Lattice PLD's:
Are CMOS and are basically a giant PAL. They draw only a small amount of current, 0.08A. Each board has one Lattice PLD on it. The PLD does all of the address decoding, address selection, latches, buffers, outputs, etc. for each board. I choose them because the development kit is inexpensive, Lattice gives them away, in-circuit-programmable, parts are around $7.00 each and are flash, therefore reusable in other projects. Parts used are rated at 80MHz.

Additional Information about NoCan and NoCan2 (upgraded info)
NoCan is a catch-all board. It has the 2-megs interface, 1 serial port - 6551, 1 parallel port - 6821, 4MHz circuit with on/off switch connection. This board is designed to go over the top of the metal RF can. It installs similar to the RamZilla board, it uses IC1, IC3, CN4, CN5 and CN6 for the connections.This puts it kinda high inside of a coco-3 case, and may be better for a re-pack. You could leave out the serial and parallel stuff and just use the 2-megs, that would be a choice. The serial and parallel ports are set to be in one of four address ranges, $FF10-F, $FF30-F, $FF60-F or $FF70-F. First you select, with jumpers, one of those previous ranges. You then select which chip gets one of four inside of one of those address ranges. For instance, if you select range $FF60-F, then the seial port could be one of the four addresses, $FF60-3, $FF64-7, $FF68-B or $FF6C-F. The parallel port will be only one of the three leftover addresses of $FF6X. You can't put both chips on the same address by standard jumpering. This board needs 3 connections as flying leads, HSync & VSync (IC15, pins 13 and 11), and the front of (keyboard side) R63 (28MHz).

NoCan2 Rev E1 has a 16550 serial port instead of the 6551 and a different address selection scheme. The coco-3 power supply does handle this board with everything installed but the heatsink gets really hot! Needs a FAN. The parallel port is set to be in one of four address ranges, $FF10-F, $FF30-F, $FF60-F or $FF70-F. First you select, with jumpers, one of those previous ranges. You then select which chip gets one of four inside of one of those address ranges. For instance, if you select range $FF70-F, then the parallel port could be one of the four addresses, $FF70-3, $FF74-7, $FF78-B or $FF7C-F. The serial port (16550) will be addresses of $FF1X or $FF3X with another jumper selecting which half, i.e. $FF10-7 or $FF18-F. I have four prototype boards that all run and boot NitrOS-9 at the "4MHz" setting. Two are currently spoken for and the other two I keep. Since this board has errors, I have gone to the next revision, E2.

The included "4MHz" circuit may not work on all coco-3's, although I've been using it daily, booting NitrOS-9 from a floppy at "4MHz". It doesn't work with 68B09E's, but does with 63B09E or 63C09E's. It possibly won't work at all with the "32MHz Crystal Hack". If it did work, the 63x09E would then be running at about 5.33MHz burst. A small circuit board has been added to one of my coco-3's so that a DIP-14 oscillator clock can be added to check out various frequencies, will give datails later. The original coco-3's 28.63636MHz clock is used to generate the "4MHz" and actually comes out to 4.77MHz and has Q and E at 66% overlap. Creating the "4MHz" was easy to do it using the coco's clock, because it keeps the signals all in "sync". The Lattice chip will handle 80MHz input, so the slug is the CPU. The "4MHz" circuit actually puts out a single Q & E burst between system Q & E's, but only when memory is not going to be used next.

NoCan2 Rev E2 is the next one to be produced, stay tuned.

Additional information about IdeZilla and Pocket-IDE
IdeZilla is an IDE interface to coco's 1 2 or 3. Pocket IDE is for coco-3's only, although it will work on other coco's if you have the correct connection. Pocket-IDE requires a custom installation. You can use NitrOS9, OS9 or any other operating system you choose, even RS-DOS. However, RS-DOS doesn't know about IDE and you'll have to create your own driver, or get one from a website. This board can be set-up to look like (software-wise) the Glenside IDE board addressing, with maybe the exception of the alternate IDE registers. It has a standard TIN card edge connector. The board can be modified to be "remote" by cutting loose the front header part of the circuit board, (between the two headers) and installing any length (up to 8") 40 pin cable. I used an IDE cable for this extension because there are plenty of them to go around and they are cheap and already have 40 pins. Also, if you choose to implement the "bottom connector" scheme, then this board is ready to go by connecting to the second header on this board and ignoring the card edge front end header.

FAQ:

1. What is this board compatible with?
    A few folks have asked how fast this board is and is it compatible with "what" drives. So far, it's compatible with any ATA drive, including older drives that are CHS mode only. It even works with those Kittyhawk drives, I have two. How fast is a function of how fast your driver or OS is. The Lattice PLD is a 80MHz part, so I don't expect that any coco can get anywhere near that speed. For myself, I have had megaread times (TIME;MEGAREAD </DD@;TIME) of approximately 13-16 seconds (even shorter with "4MHz"), BUT --  the drive had a driver that only used 256 bytes of every sector, ignoring the upper 8 bits of every IDE word. Megaread times of approximately 25-30 seconds are common (for me) under NitrOS9 v200n using a 512-byte driver.

2. How much does this board cost?
    A. The chip is about $7.00. I program it. Available through Allied, at a $25.00 min charge.
    B. So far, the cost is related to the actual board itself. A place in Canada makes prototype style boards without soldermask or silkscreen. Each board is roughly about $32. 4 boards, each is about $21. For 8 boards, each is about $15. But, you must know that these are prototyping prices. Production board prices are always lower.
    C. Lattice PLCC-84 sockets can be gotten from Digikey as well as the 20x2 headers (3 needed) and the extra parts, resistors, caps, LED's, etc. Ask for a complete parts list.

2. When is this board available?
    For now, it is not. It depends on how much interest there is. So far about 5-6 folks are interested.

