TI 99/4a Mitsumi to Cherry MX Keyboard

When I purchased the TI 99/4a Computer a few years ago, the keyboard was not working properly. I had ended up with one of the Mitsumi based keyboards with the membrane that degrades. After taking it apart to separate the stuck membrane on the keyboard, it still wouldn’t work properly. The best option for me at the time was to get another TI 99/4a that had one of the other keyboard types in it as many of these Mitsumi membranes have failed.

There were other modern replacement keyboards for 80s era (well late 70s for the TI 99) computers, but there wasn’t one for the TI 99 that I had been able to find at the time. I looked over the Mitsumi keyboard, it is very similar to the Commodore 64 Mitsumi keyboard in construction. The plungers look to be identical to the Commodore 64 except they are white plastic instead of black. The “pads” on the bottom of the plungers are different though, as these work with the Membrane not with carbon pads like on the Commodore 64. The frame is the same style, and I expect it is possible the springs are the same or close enough to use. The Spacebar spring is not the same as the rest. There were projects for the Commodore 64, so it needed a PCB design, and maybe adapting the key “adapters” that were used for the C65. That was well beyond my current abilities (and probably still is).

I decided to buy another TI 99/4a with one of the “other types” of keyboards. I ended up with a TI 99/4a with a Hi-Tek brand Stackpole keyboard. The Stackpole uses square tubes that the keys insert into that tend to split at the corners. I do have some split tubes in my keyboard, but it was working fine. I did have concerns the stems would split worse with use. Shelby of the Tech Tangents (at the time still called AkBKukU) on Youtube worked on restoring another Stackpole type keyboard and released a 3d model to print new tubes for the keyboard. I printed one on my 3d printer and it turned out great. So there are options to extend the life of the Hi-Tek Stackpole keyboard. The keys may not work as good as new, but they should be reasonable.

With the keyboard issue sorted, I shelved the idea of coming up with a new Keyboard PCB. I didn’t like I had an otherwise completely functional TI 99/4a sitting unusable on the shelf though.

I recently came across a project to rebuild the Mitsumi 99 keyboard with Cherry MX Style mechanical key switches. https://github.com/visrealm/keyboard4a99 I watched the Youtube build guide. I have worked on various keyboards, and the project is quite strait forward. It is designed specifically to rebuild the TI 99/4a Mitsumi keyboard, which is what I needed. It reuses the keycaps, the spacebar supports and clips, as well as the keyboard cable and some support rails and screws.

I expect with a bit of work, that it could be adaptable to use the Hi-Tek/Stackpole keycaps. It would take a different Stackpole to Cherry MX Key Adapter (I found one, but I don’t know that key height would be correct https://www.thingiverse.com/thing:4735119) You would also require making some metal support rails. I believe the keyboard “cable” is shorter on the Hi-Tek, so you would probably be making a new one. The visrealm project has an alternate spacebar support clip, which may work with it, depending how the spacebar works.

He has the Bill of Materials in the Github post. As of right now.

We need the original Mitsumi type keyboard to get the Keycaps, and Metal support rails with screws. Then optionally the keyboard cable (which I used), and the Spacebar Clips to put into his inserts. There is an optional version of replacement Spacebar Clips, but there is no point in me leaving those on the old keyboard, the same with the cable it can be recreated but there is no point in that for me.

Then we need the Keyboard PCB. I downloaded the Gerbers from Github and uploaded them to JLCPCB. He recommends black to reduce seeing the PCB through any gaps between the keys, so I went with black.

We need 47 Cherry MX Compatible switches. I wanted Brown Switches, as they are my preferred type. I know there are now many other “colors”, but I don’t know anything about the colors other than red, blue and brown. I like a reasonably quiet switch, and I like the tactile feedback of the brown switches. I went with the cheapest option for me to get enough brown switches, which was to buy one of those 60% keyboards for $20. It was cheaper to buy that keyboard on sale than it was to buy 50 loose key switches.. I was sure to buy a keyboard with removable switches. It was a bit of a pity to strip the keyboard. It seems to be made well enough, but I don’t like the form factor. I will keep it for the few remaining switches, or incase one of my kids wants a keyboard of that style. It will be easy to repopulate it.

I needed a 8mm latching switch (compatible with the GPBX-800L at Mouser) for the Alpha lock, which again I went to Amazon, as I don’t have a need to place an order with Mouser. I now have 19 spares, and 20 of some three other variants. There is also a 1N4148 diode for the Alpha lock mod to keep the Alpha lock from letting the joysticks work. Which I have plenty of in my stock.

