Presentation is loading. Please wait.

Presentation is loading. Please wait.

Particles & Fields Data Processing Unit (PFDPU)

Similar presentations


Presentation on theme: "Particles & Fields Data Processing Unit (PFDPU)"— Presentation transcript:

1 Particles & Fields Data Processing Unit (PFDPU)
Electronics Box Mechanical Evaluation Bill Donakowski, SSL David Pankow, SSL Paul Turin, SSL Scott Tucker, et.al., LASP NOTE: Maven has requested in depth review support from GSFC on this topic. Our point of contact is Seke Gordo (code 542). This review is ongoing and still has some open items.

2 PFDPU Mechanical Evaluation
Evaluation Approach Experience has shown that the largest components offer the greatest risk Ceramic substrates are quite stiff as compared to flexible circuit boards Most likely failure mode is (KOVAR) fatigue of corner leads on large components Fatigue strains are dictated by the accumulated flexures of the PCB. Steinberg’s methods predict max. board deflection for infinite fatigue life He’s fitted his accumulated PCB test data to provide useable empirical relations These results may also be applied to yield & finite fatigue life predictions Infinite life is appropriate for many terrestrial applications, while the typical NASA launch vibration (& testing) exposure is very brief. PCB (four sided) edge support naming conventions Pinned – edges constrained in translation, but not in rotation Fixed – edges constrained in both translation and rotation Working Reference Steinberg “Vibration Analysis for Electronic Equipment” 3rd Ed. J.Wiley, 2000

3 Summary of Evaluation Relations (after Steinberg)
PFDPU Mechanical Evaluation Summary of Evaluation Relations (after Steinberg) First Mode of a given PCB design (fixed edges): Fn = (2p/3) (D/r)½ [3/A4 + 2/A2B2 + 3/B4]½ • D = Eh3 / [12 (1-m2)] • r = W/g / [AB] A & B – PCB dimensions • h – PCB thickness • W - weight Max PCB deflection for Infinite Fatigue Life: Z∞ = B / [C h r √L] • C – component factor (table) • r - position factor (table) L – component length Maximum PCB test deflection of a given PCB design: Z3s = 36.8 √P / Fn5/4 • P - ASD test level (G2/Hz) near Fn Accumulated fatigue damage using “Miner's Rule”: Rn = (Fn ST) [0.683/N1s /N2s /N3s ] • Nns = [Z∞/Zns]b x 1E7 cycles ST – accumulated exposure time Yield Strength Fraction: Fy = (Se / Sy)(Z3s / Z∞) • Se - endurance strength • Sy – yield strength

4 PFDPU Mechanical Evaluation
OBSERVATIONS GSFC-STD-7000 (GEVS) and NASA-STD-5001 pose two design constraints 1.6 min. safety factor on metallic yield in vibration from GEVS table 2.2-3 4X exposure time calculations to account for scatter in typical life data Designing for “Infinite Life” is an admirable goal, but overly restrictive Accepting modest fatigue damage (perhaps < 25%) poses minor risks Plot below illustrates design reserves for retesting & programmatic unknowns 1.6 min. GEVS yield safety factor → 62% of yield (red line) Finite life only in the box level PFM test Subsequent tests are at lower levels FM Level Req’d PFM Level

5 PFDPU Mechanical Evaluation
PFDPU Box Construction Each PCB is bonded into a ledged aluminum frame Eleven slices stack together & bolt to base plate Internal stacking connectors provide “backplane” Two threaded “skewers” hold top corners together

6 PFDPU Mechanical Evaluation
PFDPU PCB Artwork MAG PCB DFB PCB REG PCB – no concerns DAP PCB NOTE: No large components on BEB board

7 PFDPU Mechanical Evaluation
IIB Board: includes two perpendicular daughter boards w/ frame (and no large components) DCB Board: includes a partial daughter board on standoffs

8 Original PCB Evaluation Results
PFDPU Mechanical Evaluation Original PCB Evaluation Results COMMENTS * The DCB board has a daughter card and many other features that make a proper (FEM) evaluation questionable. Having the EM board previously shown, the plan of action was to measure this board frequency, and make adjustments to obtain a first mode of > 200 Hz • These evaluations assume good workmanship practices in building these PCBs.

