Program memory (flash) data retention

Log in or register to post comments
Go To Last Post
13 posts / 0 new
Author
Message
Nuno
Nuno's picture
Level: Resident
Joined: Sat. Nov 1, 2003
Posts: 687 View posts
Location: Amadora, Portugal
#1
Posted by Nuno: Mon. Apr 23, 2007 - 03:51 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Fellow Freaks,

I've trying to find out what's the program memory data retention with no success. A search in the forums reveals only answered similar questions. Then I did a search on google and found some documents from ATMEL, but I really can't get any meaningful value out of them!
I got to these documents:
http://www.atmel.com/dyn/resourc...
http://www.atmel.com/quality/acr... (this one from 2007)
But these seem to be just test reports without values that, at least I, can't translate to the usual "years".
I also took a look at the FAQs in ATMEL support center but nothing.

Has anyone any information of this?

Thanks

Embedded Dreams
One day, knowledge will replace money.

Tags:

AVR Microcontrollers, megaAVR and tinyAVR
Top
ka7ehk
ka7ehk's picture
Level: 10k+ Postman
Joined: Fri. Nov 22, 2002
Posts: 20302 View posts
Location: Tangent, OR, USA
Posted by ka7ehk: Mon. Apr 23, 2007 - 05:33 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Normally, EEPROM has been lower (less time) data retention than flash by 2x or more.

The functionality of a programmed chip for a given applciation would then be limited by the data retention of the EEPROM, not flash. Maybe they figure that they only need to give EEPROM retention since that would be the limiting factor.

Jim

Jim Wagner Oregon Research Electronics, Consulting Div. Tangent, OR, USA http://www.orelectronics.net

Top
theusch
theusch's picture
Level: 10k+ Postman
Joined: Mon. Feb 19, 2001
Posts: 39512 View posts
Location: Wisconsin USA
Posted by theusch: Mon. Apr 23, 2007 - 06:11 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:

Then I did a search on google and found some documents from ATMEL, but I really can't get any meaningful value out of them!

There is an indirect reference to 20 years in
http://www.atmel.com/dyn/resourc...

Quote:
If the number of E/W cycles performed on a cell exceeds the guaranteed
10K E/W cycles, the data retention will decrease below the expected 20 years.

If we go under the assumption that AVR flash memories and their retention is the same across models, then the AT90SC summary is pertinent, and quotes 40 years:
http://www.atmel.com/dyn/resourc...

Quote:
Data retention for a minimum of 40 years

Beyond that, I think you will need to write Atmel a note with your question.

Lee

You can put lipstick on a pig, but it is still a pig.

I've never met a pig I didn't like, as long as you have some salt and pepper.

Top
Nuno
Nuno's picture
Level: Resident
Joined: Sat. Nov 1, 2003
Posts: 687 View posts
Location: Amadora, Portugal
Posted by Nuno: Mon. Apr 23, 2007 - 07:59 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Jim, where did they give EEPROM data retention time? I'm actually more interested in program memory retention.

Lee, I know that AN. But didn't show up in my googling :(. Anyway, I think we "deserve" to know for sure... I've just wrote Atmel asking about it.

Thanks for the answers, I'll post back when/if I get any answer from ATMEL.

Embedded Dreams
One day, knowledge will replace money.

Top
sgomes
sgomes's picture
Level: Resident
Joined: Wed. Feb 19, 2003
Posts: 920 View posts
Posted by sgomes: Mon. Apr 23, 2007 - 08:23 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

How long has flash memory of this type been around? Wouldn't ANY number from Atmel be theoretical at best? What effect does extreme solar flare activity have on it? How is it spec'd? a single bit? a byte? a word? 50%? recoverable or non-recoverable? device failure?

Go electric!
Happy electric car owner / builder

Top
Nuno
Nuno's picture
Level: Resident
Joined: Sat. Nov 1, 2003
Posts: 687 View posts
Location: Amadora, Portugal
Posted by Nuno: Wed. Apr 25, 2007 - 01:49 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Ok, I got the answer form Atmel!

Quote:
Atmel guarantees 20 years FLASH and EEPROM Data retention with zero failure for all its devices. That means as long as the, in the datasheet for the Tiny26 mentioned, 10,000 Write/Erase Cycles are not exceeded, the 20 years are guaranteed.

Embedded Dreams
One day, knowledge will replace money.

Top
sgomes
sgomes's picture
Level: Resident
Joined: Wed. Feb 19, 2003
Posts: 920 View posts
Posted by sgomes: Wed. Apr 25, 2007 - 02:44 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Interesting. So what is the guarantee? Will they replace the IC? What good would that do?

