PAL users and buyers read on.

I am recommending that Karen Bjornquest stock LATTICE GALS instead of PALs for
most applications in engineering. It may be beneficial to also use them in
production instead of PALs.

A GAL is an electrically erasable multifunction PAL. The G stands for 
GENERIC.
Basically there are three GALs : GAL16V8, GAL20V8 and GAL39V18. The GAL16V8
will accept the fusemap for a PAL16L8 and most other PALs in the 20 pin
series. They can be programmed with our current software and hardware. The
DATA I/O configures the GAL to look like the PAL you intended. You just enter
the GAL family pin/code. (The JEDEC file has the pal type you intended). The
GAL20V8 is a generic form of the 24 pin PAL series (8 outputs only). The
GAL39V18 is a generic form of the PAL22V10 and IFPLs (Signetics 82S153 etc.)
and includes user selectable latches or flipflops on the inputs and outputs
and 8 buried state registers. 

If you want the change the logic in a GAL, simply take the same device back
to the programmer and program it again. 

Available speeds are 10nS, 15nS and 25nS for the GAL16V8,
		15nS and 25nS for the GAL20V8 and 30nS for the GAL39V18

Costs / availability
Jerry Greiner, sales manager for Lattice and Joan Alpen from Merit 
Electronics (distributors) indicated that Atari could buy GALS for production 
(at least the GAL16V8-25) at the current going rate for PALs of $0.95 even 
though the the "street" price is around $1.40 
The Atari special on GAL20V8-25 would be around $1.60, (street price $2.25)
The GAL39V18 (soon to be renamed GAL6001) has a current price of around $9.00 
moving towards $6.00 by the end of the year.
For production purposes these parts can be programmed at the distributor, and 
should we need to change the program, or over-order a part, they can be 
returned to the distributor for reprogramming.

They have a definite cost advantage for engineering prototype development, 
since they are reprogrammable, and since we buy our PALs for engineering use 
at anything up to $2.50 at FRYs. Instead of stocking 16L8s, 16P8s, 16R8s, 
16R6s, 16R4s and 16RP8s, we could just stock 16V8s. 


Footnote. 
Lattice are developing a 2000-3000 gate LCA (XILINX look-alike) that will use
EEPROM technology for configuration instead of RAM as Xilinx uses. 

The DATA I/O logic programmer is now upgraded to version 9.1 (it was at
version 2.0). It can program GALs from Lattice, National, and apparently
several other companies making GALs (I didn't know that there were others), 
as well as a bunch of other newer programmable logic types.
We have 500 GALs in stock in engineering (25nS 16V8 variety).
A new wall chart is up showing family/pin codes.

ABEL should arrive next week. 

					Pat.

If you are trying to program a GAL using a PAL jedec file (this is called
cross programming) you must supply the Data I/O with the RAL family pin 
code of the device you are emulating. (RAL is reconfigurable array logic.) 
This tells the Data I/O what configuration to use.

e.g. You have a Lattice GAL. You have a jedec file for a PAL16L8.

Give the Data I/O programmer the Lattice 16l8 RAL f/p code of 36/17 instead of 
the Lattice GAL f/p code of 36/55. That is all that is required.

The new chart does not give RAL numbers, but the old one does and the 
manual does.

						Pat.

Watch those GALs...at the RACES.

Three times now I have come up against timing problems with GALs on
production game boards. The problem ? The GALs are too fast compared to
pals. I am comparing 25nS GALs to 25nS PALs. 

The senario is that a PAL is used to generate some timing signals. In all
cases so far the development boards used PALs because we hadn't started
using GALs yet. Come production release time we call out GALs because thats
what we do these days. 
Then manufacturing calls in with problems. (Meantime the responsible
engineers have vanished.) 
The problem is caused because the design relied upon a minimum delay
through the PAL. But the GALs are faster than the PALs and so are not
providing the same delay. So we have a classic race condition. 

It is worth noting that the problems were voltage dependent. This is most 
likely due to the fact that GALs are CMOS devices, while PALs are bipolar 
devices. (Remember when CMOS used to be SLOWER than bipolar.)

Bottom line. Test your GAL circuits at 5.25 volts cold and 4.75 volts hot 
to check for race conditions at both ends of the performance spectrum.

								Pat.

I hesitate to send this message since, considering how the universe works,
I'm probably cursing myself to be bitten by a race condition. HOWEVER...

	The implications of Pat's message was that it was O.K. to depend on
minimum delays as long as you test your circuit with the same parts as used
in production. (Pat- I realize you may not have meant to convey that 
impression.) I believe that unless a minimum delay is SPECIFICALLY mentioned
in the data book it is a grave mistake to rely on delays. It is especially
easy to fall into this trap here at Atari since we all know that tolerating
some "benign" race conditions can save some money in the form of registers
you don't need to resynchronize your signals. Tolerating races is addictive
and slowly erodes your standards of what constitutes good design the more
times you get away with it. I know I'm running this into the ground but please


	JUST SAY NO!   REGISTER TODAY!