9 PFDPU Mechanical Evaluation
PFDPU Box Evaluation Steinberg calls for an octave (2X) separation between boards and box Charts below show potential board – box resonant interactions (using Miles Eq’n) Lead fatigue is governed by board displacements shown in the right chart. The 2X criterion reduces box-board mode interactions to quite modest levels Acceleration Displacement

10 PFDPU Mechanical Evaluation
MAVEN PFDPU Design Details 2X Skewer Rods 11 X Individual Frames (Al 6061 T6, .100” thick) SolidWorks model of PFDPU Attach Frame Bolted to S/C Each frame interlocks with adjacent

11 PFDPU Mechanical Evaluation
MAVEN PFDPU FEM Analysis Details Fixed Constraint at Box Base (no other constraints in model) 11 x Individual PCBs/11 Individual Frames Exploded View

12 PFDPU Mechanical Evaluation
Top of Box Frames Bonded (Skewer Rods hold frames together) FEM software SolidWorks ‘Simulation’ 2011 SolidWorks Premium 2011 x64, SP 1.0 Includes FEM Simulation 2011 x64 SP 1.0) Analysis Details 2-D Shells and Mesh All 11 Frames and PCBs modeled Frames fixed at bottom (mounting to S/C Deck) Frames bonded to each other at Top (skewer rods) Bottom of Frames Fixed (PFDPU Base Plate not included - mounts to spacecraft) Material Properties Component Material Density (#/in3) Modulus (psi) Poison’s Ratio Frames Al 6061 T6 .100 10,000,000 .33 PCBs Modified FRF .225 70,000,000 .14

13 PFDPU Mechanical Evaluation
First Mode (side view of PFDPU) Second Mode (top view of PFDPU) Numerical Results Mode Frequency (Hz) Description 1 581 Out-of-plane to PCBs, shear in relation to fixed bottom 2 684 Flexing of Side Walls 3 687 4 695 (modal mass was ~68% of total mass) (many similar modes)

14 PFDPU FEM Modeling Discussion
PFDPU Mechanical Evaluation PFDPU FEM Modeling Discussion Box Modes (only) are the desired results in this effort Software can provide up to 25 modes starting from the lowest value Using actual PCB physical properties will only yield a plethora of board modes PCB modulus (E) was increased 20X to avoid the box modes (i.e. very stiff boards) Higher order box modes were frame flexing or (Al) end panel diaphragm flexing A subsequent test run with the actual E(pcb) did not provide any box modes This did provide a useful result where PCBs clustered around 176 Hz vs. 185 Hz The 5% decrease in Fn is from the “less than rigid” nature of the stacked frames This is a conservative result because stiff boards contribute to the modal mass Actual ‘floppy’ boards will contribute less to the modal mass at these frequencies PCB modes can be increased to ≈300 Hz before interactions are a concern

15 Review of, and Modifications to
PFDPU Mechanical Evaluation Part 2: Review of, and Modifications to STEINBERG’S APPROACH Issues to be considered: EM – DCB board test results Validation of PCB frequency predicts PCB mode shape vs. fatigue amplitude discussion PCB material damping

16 EM-DCB BOARD TEST (¼ G sine sweep)
PFDPU Mechanical Evaluation EM-DCB BOARD TEST (¼ G sine sweep) The partial daughterboard & many connectors defy proper analysis TEST RESULTS Fn = 211 Hz (185 Hz Steinberg predict for a nominal – single board) Qmax = 46 (measured near the board center) • Qcg = 4/p 2 Qmax A measured Q at some arbitrary point is not meaningful, Qcg is needed Qcg = 18.6 (14.5 predict using Steinberg’s QSt = √Fn)

