7

Why isn't there any EEPROM in the STM32F4 series MCUs?

Mostly I have Microchip MCUs and they have EEPROM available in them, but I just found out that it is not available in the STM32F4 MCUs... And it looks like not in other families as 'F0, F1, F2 and F3 either.

Is there a way around to save parameter values in the absence of an EEPROM?

Peter Mortensen
  • 1,676
  • 3
  • 17
  • 23
scico111
  • 857
  • 16
  • 39
  • 5
    "What could be a good reason?" questions do not fit within the Stack Exchange mission, and "such an important memory area" is very **application-determined**. Looks like no EEPROM in the STM32L4's either, but the L0's and L1's have it. Or you can add an extra chip if you have a need for which emulation won't work. – Chris Stratton Feb 19 '19 at 22:28
  • Is it safe enough to use an emulated eeprom vs external eeprom chip? – scico111 Feb 19 '19 at 22:29
  • 7
    That would be entirely *application dependent*. Since you've said nothing about what you are trying to do for a question which would have to consider the specifics in extreme detail, no one can help you. – Chris Stratton Feb 19 '19 at 22:48
  • 5
    The most likely explanation is that the application(s) for which the chip was initially developed did not require it. Remember, EVERY chip ever developed was designed for a specific large-volumne application, and only later added to the manufacturer's general catalog. The overhead of a new chip design is just too high to allow designing chips speculatively. – Dave Tweed Feb 19 '19 at 23:14
  • 5
    "such an important memory area" - important to who? I'm currently working on a project using an STM32F4 device and I would have no use whatsoever for a little bit of internal EEPROM. The extra cost it would add to the device would certainly make a difference though. – brhans Feb 19 '19 at 23:14
  • Small amounts of anything may not be expensive, but at scale, in counts of millions, in terms of _total_ unit-cost, it can be very significant. –  Feb 20 '19 at 14:48
  • You need L0 series for that. Emulated EEPROM in program flash is really tiresome to work with, I would strongly recommend to stay clear of that. – Lundin Feb 20 '19 at 16:03
  • Perhaps manufacturers reason that it is better to allow customers who need EEPROM to incur the small cost of adding it externally rather than burden all customers with the cost of internal EEPROM that many or most customers do not need. – user57037 Feb 21 '19 at 01:23

3 Answers3

25

All STM32 MCUs have self-programmable flash memory. If you need to store user settings, you can store them in an area of flash.

ST provides a library to perform EEPROM emulation on the STM32F4. (There are similar libraries for most of their other parts as well.) Even if you don't plan on using that library, their application note explaining how it works may be interesting to read.

  • 7
    This is correct but it is a poor substitute. The CPU cannot execute while the flash is being written or erased, and erasing takes a long time. There are tricks (multi flash banks, ram functions) but none are as tidy as just having an internal eeprom like AVR and PIC. – Jon Feb 19 '19 at 22:33
  • 4
    @Jon of 577 Microchip PIC32 and Atmel 32 bit MCUs currently in production, only 63 have Data EEPROM. https://www.microchip.com/ParamChartSearch/chart.aspx?branchID=211&popular=1 – Bruce Abbott Feb 19 '19 at 23:59
  • 3
    @Jon, Not sure why you believe "The CPU cannot execute while the flash is being written or erased..." I have used this method on an F1 device. If the CPU had stopped during a write to Flash, I think I would have noticed – user28910 Feb 20 '19 at 13:26
  • 2
    @user28910 it’s not a matter of opinion. It is in the product documentation. Perhaps you were using a dual bank device (the larger sized ones). These can execute out of one bank while writing to another but this means your need to keep all the relevant code in the correct bank. – Jon Feb 20 '19 at 15:38
  • 1
    @Bruce Abbott interesting point. I knew there were chips without eeprom but didn’t realise it was so extensive. – Jon Feb 20 '19 at 15:39
  • @user28910 He means cannot execute code from the same flash bank that is being written/erased. Which is true for any MCU on the market and you will definitely notice, as it will hang-up or go haywire. It is perfectly possible to execute a flash writing driver from one bank while programming another. Or indeed to execute the code from RAM. It's still a poor man's EEPROM and a crappy solution. – Lundin Feb 20 '19 at 16:06
  • While this is a useful practical solution to the OP's real problem, it does not answer the question "why", and should not have been accepted as an answer. IMnsHO. – mlp Feb 20 '19 at 21:33
11

EEPROM is very expensive in terms of cell size (leading to a larger die and hence higher cost). Manufacturers started trying to get rid of EEPROM as soon as the first Flash based controllers were released.

Especially when you consider the varying user requirements for the amount of EEPROM, it makes more sense to emulate in Flash, despite the limitations. As opposed to (for example) having a fixed 512 bytes of EEPROM, when one customer is only using 20 bytes, but paying for 512.

elchambro
  • 796
  • 4
  • 13
  • 1
    There is however data flash, which behaves quite similar to EEPROM, in terms of write cycles etc. But often requires many bytes to be erased, such as chunks of 8 bytes at a time or so. Whereas program flash will have huge erase sizes. – Lundin Feb 20 '19 at 16:09
2

If having non-volatile memory which is programmable and erasable to the byte level is important (as opposed to Flash memory, which must have entire sectors/pages erased), have a look at the STM32L0 and STM32L1 series. They have true data EEPROM embedded on it.

Jarhmander
  • 443
  • 2
  • 7