End of Term Coop Presentation
Ben Magstadt
May 9th, 2011 – August 17th, 2011
What I Accomplished/Learned While at Texas Instruments
• FBB Analysis for OMAP4430 & OMAP4460
• Continuity Test Time Reduction for OMAP4430
• Smart Reflex & Vmin Analysis for OMAP4460
• TDL Debug for OMAP4460
FBB Analysis• Body Biasing Effect
– A voltage difference between the body and source in a transistor causes a change in the threshold voltage
– In other words, it is just a difference between the Voltage on the domain and the VBB level
– Forward Body Bias• Lowers the threshold voltage• Higher performance & higher power
– Reverse Body Bias• Increases the threshold voltage• Lowers performance & power
FBB Analysis for OMAP4• Problem
– Many patterns at speed corners need some extra help to pass at the spec Voltages
• Solution– Increase in Forward Body Bias gives that extra help
and reduces the Vmin level for the patterns– Implement FBB versions for the speed corner
patterns for them to pass at spec Voltages– For 4460, both IVA and MPU patterns that run at
speed corners need FBB– For 4430, only MPU domain that run at speed corners
need FBB
FBB Analysis for OMAP4460Vmin(Volts) vs. Temperature(C)
1.16
1.17
1.18
1.19
1.2
1.21
1.22
0 20 40 60 80 100 120
Temperature(C)
Vm
in(V
olt
s)
VMIN_VC_VSPEEDBIN_PMT_LED_N_ST_BST_FB
VMIN_VC_VSPEEDBIN_PMT_LED_N_ST_BST
SPEEDBIN SPEC
FBB Analysis for OMAP4• Yield problems with introducing FBB
– Can’t implement globally across all units– Variation among die (N2P)
• Hotter Material– Needs no help to pass Voltage Specs
– With FBB it exceeds Current Specs
– Vmin changes little with change to N2P
• Colder Material– Needs help to pass Voltage Specs
– With FBB it is passes Current Specs
– Vmin changes more drastically with change to N2P
FBB Analysis for OMAP4430• Problem
– Find the amount of FBB needed to have a good trade-off between Power/Vmin
– Need a dynamic implementation w.r.t the die parametrics (N2P)
• Project– Sweep the amount of FBB applied and collect
Vmin/IDDQ data– Analyze
• Change in Vmins• Change in Power
FBB Analysis for OMAP4430:MPU: 1.3V VBB: 1.05V FBB: 250mV
FBB Analysis for OMAP4430:MPU: 1.3V VBB: 1.35V RBB: 50mV
FBB Analysis for OMAP4
• Future– Large Data collection currently going on to
find these ‘sweet spots’ for OMAP4430– This process will be repeated for OMAP4460
once this is finished– Implement FBB on a chip to chip difference to
optimize performance and power
Continuity Test Time Reduction
• Problem– Time is money– Generic wait time used for all analog pins
during continuity tests
• Project– Used oscilloscope to find the specific rise time
for each analog pin in continuity test– Optimized the wait time based on the rise time
for each pin
Continuity Test Time Reduction
• End Result– Saved 1.8s on Single Site– Saved 1.9s on Octal Site
• Implemented a feedback procedure that waits more on certain devices that fail to rise in the specified time
Smart Reflex & Vmin Analysis
• Problem– OMAP4430 & OMAP4460 are creating larger
power needs– Need a way to run different units at different
voltages to save power
• Project– Collect & analyze data on the Vmins for
OMAP4460
Smart Reflex & Vmin Analysis
Smart Reflex & Vmin Analysis
• Project (continued)– Used the data collected to find the worst case stable
patterns for Vmins– Units then trimmed to these SR Fuse Values
• Future– Smart Reflex process being revamped so it searches
thru all patterns to find worst case Vmins– Needs more stability in patterns to be more effective
TDL Debug
• Problem– Failing and instable patterns that need to be tested for
customer coverage
• Ongoing Project– Take fail logs, shmoos, perform experiments, & talk
with designers to find solutions
• Future– More debug to be done to reach more stability in
patterns
THANK YOU!