17 VIABILITY OF PCB FREQUENCY PREDICTS
PFDPU Mechanical Evaluation VIABILITY OF PCB FREQUENCY PREDICTS NASA-SP-160 (1969) is one of the most complete analytic works available <available on line by searching ntrs.nasa.gov> A 5% frequency reduction is needed to reflect the less than rigid stacked frame FEM. Minor corrections are needed for the EM – DCB test results The test board had eight screws, but was not glued into it’s board frame DCB test board was FR-4 while FMs will use higher modulus polyimide PCBs Isola P95 is one example of a higher modulus board material

18 Steinberg on PCB Edge Conditions
PFDPU Mechanical Evaluation Steinberg on PCB Edge Conditions PCB mode shape descriptions Pinned edge mode shape is a half sine wave: Z(x.y) = [sin(px/a) sin(py/b)] Fixed edge mode shape is a full sine wave: Z(x,y) = ½ [cos(2px/a) -1] [cos (2py/b) -1] Local curvature differs by a factor of two at the board center In his text, on pages , this curvature difference is dismissed as small Section 8.6 discusses edge wedge clamps (fixed edges) His fatigue database presumably includes these types of PCBs

19 Steinberg on PCB Edge Conditions (con’t)
PFDPU Mechanical Evaluation Steinberg on PCB Edge Conditions (con’t) Pinned vs. fixed PCB edge comparisons – an even more conservative approach Heavy solid lines are the mode shapes – dashed lines are the local (2nd derivative) curvature Rust line is the |fixed to pinned edge curvature ratio| Fatigue amplitudes are defined by curvature between any two chip ends (arbitrary positions) Red 1.26 is the average of this curvature ratio between 20% & 80% of board dim’s Large parts typically aren’t put in board corners (centered can be a smaller issue) We introduce a modified (x , for x&y) location factor: r’ = 1.60 r(Steinberg)

20 PCB Material Damping Properties
PFDPU Mechanical Evaluation PCB Material Damping Properties Inherent PCB material damping is proportional to 1/Q test results Damping is the least understood material property, subject to many factors SDOF models assume viscid damping, while data often shows hysteretic trends Test data often reflects a weak amplitude dependence Actual DCB test result was close to Steinberg predict ( Q = √Fn) Design margin predicts are tolerant of variations in Q (below) test predict

21 PFDPU Mechanical Evaluation
DCB Daughterboard has only a CCGA (ceramic column grid array) Actel Considered a risk, but not as bad as the BGA (ball grid array) Maven Thermal Expectations Actel has provided an extensive thermal cycling test report Typically 1000 cycles of +105 to -55C needed for failure Coffin Manson predicts provided for other environments PFDPU Thermal environments are expected to be mild > 10C variation predicts for Martian orbits CCGA Contacts Detail Maven Vibration Expectations Small (2.5” x 2.7”) pinned edge board ~ 1100 Hz first mode predict (DB only) Z3s deflection ~ 0.004” predict Z∞ deflection ~ 0.006” for infinite life DB’s input will be rolled off ~ 12dB from the ~300 Hz first mode of its motherboard. ~ 0.007” deflection at DBs center [R] from MBs 300 Hz 1st mode (~ 0.003” under CCGA)

22 Revised Design Margins
PFDPU Mechanical Evaluation Revised Design Margins The revised results table below reflect: Corrections based on the EM-DCB reduced vibration test data Scaled DCB PCB frequency predicts The 5% frequency reduction to account for the frame stacking Enhanced location function (r’ = 1.60 r) accounting for larger board curvature

