The awakening of the Spectrum 128K

In 1985, the 128K version of the Sinclair ZX Spectrum was designed and sold first in Spain by Investrónica S.A. Investrónica was the official dealer for Sinclair products in Spain until Amstrad bought the rights on the 80s successful computer. Investrónica was also part of the El Corte Inglés group, a big Spanish corporation, whose most known company was (and is) “El Corte Inglés” department stores.

A boxed Spanish Spectrum 128K

Investrónica was as technology company with a wide experience in building electronic machines, the core of their business was creating an providing expensive machines for tailoring companies, but they also were investing in many areas, and consumer electronics was one of them.

According Jose Antonio Landrove, Chief Procurement Officer at Investronica in 1985, early that year Sinclair was not interested in developing the Spectrum further than the 48K+, as their efforts were concentrated in the QL and the C5.

Investrónica had already a large experience in building electronics, and also had permission from Sinclair to create their own keyboards. Spanish regulation forced all computers sold in Spain to have spanish special characters in their keyboard, so they were already purchasing 48K+ cases to swedish company Celluloid AB (also a provider of Sinclair Research), and membranes to a british company located in the Isle of Wight named NFI. The 48K boards were of course bought to Sinclair and then all pieces were properly assembled in Spain.

As Sinclair was not interested in building a 128K Spectrum, but Investrónica foresaw some good income from such a model, they asked Sinclair for permission to design a 128K Spectrum themselves, and Sinclair agreed. The project was internally nicknamed “Derby”.

So basically Investrónica had already a provider for cases and membranes, and just required to design the board. It happened that by that time Martin Brennan, John Mathieson and Ben Cheese, all of them Sinclair Research ex-employees , have founded their own company, Flare Technology, and they knew a lot about the ZX Spectrum, so the motherboard design was
comissioned to them, with the exception of the 128K ULA, whose design was made by Texas Instruments.

There was another provider involved: the Chilean industrial designer Guillermo Capdevila updated Rick Dickinson’s case design, and it’s because of him that the 128k+ model is today nicknamed “the toastrack”, as he was the one adding that quite visible feature. He was most likely also the designer of the numeric keyboard that was bundled together with Spanish 128K+, although for obvious reasons the design was strongly based on the original case design made by Dickinson.

Once the board and case designs were made, the boards were built in Madrid by Spanish company Sagitron, a company already working for Investrónica for many other electronic projects. Contracting Sagitron instead of Timex was a reasonable option, as they were reliable and at that time, when Spain was not yet in the EU, using Timex as provider would have meant adding some import taxes.

Contents of the Spanish 128K Spectrum box

The Spectrum 128K was marketed in 1985 in Spain, ready for the Christmas campaign, and not until 1986 in UK. Apparently Sinclair had too much stock of 48K boards to sell and that forced them to delay launch in UK, although that may be just a rumour, as maybe it’s because by Christmas Sinclair was already having financial problems due to the big failure with the C5, and their debt with Timex made asking for new boards impossible.

So in the end, Investrónica pushed so a 128K version existed, and took the leading place as Sinclair was not interested (or just unable) to do it. Should Investrónica took another path, there may had not been a 128K Spectrum, and maybe Amstrad would also had had less interest in the deal that less than one year later (April, 1986), as the 48K designs were clearly outdated in 1986 when compared with other competitors in the market.

The Spanish ROM in the 128K board, with the project nickname “Derby” on it (on a sticker)

The knowledge adquired by Investrónica while building the 128K Spectrum lead a few months later to the design of the Inves Spectrum, probably one of the first, or maybe the first, western Europe clones of the Spectrum 48k. Investrónica was replaced by Indescomp, another Spanish company, as official Sinclair dealer once Amstrad bought the rights on the Sinclair computers, so they designed their own clone, an action that ended in a trial when Amstrad sued Investrónica because of that, but well, that’s another story.

So what’s different in the Spanish 128K Spectrum models?

The English engineers who received the original design made some changes to the 128K board and ROM, so there are changes between both versions.

