The LF871 and the F871 are the same silicon with different test limits, and in production they get mixed up. A board built during a shortage may carry either, and the marking on the package is the only thing that says which. For a read that matters less than it sounds; for a rebuild it decides the supply range the replacement has to meet.
Same Die, Different Spec
Both parts have 2048 words of flash, 128 bytes of RAM, 64 bytes of EEPROM, 33 I/O and eight 10-bit channels. The LF version is specified down to 2.0 V and the F version to 4.5 V, and the difference is in the test binning rather than in the layout.
We report the device ID and the package marking together, because a customer who sends a board described as an F871 sometimes has an LF871 on it, and the reverse happens just as often when a product was re-binned during a shortage.
Why Two Volts Changes The Firmware
A design that may run at 2.0 V cannot assume anything about the analogue front end. The converter reference has to track the supply, the oscillator has to be stable across the range, and the brown-out detector has to be set at a level that is meaningful at both ends.
Those decisions appear in the image as configuration bits and as constants. A reconstruction that copies an F871 image onto an LF871 blank is usually fine electrically and occasionally wrong at the low end, where a threshold that was tuned for 5 V trips at the wrong rail voltage.
Eight Channels On A Shared Rail
Eight analogue inputs sharing a package with 33 digital functions means a lot of pins change role. On a low-voltage design the switching thresholds are lower too, so a pin configured as an analogue input while the firmware drives it as an output can inject a signal into the converter that the design never anticipated.
The channel record we deliver lists which channels the firmware selects and which pins are driven during a conversion. It is the document that catches this class of error.
Mixed Rails On One Board
A product specified for 2.0 V often runs at 3.0 V in practice, because the battery stack decides it. The firmware therefore cannot rely on the LF part’s extra range being used, and it may not be tested at the low end until a customer’s unit goes flat.
For a read that is irrelevant. For a rebuild it is decisive: we ask what the product actually runs at and verify at that voltage rather than at the part’s minimum, because a reconstruction verified at 2.0 V on a 5 V board tells nobody anything useful.
Why Binning Is Not A Layout Change
An LF part and an F part come off the same die and are separated by test. There is no mask difference, no metal option and no extra circuit. The consequence is that a dump from one is a valid reference for the other, and the only thing that has to change in a rebuild is the supply specification of the replacement.
That makes the LF/F distinction one of the few in this family that can be settled from the package marking alone, provided the marking is legible. On parts that have been through a reflow oven twice, it often is not, and then the supply range gets measured instead of read.
Reading The Part
Configuration word decoded first, ID recorded, array swept twice. The 40/44-pin package is straightforward to open and the die is large, so the physical work is not the constraint; the constraint is the verification, because a pin-heavy low-voltage board has many states to exercise.
We test at both supply extremes where the customer’s product runs at both, which for a battery product is the only meaningful test.
Delivery
Binary, HEX with configuration, a listing, the channel record, the input-output table, and a note stating which supply range the image was verified at — F or LF — so the customer can specify the right replacement part.
The standard-voltage version is the PIC16F871 page, the 28-pin low-voltage part is the PIC16LF872 job, and the 28-pin low-voltage part without the serial port is covered at the PIC16LF870 page. Quotes start at the services desk.