I used some M3 Brass inserts. It mentioned 10, where there are 3 for each of the “PCB mount rails”. I take it the other 4 are for the Spacebar support mount, but the ones to reuse the Mitsumi spacebar clips do not have the larger holes to accept the Brass inserts. I used M3x8 machine screws, for in the Brass Inserts as well as for the Spacebar mounts. The holes for the Spacebar mounts are for self-tapping screws, but the M3 machine screws went in fine, and actually were nearly too tight.

We need to 3d print the PCB mounts for each side, the Alpha lock, the PCB support backers (2), the Spacebar supports (2). The Keycap adapters (46?), there is a model that you can print them all at once, I did find them annoying to get apart but they were usable. There is also a unique Keycap adapter for the Spacebar (supposed to be a little different), and a unique one for the Right shift. I printed the parts all in PETG. I did make sure they were not brittle, during assembly and testing I had no issues with any breaking in my case. I did print “outer walls” first, this kept the cross on the top sized properly and kept it from stringing between the crosses.

I have to have the Mitsumi keyboard pulled from the TI 99. Which is 7 screws in the bottom to open the case. Then 2 to release the power supply board, 3 to release the main computer pcb and housing, and 4 screws to remove the keyboard itself after unplugging it from the computer pcb. The guide on Github shows all the screws and steps.

Next I striped down the Mitsumi keyboard. Desoldered the wires from the Key lock switch. Then I unscrewed the metal support rails putting them to the side to reuse later.

Starting here. First I will remove the metal rails to reuse. Then the remaining screws.

Then I took the rest of the screws out to release the PCB. Next I used my desoldering iron to remove the keyboard cable. Just a little note on the cable, “pin 1” is Not the Red side of the cable. Keep an eye on how it is installed and reinstall it in the direction it was on the original pcb.

While the PCB was off, I removed the spacebar and took out the Spacebar support clips.

Spacebar, Spacebar clips, and Keyboard cable. (and 8 key)

While I had the Mitsumi keyboard in storage, I had broken the 8 key stem. I had to get the broken stem top out of the 8 key without damaging the key. I used a 1/16th” drill bit to drill a hole in the broken stem top. Then I screwed a small screw into the 1/16th” hole, and used a pair of pliers to carefully pull the part out of the keycap. The screw was only screwed in a short distance, if it is too big, or in to far, you may wedge the part in tighter.

Getting the broken stem top out of the 8 key

I then took the screw out and used some super glue to glue the key stem back together, it may or may not hold up, but I also don’t expect the keyboard to get keycaps again.. After that I screwed the old keyboard back together. I will keep it for parts. If you don’t put the PCB back on, when the keycaps are pulled off the key stems will just fall out everywhere.

To start building up the new keyboard. I mostly followed the video guide. First I took the Metal support rails and added Kapton Tape to the bottom for insulation. The guide he used clear tape, and suggests maybe 2 layers. The Kapton Tape is much more durable, and is a proper insulation tape it you happen to have it. Note that the rails are “turned” compared to on the Mitsumi, using the “other holes”. I lined them up with the white lines on the PCB Silkscreen.

Once the rails are insulated they get installed with the 6 original screws, into the 3d printed support rails that are basically just to hold the screws.

Then I installed the Spacebar Supports. You can also see the PCB Mounts with their brass inserts installed. I was checking that they were going to fit properly. The PCB Mounts were added after all the key switches were soldered in.

I soldered on the keyboard cable next.

I started soldering in the switches. It is recommended on Github to get the key switches with the extra support pins in line with the center posts. The switches I purchased did not have those. It did make soldering them in strait a bit more difficult. I did work to solder them in strait as possible, which I did have some issues with. If they are perfectly strait, well the keycaps will still be crooked… well in my case they were. While I don’t know the reason for them being crooked, once installed, it has something to do with the 3d printed adapters.

To solder the switches, I started with the spacebar and went across the rows from there. I didn’t do it the way the guide showed. I am sure they would have been very crooked, maybe if I my switches had the additional alignment pins it would have been too bad. I ended up using Silly Putty to keep the keycaps in place and strait when I flipped the keyboard. With just one Silly Putty egg, I could do a full row. Though when I started I was using larger clumps of it as seen in the photo below.

The 0 key switch ran into the edge of the one 3d printed screw rail. I had to shave just a tiny bit off the rail to get it in strait.

So now the soldering is completed.

The soldered keyboard. The Spacebar Clips are installed. Tested the 8 Key.

So we did the key adapters. I printed out the grouped file. Then I had to cut them apart. I used Cura which did something weird to the one layer. Cura was reliable, but Orca indicated a better outcome, but I couldn’t get it to print successfully.. So It came down to using a new sharp Xacto blade to cut the adapters apart. When I had rough edges, I cleaned them up with the side cutters or Xacto blade.

