mipiweb.de Berlin, Germany
| homepage -------------- of -------------- Michael Dietmar Pierschel |
----------------------------------------------------------------
1. If you get any measurement values (measured by yourself or anyone else) check carefully
the used measurement circuit configuration
the measurement setup
the measurement device switch or internal settings (data range)
the measurement setup, probe card and device ground connection configuration
the data units you get together with the data
the DC signal polarities of all pins incl. bulk connection
do not forget to check the DC polarities and actual DC values of AC signal pins as well as their signal coupling status (direct/capacitive/inductive)
check for a DC path short or unintentional low ohmic connections of AC signal pins, the AC resistances should be commonly matched at interface pin connections
check for stacked DC sources and the polarities of these sources (are there ground potential free outputs available ?)
the absolute pin number and all signal directions at all interfaces (measurement setup/probe card/device under test)
the hidden data (killed/filled in/changed/exchanged/substituted values ect.)
the measurement standalone test procedure values (deembedding measurement data)
the completeness of the derived information [lot, probe or wafer number and history, measurement structure number and field, actual date, temperature, air pressure etc.] (this enables crosschecks later on and lead you to detect actual data inconsistencies)
do not forget to include the real used units to all these background values
try to compare the data to any other independent measurement method
try to compare the data to a circuit simulation which is as close to the whole measurement setup as possible
make sure, that the device is driven in exact the same circuit configuration as used for measurement
try to apply your common sense to the derived measured data, could it be or not ?
and (if available) to the estimated values in the test documentation
do not forget to consult this checklist point for point with a written sign to permit the proceed procedure to the next point
----------------------------------------------------------------
2. Commonly the measured data do not match the theory => DO NOT LOOSE YOUR TRUST IN PHYSICS.
Either the data is wrong or the applied model, but probably both.
Work on it, things will change.
A T T E N T I O N : never trust any output of any program, even if the plot looks pretty good.
Try to get a similar output with any other software or experimental measurements.
Remember that optical mirrors have two sides! Check careful if the right one is used in the beam path.
Check it for every single mirror separately.
----------------------------------------------------------------
3. Try to put any information you get or you derive to the computer.
Paperwork tends to sublime and should be for other people only.
These data may be searched by the machine which will save your time.
----------------------------------------------------------------
4. N E V E R try to check two numbers in a computer program to be identical, even if everybody believes to work with digital machines and no compiler prints out any warning for possible errors.
Do not trust any routines who should deliver a rounded value before the identical test.
If you print out or draw to screen two numbers and you check this output to be identical character for character, the numbers stored internal in the computer may be different in value.
----------------------------------------------------------------
5. Measurement data, regardless if self measured or measured by others, tend to mix-up.
- Wafers or parts of wafers may be easily mixed-up. Especially if there are only marginal differences in layouts.
- Normally the measured data, stored in a computer, get names before storing. This naming is often false and in most cases without the required parameter set (temperature, measurement conditions ect.). People tend to use any system(s) for the naming but the human is no machine and the system changes systematic. This means that after a certain time a unique mapping is impossible.
- Measurement data are commonly not only stored but formatted in a computer. Here a lot of intermediate files are required for changes of commata, dots sequences ect. Mix-ups are very easy introduced here. Most often some lonely additional information’s which have been stored with the data is killed here.
- Is the data ready to produce a graphic, most often title, axis label, units and measurement conditions as well as date of measurement an graphic creation will not be included.
- That rapid produced graphics are most often basis for long time decisions.
----------------------------------------------------------------
6. Do not believe anybody who assures you that nothing have been trashed. If all pieces of a wafer should be archived, exactly those who shall be measured will indeed be trashed since no visual structures can be found upon with pure eyes.
- Make sure to get the trash available for you.
----------------------------------------------------------------
7. An electrical measurement system is usable only if the devices will not be destroyed while contact them to the system.
It does not matter to the devices if the destroy is made electrically or mechanically.----------------------------------------------------------------
FazitEverything which is not machine proofed will contain errors. Everything which is machine proofed may also contain errors.
