SRAM based NVRAM

[vlad01]

Actually I found in the description that the AS7C1024 will retain data down to 2v in standby, but they only guarantee that for the 3.3v model that the consumption will not exceed 36mw at the nominal voltage.
So the specs are still vague even with that min 2v I found in the description. So it still sounds like it uses more power than we ideally want.

I think you'd want to get into the low single mW or ideally into the uW range if possible, then you will get good life out of coin cells.

I do believe that the ECU provides additional power for the memory retention, but having long life when completely unpowered like the NVRAM can, would be the best scenario.

My RT ECU never lives in any car other than during tuning. I have all my cars run a normal memcal when I am not doing a lot of tuning back to back. So my NVRAM lives without power 99% of the time, relying on it's own battery. It makes it even more amazing that it has lasted as long as it has, 15 years!

[Staggie]

Yeah that other chip does look a fair bit nicer on the battery life, as speed isn't really an issue it is probably the better choice. I have rigged something up so I can measure the actual current draw of the chip in circuit. currently the AS7C1024 is pulling 104 uA and slowly decaying, which makes sense as if it stayed at that I would have 2 rather flat batteries right now. Will keep an eye on it over the next 24hrs and see what it does.
I do need to optimise the WE line pullup as it is rather strong and would not be helping the current draw on the Vout line of the BQ2201.

Could always sub in ML1225's and build in a charger

But yeah you are right with it living out of the car, would be a pain in the ass having to put new batteries in every time want to use it, and 1225's aren't exactly something you generally find in the kitchen junk draw at 6pm on a sunday.

[vlad01]

Can the pull up line be be tied to the external supply so that only pulls up or flows a current when the ignition is on, if that's possible? Since we are not writing when the ECU is off, to me that makes sense, and it should reduce the load on the batteries.

[Staggie]

My only concern with that is what state is the WE pin in when it does power up, probably wouldn't take much to corrupt it.
Is certainly worth a test though. I did measure the current with pullup on and off Vout and the difference was negligible.
When the chip is fully asleep (ie. batteries have been completely removed and reconnected) the whole circuit was only pulling 1.4uA.
More testing is definitely required, that other chip does certainly look like a winner, need to compare the pinout and physical spacing to see if can get the board to support both for testing purposes.

[vlad01]

Assuming what you did there clears all the content, I'd say that's not really asleep, rather that all the transistors are unlatched being is a zero state across all memory bit cells, therefore the current shows up as low. So when there is actual content in the cells, the latched transistors will always pull some current.


Anyway, did a bit of digging and turns out there is an SRAM for exactly this type of scenario. A model range in the MoBL (more battery life) from Cypress/Infineon.

It has a special auto power down mode for when there is no activity on the addresses even when powered. My understanding is that further power reduction when it is actively put in standby.

According to google ai this is the 128x8 SRAM with the lowest data retention power consumption, although the one 202 mentioned has a slightly lower standby of 0.6uA. this model is 1uA, but has all the more advanced power saving features. In practice, it might come down to the individual chip itself. Some are better samples than others. Both this and the one 202 listed are both stated at max 4uA for the best variant.



Data sheet.

https://www.infineon.com/assets/row/pub ... eet-en.pdf

[Holden202T]

hmm i think i have that chip on one of the boards i made first time around!

[Staggie]

I might need to borrow that 202 for "testing purposes"
Saves a digikey order for now.

[antus]

Just off the top of my head I think there is also fram, which is like sram but without the battery. viewtopic.php?t=4574
Looking around there seems to be compatible options now, might be better than sram for new builds and completely avoids the battery. I havn't checked all the specs for gotchas. But the FM18W08 looks like it might suit.

[Staggie]

That would certainly be ideal antus, after a quick search i think the issue might be finding one of a bigger capacity in 5v that isnt EOL
Definitely worth adding to the research list.

[antus]

rs components have stock now and mouser is showing an eta. I also see them on ebay. So it appears they are available. As far as capacity goes 1 of them would be OK for 12P and you could have more than one with some digital logic on the high address lines and/or chip enable lines to switch which chip is active to map them out next to each other in the address range.