Summary of the "Yield after Packaging" with QIE7b Two QIE7b wafers from lot T304, were wafer tested with bad die being inked. The wafers were sent for packaging by ASAT. Only good die were to be packaged. We find 99.4% good die from the after packaging testing after screening by wafer testing prior to packaging. We find an additional 0.5% of the devices being "Flakey" which means they test both good or bad in the after packaging testing. From this experience, we expect 98.9% of the devices that pass as good from wafer probing to be truly good devices. Finally, we find a class of suspect devices from wafer probing information that were NOT found in after packaging testing. These devices lie on position 2 in a twelve position reticle and have pedestal and analog outputs shifted substantially lower than other devices. The after packaging testing can be modified to find and screen these devices. Wafer testing results: Lot T304, Wafer 05: 635 / 699 good die Lot T304, Wafer 08: 628 / 699 good die Total good die: 1263 Total inked die: 135 Packaging Returns: Packaged parts: 1248 Good die left on skeletons: 2 (one did have a tiny ink dot) Missing good die that weren't packaged: 1263-1248 = 15 Missing bad die not on skeletons: 37 After packaging testing results: Wafer testing good die: 1248 After packaging testing good die: 1200 48 devices that test bad after packaging when expected to be good: After packaging failures understood (different limits): 35 After packaging failures not found due to higher wafer probing noise: 3 After packaging failures that appear to have arisen after packaging: 4 After packaging failures that are "Flakey": 6 Some additional notes: It is thought that missing parts were used for machine set-up or other activities but this is to be confirmed. ASAT provided sheets showing "Parts In" and "Parts Out" of all the processing steps involved in packaging. The initial "Parts In" shows 1248 devices. The 1248 packaged devices were re-tested after packaging using the Baumbaugh/Knickerbocker test set-up. Requirements for good/bad die were not identical between the wafer probing and after packaging testing with the after-packaging testing being more restrictive. The after-packaging testing found 48 devices that had measured values outside limits determined from an initial small sample (116 parts). These 48 parts were examined in detail (six of these were re-tested good with the after-packaging set-up and are called "Flakey"): A total of seven devices were found that in principle should have been screened by the wafer probing. Of these, three are understood because the failures showed up in exponent 0 or 1 readings that were masked by slightly higher noise levels in wafer probing compared with packaged part testing. Three additional devices also had bad exponent readings after packaging that presumably tested good on the wafer probing. The final device failed the reference input test and presumably tested good on the wafer testing. In short, 99.4% of the die flagged as good by the wafer probing are also found good by the packaged parts testing. If we include "Flakey" parts which sometimes look bad and sometimes look good, 98.9% of the die flagged as good by the wafer probing have every indication of being truely good devices. A detailed breakdown of the 48 problem devices is shown below: Num Problem Comments 4 Cap-ID's not rotating All retested as good parts. Flakey parts. 13 Pedestal out of range All within probe station limits. *** 2 masked by noise, 1 now with bad exponents *** 6 AC/DC impedance range All within probe station limits. 1 Reference input test *** should have been flagged as bad *** 5 Aout out of range All within probe station limits 5 Exponent errors 2 Flakey, *** 1 masked by noise, 2 now with bad exponents *** 4 Digital LVDS voltage 3/4 passed on retest. Others in probe station limits 10 Current out of range All are within probe station limits The Pedestal out of range devices were studied in further detail. It was found in the wafer scale testing that position 2 in a 12 chip reticule gives unusually low pedestal values. This is evident only when 10's of wafer results are compared -- it was not evident with only these first two wafers.These devices will be rejected for all lot T304 wafers. There are approximately 100 devices expected to have this unusually low pedestal on these two wafers already packaged. New settings have been incorporated into the packaged tester box to screen out these 100 or so devices.