Mike's Hobby Blog

preview my next (video) podcast to see first ever power on and the SCELBI equivalent of “hello world”

http://www.willegal.net/podcast2-2-13.mp4

Here is the SCELBI input card – ready to go. The SCELBI instructions were fine. The 65 10K ohm resistors take a while to solder in, but otherwise it is a very straight forward process. It draws around 200 milliamps in static, standalone condition.

Next up is 1 bank of 8 SRAM sockets on an SRAM card. Soon after that, I’ll be manipulating toggle switches!

Since the SCELBI was sold as assembled, kits or bare boards, what you do about sockets is largely up to you. However based on observations of original SCELBI mini-computers, here are some suggestions to make your SCELBI more authentic.

Suggestions for the Front Panel, CPU, DBB and Input Boards:
Many of the boards that I’ve seen, use the Molex strip connectors mentioned in an earlier post for all ICs on the board. SCELBI instructions only mention using these for the 8008 and soldering the rest of the chips directly to the PCB. I have seen one original system set up this way. Since the 8008 is the only rather expensive chip used on these boards, I have chosen to follow instructions and solder all parts to the PCB, except the 8008. If I need to pull a 74xx part, I’ll cut the legs off and remove the legs individually. A friend of mine managed to obtain a roll of the MOLEX pins and I installed those for the 8008. They are rather fidgety to install, and I suspect reliability might be an issue. Even if you can find enough stock to do an entire system, I wouldn’t recommend doing a whole system with those MOLEX pins. I have also seen one 8B that used regular 70s style closed frame solder tail sockets. You should be able to find a bunch of older closed frame sockets at surplus dealers, if you want to socket your boards. I haven’t seen any SCELBIs with machine pin sockets.

Suggestions for the Memory Cards
Though the instructions don’t call for sockets, all the memory cards I have seen, use either MOLEX strip sockets or old style closed frame sockets. Apparently memory was expensive enough and perhaps failed often enough in those days that soldering in memory was not considered wise.

Example Closed Frame Socket Used on SCELBI Memory Board.

I believe this one is made by SCANBE. SCANBE’s were also used in arcade machines and have a terrible reliability according to arcade system restorers. I’ve seen one page that suggests replacing all SCANBE sockets before proceeding further.

More on Molex Strip Sockets
Here is a page from an old (1973) MOLEX M-100 catalog

Following instructions in the manual, I wired the front panel switches. I also added a temporary jack for power, until I get something more authentic.

With this setup, once I complete the memory and input boards, I should be able to toggle in a tiny program. Actually I may be able to toggle some instructions directly into the CPU, even without the memory card.

The push button switches should be normally off, with one terminal connected to ground and the other to the control signal. Pull up resistors on the CPU board keep the signal high, until grounded. The toggle switches are connected in a similar fashion. Connect the center terminal to either ground or the data bus input signal and the other to the terminal such that when the switch is in the “up” position, pull up resistors on the input board keep the signal high. When the switch is in the “down” position, the signal should be connected to ground.

Here it is resting on top of an iPad (which is on top of a powerbook).

Only thing left to do is to install the +5 volt bus workaround jumper between resistors R37 and R38 and add the 6.3 volt zerner. For some reason the +5 volt jumper wire isn’t mentioned in the SCELBI instructions. I got a bad batch of Zerners, so I’ll have to install that later on. It draws about a 1/2 amp when applying +5 volts to this board without the rest of the system connected.

Except for the omission of the jumper, the standard SCELBI instructions seem accurate for the DBB board.

In case you didn’t remember from my previous post, here is where the +5V jumper goes.

Prototype Chassis with backplane installed – top view

This is the result of some hacking of a backplane into a BUD AC413 chassis. The AC413 is the same width (12″) and depth (10″) as an original SCELBI chassis. However the BUD AC413 is 3 inches tall, while the original SCELBI chassis 3.5″ tall. During this prototyping/hacking effort, I found out a few important things.

Cutting aluminum is not hard, but it’s best to drill holes with a jig or at minimum use a punch to mark the center of the holes in the aluminum. I eventually used one of the scrap backplanes from the first delivery to create a jig for drilling holes.

