That's a great tip for beginners. For more advanced users, I personally don't like relying on input masks because I find them quite limiting. For something like this, at a developer level, I would recommend using an After Update event with the UCase function.

I recently did that for file extensions. I wanted them all to be uniform in the Table. I make sure they were all lower case using the LCase function.

The biggest place I provide input masks is for Tax ID Number (TIN, EIN, SSN) and Dates. They give a nice visual representation to the user for what is expected. However, with dates, I also like to be able to have the Date Picker as an option but unfortunately, if you have an input mask, the date picker is disabled so I had to do a couple little subs so that if the user want the date picker, they can double click and it will clear the mask and popup the picker and when they leave the field, it restores the input mask and formatting so it ensures it's displayed properly.

Thanks guys. This demonstrates the great heights that Access can reach. I'm really a novice down in the foothills and you guys are up near the summit of Everest.  But it also demonstrates that Access has to be this difficult (powerful), because what we want to achieve can be deceptively difficult.
I used an input mask to get users to input consistent data, (and I didn't know better), but this apparently simple data input has a couple of complexities. I've already quoted the London exception - for most of the country 2 letter (BH signifies my part of the country (Bournemouth) which is split into about 27 geographic areas, numbered 1-27. Each of these areas is further split down to quite small areas containing 15-20 addresses using a number and 2 letters so my PostCode is BH3 7HZ.
The first problem for the input mask is having one or two placeholders. Using 1 only allows you to get to BH9, and using 2 means that you need always to input BH03 in my case.
Then as I say, some London addresses break the pattern for the first part anyway, e.g.  SW1A 2AA.
I'm probably going to spend some time now exploring Richard's UCase function to make something that handles these different inputs properly.