Additional information about RamZilla
This circuit board can be used for OS9 or NitrOS9, that's what I do. You must replace IC1 and IC3 with sockets plus add a 2-pin header to IC15, that's all of the mods needed to the motherboard. After that, it's just a plug in memory board. You may need to cut loose two caps (C10, C11) on the motherboard for RAS and CAS, and maybe two more caps (C65, C66) for E and Q, depends on your motherboard. Mine, I cut all four, in addition to those above mods. My coco-3 is standard, no other chips replaced, not needed, especially the three 74ls244's and the 74ls374 in the data path to the GIME. Because I use 3-chip 1-meg SIMMs, the actual load is less than a (16 chip) 512k board. Actually the same load as a 128k board, only having 4 chips total, or 2 chips per SIMM. Named by: John Strong of StrongWare. John actually has one of the early RamZilla PLD's. How's it going John, is it built yet?

Also, a portion of SockMaster's 4MHz circuit (with his permission) is now inside of the RamZilla PLD. It's actually about 4.77MHz, so not all 6809E's will work. There is a switch header, so that the user can turn it on/off. This way a coco could be running at a slightly increased speed and have 2-Megs, with a serial port. One more flying lead must be added, the 28MHz oscillator output, at the front end of R63 next to the GIME chip. This would "splice in" an additional cycle (fast mode only) in the time when E and Q are both low, according to SockMaster's specifications about memory accesses.

FAQ:

1. Does this board work everytime that I turn my coco on ?
    So far, everytime that I run NitrOS-9, it works. Haven't had any problems with it not running from power-up. Don't have to hit "RESET" to get things to work.

2. How much does this board cost ?
    The cost is related to the Cypress SRAM, Lattice PLD and the circuit board. For now, there's two versions, one with the serial port, one without. It seems that folks want the serial port. Cost is about $80.56 for two boards. But, you must know that these are prototyping prices. Production board prices are always lower. Also, there's a few special sockets, one for IC1 and another for IC3. What is needed is a really tall socket that will fit through the holes in the board and reach down to where the socket is on the coco, and plug into that socket. Mill-Max makes such  parts (www.mill-max.com). Otherwise, the rest is standard sockets, etc. The chips hardest to get is the Cypress SRAM, Cy7c122-15, the replacement is the Cy7c150-15 and is available. the board accepts either chip. The Lattice part, which I program, is in stock at the local distributors, Allied for instance. I currently do not plan to provide the 6551(A)'s, or the MAX232 chips or the 1.8432 MHz oscillator can.

3. When is this board available ?
    Whenever there's enough interest. I think there's only two or three interested. Maybe the addition of SockMaster's 4MHz circuit may help.

Additional information about CoCoZilla
CoCoZilla, recently named by Bill Clemons, was created around 1985 or so. Didn’t have the bucks to buy a real one, but had the tools and services to create one from scratch, which was at the time, cheaper. Procured a custom wirewrap board and socket pins for this project. Next was to create a schematic that could be yellow-markered as I went along to mark progress. Started with the coco-3 and added full buffering. Then the pseudo-MPI was added. Didn’t want the “switch” but wanted functionality, so the MPI got full buffering also. Then the next challenge was how to physically address the MPI bus separate from the coco bus. I used 20L8 pals for that job, as I wanted the $FF1X and $FF3X ranges added to the MPI. At the time, the 2-megs circuit was already in progress and since that would be a nice addition, started conversing with Kevin Darling for his knowledge of the same. With that running, next began the idea of mass storage. First to be added was the floppy interface, worked right away, amazing. I used to think on how nice it would be to have IDE drives instead of those MFM drives, but started on the MFM’s, as they were available. Came up with a WDXT PC circuit board, added an IBM XT slot to cocozilla along with another PAL for system addressing of peripheral equipment. I kept the idea of IDE for later. Then came OS9, something that was purchased from a Tandy. Then got the Development Pak from Alan Dages. Upgraded the cocozilla memory from 512 to 1 megs, everything worked so far. Just this (1998) year, a chassis was found that could take that wirewrap board and all of the MFM, IDE, floppy drives that could be thrown at it, and upgraded to 2-megs. I later found out that it was an Intel Magellan chassis, from the designer. The chassis has room for as many as 8 drives of any kind, 4 visible and 4 inside.

   Even more amazing is that in the past, I had layed out this board in 2-layer style, and it fit into a PC-XT board size. So that it could drop it into any slant-front PC chassis. The MPI section was unfortunately straight up, and part of the board. I toyed with the idea of making the slots on a cable to be mounted on the inside of the left of the chassis. with the electronics on the main board. Now with surface mountable parts, this whole thing could be in a mini-tower. Check out PaPaZilla.

Information on PaPaZilla
PaPaZilla is another coco-3, EXCEPT it is a fully buffered CPU. Has full address decoding for all of the I/O space, with headers for hackers. A second system-EPROM socket, jumper selected. A 6821 wired as a parallel port. A 6551 wired as a serial port and am considering changing that to a 16550. The 2-megs (RamZilla) circuit, uses 2 each 1-meg SIMM's. Separate +5 voltage regulators, one for this board, one for the game-slot. Uses the standard keyboard 6821 connection and general location. Has pins brought out for the 4 unused pins of the other 6821, user definable. The keyboard and other 6821 are re-mapped so that the shadowed places are missing. For instance, the keyboard 6821 appears at $FF00-$FF03 and $FF0C-$FF0F. The other 6821 appears at $FF20-$FF23 and $FF24-$FF27. This allows the second 6821 space to be used along with the hole in the keyboard 6821 mapping. There's several 74F139's to get to these new (now un-used) address blocks.
 

Paul T. Barton
  1