converted into test programs for the test system. In addition the layout scripts create a coordinate list for the test lines, so that these coordinates can be appended to the wafer definition required by the test system.

5.3.3 Documentation

The whole test-chip is documented in HTML format by a table of test lines, each of which has a link to the layout and to the test program definition file. This documentation is also created from the arrangement list resulting from the transistor variation, and from a manually created arrangement list with those design rule structures that are only available on single test lines. However, the conversion to HTML is performed with a script and therefore is not prone to human errors.

5.3.4 Conclusions

The creation and the maintenance of test programs for design rule and device evaluation test-chips has been drastically facilitated by the introduction of an automated test program creation suite. The time to create a test program for device screening could be reduced from several weeks when the programs would have been created manually to a few days for the initial working templates. Any changes in the measurement sequence could be done within hours instead of weeks. It was guaranteed also, that all devices were measured in the same way without the risk of typing errors in the measurement programs.


next up previous contents
Next: 5.4 SPICE Modeling Interface Up: 5. Implementation Previous: 5.2 Process Flow Interface

R. Minixhofer: Integrating Technology Simulation into the Semiconductor Manufacturing Environment