How can a 51822 QFAB have more than 128kb?

[Flagada15]

Hi all!

I recently ordered BLE tag with nrf51822 QFAB chips.

When I read back the memory content (with mem8 jlink command) I manage to read up to the address 24A60 which means 150112 bytes.

So my question is how can it be possible to have more than 128kb in a QFAB chip? I am wrong somewhere?

Thanks in advance for your answer

[psharma]

I am not a Nordic employee, so I cannot say for sure, but I would thank that chips are all made the same, and stamped according the QA test. So If the flash of a sample from a particular batch of silicon tested between 128kb and 256kb, the entire batch was stamped as AB. Hence the chip you have at hand may be a part of such a batch, and may have more flash than specified. A similar case may also be true with AA and AC variants where the AA were found to be usable with less that 32kb of RAM stable, so they were marked as 16kb RAM variant.

That's my wild guess based on the fact t hat a similar thing was done with the Atmega microcontrollers, where those ICs that were not stable at the full 16MHz were stamped with L, marking them as low power variant, and specifying their rated max speed at 8MHz. Even though those chips may do well even at 5V 16MHz.

Please correct me, of course, if I am wrong.

[psharma]

This question deserves to be upvoted!

[Roger Clark]

Again. I'm not sure if this applies to Nordic devices, but the same thing occurs with STM devices. The STM32F103C8 device is only supposed to have 64k flash but in reality every STM32F103C8 device is really a 128k flash device (The 8 on the end is really a B (bravo))

This only seems to apply to Flash and not RAM. i.e I have some nRF51822QFAA devices and they do not have 32k RAM like the QFAC does.

[Flagada15]

Thank you for your answer this sound good to me :) But on the other side I also have QFAB chips that stops exactly at 1FFFF (128kb) So it seems that some chips are made to be "true" QFAB with exactly 128kb and others were intended to be QFAA but where "downgraded" to QFAB after the tests. Is that correct you think?

[Roger Clark]

What you are describing is the normal yield tests that semiconductor manufacturers do on each chip.

If parts of the 256k flash are not working the MCU is sold as a QFAB instead of a QFAA

Bear in mind, that this means that Nordic don't think the full 256k is available or possibly not reliable

The only way to test it would be to fill every byte with 0x55 and then again with 0xAA and confirm that when you read back the flash that the data pattern is correct - No doubt there are more sophisticated tests.

Also bear in mind that it could be that the Flash does not work when the device gets hot, as they may test the Flash at the upper and lower limits of the operating temperature as defined in the hardware spec for the MCU

PS The same sort of thing applies to processor speed, some manufacturers release different speed versions of MCU's and they pick them based on how fast each MCU will run