• As it already happened with first issues of 48K boards, the Spanish boards had bugs: the most important one is the CLK signal of Z80 CPU is not routed to the edge connector, so some devices that require that signal, doesn’t work or doesn’t work well (i,.e. DivMMC). It’s easly fixable by just soldering a wire to the proper edge pin anyway.
• Carrier signal in the RF (TV) output was 5,5Mhz in the Spanish version and 6Mhz in the English one.
• Pinout in the RGB connector is different (same signals, different position) so SCART(peritel) made for the English toastrack won’t work in the Spanish toastrack and viceversa.
• Also, the composite signal in that connector, it’s a B/W one instead of color. While that doesn’t matter if you use the SCART cable, it would affect when trying to connect via composite video connector. There is a procedure to fix that though.
• There are several components changed in the motherboard, but none of them makes a big difference.
• The numeric keyboard was a standard in the Spanish 128K model, and was sold separatedly in the English one.
• The ROM is actually a EPROM, and the sticker covering the erase window reads “Derby” in hadwriting, as for the project nickname.
• The keyboard has Spanish characters like Ñ or Ü. In fact there were not any additional keys, just some symbols were replaced in the ASCII table so they looked like that Spanish letters and accented letters. Also, some keys are translated (like BORRAR instead of DELETE)

Spanish keyboard

• The ROM was also updated so all messages are in Spanish. The basic tokens were not translated (what would had been a very bad idea anyway).

Spanish error message for “D BREAK — CONT repeats”

• The computer starts with a copyright notice “©1985 Sinclair Research LTD ESPAÑOL” instead of using the usual start menus you can see in the English one, and in the Amstrad models. It’s a message closer to the original 48K models “© 1982 Sinclair Research Ltd”.

Copyright message

• Updated charset in the ROM so it supports Spanish characters (¿,¡,Ñ, Ü)

Spanish charset

• The Spanish version includes a full screen text editor, that you could call by using the new token EDIT (i.e. EDIT E$). That was a built-in competitor for Tassword and other word processors at the time, but being limited to 32 columns made it only usable with zx printer, so it was not very successful. Still, it has a lot of features that could be invoked by using the numeric keyboard, that made it quite good (but for the 32 columns issue).

The Spanish 128K word processor

• The BASIC editor was improved, but engineers at Sinclairs restore the original one in English version. It had a full screen editor (instead of line by line), and the alignment when the code was listed was better than previous and later versions.

Full screen editor

List of basic code (Naulen, by Mariano Chiaverano)

• Additional tokens. Apart of EDIT, there were two more: RENUM (which allowed renumbering code, which properly updated GOTO and GOSUB calls) and DELETE (allowing deleting a range of basic lines). The Spanish version of RENUM allowed selecting start line, new number for that line, and step between lines. The English version had no parameters and automatically renumbered all lines in steps of 10. DELETE was removed from the English versions. Of course, basic instructions PLAY and SPECTRUM, were implemented first in this version, but those ones remained in the english one.
• The calculator remained in the English version, but in the Spanish version you didn’t need to select any option in the menu, you could just type operations in the basic prompt.

The Calculator

In the end, the Spectrum 128K become the most compatible with previous software from all 128K Spectrums, probably cause people at Flare and Texas Instruments did know what we were doing very well. The first Amstrad model kept same compatibility cause the chief engineer at Amstrad was given a 128K Spectrum and asked to clone that before Christmas campaign started, so not many changes were done.

Later models, the black models (+2A/B and +3) were less compatible cause they were rebuilt by Amstrad engineers on their own.

Disclaimer

This article is based on what is possible to check today, and what has been told by Jose Luis Landrove from Investrónica, if you have further or different information please don’t hesitate to let me know, so the article can be completed, or different point of views shown.

Greetings

For writing this article I have used the following sources, apart of my own experience:

• Article and images about software differences in the Spectrum 128K by baltasarq at Zona de Pruebas Spanish forum. Some images were also made by baltasarq.
• El mundo del Spectrum Podcast 7x05, interview with Jose Antonio Landrove, by Jesús Martínez del Vas and César Hernández Bañó.
• The ZX Spectrum 128K Spanish and English manuals
• “What you see is what you get” book — by Alan Sugar (autobiography).
• Thanks to Mcleod_ideafix for the details on Spectrum 128K hardware changes
• Thanks to Brendan Alford for the details on the composite output difference.

Note: I’ve also used several photographs from the 128K model and box taken from the internet. I could have made myself most of those photographs as I own a Spanish 128K Spectrum, but I didn’t consider losing time doing them with so many photographs available. It’s difficult to list authors here but if you recognize some photograph of yours and want to be acknowledged please let me know. If you want me to remove your photograph please let me know as well.

