Read/write limitations of Flash memory?

[Guybrush]

I remember reading that compactflash could only be read/written to so many times, like the MTBF stat for hard drives. What is it, and is it true for other forms of flash? Anyone know what the "best" is in terms of lasting the longest? Thanks,

 

[blahblah99]

I'm not sure about other components, but the PIC microcontroller's I've used specs out the rewrites to about 10,000 times.

 

[Peter]

Flash memory cells survive around 10000 to 100000 write cycles. Better drive controllers count and balance sector usage to stretch the life cycle. Cheaper ones, where e.g. directory sectors get written all the time while those further back aren't being used at all, die much earlier.
SanDisk for example claim they have the most intelligent algorithm for this. They've been in the FlashDrive business long before it became a commodity, so if true, it'd not be very surprising.

 

[dejitaru]

How do the memory cells die?

 

[Eskimo]

Flash memory utilizes a dual gate technology. The bottom gate is the storage gate while the top gate is the control gate. Without going into too many specifics the storage gate is used to store electrons within it, even after power has been shut off to the device. In order to get the electrons into the storage gate a phenomena known as Fowler Nordheim tunelling (FN) is used. This is a quantum mechanical mechanism that occurs when an applied gate bias is sufficient to bend the conduction band of the substrate towards the gate oxide to form a triangular quantum well. This well serves to accelerate the electrons to a point where they can physically transverse the oxide barrier between the channel and the storage gate.

Unfortunately FN tunneling is destructive to the critical tunnel oxide which seperates the channel from the storage gate. The highly energized electrons have sufficient kinetic energy to break some of the silicon-oxygen bonds within the oxide. Where this occurs point defects in the lattice can form which are energetically favorable sites for positive trapped charges to reside. Eventually these positive charges will reach from the channel to the gate and form a conductive path which shorts your insulating tunnel ox layer.

AMD has a great flash tutorial on how Flash (heh) works. Which will better explain the concept to you. Should be able to find it HERE
Click on the section called device operation

 

[Evadman]

I was thinking about going back to college and getting a degree in EE, but I think my brain just melted after reading Eskimo's post.

 

[Eskimo]

Originally posted by: Evadman
I was thinking about going back to college and getting a degree in EE, but I think my brain just melted after reading Eskimo's post.

If it makes you feel any better I wasn't an EE, and I got C's in a lot of the EE courses I did take 😉.

 

[Evadman]

Originally posted by: Eskimo

Originally posted by: Evadman
I was thinking about going back to college and getting a degree in EE, but I think my brain just melted after reading Eskimo's post.

If it makes you feel any better I wasn't an EE, and I got C's in a lot of the EE courses I did take 😉.

well, normaly if I do not understand something, I just type it into google and find out what it means in simpler terms that I can understand. I do not think i can type entire paragraphs into Google though. I am dead set on understanding what you said though, it will just take me quite a bit longer than the average HT post. Thank you for posting it. My brain needs to be melted every once and a while to work 🙂

 

[Eskimo]

You might find http://smithsonianchips.si.edu/ice/cd/MEM96/TITLE.pdf useful. Click on section 10 for more on flash technology including some of the different types of Flash memory (NOR, NAND, MLC, etc..)

Also I should have mentioned that the more common method of putting electrons into the floating gate was channel hot electron injection. Basically exposing the electrons in the channel to a transverse electric field strong enough to sweep them up into and through the tunnel oxide. FN Tunneling is more common in the erase operation and CHE still dominates the programming although some Flash technologies use FN for both.

 

[Guybrush]

Hm, ok thanks. I was talking to a friend about making a notebook boot/run off a CF card instead of a hard drive (you can boot off a USB key if the board supports it).

Now i'm not so sure it's practical since 2.5" hard drives seem relatively light. Wonder how much heat you'd save off it. I guess there's also issues of data throughput. Was hoping you could at least save a lot of power though.

 

[Peter]

Notebook HDDs typically are 5V 500 mA (maximum, idle is much less), so there's not too much power to be saved. Besides, FlashROM drives typically are in the 1 to 3 MB/s speed range on reads, much slower on writes, so you really do NOT want to have one as the system drive to standard Windows.

If you want to save power by accessing the HDD less, give the thing lots of RAM.

 

[Howard]

Didn't Samsung invent flash memory that could be read/written an infinite number of times? IIRC...