23 PFDPU Mechanical Evaluation
BOARD BY BOARD SUMMARY DCB 1&2 (UCB) stiffened by partial daughter board: 0.062” PCB with glued edges Test measurement of EM DCB resonant frequency proved adequate IIB (UCB) w/ 2 T daughter boards and no large parts: 0.062” PCB with glued edges REG 1&2 (UCB) NO large parts: 0.062” PCB with glued edges DAP 1&2 (UCB): one large part: 0.062” PCB with glued edges MAG 1&2 (GSFC) one large part: Considering (CAD-FEM) a local board stiffener Fallback Option: ” PCB with glued edges BEB (LASP) NO large parts : ” PCB with glued edges DFB (LASP) Investigating (CAD-FEM) a local board stiffener Fallback Option: 0.093” PCB with glued edges, or DFB-BEB-end plate partial skewer PCB Edge Bonding (gluing) Plan All board frames have a 0.080” wide shelf (all around) for good board support Generous structural epoxy (Hysol EA-9309) fillets on both sides, before conformal coating

24 APPENDIX LASP’s Board Stiffener

25 Structural Analysis/Redesign EUV Mechanical
Appendix: LASP’s DFB Stiffener Digital Filter Board Structural Analysis/Redesign EUV Mechanical David Normen / David Braun Structural Analyst / Mech. Design Eng.

26 Appendix: LASP’s DFB Stiffener
LASP provides DFB and BEB board to Berkeley. DFB BEB

27 Appendix: LASP’s DFB Stiffener
DFB’s FPGA did not meet LASP’s criteria for infinite life. Steinberg Analysis for Electronic Equipment Trade study to improve the FPGA’s structural performance. Improved mounting does not predict infinite life, but should be acceptable. Board w/ CQ352 Board Thickness Max Board Disp (inches) Lead Wire Stress (ksi) Board Freq (Hz) Steinberg 1/16” .006 N/A 223.3 Initial Mounting. (8 fasteners) .013 92.7 176.1 Improved Mounting 14.5 394.3

28 Appendix: LASP’s DFB Stiffener
Convert EMI Shield to stiffen board. Add a center support. Add simple support around the edges of the board. Mat’l = Al 6061-T6 FPGA add Corner staking with Scotch Weld 2216. Standard LASP lead wire forming. Frame Note change to 8 edge fasteners. Board No layout changes. EMI Shield and Stiffener Lead wire and corner stake elements

29 Appendix: LASP’s DFB Stiffener
CQ352 FPGA

30 Acceptance(RV + Acoustics) / 1. min
Appendix: LASP’s DFB Stiffener Criteria: DFB/FPGA must survive all vibration tests + flight w/ margin. (≠  life) ≥ 9 minutes of vibration. < 100% Cumulative fatigue damage. Vibration Test # Axes Level/Time per axis # of runs Duration (min) Lead Wire Stress (ksi, rms) Cumulative Damage (3) Time to Failure(2) PFDPU Test 1(1) Proto-flight/1. min 2 2. 14.5 1.3% 39.5 Space Craft Level 3 Acceptance(RV + Acoustics) / 1. min 6. 10.3 .0% Launch 1 1. Total 5 9. --- (1) Assume only vibration normal to the board causes damage to lead wires. (2) Time to Failure includes a 4x uncertainty factor for fatigue life per GEVS (3) Failure occurs when damage ≥ 100%. Alloy 42 lead wires, use Kovar properties.

31 Appendix: LASP’s DFB Stiffener
Stiffener and EMI shield are integrated as one part DFB layout will use “EM” board-to-frame mounting hole pattern LASP supplied Stiffener LASP supplied DFB PWBA LASP supplied Frame Berkeley supplied Vent Assembly

32 Appendix: LASP’s DFB Stiffener
Fasteners pass through frame and PWBA and threads into stiffener Stiffener sits within the 0.070” keepout of PWBA EMI groove Stiffener, chromated aluminum, 178g PWB Frame, chromated aluminum

33 Appendix: LASP’s DFB Stiffener
Copy of frame drawing so we can duplicate size and tolerances Locate (2) clock signal connectors Determine fix ~.050 height discrepancy of stackable connectors Check on fastener head clearance between DFB & BEB

