| ??? 10/19/03 04:32 Read: times |
#56860 - Sure memory? -> Statistics! Responding to: ???'s previous message |
Manfred wrote:
I found: 20 years bits are sure in memory, it is ok. Christian wrote: The retention time is more then 20 years ( but you can expect the twice ). You both are wrong! When treating 'data retention' and 'endurance' you must think in terms of statistics. So, you don't have any guarantee, that your 'bits are sure in memory for 20 years'. And it's wrong to say, that 'you can expect the twice', namely 40 years. Data retention is defined as following: Take an extreme (theoretically infinite) number of identical chips, all being programmed with some information in memory. Then, after months and years flow by, more and more chips undergoe a memory failure, means, at least one memory cell of chip cannot keep it's information any longer. When a certain percentage of chips (which has to be clearly defined!) has shown such failure you stop the 'clock'. The time needed for reaching this percentage of failures is your 'data retention' time. Three points additionally to note: 1. Consequently statistically spoken also this testing procedure must be repeated for an extreme large (theoretically infinite) number of times. The average taken of all these tests will then yield final 'data retention' time. 2. Data retention highly depends on crystal temperature of chip. So, also testing temperature must be clearly defined. At high temperatures data retention time heavily decreases. How much you will find in corresponding reliability report. Arrhenius equation is the keyword here. 3. Chi square distribution shows, that you can extremely manipulate actual 'data retention' time, when slightly changing this 'percentage of failures'. You see, 'data retention' is a statistical phrase. It doesn't tell you anything concrete about YOUR chip, but only about an infinite number of chips. You can be lucky and YOUR certain chip keeps it's information for 20 years, or 40 years, or even 17903 years. But may be your chip shows the first failure already in 10 years, or 5 years, or even today evening! All this can happen, although your chip follows defined behaviour of datasheet, namley showing a data retention time of 20 years!! Critical applications like aeronautics, spacecrafts, etc. must always be prepared for such accidential memory loss. Especially if environmental temperature is higher than about 40°C. 'Space shuttle' tries to solve this problem by using an array of 5 independent computers. And I remember a certain flight, when 4 of them failed because of a fabrication problem... Kai |
| Topic | Author | Date |
| sure memory | 01/01/70 00:00 | |
| RE: sure memory | 01/01/70 00:00 | |
| RE: sure memory | 01/01/70 00:00 | |
| RE: sure memory | 01/01/70 00:00 | |
| RE: sure memory | 01/01/70 00:00 | |
| Sure memory? -> Statistics! | 01/01/70 00:00 | |
| RE: Sure memory? -> Statistics! | 01/01/70 00:00 | |
| RE: Sure memory? -> Statistics! | 01/01/70 00:00 | |
| RE: Sure memory? -> Statistics! | 01/01/70 00:00 | |
| RE: Sure memory? -> Statistics! | 01/01/70 00:00 | |
| RE: Sure memory? -> Statistics! | 01/01/70 00:00 | |
RE: Sure memory? -> Statistics! | 01/01/70 00:00 | |
| RE: sure memory | 01/01/70 00:00 | |
| RE: sure memory | 01/01/70 00:00 |