The spacebar has a slightly different adapter. The Alpha Lock has a different type of cap. Then the Right Shift has a special centering double pin adapter.

So then I went through all of the keys separating the adapters and installing them. The key switches are quite strait, although not perfect. Though once installing the keycaps with the adapters they end up not aligned very well. They look to have printed well. I expect it is due to variations in the inside stem that goes into the Cherry MX type switches. The seam on the outside may have also affected them and made them rotate a bit. The stems are rectangular, so you have to be careful to not have them rotated 90 degrees, as they may “force on” but will bind. They may have been more consistent if I had been careful to not rotate them 180 degrees, but that was hard to tell.

Below is the completed unit ready to be installed. If you click on the front view, you can possibly see the P key is obviously twisted. The angle makes it hard to tell how numerous of the keys are twisted a bit.

Next I installed the keyboard. The keyboard is the lowest part of the 99 or rather the highest part. The keyboard must be installed before the Computer PCB and under the Power supply PCB. First I lifted the loose computer out of the way just enough to drop in the keyboard behind it, so that I could get the keyboard screws in properly. Then I lifted the computer up to plugged in the keyboard to the computer assembly. Then I screwed the computer and finally the power supply PCB. After that the bottom gets screwed back on. Finally the outer part of the Power Switch piece slides back into the case.

Finally the TI 99/4a has the new Cherry MX style replacement keyboard installed.

The P is kind of obvious from the angle again. There are other keys twisted a bit. I tested the keyboard before reassembly of the computer. All the keys work as expected including the Alpha Lock. The only issue, is the Enter key must have a clearance issue, it is partially press. I may pull the 99 apart to recheck it. It is only partially pressed, so it isn’t stuck down, and it still operate normally except not raising fully to the expected position.

As far as the project, it is great. The oddity is with the 3d printed key adapters. Others may have less problems with that and not have keys twisted noticeably. The key switches with the extra alignment pins may be a good idea and speed up assembly. The Alpha Lock switch doesn’t have “much” travel, it is not very obvious that it is pressed by looking at it. It does work fine.

I did have an issue with the Spacebar initially not wanting to raise up, it was binding a bit. I loosened the 3d printed supports and managed to get the alignment better. This may also be due to “twisting” related to how my adapters printed.

The keyboard feels great to type on. I was able to put the Hi-Tek Stackpole type keyboard back in my spare TI 99/4a making it complete again.

I hope to make some time to use the TI 99/4a. I think I’ll put away the Commodore 128 for a while so I can have the 99 hooked up and see if I can get familiar with it.

TI 99/4a 32k Memory Sidecar Alternate Memory IC Adapter

I recently built a Jedimatt 32k Memory Sidecar for my Ti 99/4a computer. https://hobbytronics.home.blog/2024/05/04/ti-99-4a-32k-memory-sidecar-build/
This was after having a Backbit 32k Memory Sidecar module that was not working properly.

When buying the parts for the Jedimatt 32k build, I purchased a new SRAM IC for the Backbit 32k unit. I replaced the ram IC on the Backbit 32k unit and it still wouldn’t work, reliably. That indicates some other fault on it, either the CPLD or somehow the Edge Connector on it not making proper connection. It will “randomly” work for a short time then quit again.

The Static Ram I purchased for the Jedimatt 32k unit uses a Low Power CMOS 28pin 256-Kbit 32k x 8 Static RAM that is TTL Compatible that runs at 2.7V to 5.5V. The Backbit 32k unit uses a CMOS 28pin 256-Kbit 32k x 8 Static RAM that is TTL Compatible that runs at 4.5V to 5.5V.

I looked at the Datasheets for the two of the Static RAM ics. With one being branded Alliance Memory and the other Cypress, although both list Alliance Memory Inc as the manufacturer. I looked over the pinouts, and they are pin compatible, short of the Address inputs are numbered differently. The primary difference is the footprint. The AS6C62256-55PCN that I purchased for the Jedimatt 32k build is a 28PIN 0.6″/15.24mm Width DIP Package. The CY62256NLL-55SNXIT that is on the Backbit 32k is a 28PIN 0.295″/7.5mm Width SOIC Package.

I purchased some SSOIC to DIP Adapter PCBs. I had assumed they would be sized to adapt to the 28PIN 0.6″/15.25mm Width. I was wrong, they are wider. They are also the narrower SOIC Footprint, which happens to be the same footprint on the Backbit 32k PCB. That meant the CY62256NLL already had the legs bent to go on the narrower footprint, which by the way, is no fun doing. I guess it would have been to easy to just solder the CY62256NLL-55SNXIT to the Adapter and put on the Round Pin headers.

