PDP-8 /e/f/m memory

[Don]

Allison wrote:
>> Subject: Re: PDP-8 /e/f/m memory
>>   From: Don <THX1138 at dakotacom.net>
>>   Date: Wed, 16 Aug 2006 11:41:47 -0700
>>     To: General Discussion: On-Topic and Off-Topic Posts <cctalk at classiccmp.org>
>>
>>> I thought I'd said that.  Bit of work history, engineer and product 
>>> engineer for a semi company that sold micros and ram.
>> Work history doesn't help list readers since they can only read
>> what you've *written* (except for those few clairvoyants sitting
>> in the back row...)  :>
> 
> Most will not build unless supplied as a kit for $49.95.

Presumably, (*some*) people may still be *interested*
or want to *learn* something -- even if they are not
going to *build* it.

>> You neglected to note that Icc grading *between* CMOS static
>> RAMs AT DC varies SUBSTANTIALLY.  Power consumption at operating
>> frequency isn't an issue (here).  Rather, the difference between
>> Icc(standby) on a "regular" 62256 (e.g.) and a "low power"
>> 62256.
> 
> Yes, I know.  Does anyone care?

Anyone who is thinking about the problem (e.g., "miniature
PDP-8" thread) may.

>> You can buy even LARGER devices (e.g. > 1MB -- B not b)
>> that will idle at *2* uA.
>>
>> I.e. the data retention time of a well designed BBSRAM
>> circuit *is* limited to the shelf life of the battery
>> powering it (during standby).  *But*, only if you select
>> the right grade SRAM.
> 
> At 2ua a 3V 30mAH Li cell has a life of how many years?
> At 20ua a 3V 30 mAH Li cell has a life of how many months?
> At 2ma using two AA alkaline cells (Duracell) in hundreds of hours?

The point was, design a good BBSRAM circuit, put an off-the-shelf
"consumer battery" on it and it will last the shelf life of
the battery.

Keep that item out of nasty environments -- which would also
be harmful, long term, to your '8 (or whatever) and you have
a nice, simple, solution.

>> In my reread of your comments, I don't see *that* mentioned.  :>
>> (much of the 32KB devices you'll find in PC's and their ilk
>> are not chosen for this very low Icc(standby) -- *especially*
>> cache RAM!)
> 
> Doesnt matter!  if battery life were an issue I'd have brought 
> it up.  However I specifically metnioned in another post that 
> battery life or even battery backup was not an issue.

Sorry, I didn't see the sign that proclaimed "This is Allison's
post; only replies that she deems acceptable are accepted here" :>
*Other* people may be interested in this, even if you are not.

My apologies if your thread has become intertwined with the
"miniature PDP-8" thread.  But apparently there *are* people
interested in different approaches to semiconductor memory
in a PDP-8 -- or mini-PDP-8.

> FYI: nearly all the previous commercial designs the ram array
> was usually 2102s and even the LP version at DC was both heat 
> and power intensive. Their power drain was measured in amps
> for the array (96 2102s for one board).  B elieve it or not
> they did provide for backup  (using large SLAs).
> 
> I'd point out that I have a bag full and some are very low power
> I measured several for this and got less than 1ua at 3v and room
> temp, outputs floating inputs grounded CS/ and OE/ negated.  You 
> forget guarenteed specs vs whats likely supplied as yeild is often

I only design for worst-case.  The numbers I have previously
cited are actually *worse* than worst case.  But, they vary
greatly from manufacturer to manufacturer.  E.g., Cypress's
super low power 62256's idle at 1/20th the power of Hyundai's.

If you are building "one off" and can afford to select parts,
then perhaps you could chose to pick one manufacturer's
parts over another.  If I were designing a *kit* and
expected some volume of sales, I would opt for the real
worst case, vendor independant approach which means some
folks would get pleasant surprises, others would get what
they paid for.

But, we already know *you* aren't interested in this.  :>

> to the better part.  That was only the CMOS cache rams pulled 
> from old 486 boards of late generation (green). I have a very deep
> new parts stock with a tubes of graded 62256 and other 
> low power rams as well as CMOS logic.

Great!  Build it however *you* want!  Since it is *yours*.

>> To put things in perspective, a 1F supercap charged *nominally*
>> to 5V (use care here since supercaps typically don't have much
>> margin for overcharge  :>) would discharge to 2V (the typical
>> data retention voltage of a CMOS SRAM) in just 3000 *seconds*
>> (less than an hour) with a 1mA retention current.  This could
>> be extended to ~30hours using 25uA devices.  Using a *2*uA
>> device can extend this to 2.5 weeks...
> 
> Read that as a battery would do far better. 

I mentioned the numbers for a supercap as a delayed reply
to someone's *explicit* question about it.  It was a great
way to put the numbers "in perspective".

Good luck with your design!