Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Q: 22v10 GAL for address latch/chip select on uC system

173 views
Skip to first unread message

Peter J. Kerrigan

unread,
Feb 6, 1996, 3:00:00 AM2/6/96
to
I'm going to do a 8031 project with external RAM/ROM/Memory Mapped Stuff.
Instead of using the '373/'138 combo for address latch/chip select, why
couldn't I do it all with a 22v10 or 26v12 GAL? Other than the expense
of the part, and the hassle programming it, why would I not *want* to do it?

This is for a hobby project, so production costs/etc don't apply.
--
Peter J. Kerrigan
p...@mcs.com

Mike Diack

unread,
Feb 7, 1996, 3:00:00 AM2/7/96
to
In message <<4f8d5b$o...@Venus.mcs.com>> p...@MCS.COM writes:
>I'm going to do a 8031 project with external RAM/ROM/Memory Mapped
> Instead of using the '373/'138 combo for address latch/chip select, why
> couldn't I do it all with a 22v10 or 26v12 GAL? Other than the expense
> of the part, and the hassle programming it, why would I not *want* to do it?
>
A fine idea - only 3 good reasons that i can think of for not doing it
current consumption, Current Consumption & CURRENT CONSUMPTION
what the world needs now is a 22V10 that is as economical as a HC573
M

Falstaff

unread,
Feb 7, 1996, 3:00:00 AM2/7/96
to
p...@MCS.COM (Peter J. Kerrigan) writes:

>I'm going to do a 8031 project with external RAM/ROM/Memory Mapped Stuff.


>Instead of using the '373/'138 combo for address latch/chip select, why
>couldn't I do it all with a 22v10 or 26v12 GAL? Other than the expense
>of the part, and the hassle programming it, why would I not *want* to do it?

>This is for a hobby project, so production costs/etc don't apply.

Perhaps power drain -- a 22V10 may use up to 200mA for some versions,
while a HC373/HC138 could use less that 1mA (depending on the speed you
run them at).

Frank
--
"Life without a backbone is hardly worth consideration."
------------------------------------------------------------------------
Frank A. Vorstenbosch +31-(70)-355 5241 fals...@xs4all.nl

Rene Putinier

unread,
Feb 8, 1996, 3:00:00 AM2/8/96
to
If I remember correctly, the 8031 is multiplexing its lower 8 address
and data lines (this is what the 373 is for?). If so, even with a
26v12, since you are using eight lines for this with only four outputs
left for decode. If you don't need more than four, seems like this
would be OK. The 22v10 probably wouldn't cut it since you only have
two outputs left.


p...@MCS.COM (Peter J. Kerrigan) wrote:

>I'm going to do a 8031 project with external RAM/ROM/Memory Mapped Stuff.
>Instead of using the '373/'138 combo for address latch/chip select, why
>couldn't I do it all with a 22v10 or 26v12 GAL? Other than the expense
>of the part, and the hassle programming it, why would I not *want* to do it?

>This is for a hobby project, so production costs/etc don't apply.

Gary A. Crowell Sr.

unread,
Feb 8, 1996, 3:00:00 AM2/8/96
to
p...@MCS.COM (Peter J. Kerrigan) wrote:

>I'm going to do a 8031 project with external RAM/ROM/Memory Mapped Stuff.
>Instead of using the '373/'138 combo for address latch/chip select, why
>couldn't I do it all with a 22v10 or 26v12 GAL? Other than the expense
>of the part, and the hassle programming it, why would I not *want* to do it?

>This is for a hobby project, so production costs/etc don't apply.
>--
>Peter J. Kerrigan
>p...@mcs.com


You might want to compare the power consumption, if that's an issue.