I started with soldering the CY62256NLL-55SNXIT to the Adapter PCB. I then installed a row of Round Pin Headers in Pins 15-28. This is because of where it needs to fit on the Jedimatt 32k PCB. I next need to install a row of offset Round Pin for Pins 1-14.

I went back to the same Round Turn Pin Sockets that I had used on the Jedimatt 32k Edge Connector. I installed the Round Turn Pin Socket row into a Round 28Pin IC Socket. This is why I used the Round 28Pin IC Socket instead of the standard dual wipe 28Pin IC socket that I specified in the parts list for that build. I then inserted the adapter board above to find the spacing I needed to get everything aligned.

Once I had it held in the spare 28Pin Round Pin IC Socket, I took some Solid Copper Wire (a strand from some Cat 6 Solid Core Copper Network Cable) and stripped the insulation from it. I shaped it to fit into the Round Pin Socket. I made up a total of 14 pieces.

Above you see the little copper wires as fitted into the Round Pin Header. I then placed the PCB onto the wires to Hold them in proper alignment. Once they were held in place, I soldered them into the Round Pin Header Socket.

Once the copper wires were soldered, I installed the Adapter PCB into the spare IC Socket and soldered the wires into the PCB to get the proper alignment for everything.

With it upside down, I then made sure to have good soldering on the bottom to help the Copper Wires keep secure. The bottom of the PCB had Kapton Tape on it to keep the Copper wires from wearing into the solder mask and shorting over time. Now the adapter board is all soldered up. It was time to see if it will fit.

Below you can see it installed into the PCB. I guess now the Green PCB is fine as they look good together.

It is just barely up over the case. The Top Cover has a hollow in it though so it will still fit.

It was time to power it up and see if it works.

It seems to be stable and working properly. I am going to leave this CY62256NLL-55SNXIT on the adapter PCB installed into my Jedimatt 32k Memory Sidecar. I will put the 28PIN DIP Static Ram IC back in my spare parts. I have 4 more Jedimatt 32k PCBs that I could build up. I may still try to revive the Backbit 32k Memory Sidecar, but if I don’t, I could at least salvage the other CY62256NLL-55SNXIT and the Edge Connector to build another of the Jedimatt PCBs.

8/31/25 A little update on the Backbit 32k Memory Sidecar. I did buy a new CPLD for it a while ago, but before installing it, I decided to strip down the whole board. I removed the 5 Caps, the Edge Connector, the ram and the cpld. I cleaned the pcb. I reinstalled the Edge Connector flipped 180 degrees, I felt it potentially was a contact issue, and flipping the connector may make a difference. With the edge connector “pins” being flush with back holes on the pcb, it may have had a bad solder joint in one of the pins where the solder didn’t flow properly to the pin? I then soldered the Ram IC, CPLD and finally the capacitors back on. With practice I have been doing much better with soldering the CPLD. Also with good flux, and a new microscope for inspection. The primary issue with the board was solder bridges, some very small bridges kept hiding between the CPLD pins, but the new microscope made them much easier to see. After cleaning and reinspecting the pcb, it is working in the TI 99/4a just fine. I tested a couple 32k programs and they started fine, I have been running Jedimatt’s Expansion Memory Test Burnin as seen above for over 20 passes on it. I’m glad it is finally working reliably. This does mean I now have two 32k sidecars, for now I think I’ll use the Backbit one. I also have a spare CPLD, the primary reason I reinstalled the original one is I really don’t know how to program the new one, as to if I can use the same method I have used for the ZX Nuvo 128 board, or if I would need a different programmer. The board also does not have any traces going from the JTAG pins, meaning soldering on very small wires to very small pins to even attempt it. There may have been a bad solder connection or bridge, or it may be the edge connector in some way. I did find that the Andonstar 246S is a great upgrade over the little lcd microscope I had previously. It is much clearer and I could see the solder points and solder bridges that needed to be cleared. It has enough clearance to the microscope that I can use it to view the screen while soldering.

TI 99/4a 32k Memory Sidecar Build

I have had the TI 99/4a for awhile now, and posted the Recapping of it. I actually have two TI 99/4a computers as the first one had a bad membrane type keyboard. The second one had a Stackpole type keyboard, which is is working order (although some of the square tubes are split). The Joysticks are awful, and I built up a Joystick Adapter to plug in Atari 2600 compatible joysticks to it. That made things much better for it. I also built up a Pitfall Cartridge for it and 3d printed a shell. Still I haven’t used it much. The Ti 99/4a was the first computer I had, it was old at the time. I had a few cartridges for it, and used BASIC in it at the time. It was quite limited with all the more software I had available.

