Using an ICL7660 to feed -5VDC to 4116 DRAMs?

[Ethan Dicks]

On Wed, Jul 02, 2008 at 08:14:32AM -0700, gil smith wrote:
> Hi Ethan:
> 
> I use the 7660 all the time -- it needs two 100uF caps -- works 
> great!  Can't draw a lot of current, but if the spec fits it is a great 
> solution.

Right.  I've used them before, but mostly for negative LCD bias voltage
for HD44780-type LCDs - they don't need much current at all (I think I
was using 10uF caps, but 100uF is no problem to find).
 
> There is a bit of switching noise (at about the switching freq of 5 to 10 
> KHz, IIRC), but it is usually not a problem.

Probably won't be an issue for this application.
 
> There is some loss, which goes up as you load the device;  ie:  if you 
> drive it from +5.0V and load it lightly, you will get close to -5.0V, but 
> as you draw more current it will drop down to almost -4.0V. 

Right.  That was what I was a little concerned by - 16 DRAMs do add up
to some measurable current, even at 100-200uA each.  What I wasn't sure
about is how quickly the ICL7660 drops off in a real application (vs a
nice clean curve on a spec sheet).

> You can 
> compensate for this by driving it with about +6V instead (eg: a 78L05 with 
> a diode or two stuck in the ground lead).

Hmm...  Hadn't thought of that trick; that might be handy if I need
a little boost.

> Another was to go is to use a real switcher like an LM2674M-5.0, but 
> configured for -5 instead of +5 -- I use this chip all the time too for 
> both pos and neg 5V -- very efficient.  Drive from unreg volage (eg: +12 to 
> +20 or so); you also need a 100uH inductor, a schottky diode, and a few 
> caps.  Pricier, but can efficiently source up to an amp.

I've seen mini-switchers like that in the past, 5 lead T-220 package,
IIRC, but I don't happen to have any handy, and I wouldn't be able to
get an order from Digikey for nearly 4 months in any case.  I have
to use what's on hand.

Tonight, in preparation, I've been scraping off the broken wire stubs
and such from a previous owner's attempt to hack more memory onto it.
It looks like he mostly added external wires and only cut two or three
traces, but after I hit the old flux with some cleaner, the board should
look nearly good as new.

One thing I hope to find on funet/zimmers.net tomorrow when the 'net
comes up is a copy of the diag ROM at $9000 that's part of the offical
C= diagnostic dongle - I have one at home, a 40-pin clip for the CPU
that you clamp on, then press reset, then unclamp - inside is a '138
decoder and a 2716 EPROM - according to the docs, the clip forces the
CPU to execute at $9xxx not $Fxxx so it can pick up the ROM contents,
then it copies itself somewhere (zero page?) and performs some
rudimentary diagnostics.  It's not as good as plugging in a Fluke 9010A,
but for a machine that mostly works (CPU, low mem, video, decode logic...)
it's a quick way to test things.  Fortunately, there's room in the
RAM/ROM board for several sets of ROMs (it uses a 29F010 FLASH with
a DIP switch and a PAL), so even if I don't find the official C=
diagnostic pod code, I'm sure I can bang something out pretty quickly
and drop it into the FLASH instead of the official ROM code at $Fxxx.

If anyone has any comments on 4116 voltage "gotchas", I'd still love
to hear about them.

Thanks for the reply, Gil,

-ethan
	
-- 
Ethan Dicks, A-333-S     Current South Pole Weather at  2-Jul-2008 at 15:30 Z
South Pole Station
PSC 468 Box 400       Temp  -79.1 F (-61.7 C)   Windchill  -114.0 F (-81.1 C)
APO AP 96598          Wind    8.7 kts Grid 33   Barometer 680.4 mb (10613 ft)

Ethan.Dicks at usap.gov            http://penguincentral.com/penguincentral.html