The problem
A BCD (binary-coded decimal) input uses 4 bits to represent one decimal digit, 0 through 9. Bit patterns 1010 through 1111 (10-15) never occur in valid BCD — they're not just unused, they're guaranteed impossible if the rest of the system works correctly, which makes them genuine don't-cares rather than something to just leave as 0.
Segment "a" is the top horizontal segment. On a standard display, it's lit for every digit except 1 and 4:
| Digit | B3 B2 B1 B0 | Segment a |
|---|---|---|
| 0 | 0000 | 1 |
| 1 | 0001 | 0 |
| 2 | 0010 | 1 |
| 3 | 0011 | 1 |
| 4 | 0100 | 0 |
| 5 | 0101 | 1 |
| 6 | 0110 | 1 |
| 7 | 0111 | 1 |
| 8 | 1000 | 1 |
| 9 | 1001 | 1 |
| 10-15 | 1010-1111 | X (don't-care) |
The minimization
Using the tool below (labeling the BCD bits A=B3, B=B2, C=B1, D=B0), all four groups turn out essential — no Petrick's method tie-breaking needed here, every group is the only one covering at least one required cell:
Each term corresponds to a real grouping: B′D′ covers the four-corner wraparound group {0,2,8,10}; C covers an entire 8-cell half of the map {2,3,6,7,10,11,14,15}; BD covers {5,7,13,15}; and A covers the whole upper half {8-15}. The don't-cares at 10-15 get absorbed into whichever groups they help enlarge — notice three of the four terms include at least one don't-care cell in their coverage.
Try it live
The solver below is pre-loaded with this exact problem. Toggle cells, switch to POS, or try a different variable count to see how the same engine handles it.