On a “real” SCELBI chassis, the cutout for the edge connectors fits very tightly around the perimeter of the edge connectors. Making the cutout any larger than necessary, makes it very difficult to locate the support screws on the sides of the backplane. The edge connectors extend 8 7/8″ left to right and 5 5/8″ front to back. The backplane is 9.5″ wide and if the hole is made 9″ wide, that leaves only 1/4″ overlap per side to drill a 9/64″ hole for the #6 mounting screw. If possible, it would be best to cut the hole for an exact 8 7/8″ side to side fit. Though it doesn’t look like it, my prototype left too much space between the edge connectors and the chassis, making fitting the side mount screws a challenge.

The cutout starts 1/2″ from the front of the chassis and is centered between the edges.

A view from the bottom.

Screws holding the edge connectors/card guides have both a washer and star washer, but no lock tight on the original SCELBIs. I don’t have card guides and have used 1/2″ #6 machine screws.

Screws holding the backplane to the chassis have neither a washer or a star washer, but a held in place with red Locktight on original SCELBIs. I haven’t used Locktight, because I want to find shorter screws and I have more work to do on the chassis with the toggle switches and I/O connectors. The 3/8″ screws are a bit too long on my reproduction. A 1/4″ screw might be a better match to the original.

The screws at the front of the backplane do no go through holes in the backplane, but the edge of the nuts hold the backplane in place.

When soldering the edge connectors, only the pins with connections and pads are soldered.

One more thing – I positioned one of the holes for the edge connector mounting screws 1/16″ too far back. This hole will need to be drilled out to properly fit a #6 screw. However the error is so small that the enlarged hole cannot be seen under the nut of the screw or the ear of the connector.

An insulated stiffener running down the center of the board between the connectors would have been a good addition to this design.

Figured I’d share a picture while I still had a pile of them

update: this clock setup is incorrect!, the second clock starts too late and last too long. It must finish before the first clock rises.

The clocks are adjusted to run at 500 KHZ. Note that the build manual mentions that clocks are supposed to be self starting, but may not start with some user power supplies. Sure enough, I’m seeing this issue with my bench supply. It is easy to kick start by shorting both ends of electrolytic cap C9 (located near trim pots). Note that adjusting these clocks takes a bit of twiddling.

Here it is, without the CPU socket, which will be made up of old fashioned Molex strips of pins. 7400 series parts are soldered directly to the PCB, without sockets. I will be socketing SRAM parts on the SRAM board, when I get to that.

I’ve been following the old SCELBI documentation which can be found here:
http://www.olson-ndt.com/Scelbi/Schematics2/

The documentation is actually quite good, which saves me a ton of effort in writing my own. However, I’ll share a few additional tips here.

The CPU card contains both 7404 and 74L04 parts. Don’t mix them up like I did.

There are 8 1K resistors

Actual boards that I have seen, use 3K instead of 3.3K resistors

The 5K trimming resistors have the pin that is offset from the other 2, offset by 10 mills. Bourns offers two spacings, either 20 mills or 10 mills. Somehow, though I knew the difference, I ended up with the wrong version. I was able to bend the pin under to fit, for now, until I am able to replace them with the correct part.

The are 8 1N914 diodes and 2 zener diodes. The document only mentions 9 diodes.

I’m using modern 6.2 volt and 12 volt, 5 watt zener diodes for the over voltage protection circuit. Actually, I don’t have the 6.2 volt ones on hand, at the moment, so they will be installed later. The leads on the 12 volt zeners are too big for the holes, so I drilled out the hole with a 3/64″ bit mounted in a pin vise. I drilled in a bit in from each side, rather than straight through from one side, so I didn’t risk delaminating the pads. This effectively removes the plating, so I soldered the end that had a trace on the top side on both sides of the PCB. This isn’t noticeable.

After soldering, I clean the flux from the board with isopropyl alcohol.

I obtained parts from http://unicornelectronics.com, http://www.onlinecomponents.com and http://www.mouser.com. It is possible that Unicorn may be able to supply complete parts kits in the future, as I am going to provide them a complete BOM, once I get the prototype working.