Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!uwm.edu!mailrus!cornell!uw-beaver!uw-entropy!dataio!aez
From: aez@Data-IO.COM (Adam Zilinskas)
Newsgroups: sci.electronics
Subject: Re: PAL's vs gates
Message-ID: <2320@dataio.Data-IO.COM>
Date: 5 Feb 90 23:00:53 GMT
References: <358@poppy.warwick.ac.uk> <8590@ingr.com>
Reply-To: aez@Data-IO.COM (Adam Zilinskas)
Organization: Data I/O Corporation; Redmond, WA
Lines: 80

>>     I'm thinking about building a board for my A500, however some slightly
>>involved logic is required. Is there any intrinsic benfit from using PAL's
>>rather than separate logic chips - apart from the obvious board looking
>>nicer/reduced chip count. I have no facilities for programming pals and so do
>>not want to go to the trouble if they have no advantages - other than
>>mentioned above, which I can live with.
>

(please note that I come from a biased position, I make software for 
programming PALs, PICs, FPGAs and whatever alphabet soup chip that needs
logic synthesis).

Reasons for using PALs:

1. Inventory. Many of the larger PIC (Programmable IC) chips can emulate 
  any TTL functions that you could use in a design. Its your choice to
  keep several 22V10's around or have 3-4 7400, 2-3 74163, 2-3 74151 ...

2. Board changes (only matters when making PCB, only slightly matters
  if doing wirewrap). A new PIC and be programmed and patch a circuit 
  within hours/minutes compared to cutting traces on boards and patch 
  wiring around. (this really depends upon how big is the screwup, changing
  a logic function is easy, getting a signal to a chip that was not there
  befor still needs a patch wire, but maybe no new chips).

3. Speed. Unless you have TTL functions that exactly meet your needs, you
  have to build composite circuits, having signals pass through multiple
  levels of logic and going between chips slows things down. Most PIC chips
  do their functions with 2-level logic (typically ORs of ANDs with some
  inverters, there are exceptions) and the going rate on the fastest PIC
  and like 2-7 nsec for a function, you win even more if the signal uses
  internal feedback and doesn't have to go through those slower pin buffers.

4. Density. 14->18 pin chips are fine, but when a design takes 10 small
  chips or 2-3 20 pin PICs, the PICs are cheaper in board real-estate.

Reasons for not using PALs:

1. You have to program them. For productions lines of making thousands of 
  anything, you want stuff in bulk. Each PIC has to be individually programmed
  and inserted into their right spot. Many companies [like mine :-) ] provide
  gang programmers to make this easier, the programming step essentially
  replaces the problem of getting the right TTL part out of inventory.

2. You have to program them. Most of the programming languages for PICs are
  more like computer programs and some hardware designers don't know how
  to program. Its one thing to understand a 7400 2 input NAND and another 
  to figure out how to use the macrocell in a 22V10 to implement one 
  bit of a shift register. (I am not ragging down on all hardware engineers,
  its just that the difference between PICs and "glue logic" (74xx parts) is
  akin to the difference between vacuum tubes and transistors, some engineers
  make the transistion, some do not, engineering darwinism :-)

3. You have to program them. A hobbyist can go to Radio Crack and get glue 
  logic. Getting a PIC, getting the software to program it, then getting the
  box to program the chip is not at RS (at least not now :-). You can go 
  to some distributers like Hamilton Avnet and if you talk to them nicely,
  they may let you use their equipment. For a hobbyist (I define that as
  anybody who is not making a living off their design, or can't afford the 
  $$ for the equpiment so they still can't make a living off their design), I
  suggest that you become good friends with somebody who has a PIC programmer,
  you may be able to get to use his software and hardware.

4. Seat of the pants debugging falls by the wayside. When your circuit is
  packed into a 20 pin chip (18 points to put a scope probe on), just slapping
  together a design and changing it until it works becomes near impossible.
  You have to DESIGN (nasty thought %-), most PIC programs provide some
  sort of simulator that helps but in general, you have to do the right
  thing for the right thing to happen!

I have tried to keep out any overt advertisements about what Data I/O provides
compared to our competitors. If you ask, I can give you our toll free number
and you may be able to ask for literature (and maybe demos) for both the 
hardware and software
we provide. For our competitors, look them up in the various engineering
magazines.	

				Adam Zilinskas
				Data I/O corp Logic Synthesis
				Where blowing a fuse is sometimes a good thing.