Go electric!
Happy electric car owner / builder

Top
lfmorrison
lfmorrison's picture
Level: Raving Lunatic
Joined: Wed. Dec 8, 2004
Posts: 4664 View posts
Location: Nova Scotia, Canada
Posted by lfmorrison: Wed. Apr 25, 2007 - 02:50 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I think that by "guarantee" Atmel means that a reasonable Engineer should feel secure in allowing an AVR to be used in situations where the product specifications require a certain operational lifetime of 20 years or less.

I would look at it in the same light as Atmel's "guarantee" that an ATmega128L will function at up to 8 MHz from 2.7V up to 5.5V, whereas Atmel "guarantees" that an ATmega128 will function at up to 16 MHz from 4.5V up to 5.5V.

Sure, some ATmega128L's might function at 10 MHz some of the time given the right supply voltage, and some ATmega128's might function at 4.0V some of the time given a slow enough operating frequency. But Atmel won't make any "guarantees" about those cases.

Top
sgomes
sgomes's picture
Level: Resident
Joined: Wed. Feb 19, 2003
Posts: 920 View posts
Posted by sgomes: Wed. Apr 25, 2007 - 03:56 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

True but in that case if I had a product that was not able to run at xMHz at yVolts then I could get Atmel on the line and get some engineer's attention. In the case of a Flash failure 20 years from now, what then? Why not say 50 years! I guess my point is the same problem I have with all those "bathtub curve" predictions. It might give you a warm comfy feeling but I wouldn't base any hard engineering decisions on it.

Go electric!
Happy electric car owner / builder

Top
Nuno
Nuno's picture
Level: Resident
Joined: Sat. Nov 1, 2003
Posts: 687 View posts
Location: Amadora, Portugal
Posted by Nuno: Wed. Apr 25, 2007 - 04:39 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Have you tested an AVR over temperature range? For how long would you test one at, let's say, 65ºC, to make sure they'll still be working next year? How do you know that they don't degrade to the point of failure after 1 year at this temperature? Or after 5? What would you do in 5 years if one failed on you?
They know their materials very well, they know their processes very well and they know the theory also very well. Then they do the math, do some "scaled-down" tests, and add a big tolerance factor. Then they give you the numbers! They're saying that the product will keep up with *at least* those numbers. Non-volatile memory isn't new, so the industry already has "many" years of experience. You'll have to thrust something!

Embedded Dreams
One day, knowledge will replace money.

Top
lfmorrison
lfmorrison's picture
Level: Raving Lunatic
Joined: Wed. Dec 8, 2004
Posts: 4664 View posts
Location: Nova Scotia, Canada
Posted by lfmorrison: Wed. Apr 25, 2007 - 05:09 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

sgomes wrote:
True but in that case if I had a product that was not able to run at xMHz at yVolts then I could get Atmel on the line and get some engineer's attention. In the case of a Flash failure 20 years from now, what then? Why not say 50 years! I guess my point is the same problem I have with all those "bathtub curve" predictions. It might give you a warm comfy feeling but I wouldn't base any hard engineering decisions on it.

By that argument, I guess, if the product needs to be still functioning for any non-zero timeframe, the only acceptable solution is to not use an electronic computer in the first place. Because every type of memory will cease functioning eventually.

Top
lfmorrison
lfmorrison's picture
Level: Raving Lunatic
Joined: Wed. Dec 8, 2004
Posts: 4664 View posts
Location: Nova Scotia, Canada
Posted by lfmorrison: Mon. Aug 6, 2007 - 11:53 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

FYI, the ATmega8 datasheet (at least) now contains a specific reference to Flash and EEPROM data retention times:

Quote:
– Write/Erase Cycles: 10,000 Flash/100,000 EEPROM (1)(3)
– Data retention: 20 years at 85°C/100 years at 25°C (2)(3)

Notes: 1. At 85°C. Guaranteed after last write cycle.
2. Failure rate less than 1 ppm.
3. Characterized through accelerated tests.

http://www.atmel.com/dyn/resourc...

Top
theusch
theusch's picture
Level: 10k+ Postman
Joined: Mon. Feb 19, 2001
Posts: 39512 View posts
Location: Wisconsin USA
Posted by theusch: Wed. Aug 15, 2007 - 07:30 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Newly-released Mega88-family sheet also. I'd guess that the decided to air that deep, dark, secret. ;)

You can put lipstick on a pig, but it is still a pig.

I've never met a pig I didn't like, as long as you have some salt and pepper.

Top