In anticipation of getting a FinalGROM Cartridge or Backbit or something, I had purchased a 32k Memory Expansion that was for sale from Backbit. After I received it, I designed a 3d printed case for it. It turned out really well and has been sitting with my TI 99/4a since then, but I was unable to use it at the time.

I like the small and clean look of the Backbit 32k.

The FinalGROM Cartridge came in, I made up an SD Card for it. I tested it with some games that did not require the 32k Memory Expansion. Then I plugged in the 32k Memory Expansion and the 99 wouldn’t power up properly, it just made an audio buzz. I checked the contacts, checked the Memory Expansion. I noticed flux residue on the ICs of the 32k Memory board, and cleaned that off just incase it became contaminated. I ended up removing the CPLD from the board, and inspecting it. I soldered back on the CPLD, and the buzz was gone, and I got it working just one time. Then I powered it off and tried another 32k program and it quit working again. I rechecked the soldering on the CPLD, as well as the ram IC and edge connector and touched it all up. It would randomly work, it may pass Jedimatt’s Burning Ram test for a few passes then fail. Then it generally won’t detect the 32k memory until it is powered off a bit and reseated.

Backbit has quit selling the 32k Memory Expansion, and has released on Github as a project to allow you to build your own.

The Backbit 32k is more compact, and should be cheaper to make than Jedimatt’s design if you have all the required equipment to program and assemble it. Jedimatt’s 32k which is available to build by ordering PCBs with the provided Gerber files and standard components. Jedimatt’s 32k requires no programming of components. There are a few issues with building the Backbit 32k. It has a CPLD and that requires programming with a JTAG programmer. The board doesn’t have a JTAG Header. I believe this is simply due to it not being designed as a DIY Project, Backbit is a business and she makes and sells her products. I later found the RAM IC Footprint is wrong, at least for the ram IC that was fitted on mine when I bought it. The CPLD is on the top edge of my ability to solder with the super fine pin pitch. The Backbit 32k has not been around nearly as long as Jeditmatt’s, I can’t be sure that the problem is a defective part. It may also just not like my specific Ti 99/4 for some reason.

Jeditmatt’s 32k design has been around for many years. It is all through hole style components and easier to solder together. Surface mount soldering is not that difficult with some practice. That is at least for the larger parts, the smaller pin pitch like the CPLD gets far more difficult for me.

https://jedimatt42.com/4a/ti32kmem/

This left me with a problem. I know I can build Jedimatt’s design, I know it is thoroughly tested. It takes way more parts and time to assemble, and is also more expensive to build. I just love how the Backbit 32k looks. I also currently have no interest in a TIPI which does require Jedimatt’s type of expansion.

I looked around and found a Jedimatt type expansion for a reasonable price with shipping. I would rather build it though. I looked into it. I found all the required parts at Digikey. I also found the Ram and CPLD for the Backbit 32k there.. I also found a more “stylish” case for the Jedimatt 32k board on Thingiverse.

I decided to order the parts for Jedimatt’s board. While I was at it, I did order a replacement ram chip for the Backbit 32k board. I could have picked up the CPLD as well for $3.50 or something. I wasn’t interesting in trying to hand wire that for the JTAG programmer at this point (I may next time I place an order).

Jedimatt’s 32k Parts List. I have provided links for more specific parts. The most generic parts can be purchased easily. I included a link to the type of Round Female Socket Pins that I used, Digikey does have them as well if you are willing to look for them.

Jedimatt’s PCB Design. Download his Gerber files and order from your favorite PCB manufacturer.
32k Ram AS6C62256-55PCN 28DIP 0.6″
SN74HCT245N IC TXRX NON-INVERT 5.5V 20DIP
CD74HCT21E IC GATE AND 2CH 4-INP 14DIP
SN74HCT138N IC DECODER/DEMUX 1X3:8 16DIP
2N3904BU Transistor NPN 40V 0.2A TO92-3
ICS-314-T 0.3″ IC SOCKET, DIP, 14P 2.54MM
ICS-316-T 0.3″ IC SOCKET, DIP, 16P 2.54MM
ICS-320-T 0.3″ IC SOCKET, DIP, 20P 2.54MM
ICS-628-T 0.6″ IC SOCKET, DIP, 28P 2.54MM
CONN EDGE DUAL FMALE 44POS 0.100 395-044-524-202 (Requires Modification)
44 PIN (Two Row 22PIN 2.54mm Pin Header) (Made with Smaller pieces)
3 PIN 2.54mm Pin Header (For Power Jumper)
2.54mm Jumper
DC 2.1×5.5mm power jack
3mm RED LED
680 Ohm 1/4W Resistor
2 ) 1k Ohm 1/4W Resistors
10k Ohm 1/4W Resistor
4 ) 0.1uf (104) capacitors
2.54mm Round Female Socket Pin Header Strip (88Pins are required)

