Presentation is loading. Please wait.

Presentation is loading. Please wait.

12004 MAPLD/1002Katz Propagation Delay Stability in Logic Devices Richard B. Katz NASA Office of Logic Design 2004 MAPLD International Conference September.

Similar presentations


Presentation on theme: "12004 MAPLD/1002Katz Propagation Delay Stability in Logic Devices Richard B. Katz NASA Office of Logic Design 2004 MAPLD International Conference September."— Presentation transcript:

1 12004 MAPLD/1002Katz Propagation Delay Stability in Logic Devices Richard B. Katz NASA Office of Logic Design 2004 MAPLD International Conference September 8-10, 2004 Washington, D.C.

2 22004 MAPLD/1002Katz Abstract This paper will present data on propagation delays in logic devices, examining their distributions, the effects of life on propagation delay, and the characteristics of the changes in delays. Factory life test/qualification data will be presented along with recently taken in the evaluation of damage to programmed antifuses. Analysis of the data will be used to examine existing design rules, but for min/max analysis as well as those pertaining to the relative changes in delays.

3 32004 MAPLD/1002Katz CD4000B Propagation Delay Data t PD (ns)

4 42004 MAPLD/1002Katz RH1280 Life Test Delta Data

5 52004 MAPLD/1002Katz RH1280 Life Test Delta Data

6 62004 MAPLD/1002Katz RH1280 Life Test Delta Data Tested at V DD = 4.5V; T = 125 ºC Binning Delay Delta Data Summary (ns) T0 168 hrs T1 T1-T0 500 hrs T2 T2-T0 1,000 hrs T3 T3-T0 Average 106.5105.5106.1-0.5104.0-2.7 Maximum 120.1113.74.2113.65.3119.77.3 Minimum 93.496.3-10.796.1-9.093.4-9.9 StdDev 4.93.72.23.92.44.62.9

7 72004 MAPLD/1002Katz RT54SX72S: Propagation Delay vs. Life RTSX32S t PD sensitive to total ionizing dose in a Co-60 irradiation chamber. Design changes in the RTSX72S reduced the sensitivity to acceptable levels. These design changes were incorporated into the next revision of the RT54SX32S.

8 82004 MAPLD/1002Katz RT54SX72S: Propagation Delay vs. Life The following chart shows aggregate data from multiple lots of RT54SX72S FPGAs, totaling 1,040 individual devices. The speed data, with a mean delay of 73.8 ns, is measurements of the binning circuit, representative of logic paths, and is used to determine the device speed grade. Note that the delays over life do not necessarily "track," as there are both differences in the changes of delay as well as some differences in the sign of a delay change. Construction of the RTSX-S Binning Circuits 5 segmented vertical routing tracks are allocated to the binning column Fixed length horizontal tracks 3 Fuses per connection Output Cross connect Input Number of rows per device RT54SX32S has 30 rows RT54SX72S has 48 rows Number of fuses = ((N-1) x 3) + 2 RT54SX32S => ((30-1) x 3) + 2 = 89 RT54SX72S => ((48-1) x 3) + 2 = 143

9 92004 MAPLD/1002Katz RT54SX72S: Propagation Delay vs. Life

10 102004 MAPLD/1002Katz “fdelay” Measurements 1144 logic modules configured in a ring oscillator configuration Test Conditions: –Temp= +25 °C –V CCA = 2.5 VDC –V CCI = 3.3 VDC –Test= 7 (I/Os static) This is not a typical accelerated test; it is “de- accelerated” RTSX32S and RTSX32SU Devices Ring oscillator and nominal conditions - YUCHH!!

11 112004 MAPLD/1002Katz “fdelay” Circuit I hate ring oscillators; use a clock..

12 122004 MAPLD/1002Katz RTSX32S Distribution @ 24 Hours Intrinsic distribution is tightly grouped. Outliers are devices with damaged programmed antifuses.

13 132004 MAPLD/1002Katz RTSX32S Distribution Statistics Performance Over 600 Hour Test, Nominal Conditions All delays are in ns. StatisticMeanStd DevMinMaxNum Samples 0 Hours 929.034.1867.81269.6499 24 Hours 929.552.6870.81698.7489 168 Hours 928.229.0868.81285.4489 336 Hours 928.126.2868.41289.8489 600 Hours 929.727.6870.21315.0489

14 142004 MAPLD/1002Katz RTSX32S 24 Hour Outlier Performance Damaged programmed antifuses show signs of instability over the course of the test. Note S/N 34319 first decreases in speed and then returns close to its initial value.

15 152004 MAPLD/1002Katz RTSX32S t PD Delta Analysis 24 to 168 Hours Intrinsic population is stable over the course of the test. Outliers are devices with damaged programmed antifuses with delta delays ranging from 10’s of ns to hundreds of ns. Other data shows delays from damaged devices may exceed 1 µs. Mean delay is ~ 930 µs.

16 162004 MAPLD/1002Katz RTSX32S t PD Analysis Startup Transient Frequency measurement at startup. Ring oscillator period increases for a considerable period of time before stabilizing, likely a result of self-heating, as digital CMOS slows with increased temperature. One of the two “practice parts” shows a discontinuity.

17 172004 MAPLD/1002Katz RTSX32S t PD Analysis Startup Transient Frequency measurement of three devices with damaged programmed antifuses. Note the discontinuity, showing instability of the damaged element. Additional test data confirmed the instability of damaged programmed antifuses in a different S/N device.

