Optimizing Automatic Parametric Test (APT) in Mixed Signal / Mems foundry

Transcription

1 Optimizing Automatic Parametric Test (APT) in Mixed Signal / Mems foundry Authors: Steffen Richter, Group Mgr PCM-Member Technical Staff, Xfab Silicon Foundries Alex Pronin, Ph.D, Lead Applications Engineer, Keithley Instruments

2 Motivation Introduction of a APT hardware supporting UHV PCM testing current equipment up to 100V, well suitable for 5V and below CMOS processes, but with drawbacks up to 200V. APT hardware to support default low voltage tests as well as UHV tests within one test run to increase throughput enable dedicated PCM tests using standard 200V SMU s and 1000V SMU together, at same time establish a future common APT standard: one tester and one software platform, but dedicated modular solutions possible to support various PCM test requirements with respect to our product portfolio (MEMs, UHV, Low voltage / low leakage, High power) replace out-of service equipment Support by tester vendor to accelerate production release of new APT hardware

3 About X-FAB The More than Moore Foundry. 20+ years of experience in pure-play foundry services for analog/mixed-signal semiconductor applications Best-in-class design and engineering support Technologies interfacing the real world Technologies for sensors, actuators and SoCs on modular CMOS, SOI and MEMS processes engineered in Germany Flexible combination and integration of power, high-voltage, analog, sensors and non-volatile memory features Dedicated MEMS foundry operations Manufacturing excellence 5 wafer fab facilities in Germany, Malaysia and US Capacity: 72,000 eight inch equiv. wafer starts per month All production sites are automotive qualified 2,400 employees worldwide

4 About Keithley A world leader in precision DC electrical test instruments and integrated systems Precision measurements for scientists and engineers Materials research Semiconductor device development and fabrication R&D and production test of electronic systems Measurements that determine properties, yields, and performance 60 years of measurement expertise and a rich understanding of measurement applications

5 Test platform change Test Platform/Environment has finite life time and has to change to: Accommodate new test requirements or replace an obsolete instrumentation Adapt a technology upgrade from existing hardware vendor Usually minor change in test environment Incremental upgrade (new functionality) Selection of different vendor Major change (has to be justified) in the environment Upgrade to novel functionality

6 Test program migration requirements Data Correlation Various criteria Simplest one: 95%/5% Test speed optimization Requires throughput analysis Handle some standard test scenarios Test flow integration Data format, Data base integration Operator procedure (!)

7 Data correlation, issues Data does not always represent physical properties of the DUT Data has contribution of the instrumentation, and this contribution needs to be understood Data cannot be always easily correlated: Leakage Capacitance, etc. Even simple cases have exceptions example: metal line resistance

8 Data correlation, Support Numerical and graphical overview of Parameters and their matching with 3%- and 5%- deviation limits Overview in Numbers suitable for status update and to estimate the remaining work. Easy to see which parameters need further optimisation

9 Data correlation, test cases Leakage SMU usually measures 4-5 digits of the current range Smaller the range (required current) longer the test Rarely leakage is actual leakage of the DUT, in most of the cases Leakage data represents interplay of Test setup (lowest allowed current range) with device capacitance, and used delay As a result leakage is dial-able measurement Does not make sense to wait very long be reasonable with the selection of the range

10 Typical Sensitive vs. Test Time Trade-off see also S530 Brochure (

11 Data correlation, test cases Capacitance It is an IMPEDANCE measurement Model dependent, Parallel or Serial Compensate for offset, chuck/up/down DUT is Distributive, not Lump Chuck/Substrate HAS to be AC driven don t forget frequencies dependence

12 Data correlation, test cases Resistance issues Contact resistance can be avoided by Kelvin/4-terminal measurements ECD measurements of metal lines have to include compensation for thermo electrics and SMU offsets Selection of the test current: high as possible but not too high (SOA) Contact resistance can NOT be avoided for high current MOSFETs Impacts: gm, Idon, Vth

13 Data correlation, test cases Breakdown test Capacitive Load Charge Consideration time > CV / Itrigger Voltage sweep SMU is in low impedance mode Force Current SMU is in high impedance mode Destructive effects Protection modules V sec.

14 Data throughput analysis Collect test time for each type of test on the wafer Create Pareto chart Prioritize an optimization effort with usage %

15 Throughput optimization, test cases Vth, threshold measurement from intercept of the tangent at maxgm Vth is calculated only from 2-3 point Key: limit lowest allowed current range!

16 Throughput optimization, test cases Leakage Leakage test often is pass/fail test Requires low current range, therefore time-consuming In many cases actually returns instrumentation settling Optimization approach: Control noise floor, lowest allowed current range Continuously measure current, return measurement if data settled

17 Throughput optimization Recommendations Run fixed Current range whenever possible Use Limited Auto ( Auto Range with specified lowest allowed current range) Do not add blanket delays Control size of the voltage sweeps Use force I for breakdown tests, instead of voltage sweeps

18 Protection Modules for S530 System Protection of system up to 1000V No compromise of low-level measurement capabilities Protection of probe needles and wafer Keithley IP (patent pending)

19 Handling of HV charge build up S530HV tests up to 1000V in production Unintended charges HV Matrix and System design (patent pending)

20 Key to success Team effort Cooperation between customer (XFAB) and vendor (Keithley) - exchange of measurement expertise with device knowledge was key to success Multidisciplinary approach XFAB PCM (Process Control Specialist) and XFAB device design experts Keithley system engineering and application team

21 Thank you for your attention!