I downloaded the Gerber files from Jedimatt’s site. Then I went over to JLCPCB and ordered the boards as that is where I order my PCBs from. I managed to order 5 of them and have them shipped to me for a total of $4.11 ($2.00 for the boards and $2.11 for shipping and taxes). I did kind of mess up, I didn’t want them in Green, but forgot to change it. The other colors did list they would take 2 more days to ship though.

I placed my order with Digikey for the parts I required. Between the PCBs and the Digikey order, I paid less than I would have for a completed 32k Memory Module on Ebay. I did have the 74HCT138 already, as well as the Resistors, Capacitors, 2.54mm Pin Headers, 2.54mm Jumper, DC Power jack, LED. I am also 3d printing the case. I ordered 5 of each of the 245s and 21s as well as two of the ram ics (and a replacement ram ic for the Backbit 32k incase that was the fault there).

First the Backbit 32k. I replaced the ram ic. The one I ordered was the exact part number that had been on it. I had the same results. Initially I couldn’t get it to work at all, once it made the buzz on power up, on removing it and trying it a few more times, it just failed every time I ran the RAM Test on it. It then worked for a little one time.. So it is on the shelf incase I want to buy another CPLD for it at some point.. I am wondering if it may be the edge connector is for some reason not making good contact. I am thinking I may get out the other TI 99/4a to see if it works on it. That one has no keyboard in it though, and I can’t start the ram test without one. The keyboard for it may be just working well enough to do that though if i reinstall it..

Next I moved on to the Jedimatt 32k build.

The boards came and looked great as usual.

The part that needed the most consideration was the Edge Connector. The edge connector that I purchased is of the same series as the one I purchased to replace the Cartridge Connector on my Commodore 128. Being that the connectors are the same 44pin, the difference between is the one for the Commodore 128 was a Right Angle connector. Being how it worked with the Commodore 128 I figured it should work for the TI 99/4a. It needed modification though, to remove the “mounting hole ears” on the side just like it had with the Commodore 128. This was a bigger deal here due to fitting in the 3d Printed case, the opening is very tight, and I needed to file it down a bit smaller than I thought I would.

There is a difference between the Edge connector between the Jedimatt and Backbit 32k. Jedimatt recommends some “individual pins” to use as extenders on the Edge Connector. It doesn’t seem they are available. I used Round Pin Female Socket strips. This is required because the “legs” of the Edge connectors are too short to reach deep enough into the side of the Ti 99/4a, or into the pass through socket of other Sidecars like the Speech Synthesizer.

I printed out the case and fitted the board to check how far the connector came out. With no pin headers in between it was way to short. I then tried with 1 pin header to see how that fit. It was still to short, compared to the Backbit 32k. The Backbit 32k didn’t use pin headers to extend the connector, it was cheated a bit by not putting the edge connector pins the whole way through through the PCB. The 3d printed case I made for the Backbit 32k is much thinner than the cases designed for the Jedimatt 32k (or at least the one I am using). I could rework the 3d printed case for the Jedimatt 32k, and maybe I could have managed then with a single pin header to extend it. That was not something I wanted to do though, it would have taken a good bit of work on the model.

That meant I needed to stack 2 of the Round Pin Female header pins for each of the 44pins of the edge connector. Which required stripping out 88 of the Round Pin Female header pins. The next step was to insert them into stacks of two each. The stacked pins were then soldered together, holding them with my Helping Hands. I went to stacking 5 pairs at a time and standing them up in holes in the PCB by the time I was done. While it was a bit awkward being they could spin around and even be picked up by the solder iron, that was quicker than trying to get the alligator clips in the Helping Hands to hold them properly.

The pin header strips before removing, then the individual pins, a pair of stacked pins and in the middle a pair soldered.

Starting to put the pairs of pin headers on the connector.

That was a lot of time consuming work. Stripping the 88pins out of the strips without loosing or damaging them. Then stacking and soldering them together in pairs without making an awful mess of it. Then installing them onto the modified Edge connector that I had taken the wings/ears off and sanded down a bit more to fit through the 3d printed case.

