There is no single number, and a supplier who quotes one without the test profile is quoting nothing you can use. Solder joint life falls roughly with the temperature swing raised to the power of two, so a joint that passes 1,000 cycles between 0 and 100 C does not pass 1,000 cycles between minus 40 and 125 C. Run the same joint on the wider swing and the arithmetic puts it near 390 cycles. The useful question is not how many cycles, but under which profile, with which criterion, on which board.
Everything except the profile is second order. The strain that cracks a ball comes from the difference in expansion between the package and the board, multiplied by the distance from the neutral point, divided by the standoff height. Take a 25 mm body BGA on 1.6 mm FR-4. The corner ball sits about 17.7 mm from the neutral point, the mismatch between the package substrate and the board is about 6 ppm per degree, and a swing of 165 K produces 17.5 um of lateral displacement. Across a 0.45 mm standoff that is 3.9 percent shear strain, about 39,000 microstrain, every cycle, in a material that yields far below that figure at these temperatures.
That one calculation is why the answer is a slope, not a constant. Double the swing and damage per cycle roughly quadruples; halve it and the same joint outlives the product. It is also why a board that passes a bench check at 25 C can still fail in a cabinet that swings 60 K twice a day for five years.
The three numbers behind any cycle count
Peak temperature. SAC305 melts at 217 to 220 C, which is 490 K. At 125 C the joint sits at 0.81 of its melting point, at 85 C at 0.73. Above roughly 0.5, solder creeps: it deforms under a stress it would carry elastically at room temperature. Tin lead at 183 C starts even higher on that scale, which is one reason older parts look more robust in the field than their cycle count suggests.
Dwell. A joint held at the top of the swing for fifteen minutes relaxes by creep and returns with a coarser grain structure and less of its initial strength. A joint that passes through the peak in thirty seconds does not. This is why a thirty minute test cycle and a 24 hour field cycle are not interchangeable, and why an acceleration factor built only on the swing overstates field life.
Rate. The standard correction, Norris-Landzberg, adds a cube root frequency term and an Arrhenius term on peak temperature to the Coffin-Manson swing term. Those two terms are the difference between a report that predicts the field and one that merely ranks two samples.
| Profile | Swing | Relative life, same joint | What it represents |
|---|---|---|---|
| 0 to 100 C | 100 K | 1.0 | Consumer and indoor telecom enclosure |
| minus 40 to 85 C | 125 K | 0.65 | Outdoor unit, automotive cabin |
| minus 40 to 125 C | 165 K | 0.39 | Under hood, defense, base station |
| minus 55 to 125 C | 180 K | 0.33 | Airborne and shipboard |
The table is arithmetic, not test data: each figure is the swing ratio raised to 1.9. It shows why two suppliers can both say 1,000 cycles and mean rather different hardware.
Why two 1,000 cycle reports can describe different joints
IPC-9701 sets out the conditions that get quoted most often: TC1 from 0 to 100 C, TC2 from minus 55 to 125 C, TC3 from minus 40 to 125 C. JESD22-A104 condition G is the same minus 40 to 125 C band with its own ramp and dwell rules. A report that names neither the standard nor the condition is a number without a unit.
The second variable is the specimen. A daisy chain coupon has no die shadow, no copper plane under the balls, no void in a thermal pad and no stiffener, and all four change the answer on the real board. The third is the failure criterion: a 10 percent resistance rise and a 20 percent one give different cycle counts from the same test, and the looser criterion can look 30 percent better with nothing changed.
Three things that get read wrong
A cycle count without a profile. The most common. A quotation that says qualified to 1,000 thermal cycles says nothing until the swing, the dwell, the ramp and the criterion are attached, and a buyer comparing two suppliers on that sentence is comparing two different tests.
Treating the coupon as the board. Coupon data ranks materials well and predicts a product badly, because the coupon has none of the features that concentrate strain: the die shadow in the middle of a large package, a plane that holds heat during the ramp, a via in pad under a land, or the stiffener that halves the board's movement and doubles the joint's. Ask for the coupon result as material evidence and for board level cycling as product evidence.
Underfill or corner glue as a free fix. It does raise joint life, sometimes by a factor of five on a large body BGA, and it moves the failure to the pad and to the package substrate. Both need their own qualification and neither is repairable, so a board that still needs rework should not be underfilled.
What SUPERB Assembly controls on the build
The cycling itself belongs to a reliability laboratory, and the numbers above belong to whoever runs it. What decides how the board behaves in that test is the build, and that part is ours: 7 SMT lines, placement from 01005 to 55 mm connectors, 3D SPI on every print, 3D AOI plus DIP AOI, X-ray with 2D and CT on BGA and thermal pads with void held under 5 percent to IPC-7095, a 12 zone reflow with nitrogen under 500 ppm, first article comparison against CAD data, ICT and functional test on dedicated fixtures, and final inspection to IPC-A-610 Class 2 or Class 3 as the product requires. Assembly is to J-STD-020 and J-STD-033, with bake and dry pack records travelling with the lot.
For boards that will see wide swings, the substrate side matters as much as the assembly side: Tg 150 or 170, controlled CTE, and a stack that keeps copper balanced around the neutral axis limit bow during the ramp. That is on high reliability PCB. The assembly scope is on turnkey PCBA assembly, and the automotive side of the same question is on automotive PCBA. SUPERB Tech reports the void figure per ball, because a joint that starts at 30 percent void does not get the cycle count the coupon promised.
What to send with the question
The swing and the profile name, ramp rate, dwell at each extreme, number of cycles, the failure criterion you will accept, the packages and their body sizes, board thickness and layer count, the largest copper plane under a package, void limit, whether underfill or corner glue is in scope, and any coupon or field return data you already have. With those items a supplier can answer with a test plan instead of a slogan.