34 Appendix: LASP’s DFB Stiffener
CQ352 Package on Board Board or CQ352 Mountings Board thickness Max Board Disp (inches) Lead Wire Stress (ksi, rms) Mode Freq (Hz) Q.S. Acceleration (Gs, rms) Time to Failure (min) Method of Analysis Steinberg Assumptions (Simple Support on 4 sides ) 1/16” .0055 N/A 223.3 --- Steinberg Current Mounting. (8 fasteners) 1/8” .0127 .0106 .0039 .0042 92.7 88.4 66.4 70.0 176.1 409.1 416.5 23.4 44.2 RV Q.S. – Miles Current Mounting, Corner Staking: Uralane 86.5 176.2 0. Q.S. - Miles Current Mounting, Under-Fill: Nusil CV-2942 .0108 81.5 183.8 24.2 Current Mounting, Corner Staking: S-W 2216 .0104 27.6 181.9 24.0 Current Mounting, Nusil under-fill and Corner Staking (S-W 2216) .0107 26.4 191.1 25.0 Current Mounting, Under-fill, 2216 Staking, simple edge support, 8 Fasteners .0098 .0034 .0037 25.3 28.2 27.5 207.2 480.5 26.5 49.8 Current Mounting, Add idealized center support, Q =19.1 .0061 18.6 364.8,… 52.4 DFB stiffener 01, Center Support, Q = 19.8 .0057 14.5 394.3,… 39.5 min Board damping is assumed to be fn (Steinberg) unless otherwise noted. Time to failure is 1/4th of the predicted per GEVS.

35 BACKUP MATERIAL

36 PFDPU Mechanical Evaluation
MAVEN Zone 2 –Fwd Deck Vibration Levels (derived from Atlas 5 acoustic levels) Steinberg’s KOVAR Fatigue Plot Sy ≈ ≈ Se

37 EM-DCB BOARD TEST RESULT (¼ G sine sweep)
PFDPU Mechanical Evaluation EM-DCB BOARD TEST RESULT (¼ G sine sweep) Partial daughterboard & the many other features defy proper analysis Original result assumed a single uniform density board

38 DAVE STEINBERG’s ON-LINE PROFILE MAJOR PUBLICATIONS (books)
PFDPU Mechanical Evaluation DAVE STEINBERG’s ON-LINE PROFILE Dave S. Steinberg is a well-known author and internationally recognized authority on the mechanical design, analysis, testing and packaging of cost effective sophisticated electronic equipment that must work with a high degree of reliability in harsh thermal, thermal cycling, vibration and shock environments. He has been involved in these areas related to commercial, industrial and military applications for many years. He recently retired from Litton Industries, where he was the manager of the Engineering Department for 15 years. Mr. Steinberg is a visiting professor at the University of Wisconsin and for the past several years at UCLA where he has been presenting full semester post-graduate courses on similar electronic subjects related to the reliability of electronic equipment. Mr. Steinberg has worked on many aircraft/missile/space programs, ships' electronic systems, automotive electronics and computers. These include the F-14, F-15, F-16, F-22, cruise missile, AMRAAM, ALCM, TALCM, SLCM, Titan 2 and 3, space shuttle, Mercury, Gemini, Apollo, communications satellites, DD963 destroyers, and the USS Forestall aircraft carrier. Typical companies include Lockheed, IBM, Litton, Intel, GM, Ford, Boeing, Wright Patterson AFB, Northrop, Cisco, Harris, United Aircraft, Hamilton Standard and Motorola, just to name a few. MAJOR PUBLICATIONS (books) “Vibration Analysis for Electronic Equipment” 3rd Ed. J.Wiley & sons, 2000 : 2nd Ed. 1988 Cooling Techniques for Electronic Equipment, 2nd Ed. J.Wiley & sons, 1991 Preventing Thermal Cycling and Vibration Failures in Electronic Equipment J.Wiley & sons, 2001


Download ppt "Particles & Fields Data Processing Unit (PFDPU)"

Similar presentations


Ads by Google