To get the alignment correct, I fitted the pins and edge connector into the PCB. I then soldered the 4 corner pins into the PCB. I then aligned the edge connector and soldered the 4 corners of the edge connector to the pins. This let me align it all nicely, and after that it was easy soldering the rest of the now captive pins to the Edge Connector and PCB. Doing the soldering that way worked very well. It is very solid, it doesn’t look too bad overall once it was finished. I had tried to find what other people were using to extend the Edge Connector, it looked like one example was using the same Round Pin Sockets stacked up.

Above you can see the Jedimatt 32k Edge Connector does extend just a little bit higher than the Backbit 32k Edge Connector when they are in the cases. That was the closest I could get it with using the Round Pins to extend the Edge Connector. That does make it have a bit more gap between it and the Speech Synth Sidecar and or the TI 99/4a. I am thinking of that “heatsink looking” interface bit on the side of the Speech Synth Sidecar where it meets up to the TI 99/4a. I could make one of those and stick it on the Jedimatt 32k to support the gap in a very similar way.

The Backbit 32k has a serious time advantage due to not needing to extend the Edge Connector pins. You can see below the BackBit 32k just solders in the Edge Connector, keeping the pins as barely in as possible to extend it out as far as possible. You also see it has far less components overall. (Update 8/31/25 the Backbit 32k is working now, I striped it to be the bare board, flipped the edge connector 180 degrees and reinstalled the ics and capacitors, and with a new microscope I could inspect the work properly to clear any solder bridges and see that all the pins were soldered properly)

The rest of the Jedimatt 32k build is strait forward through hole soldering. Keep in mind that you have to install the Jumper for the Power Selection. To the front it uses the 5V from the Ti 99/4a Expansion port (requires a mod to the Speech Synth Sidecar to use), and toward the back of the TI 99/4a it uses the Power Jack and you need to connect up a 5V Power supply. Be sure to only use at 5V Power supply. I am using the Ti 99/4a to power it from the Edge connector with my modified Speech Synth Sidecar. Also install all the ICs on the board in the correct orientation, Pin 1 toward the Capacitor.

The case design intends the LED to tight to the board. There is a “clear” lens bit to print. I am not sure how that is fitted and decided to mount the LED pointing out through the hole.

I had taken the top out of that circle so the LED would clear, but on trying to close it still didn’t close. On pressing it a bit, and then opening the case, I saw that the screw in the corner hit that one spot around the LED housing. On removing a bit of the plastic there, the case then closed properly. I printed it in a Silver and Black theme to fit the TI 99/4a color scheme. The 4 screws are not specified, I used four 2mm by 10mm long screws and nuts (or 2.5mm, but I believe 2mm), a little shorter screw may have cleared the LED housing..

I connected it up and ran the Ram Burn in test on it. It passed all the tests for 12 cycles, then found an error on one of the banks. I powered it off and back on, and ran the test again, it was working again. I am not sure the issue, or if it was a random failure. I did do some more testing with it today, and it worked properly. I tried out a few 32k Games and Programs all of them ran fine.

The case printed reasonably well, but has room for improvement. The Silver is a sliver PLA, the 99 4 and A had little nearly “floating bits” that are missing. They were all partly there, but due to a defect on one and the poor strength of those bits, I removed them. I am thinking of possibly reprinting it as a 2 color print with a black insert filling in the letters to give it strength to retain the floating bits. I didn’t print the “inner blocking piece” You can see the PCB and parts inside through the openings. The back is printed in Black PETG, the case itself is a snap/friction fit top. The screws hold the PCB to the Bottom, but do not hold the “top” on.

The Speech Synthesizer doesn’t pass 5V through it without a modification as shown on Jedimatt’s website. You can alternately use the external 5V Power Jack to power the 32k Memory expansion. If you have built either of the 32k Memory Expansion Sidecars, it should be no problem to add the wire to connect the 5V Pin in the Speech Synthesizer.

I did a little side project based on the Jedimatt 32k. I made an adapter to use the SOIC Static Ram that the Backbit 32k uses fit on the Jedimatt 32k. It is currently running on the 83rd pass of the Jedimatt 32k Ram Burnin Test. https://hobbytronics.home.blog/2024/05/05/ti-99-4a-32k-memory-sidecar-alternate-memory-ic-adapter/

TI 99/4a Recap

I figured I would make a post on the TI 99/4a I picked up awhile back. It was one of my first computers. I didn’t think of it at the time as a “real” computer though. I don’t even remember when I got it, but it was older and not current at the time, probably the early 90s. I wanted to get one and tinker with it like I have been with the other old systems from the 80s.

