On the shop floor, we usually talk about tolerance in discrete steps: a 0.02 mm cut, a 0.01 mm shim, one more finishing pass. But the underlying behavior is continuous. If we treat cavity depth as a function of tool position, the finite increment Δy we measure with a height gauge is only a snapshot of a continuous error curve. The differential dy is what actually propagates through the process. In practice this means a 0.005 mm spindle thermal drift at the start of roughing does not stay at 0.005 mm—it compounds through semi-finishing and finishing, and by the time we reach the cavity floor the accumulated deviation can be two to three times the initial increment. Understanding that gap between Δy and dy is the difference between chasing errors and controlling them.
Concretely, take a deep-cavity core insert in P20 or 1.2344 steel. Rough machining leaves 0.5 mm stock; semi-finishing brings it to 0.1 mm; finishing targets ±0.01 mm. Each stage has its own Δy—cutting force deflection, tool wear per pass, and fixturing shift. The differential view tells us to weight the finishing allowance by the sensitivity of the final dimension to each upstream variable, not to treat all stages equally. That is why we map error transmission before selecting cut parameters: a 0.003 mm/rev feed change in semi-finishing may matter more than a 0.01 mm depth adjustment in finishing. On the machine, this translates into tighter in-process gauging after semi-finishing and locked thermal compensation windows during the final pass.
None of this replaces the basics—rigid setups, sharp tools, stable coolant. It just gives us a language for where the error actually comes from. When a cavity comes out 0.015 mm shy on depth, the question is not "which pass was wrong" but "where did dy start drifting." That habit of tracing increments back to their source is what separates a repeatable mold shop from a lucky one. For more mold sourcing and machining know-how, visit MoldSo at www.moldso.com.
We run a lot of multi-cavity packaging molds and the tolerance stack across cavities is where we get burned. On a recent 8-cav mold the first-off parts were all within 0.03mm, but after 50k shots cavity 5 drifted past 0.08mm. Anyone tracking wear per cavity over time instead of just first-off? Curious what interval you'd set before the drift bites you.