Never blindly rely on vendor configuration control.
I once designed something using an 8-bit D/A converter that was being sold by Signetics. For one long-forgotten reason or another, I used the part at some variance from normal and doing so, I got myself into some trouble.
Figure 1 Signetics D/A converter input configuration
This now long-discontinued converter had an input configuration something like the above. For some reason, I decided that having 40 k-ohms in the chip could be useful and so I used the "Trim" pin as my analog input instead of using the "Vref" pin. Everything worked just fine and there was happiness.
Happiness came to an end approximately a year later. While my circuit had worked well, suddenly the scale factor through the D/A process went way out of kilter with readings that shot way up and into overrange. Here's why:
Figure 2 Signetics D/A converter revised input configuration
Signetics had added a new diode in parallel with the resistor tied to the "Trim" pin. I couldn't get any kind of answer as to why this was done, only a rather sheepish confession over the phone that "several" users had to reconfigure their circuits. It was small solace to know that at least I wasn't alone.
My first lesson learned was that I should have used the Vref pin for my reference to begin with. I would note at this point that a second source supplier of this part, Datel, never did put in that diode. Their parts still worked as I had intended.
My second lesson learned was that vendor configuration control was something on which I should never blindly rely. While I've seen semiconductor parts get revised over the course of their product lifetimes such as hometaxial 2N6578 power transistors at RCA getting changed to epitaxial planar in the late 1980s. Please see Semiconductor Evolution for more details.
I had never before, nor have I at any time since, seen such disregard of configuration control bring about such a functionally variant circuit topology.
Signetics is long gone. However, if something like this were to ever happen again, I would no longer be surprised.
John Dunn is an electronics consultant, and a graduate of The Polytechnic Institute of Brooklyn (BSEE) and of New York University (MSEE).