GC


Richard Steven Walz

unread,
Feb 8, 1996, 3:00:00 AM2/8/96
to
In article <4f8d5b$o...@Venus.mcs.com>, Peter J. Kerrigan <p...@MCS.COM> wrote:
>I'm going to do a 8031 project with external RAM/ROM/Memory Mapped Stuff.
>Instead of using the '373/'138 combo for address latch/chip select, why
>couldn't I do it all with a 22v10 or 26v12 GAL? Other than the expense
>of the part, and the hassle programming it, why would I not *want* to do it?
>
>This is for a hobby project, so production costs/etc don't apply.
>--
>Peter J. Kerrigan
>p...@mcs.com
-----------------------------------
Not sure if there are enough macrocells to do it, but haven't tallied
it up or tried to either. If you have the code, go for it. No reason
not to! They are fast enough and about the same price in quantity as
the glue TTL anyway. Tell us how you do and maybe publish your
codelist!
-Steve
--
-Steve Walz rst...@armory.com ftp://ftp.armory.com/pub/user/rstevew
mirrored: ftp://ieee.cas.uc.edu/pub/electronics/faqs/ftp.armory.com
and Europe: ftp://ftp.cised.unina.it/pub/electronics/ftp.armory.com
and Oz: ftp://ftp.peninsula.apana.org.au/pub/electronics/ftp.armory.com

Mark Zenier

unread,
Feb 8, 1996, 3:00:00 AM2/8/96
to
in <4f8d5b$o...@Venus.mcs.com>, Peter J. Kerrigan wrote:
: I'm going to do a 8031 project with external RAM/ROM/Memory Mapped Stuff.
: Instead of using the '373/'138 combo for address latch/chip select, why
: couldn't I do it all with a 22v10 or 26v12 GAL? Other than the expense
: of the part, and the hassle programming it, why would I not *want* to do it?

Because the 22v10 is edge triggered and the 373 is a level triggered
transparent latch. With the '373, the addresses can pass through during
the time the strobe signal is high and be valid before the trailing edge
of the ALE (or AS if you're using a Motorola 6801,hc11). The addresses
aren't valid at the leading edge of ALE, so if you use an edge triggered
part, you'd have to wait until the trailing edge, and loose that many
nanoseconds on your addressing time. Plus having to invert the ALE
signal.

I suppose you could implement asynch logic level triggered latches in
the PAL, and ignore the built in registers. The only problem then
would be that you couldn't tristate the outputs, as the 22v10 only
feeds back direct from the pin when using unregistered logic.

An = An*/ale + In*ale
An.oe = true

Mark Zenier mze...@eskimo.com mze...@netcom.com


Trevor Hall

unread,
Feb 9, 1996, 3:00:00 AM2/9/96
to
In message <<4f8d5b$o...@Venus.mcs.com>> p...@MCS.COM writes:
>I'm going to do a 8031 project with external RAM/ROM/Memory Mapped
> Instead of using the '373/'138 combo for address latch/chip select, why
> couldn't I do it all with a 22v10 or 26v12 GAL? Other than the expense
> of the part, and the hassle programming it, why would I not *want* to do it?
>
Apart from cost you cannot make a reliable, glitch free transparent latch unless you
consume extra product terms (and hence pins) :-

_____
data------------| AND |
-----------| |----- ____
| |_____| |----| |
| |OR |------DATAOUT
LE ---|>O- ------|____| |
| INV | |
| | |
| BUF ______ | |
-|>-------| AND | | |
------| |--- |
| |_____| |
---------------------------------

Assume data and dataout are high, LE transitions high to low. Because the delay of
LE to the two AND gates is not the same a glitch can occur on data out
On the low to high transition a failure to latch can occur.

T.H.


Gray Creager

unread,
Feb 9, 1996, 3:00:00 AM2/9/96
to Peter J. Kerrigan
Peter J. Kerrigan wrote:
>
> I'm going to do a 8031 project with external RAM/ROM/Memory Mapped Stuff.

> Instead of using the '373/'138 combo for address latch/chip select, why
> couldn't I do it all with a 22v10 or 26v12 GAL? Other than the expense
> of the part, and the hassle programming it, why would I not *want* to do it?
>
> This is for a hobby project, so production costs/etc don't apply.
> --
> Peter J. Kerrigan
> p...@mcs.com

here's a partial alternative...

go to Xicor's website and take a look at the X88C64 or the X88C75
devices. The 373 latching and 138 decoding functions have been brought
on-board the EEPROM memory.

