Re: Reset Memory Pattern

From: Ojala Pasi 'Albert' (albert_at_cs.tut.fi)
Date: 2003-06-14 01:06:28

> transistors). Traditional SRAM cells are made up of only 6 transistors,

True, I was oversimplifying.

> > SRAM is 
> > more stable even when you cut the power (unless you are in 
> > .18-micron with high transistor leakage currents).
> 
> This is nonsense. The contents of true SRAM is lost as soon as the power
> switches off.

If you cut the power supply, it does not immediately suck out the charge
from all of the the transistors unless you access the cells.


> When the power does not drop under this level, the contents can be read
> back after raising the power back to its nominal value.] SRAM is very
> low-power though, because the active loop is static and does not consume
> power, other than some leakage. Leakage is very small, because when the
> level of the SRAM cell does not change, there is just some charge kept
> on the insulated gate of the transistors.

This is what I was aiming at. Something I omitted: I was talking about
on-chip SRAM. You can cut the actual supply voltage of the SRAM block
as long as you don't access it. Apply the power before access.
(Some current may seep through to the SRAM from control signals etc.)
Of course this is just for fun, I would not use this for production
software.

> > This is because the neighboring cells usually have different 
> > polarities. The cell itself 'resets' to the same state, but 
> > the column-read logic inverts every other read line.
> 
> I would love to see some proof of this :)

Which one? The architecture or the reset state?

The SRAM blocks our company is using for ASIC's do not use inverted cells,
but obviously the architecture using inverted cells has been widely used.
(And then there is a simple line twisting to reduce the interference also.
 The sense (read) lines are quite long in large memories, so every bit helps.)

The knowledge of the architecture is vital in designing a proper memory
test. With inverted cells patterns like $aa and $55 do not detect shorts
between cells because of the inversion, you have to use $ff and $00.
Of course shorts are just one failure type.

-Pasi
-- 
/He wanted to weep. The well-being of Pandora had always been crucial to his
 own survival. He did not need to be near her -- better that he was not near
 her -- but he had to know she was somewhere, and continuing, and that they
 might meet again./		-- Marius in "The Queen of the Damned"

       Message was sent through the cbm-hackers mailing list

Archive generated by hypermail pre-2.1.8.