18 182004 MAPLD/1002Katz RTSX32SU Distribution @ 168 Hours Intrinsic distribution is tightly grouped. 100 devices, ~20 the number used in the SX32S test.

19 192004 MAPLD/1002Katz RTSX32SU Distribution Statistics Performance Over 504 Hour Test All delays are in ns. StatisticMeanStd DevMinMaxNum Samples 0 Hours 1123.616.41070.31154.4100 168 Hours 1126.516.41072.71156.7100 336 Hours 1126.416.41073.31156.1100 504 Hours 1127.816.51074.41157.2100

20 202004 MAPLD/1002Katz RTSX32SU t PD Delta Analysis 0 to 504 Hours Intrinsic population is stable over the course of the test. Mean delay is ~ 1,125 ns; deltas, even with measurement errors, are less then 1%.

21 212004 MAPLD/1002Katz Conclusions Distributions of t PD is fairly tight for modern devices. Changes of t PD over life tests is fairly tight for modern devices. Using ± 10% for delta t PD is appropriate. Changes of t PD over life tests is not a constant and may be bipolar. Qualifying asynchronous races by test can not be supported. Analysis of changes of t PD over time provides a strong indicator of element failure. Absolute values of delay is insufficient as a screen or test. Data is now available to show that damaged programmed antifuses are not stable over time. Thus, functional testing is insufficient as a screen for damaged programmed antifuses.

22 222004 MAPLD/1002Katz Appendix: Test Vehicle and Protocol for NASA Independent Testing of SX-A, SX-S, and SX-SU Programmed Antifuses

23 232004 MAPLD/1002Katz NASA Device Testing Strategy Temperature (-55 °C and +125 °C) Increasing V CCA core voltages, starting at V CCA = 2.75V. Increasing the number of simultaneous switching outputs (SSOs) Increasing the amount of simultaneous switching undershoot (SSU) Loading of internal nets 20% greater than the maximum that the design rule check (DRC) will permit for user designs. The use of all architectural features in the device The testing will consist of exposing devices to a dynamic operating environment with multiple stressors: Testing will proceed in a series of stress steps, the duration of each step will be 480 hours, with 240 hours at +125 °C and 240 hours at -55 °C.

24 242004 MAPLD/1002Katz Differences from “Tiger Team” Design Temperature: For NASA flight projects, testing should envelope operating conditions with considerable margin. -55 °C to +125 °C achieves that goal while being a practical test temperatures; Tiger team testing is performed at "room temperature.“ V CCA core voltages will also envelope the flight operating regime. The initial Tiger Team testing was performed at V CCA =2.5V, in the middle of the operating range, with later testing to be done at V CCA =2.75, the maximum of the operating range. The NASA testing will start at V CCA =2.75 and then increase the voltage with each step, for margin testing and acceleration. Internal Net Loading: The maximum number of logical loads for operating systems is 24. The NASA design will have a number of critical nets running with a load of 29, which is slightly greater than a 20% margin. The Tiger Team design has a maximum load of 16, below the limit of both the DRC and many flight designs. Clocking: The device under test (DUT) to be stimulated by a crystal clock oscillator through both the HCLK and one of the two routed array clocks, with functional circuits on each clock network. This is a superset of the Tiger Team design, where neither of the clock inputs are utilized (internal ring oscillators are used to manufacture clock signals) and the HCLK network is static. Reviews of many NASA designs shows that most designs utilize the HCLK and thus must be tested. Propagation delay is measures through a linear arrangement of gates, not with a ring oscillator. Hand placement and analysis was performed to ensure usage of elements such as long horizontal tracks (LHT) and long vertical tracks (LVT) for loading.

25 252004 MAPLD/1002Katz Key Characteristics of NASA DUT Design HCLK is used for the I/O shift register and CLKA is used for the array shift register (no ring oscillators are present in this pattern). The I/O shift register has logic modules between sequentially adjacent flip-flops; the array register does not, thus maintaining the design’s sensitivity to small changes in clock skew. Array shift register chains are segmented in such a manner to enable precise fanout control. There are 23 signals with a fanout of 29. The limit for user hardware is is 24. An array shift reg segment is manually placed to ensure utilization of the long horizontal and vertical tracks (LHTs and LVTs). All I/O registers are manually placed at the sequential tile nearest to the I/O buffer that it is driving. The implications of this, combined with the use of the HCLK, are both more realistic testing since HCLK usage in flight designs is high and making the SSO "more simultaneous" since the HCLK inherently has lower skew than the routed array clock, as has been used in the Tiger Team pattern for similar purposes.

26 262004 MAPLD/1002Katz Resource Utilization of NASA DUT Design SEQUENTIAL Used: 1080 Total: 1080 100% COMB Used: 1800 Total: 1800 100% LOGIC Used: 2880 Total: 2880 100% (seq+comb) IO w/ Clocks Used: 168 Total: 170 99% CLOCK Used: 2 Total: 2 100% HCLOCK Used: 1 Total: 1 100%

27 272004 MAPLD/1002Katz Antifuse Distribution of NASA DUT Design Preliminary - Not Finalized Fill in final values


Download ppt "12004 MAPLD/1002Katz Propagation Delay Stability in Logic Devices Richard B. Katz NASA Office of Logic Design 2004 MAPLD International Conference September."

Similar presentations


Ads by Google