Really I am not sure I would recommend it yet. The system is so limited, I hate to invest in it. With a Commodore 64, short of a drive solution to get to and save data the Commodore 64 is complete. With the Ti if I want to do anything interesting, it seems I have to get a 32k ram expansion for it. That or stick to Basic and original cartridges. Many of the cartridges made now look like they require the 32k Expansion to operate as well.

Anyways I did get it, and I picked up some cartridges and even a tape player that I will look at in the future.

I did check the system out when I got it and took it apart. Inside it looked pretty good. The keyboard doesn’t work properly, it is one of the membrane types.. I am afraid also if I want to do much I will have to find a donor keyboard for it.

I did buy capacitors for it, there weren’t many.

Capacitors for the Mainboard. A few spares as well.

Initially I wasn’t going to get out my desoldering iron, but I quickly changed my mind before even getting the first capacitor out. I use an ECG J-045-DS desoldering iron. It is a rather cheap device, but it has done very well for me. I do keep thinking about and looking at those fancier units like you see so much of on Youtube. The ECG is a quarter of the cost of even the cheapest of those though and has done well for me so far. I don’t get it out much, mostly it takes awhile to get heated up and does take two hands to operate.

First was the power supply, with 6 capacitors. I cleaned the power switch with some Tuner Contact cleaner while I had it open, initially I wasn’t sure if it was making nice clean contact. I also replaced the LED as it was so dim that it was hard to see. The original LED had longer legs than typical, so I extended them with a piece of solid copper wire and then put heat shrink to insulate the legs. The original LED also had some heat shrink on the negative leg. I tested the power supply and it worked properly afterward, and the LED was much brighter, but not too bright. I also tested with the 99/4a board attached and it all worked fine. The voltages tested good as well.

I did add the zip tie to the largest capacitor after the picture was taken, although it was quite secure already.

Have to crack open the board. Two holding clip things and three bolts to remove this shield.

Then was the mainboard. That has 9 total. There is a single 10uF 35V, there are five 22uF 25V, and finally three 100uF 16V. To replace them I did get the five axial 22uF caps, but the 100uF I could only get in radial from the supplier I choose. The 10uF I couldn’t get an axial cap either so I went with a Tantalum instead, I could have picked up a radial 10uF though instead. When I did the Timex Sinclair 1000 I ended up getting some Tantalums for it as well.

Here is the board before. The Heatsink compound on the video chip there is replaced as well.

The worst thing of switching out from the radial capacitors was making sure I had the new ones installed correctly. Even the Tantalum is a polarized capacitor. The Tantalum marks the Positive, where Axial Electrolytic Capacitors mark the Negative. Then Radial Electrolytic Capacitors have a stripe with an arrow on it that points toward the Negative. So I did a lot of triple checking before powering up the board. I also checked my pictures of before and after to make sure I didn’t flip anything. That is why it is good to take pictures of these things. I then cleaned up the flux from the board and hooked the bare board up to test.

It didn’t power up properly after putting the capacitors in. That was an awful feeling, this is why doing such things are questionable. I checked and rechecked the board, I looked at each capacitor, checked if any were warm and possibly in backwards. I checked all solder points. I pushed in the socketed chips. Then I checked over the board again. I found a few flecks of loose solder, and one of them looked like it may have been shorting two points on the board. I checked again to see if there was any more on it then reconnected the board to the power supply and monitor. This time it came up normally.

I put it back together and checked it, well the keyboard is still awful, but it is working. I can play some Parsec just fine, well the joysticks are still awful. So it seems I have some more work to do with this one.

I pretest the new capacitors before installing them to make sure they are in spec. I also test the old ones I check how close to spec they still were. In this case I think all of the original ones are likely just fine. I just use one of the cheap atmega based component testers.

Still in the end, I had to open it for the new LED and well I say that was worth getting in place. It looks so much better with the new LED. Sure doing just that would not have been nearly as complex a task though. I also replaced the thermal compound on the video chip as well and do recommend that. The RF Shield was reinstalled the same as original. There is the later QI board that apparently doesn’t include it, or part of it at least.

My next project was a Joystick Adapter to use the Atari 2600 joysticks with the 99/4a. It is a simple enough project just taking some connectors and diodes. It did give me a bit of trouble though initially. Possibly a solder bridge issue again, or simply an issue with a poor connection at the TI joystick port, or an issue with not cleaning the flux from the board properly. I have it working now, I am just waiting on some more DB9 connectors to finish it up. I did also make a cartridge with a new pcb and eprom that I may mix in with some post at some point. I am really not quite understanding the way to use the new cartridge boards properly yet. I did get the homebrew remake of Pitfall on it, but I am not quite understanding how to set them up for different size roms.