Gray Creager
Applications Engineer
Xicor, Inc.
http://www.xicor.com

http://www.xicor.com/xicor/pdf/x88c64.pdf
http://www.xicor.com/xicor/pdf/x88c75.pdf

to get the PDF datasheets directly... :-)

Mastrojj

unread,
Feb 17, 1996, 3:00:00 AM2/17/96
to
Other than power constraints, I see no reason for not doing it. If power
is a question, check out the Atmel AT22V10BQ or BL parts for low power
consumption. They even offer a free Atmel-CUPL version compiler. See
your area rep.

John Mastropietro
FAE Milgray Electronics

pro...@login.dknet.dk

unread,
Feb 21, 1996, 3:00:00 AM2/21/96
to
p...@MCS.COM (Peter J. Kerrigan) wrote:

>I'm going to do a 8031 project with external RAM/ROM/Memory Mapped Stuff.
>Instead of using the '373/'138 combo for address latch/chip select, why
>couldn't I do it all with a 22v10 or 26v12 GAL? Other than the expense
>of the part, and the hassle programming it, why would I not *want* to do it?

>This is for a hobby project, so production costs/etc don't apply.
>--
>Peter J. Kerrigan
>p...@mcs.com

At my EMbedded Systems class, we have some small 8031 boards, which
uses a 16v10 for address-decoding. It's very easy to make changes to
the system, as you can reprogram the PAL in a snap.
I don't see any reason, why you shouldn't use a PAL for it. It needed,
I can e-mail you some of the scripts we used for the system.

Christian.
pro...@login.dknet.dk


Stan Eker

unread,
Feb 24, 1996, 3:00:00 AM2/24/96
to
: p...@MCS.COM (Peter J. Kerrigan) wrote:

: >I'm going to do a 8031 project with external RAM/ROM/Memory Mapped Stuff.
: >Instead of using the '373/'138 combo for address latch/chip select, why
: >couldn't I do it all with a 22v10 or 26v12 GAL? Other than the expense
: >of the part, and the hassle programming it, why would I not *want* to do it?

: >This is for a hobby project, so production costs/etc don't apply.


Only one reason: if you have a heavy capacitive load on P0, and/or poor
grounding and bypassing, then having a TTL-level latch (the GAL) is a good
way to avoid sleeping at night. Try it, and look for a positive glitch on
ALE just as /PSEN, /RD or /WR fall. If you have the glitch, don't use a PAL
unless it has CMOS input thresholds.

It won't happen in all designs. In fact, it doesn't happen in MOST designs,
but it's something to be aware of. Look at Note 1 in most 8031 data sheets.

Peter J. Kerrigan

unread,
Feb 25, 1996, 3:00:00 AM2/25/96
to
Many thanks to the many knowledgeable people (including some of THE most
knowledgeable here) who responded to my post.

To summarize, most people warned me the current drain for a GAL would be
two orders of magnitude higher than standard HC parts.

My outlook on the problem changed entirely when a kind gentleman from
Xicor told me about their 8x8K EEPROM (88x64?) which has the latch/decode
(and a serial bootloader) built in. All it took then was an 'hc00 to decode
the rest of the chip selects, and I was done.

See http://www.xicor.com for details on the Xicor parts.

Neil McGann

unread,
Feb 29, 1996, 3:00:00 AM2/29/96
to
p...@MCS.COM (Peter J. Kerrigan) wrote:

>I'm going to do a 8031 project with external RAM/ROM/Memory Mapped Stuff.
>Instead of using the '373/'138 combo for address latch/chip select, why

>Other than the expense
>of the part, and the hassle programming it, why would I not *want* to do it?

Power consumption? A lot of PALs/EPLDs are power-hungry and would blow your
power budget out of the water unless you use a "low-power" version (note that
this is relatively low, not absolute). Check out the "zero power" parts, if you
need really low power.

Generally, if you can afford the power, go with the PAL, it makes your life
easier when you want to change things later.

Neil


--
Neil McGann
email: webste...@dial.pipex.com
WWW: http://dspace.dial.pipex.com/town/square/ae331/


0 new messages