Maker Pro
Maker Pro

Corrupt Flash Memory Discussion

E

Eric L.

Jan 1, 1970
0
Before I get into the nitty-gritty of the problem, I would like to know the
consensus on whether Flash Memory can be corrupted by anything other than
cosmic rays, nuclear detonation, and the occasional ray gun. All kidding
aside, does anyone have any information on EMI, and/or RFI causing bits to
flip?

Best regards,
elarson
 
E

Elvis Presley

Jan 1, 1970
0
Eric L. said:
Before I get into the nitty-gritty of the problem, I would like to know the
consensus on whether Flash Memory can be corrupted by anything other than
cosmic rays, nuclear detonation, and the occasional ray gun. All kidding
aside, does anyone have any information on EMI, and/or RFI causing bits to
flip?

Best regards,
elarson
I bet if you put a Flash Memory chip in liquid helium some of the bits would
drop out, and maybe
-20 F.. I don't know, but it could happen if it gets cold or hot enough.

I think you can exclude EMI wavelengths longer than 4 times the longest edge
of
the device from considiration as having no effect whatsoever.
That means radio waves below microwaves would have to be extremely powerful
to effect the device. Put the chip in a microwave, turn it on, then
test.... and if my guess is right, there will be problems. I think it
would probably turn out that you have to hit the chip-level electronics with
ultraviolet light to have any hope of flipping bits.. and gamma radiation
would be more effective.

If the memory is inside a metal case, it would be better off for
reliability.

If I were on a jury however, I would conclude that bit-flipping is too
far-fetched a defense.

the wavelength of the offending EMI has to be shorter than four times the
length of the structure it interacts with, to interact with it
substantially.. furthermore it has to result in enough 'kick', the rush of
phonons from the strike spot to
cause enough electron excitation to bump the voltage enough to cause
the bit to change and still the circuit has to be susceptible to register
this change.. so its an
unlikely proposition even if you try... but if you try a few billion times
as in a radioactive minute you might have some luck.

RFI is EMI limited to the RF bands.

If you record 1,000,000 flash memories then drop them all, one at a time, 3
feet onto the pavement, 50 of them will have permanent errors from breakage.

If you record 1,000,000 flash memories then read them back, one or two will
have non-repeatable errors... thats just the way it goes (without recourse
to voodoo or microsoft).
If that's true it's a one in a million chance for just bad luck with 'bit
retention'.

Another thing is ageing. Since the 'decay' of chemistry within the chip is
a function of time..it would stand to reason that with every passing second
flash memory is more vulnerable to failure.
After 100 years there is a slim chance that much data could still be
intact. 0 will still be zero, but adjusting for
inflation, 1 will be around 27 (which screws up binary), and gasoline will
be $421 per milliliter
(synthetic) - converted from yen since the dollar won't exist, having since
surpassed
itself by it's value as heating fuel and toilet paper.

hope that helps
elvis
 
E

Eric L.

Jan 1, 1970
0
Elvis Presley said:
I bet if you put a Flash Memory chip in liquid helium some of the bits
would
drop out, and maybe
-20 F.. I don't know, but it could happen if it gets cold or hot enough.

I think you can exclude EMI wavelengths longer than 4 times the longest
edge
of
the device from considiration as having no effect whatsoever.
That means radio waves below microwaves would have to be extremely
powerful
to effect the device. Put the chip in a microwave, turn it on, then
test.... and if my guess is right, there will be problems. I think it
would probably turn out that you have to hit the chip-level electronics
with
ultraviolet light to have any hope of flipping bits.. and gamma radiation
would be more effective.

If the memory is inside a metal case, it would be better off for
reliability.

If I were on a jury however, I would conclude that bit-flipping is too
far-fetched a defense.

the wavelength of the offending EMI has to be shorter than four times the
length of the structure it interacts with, to interact with it
substantially.. furthermore it has to result in enough 'kick', the rush of
phonons from the strike spot to
cause enough electron excitation to bump the voltage enough to cause
the bit to change and still the circuit has to be susceptible to register
this change.. so its an
unlikely proposition even if you try... but if you try a few billion times
as in a radioactive minute you might have some luck.

RFI is EMI limited to the RF bands.

If you record 1,000,000 flash memories then drop them all, one at a time,
3
feet onto the pavement, 50 of them will have permanent errors from
breakage.

If you record 1,000,000 flash memories then read them back, one or two
will
have non-repeatable errors... thats just the way it goes (without recourse
to voodoo or microsoft).
If that's true it's a one in a million chance for just bad luck with
'bit
retention'.

Another thing is ageing. Since the 'decay' of chemistry within the chip
is
a function of time..it would stand to reason that with every passing
second
flash memory is more vulnerable to failure.
After 100 years there is a slim chance that much data could still be
intact. 0 will still be zero, but adjusting for
inflation, 1 will be around 27 (which screws up binary), and gasoline will
be $421 per milliliter
(synthetic) - converted from yen since the dollar won't exist, having
since
surpassed
itself by it's value as heating fuel and toilet paper.

hope that helps
elvis
Thank you for your comments.

I would like to direct your attention to the center ring:
http://www.eng-tips.com/viewthread.cfm?qid=155287

I have been having a discussion with a number of engineers over a special
little topic. I think you will find this most interesting. I would be
interested in what you have to say.

elarson
 
D

Dan H

Jan 1, 1970
0
Eric said:
Before I get into the nitty-gritty of the problem, I would like to know the
consensus on whether Flash Memory can be corrupted by anything other than
cosmic rays, nuclear detonation, and the occasional ray gun. All kidding
aside, does anyone have any information on EMI, and/or RFI causing bits to
flip?

Best regards,
elarson

I have seen effects when a system was being tested for EMI/RFI
immunity.

At certain frequencies that microproc oscillator was affected, confused
the micro and cause false data to be written to memory chip.

Dan
 
E

Eric L.

Jan 1, 1970
0
Thanks for your input. What is the recommended test setup for testing for
EMI/RFI?

elarson
 
D

Dan H

Jan 1, 1970
0
Eric said:
Thanks for your input. What is the recommended test setup for testing for
EMI/RFI?

elarson

It depends on the application and what the legal requirements are - for
example

If the product is to be sold in the European Union there detailed
specifications as to how the tests must be done. These require the
equipment be tested in a special built facility with special test
equipment to do the tests.

If you just want a confidence check, you can take a hand help VHF
transmitter, hold the antenna about 6 inches from your product while it
is running and key it on and off while tuned to different frequencies
and observe the product operation for unusual events.

To do a crude test for immunity to electrical spikes wire the contact
of a relay in series with the power to the relay so that it acts as a
buzzer. Connect a wire from one end of the relay coil connected to the
relay contact thru a capacitor and then connect it one at a time to the
power leads and the input and output leads of the product. Monitoring
for abnormal operation of the product.

Without careful design few products will pass even the above tests the
first time they are applied.

And of course once you have a product which is immune to outside
influence then you need to run test to see if it is emitting unwanted
RF to the air or on the power lines. More special equipment and
facilities required.

Hope this helps

Dan
 
Top