TS1000 upgrade – log entry 4 — Oh the memories we’ll make with 16k!

---

Entries in this series:

• 1) TS1000 upgrade
• 2) More Power!
• 3) On The Big Screen!
• 4) Oh the memor­ies we’ll make with 16k!   « you are here

 

---

2023, 03 , 14

The story so far

At this point, we’re about mid-way through our upgrade jour­ney of my ori­gin­al, vin­tage Timex Sin­clair 1000. My goal is to bring it into the 21st cen­tury, but do so in a way that any modi­fic­a­tion can be totally revers­ible, bring­ing the unit back to ori­gin­al fact­ory state should we want to do that.

Cur­rent progress:

• Veri­fied that the little unit still works after 40 years and is a can­did­ate for an upgrade (Part One)
• Replaced the ori­gin­al power reg­u­lat­or and honkin’-large alu­min­um heat sink with a mod­ern effi­cient com­pon­ent (Part Two)
• Installed a com­pos­ite video modi­fic­a­tion to allow the TS-1000 out­put to be dis­played on a mod­ern TV (Part Three)

A complex job

This will likely be the most com­plex upgrade — intern­ally expand­ing the onboard memory to 16k. The pre­vi­ous pro­jects have pro­gressed in com­plex­ity / dif­fi­culty from very easy to ‘a bit more chal­len­ging’. This memory upgrade con­tin­ues that trend.

A quick search reveals quite a few memory upgrade options but I opted to go with the solu­tion in The Byte Attic video.

Survey the terrain

The Timex Sin­clair 1000 boards were a very clev­er design — in fact, they were basic­ally Sin­clair ZX-81 boards with a slight modi­fic­a­tion that upgraded them to 2k of RAM from the fact­ory. The stock Sin­clair ZX-81 only had 1k.

A look at the mother­board shows that the unit has a single 2k memory chip. Inter­est­ingly, the board has mark­ings for an altern­ate con­fig­ur­a­tion of two 1k chips — though I’ve not seen pho­tos of that con­fig­ur­a­tion yet.

Either way, we’ll be pulling that per­fectly good 2k chip out of that chip sock­et, and repla­cing it with a new 32k chip, though we’ll only be con­fig­ur­ing the sys­tem for 16k. And here’s where we notice our first little snag — the sock­eted 2k RAM chip has 24 pins. Our upgrade RAM chip has 28!

Well, the boffins at Sin­clair Research planned ahead for that. Look­ing closely, you’ll see that the board actu­ally has loc­a­tions for a 28 pin chip of this type.

But now there’s the prob­lem of that sock­et that’s miss­ing four pins. We have a few options here:

• Desolder the sock­et, solder the new chip dir­ectly to the board
• Desolder the sock­et, solder in a new 28 pin sock­et or set of pin headers
• Leave the sock­et in place and add in 4 pin headers

I’m a fan of doing the least amount of work pos­sible, so I chose the third option. Also, this meant less oppor­tun­ity for things to go wrong as I’d not be heating/cooling the board and solder points to get the chip/sockets replaced. Less to go wrong = less risk. And less work 🙂

It was a simple mat­ter to clip out some nice machine-pin head­ers and mount them to the board. First we needed to clean out the through-holes. That 40 year-old solder would­n’t mix well with the new I’d need to apply. A few moments with a 300 degree iron and a solder suck­er desol­der­ing pump and they were ready for the pins.

 

I used blue tack putty to help keep their align­ment and hold them in place. Gold looks great don’t you think?

And the fit was great!

Now things get weird (wired?)

Fol­low­ing the install­a­tion done in The Byte Attic video, and cross-ref­er­en­cing against Tynemouth Soft­ware’s post leads to this rather inter­est­ing obser­va­tion. It really does­n’t mat­ter which data lines you con­nect the RAM data pins to, as the pro­cessor logic has it all figured out.

With a RAM chip, the order of the address and data lines is not actu­ally import­ant, as it reads back using the same pins in the same order as it was writ­ten. So rather than con­nect­ing to the match­ing pins on the RAM chip, just go for the easi­est options.

Adding to this appar­ently haphaz­ard pro­cess, we now get to *bend* a few pins on the RAM chip! Yep, these lines need to be routed along a dif­fer­ent path than what the engin­eers at Sin­clair Research determ­ined. So yes, we care­fully bend up a few pins (#1, #21, #23, and #26 to be pre­cise) to provide clear­ance for us to attach wires for the new data path.

And as for the data lines, we’ll be con­nect­ing the RAM chip to them at the con­veni­ently-placed diodes, just adja­cent to the RAM chip socket.

As we’ll be sol­der­ing to the RAM chip pins and the wire of the diodes, I’ve giv­en them a bit of a clean­ing with an abras­ive fiber­glass pen, though any­thing that’ll clean off any oxid­a­tion would work; sand­pa­per, a pen­cil eraser, harsh language…

For the wire, my reg­u­lar 22 AWG pro­ject wire was too thick, so I chose some 30 AWG wire wrap wire that I’ve had for years.

First I soldered the wire to the Diodes. I’d roughly meas­ured the length of wire needed, but left a bit of slack just in case.

 

 

 

Then, after pla­cing the chip in the sock­et, com­pleted meas­ur­ing the length, trimmed and stripped the oth­er ends, and fin­ished the job. Oh, and to help keep the job look­ing neat, before sol­der­ing the wires to the diodes, I added some heat-shrink tubing to the RAM chip connectors.

 

 

Well, that was a bit of work. But all neat and tidy ‘eh?

Briefly, on the topic of memory mapping

The intern­al memory is mapped and addressed above the 16K loc­a­tion in the memory map — the first 8K is the ROM, the second, reserved for addons and stuff. That brings us to the user memory space.

On unex­pan­ded devices, the 1K or 2K avail­able would live here. With our 16K expan­sion, access­ing that chip (those chips?) is dis­abled and replaced with our chip, using the full 16K space. Only half of our 32K chip.

 

Now, in a dif­fer­ent universe/timeline (I just watched Everything Every­where All at Once — amaz­ing movie), if the memory map was dif­fer­ent, say, con­tigu­ously mapped to 32 k, then our shiny-new RAM chip upgrade would give us a 32K upgrade. But that’s not our uni­verse, so to get to 32K, we’d need to do more jig­gery-pokery. I’m not sure that’s needed as there’s not much soft­ware out there that uses 32k (or more).

Which all goes to say, now we need to test the RAM chip and our work. Of course I could just key in a pro­gram that writes a value to a RAM loc­a­tion and then reads it back, but what fun would that be?

Rather, let’s just load in Mazogs — a glor­i­ous 16K dun­geon crawl­er from 1982, and give it a good play.

Oh, you noticed that?

Yeah,  I did kinda skip over the whole grab-a-cas­sette-tape-and-load-it-thing. While authen­t­ic, it’s also a one of the things I miss the least about the early years of per­son­al com­put­ing.  So I did the thing that you likely knew I’d do, I built a thing that lets me avoid all that ‘fun’ and jump to near-instant grat­i­fic­a­tion. And for details on that, you’d best read the next entry ‘eh? (sorry if you’re read­ing this before that entry is pub­lished — con­sider that a bit of a teaser?)

 

---

Notes:

• Wire Strip­ping — it’s a roy­al pain without a really good strip­per (hmmm…). Here’s what I use, not an affil­i­ate link, makes things so much easier.
• Hi Res graph­ics. I guess that’s a thing with some 16K upgrades, but not all. I have more read­ing to do.

Bonus photo:

Because I really like the gold pin sockets 😀