UPDATE: long time ago Andrew Owen contacted me with more information. I had forgotten it for a long time, but today I’m including that information, that adds some details to it. From now on these are mostly Andrew’s words:

Investronica paid for it. The UK launch was an attempt to keep Sinclair afloat long enough for the Robert Maxwell takeover to go ahead, but of course it missed the critical Christmas sales window and the deal fell through, leading to the Amstrad takeover in 1986. Sugar was so confident of closing the deal that he had the Hong Kong team redesign the 128 into the +2 before the deal was signed. That’s what enabled sufficient machines to be produced to supply demand in Christmas 1986.

In the late 1990s and early 2000s I was researching the 128 ROMs (UK and Spanish) to find out who the initials in the credits referred to. I ended up interviewing all but one of them. In the Spanish ROM they are MB: Martin Brennan, KM: Kevin Males, and AT. I never found out who AT was, but they were an Investronica employee sent over to Cambridge to work on the ROM.

In the UK ROM the initials are MB, SB: Steve Berry, AC: Andrew Cummins, RG: Rupet Goodwins, and KM.

ROM development was all done in Cambridge. In the original 1985 ROM, Brennan wrote most of the 128 editor, including porting RS232 routines from Ian Logan’s Interface 1 ROM. Males did the AY and MIDI routines. I suspect AT’s main input would have been translating all the error messages and Spanish character set support.

As a recently discovered prototype ROM shows, development of the Spanish ROM was finished in the first half of 1985. Berry and Cummins then reworked that ROM for the UK launch in January 1986. Goodwins’ contribution was the menu system and four color swish.

Your article includes some information I wasn’t previously aware of, but it also has a number of errors in it, which I’ll address now.

• It’s not entirely true that Sinclair had never had any interest in taking the Spectrum beyond the 48. There was a planned Super Spectrum with Super Basic that was ported to the QL. And Brennan and Ben Cheese were working on the Loki project before they left.
• The 128, including the UK model, was designed in 1985. Brennan and Cheese didn’t leave to set up Flare until 1986.
• The ULA in the 128 is virtually identical to the one in the 48, the glue logic being handled by a HAL chip (with many bugs in it). It’s possible that TI did the HAL, although it’s so buggy that I kind of doubt it, but it seems very unlikely that they had any input in the ULA.
• I didn’t know that Guillermo Capdevila designed the heatsink, but aside from moving the ports around, the case is otherwise unchanged from the Rick Dickinson design. Also worth noting that Dickinson didn’t particularly like the design, but was forced to create it because of the amount of money Sinclair had spent on tooling for the QL. It’s conceivable that Capdevila also designed the keypad case, but the hardware was developed at Cambridge. This would have been necessary due to how tightly tied it is to the ROM routines. Goodwins told me that at one point they were considering adding a ball to it to enable it to be used as a mouse.
  It’s conceivable that Investronica did the board design, but it would have had to have done so in conjunction with the software engineers in Cambridge because of the Spectrum’s primitive bit-banging approach to serial communications and its hideous I/O scheme.
• The stockpile of 48K+ machines is the likely reason Sinclair delayed the launch of the 128 in the UK.
• There are a lot of hardware bugs in the Spanish machine. They weren’t all corrected in the UK version.
• The UK 128 BASIC still supports the full screen BASIC editor, but it loses the alignment feature. It also loses support for entering control codes. It also loses support for using reserved words as variables because it adds support for entering tokens in lower case. It also loses support for changing editor colors.
• Besides RENUM start value and a step, that you had to POKE on the UK machine) and DELETE (no equivalent on the UK machine), it also has a WIDTH command to set the number of columns for RS232. However, it is broken and so was left out of the documentation.
• The Spanish ROM has a running calculator which is always accessible, so the menu version is another example of how the UK ROM was a backwards step.
• While the +2 may seem superficially to be a toaster in a new case, it was completely re-engineered by a Hong Kong based contract firm and the PCB bares no resemblance to the toaster. The only carry overs are the Z80, AY, ULA and HAL. The 128 editor ROM was rebuilt from source and has a completely different set of entry points than the UK 128 ROM.
• The +3 hardware was developed by Richard Altwasser who designed the original 48 hardware. He was already working at Amstrad when it acquired Sinclair, having previously left to create the Jupiter ACE with Steve Vickers of Nine Tiles.
• The incompatibilities were because Altwasser replaced bodges in the original design that were done to keep costs down with better engineering solutions. The keyboard system is far more advanced than the one in earlier machines. But that killed support for the undocumented floating bus behavior and Sinclair joystick interfaces (less of an issue now that the machine came with joystick ports).