Current consumption on flash erase/write and timing

RE: Current consumption on flash erase/write and timing

rgrr2 — Wed, 24 Apr 2019 16:48:15 GMT

Hi cbd

thanks for your suggestions. Writing to flash is actually not the problem. Erasing a page is. And as long as partial erase is not supported by the softdevice I can see no way to get around the 4mA*80ms.

You are right: an external serial flash is also no option. There exists a single device which has comparable low erase currents. but this would be another "single source".

And yes: flash is really required for persistent parameters of the device and also as data storage for an application update.

Regards Hardy

RE: Current consumption on flash erase/write and timing

cbd — Mon, 08 Apr 2019 08:20:07 GMT

I see that you've marked one of Kenneth's replies as an answer.

Some more points to consider:

BLE listening times (e.g. if you need to wait for a message response or notification) will likely be more power hungry than flash writes.
If you perform bonding then the SoftDevice will write bonding information to internal flash. I don't know if this can be circumvented without looking.
An external flash will eat into your power budget and likely not save you anything.
For internal flash, data must be erased before the third write of the same 32bit block.
Beware of leakage currents and internal resistance with "booster" caps, they will also impose a delay on your start-up.
Consider using a low power state with full RAM retention e.g. do you really need flash?
Consider storing data in flash in small chunks over an extended period - this will allow you to use smaller "booster" caps and allow for recharge time.

RE: Current consumption on flash erase/write and timing

rgrr2 — Thu, 04 Apr 2019 19:52:25 GMT

Thanks Kenneth for your help. I'm yet not sure how to continue. I think that Nordic should specify the flash in more detail. Also a minor extension to the Softdevice flash API would be very helpful. In fact this partial erase feature could be completely hidden inside.

Kind regards

Hardy

RE: Current consumption on flash erase/write and timing

Kenneth — Thu, 04 Apr 2019 17:17:39 GMT

Then I am not sure if I have any good suggestion other than possible an external I2C EEPROM.

RE: Current consumption on flash erase/write and timing

rgrr2 — Thu, 04 Apr 2019 08:26:01 GMT

Unfortunately this is not feasible: we use the internal flash for persistent configuration data changed via BLE and also for something like file I/O also done via BLE. So the softdevice should remain switched on.

RE: Current consumption on flash erase/write and timing

Kenneth — Wed, 03 Apr 2019 19:35:07 GMT

Hi,

The datasheet does not contain any electrical data on the flash erase (or write) currents, so 4mA is basically just measurements done on a few DK while doing erase operation.

The softdevice api does not have any feature that support partial erase no, however if you don't need to erase the flash frequently, you can always disable the softdevice and erase the flash as you see fit, and then enable the softdevice again afterwards.

Best regards,
Kenneth

RE: Current consumption on flash erase/write and timing

rgrr2 — Wed, 03 Apr 2019 16:21:46 GMT

Other idea to avoid these large capacitors, could be to use the partial erase feature of the nRF52840. This would allow to split the erase operation into e.g. 10 chunks. Unfortunately there is no API in the softdevice S140 for this (Nordic? Anybody?? Would be great, if this could be added somehow)

RE: Current consumption on flash erase/write and timing

rgrr2 — Wed, 03 Apr 2019 15:25:07 GMT

yes, that's my idea as well. But where are those 4mA from?

RE: Current consumption on flash erase/write and timing

Kenneth — Wed, 03 Apr 2019 14:40:17 GMT

Hi,

In terms of timing to erase or write, you should calculate based on the worst case numbers, so if you need 1mA on average I suspect you will need rather large caps to handle approx. ~4mA for ~90ms, and stay idle for approx ~270ms before that.

Best regards,
Kenneth

RE: Current consumption on flash erase/write and timing

rgrr2 — Tue, 02 Apr 2019 16:36:39 GMT

No, we are not running from battery. We have a power source with almost constant 1mA which is buffered by a load of capacitors.

RE: Current consumption on flash erase/write and timing

cbd — Tue, 02 Apr 2019 15:09:50 GMT

If you're running on a very small battery, especially a non-lithium based one, just be aware that the peak current draw by the BLE may dip the voltage supply voltage below the chip reset level earlier than you expect. I've seen this happen with batteries that had a no load voltage of 2.7V

RE: Current consumption on flash erase/write and timing

rgrr2 — Tue, 02 Apr 2019 14:37:48 GMT

You are right with that. Fortunately one can send the MCU to sleep most of the time. There is even some spare current for BLE operation ;-)

Believe me: the device is working with the 1mA. The questions regarding flash erase current is just fine-tuning.

RE: Current consumption on flash erase/write and timing

cbd — Tue, 02 Apr 2019 14:33:49 GMT

Section 5.2.1.3 of the datasheet shows a maximum 6.3mA with an active CPU, external clock, internal flash access and no DC/DC regulator or 3.1mA with DC/DC regulator and internal oscillator.

This is obviously more than your 1mA budget (if it is a total).