Hot DRAM

[Syd Rumpo]

Apologies for the non-political post.

Does anyone know if DRAM will work at high temperatures, say above
150'C, hopefully 180'C (300F - 360F)? I know it's out of spec, and can
accept a reduced lifetime and speed, but will it work?

Cheers

 

[PeterD]

Apologies for the non-political post.

Does anyone know if DRAM will work at high temperatures, say above
150'C, hopefully 180'C (300F - 360F)? I know it's out of spec, and can
accept a reduced lifetime and speed, but will it work?

Cheers

I do a fair amount of work with 'hot' components (automotive engine
management modules), and personally I don't think this will fly. We find
that longer term exposure to 100C will cause failures in as little as a
few months, and you're talking 50 to 70% higher.

I suppose if this is a testing environment, just to see if it might
work, you may be successful.

Oh, and I've not tried DRAM at high temps, so the above comment may well
be worthless to you!

 

[Spehro Pefhany]

Apologies for the non-political post.

Does anyone know if DRAM will work at high temperatures, say above
150'C, hopefully 180'C (300F - 360F)? I know it's out of spec, and can
accept a reduced lifetime and speed, but will it work?

Cheers

If you scale the typical 64msec refresh by a doubling every 10°C to
180°C you get a refresh every 88usec, or a row every ~10nsec. Maybe if
you spent half the time doing refreshes, ran a little cooler and did a
lot of testing you might get close.

I don't think I've seen DRAM offered by the guys who sell the high
temperature qualified parts, just SRAM.

Ordinary parts also typically have high current densities in wee
narrow conductors so they can die by electromigration in a matter of
months at ~200°C.


Best regards,
Spehro Pefhany

 

[Syd Rumpo]

If you scale the typical 64msec refresh by a doubling every 10°C to
180°C you get a refresh every 88usec, or a row every ~10nsec. Maybe if
you spent half the time doing refreshes, ran a little cooler and did a
lot of testing you might get close.

I don't think I've seen DRAM offered by the guys who sell the high
temperature qualified parts, just SRAM.

Ordinary parts also typically have high current densities in wee
narrow conductors so they can die by electromigration in a matter of
months at ~200°C.


Best regards,
Spehro Pefhany

Thanks all for your opinions.

Yes, I was thinking to up the refresh rate and doubling every 10'C is in
line with Arrhenius. As for months, that would be ok, it's only 180'C
for hours or days at a time. It looks like I might have to test some -
my feeling is that the sense amplifiers will be the issue.

The very high temp SRAMs are amusing - Digikey has the HT6256 32KiB
225'C part for USD 537.50 each. That's three hundred and forty quid - a
penny a byte.

Cheers

 

[Rich Grise]

Thanks all for your opinions.

Yes, I was thinking to up the refresh rate and doubling every 10'C is in
line with Arrhenius. As for months, that would be ok, it's only 180'C
for hours or days at a time. It looks like I might have to test some -
my feeling is that the sense amplifiers will be the issue.

The very high temp SRAMs are amusing - Digikey has the HT6256 32KiB
225'C part for USD 537.50 each. That's three hundred and forty quid - a
penny a byte.

At first, I wondered, "A hot dram of what?" ;-)

I wonder if the surplus shops have all sent their core planes to the
landfill? ;-) Wouldn't a core memory work almost up to the Curie point
of the ferrite?

And whatever happened to magnetic bubbles?

Thanks,
Rich

 

[Syd Rumpo]

On 20/09/2011 21:30, Rich Grise wrote:

At first, I wondered, "A hot dram of what?" ;-)

I wonder if the surplus shops have all sent their core planes to the
landfill? ;-) Wouldn't a core memory work almost up to the Curie point
of the ferrite?

And whatever happened to magnetic bubbles?

Thanks,
Rich

Hot whiskey - an equal measure of whiskey and boiling water with a few
cloves, a squeeze of fresh lemon juice and a dollop of honey. Good for
cold days.

Core would be even more than a penny a byte, so a few megs would be
beyond my means and probably bigger than a big thing during
international big-things week.

As for bubbles, I remember going to a seminar on those. IIRC they were
256KiB devices and quite large - a good idea but too late, too expensive
and soon overtaken.

Cheers

 

[Nico Coesel]

No. Diode leakage will destroy the information stored on diode
capacitors on each memory cell. The refresh happens too slowly.

Well, refresh faster! DDR2 memory typically has a normal temperature
and high temperature self-refresh mode.

 

[John S]

At first, I wondered, "A hot dram of what?" ;-)

And I thought he misspelled Damn.