The Motorola factor
But software written in Basic wouldn't run straight off the tape because the two machines used different tokens to represent Basic keywords. And while Tandy had selected Microsoft's entry-level Color Basic, the PAT team chose an extended version with a broader array of graphics modes and other features.
PAT also contracted assembly language specialist Duncan Smeed of the University of Strathclyde, Glasgow to work on the new machine's Bios. Motorola had a strong presence in Scotland. Its UK semiconductor operation was based in East Kilbride, south of Glasgow. It still is, though the business, spun out of Motorola in 2004, is now called Freescale.
Basic programming
Smeed took the stock Motorola Bios and proceeded to tweak it for the hardware changes PAT was making to the reference designs. Smeed also made numerous optimisations, for instance recoding the keyboard scanning routines - a vital part of a system operated through Basic - to yield a 10-15 per cent performance jump over the reference hardware and the CoCo. The Bios assembly code was prepared on the University's DEC PDP-11 and then burnt to EEProm at Motorola.
He craftily dropped his initials, 'DNS' into the Bios' six-byte end-of-line sequence. Only the first two bytes of the six, Carriage Return and Line Feed are used, unless you poke an address with a '6' instead of a '2', in which case Duncan's initials print out on every line. Smeed hand-burned the code, not included in the Bios source print-out, into the final version EEProms
Motorola reveals the 6809E's basics
Smeed would leave Strathclyde in 1983 to join Dragon Data as its Systems Software Manager. He returned to academia in 1986, a year after leaving Dragon, and today he is once again teaching computer science at the University.
PAT would further differentiate Dragon's hardware from Tandy's by adding a parallel printer port in place of the CoCo's serial connector. The video chip was adjusted for 625-line PAL output rather than the US-standard 525-line NTSC. How much of this adjustment was done merely to avoid invidious comparisons between the Dragon and the CoCo, or was a genuine attempt to take a stock design and make a better will probably never be known.
As Smeed now recalls: "There were three drivers behind the Dragon's early development: time to market, time to market, time to market."
From Risc to risque: Motorola's 6809E command set prompted many a schoolboy snigger
The pressure was on to get the kit out and get it out very quickly. In those circumstance, developers have to take any shortcut they're offered.
The prototype Dragon, fitted with 16KB of memory and codenamed 'Pippin', was completed in November 1981, allowing Tony Clarke to show the PAT team's work to the Mettoy board and so gain its approval to put the computer into production.
Next page: Pippin but no togs
COMMENTS
Re: You've made a happy woman feel very old
'The Girl With The Dragon Thirty-Two', eh?
You've made a happy woman feel very old
This machine is where it started for me. If it hadn't been the Dragon then it would have been something else, but this was the first machine on which I cut any code of any sort, for which reason I feel very kindly inclined to it.
Memories of the Project Leader
I was the project leader for the Dragon 32 at PAT, or Patcentre as is was then known. I thought you might like to know what I remember of that time.
I remember spending a lot of time talking to a really helpful Motorola chip salesman called Robin Saxby. Yes, the same guy who later ran ARM and is now Sir Robin.
We certianly did not copy the CoCo. It was not really available in the UK because it had an NTSC video system which would not work on UK TVs in those days. The Motorola application for the SAM chip (synchronous address multiplxer) showed a complete home computer to which the CoCo was identical. We made numerous improvements to this app note. We included a real A/D and D/A convertors for generating the FSK signals used to store programs on tape. We added a parallel printer port and used the same chip to scan the keyboard. We had a separate power supply PCB which also contained the TV modulator. It was a single sided PCB so it saved cost but also allowed variants to different TV standards to be made cost effectively. We made a SECAM variant for France and also an RGB and US version.
I do not rmemeber Motorola suppying a BIOS. Microsoft wrote the Basic interpreter, which was essentially the same as the one they licensed to Tandy but with a few add-ons, but you were expected to create all your own peripheral drivers - a situation unchanged to this day. It was these drivers that Duncan (Smeed) wrote. I do not remeber the keyboard speed up being his alone. The nromal way to scan a key board is to activate a row and read the columns to see if a key has been pressed and repeat that for each row. Of course, most of the time there was no key pressed and this routine just wasted a lot of time getting a no key preessed result. We realised that becuase we had used the same chip to scan the keyboard as drive the parallel printer port, we could do one thing the CoCo could not, and that was activate all the rows at once. If you do this and then look at the columns, in one go you get to know if there are no keys pressed, the most common situation, and you can exit straight away. If you find a key has been pressed you scan as usual to find which one. This is what saved the time.
The PAL output had nothing to do with the CoCo. PAL was essntial for it to work on UK TVs. Few if any had SCART sockets so you had to create genuine PAL. Persuading a chip designed to make 525 line 60Hz NTSC to make 625 line 50Hz PAL instead is a non trivial exercise and needed a lot of descrete logic - ASICs were in their infancy then.
Two weeks before the official launch, the Spectrun 16K came out. The piggy back RAM PCB was designed, tested and ramped up for production in that two weeks. Later we used a bunch of Siemens 32K RAM chips that consisted of two 16K RAM chips literally piggy backed on eachother and later still upgraded the main PCB to 32K then 64K.
We then worked on the disk drive unit which was abandoned when Tony Clarke left and all development work went in house. What is probably not well known is that at the same time we were working on the successor to the Dragon, code named Draconis. This used a Motorola 68K processor and a very powerful graphics chip from NEC. Along with OS/9 as a true real time executive, this would have beaten the PC hands down as a business machine. But for the vagaries of the home computer market, we might all be using Dragons today.
Cheers
Ian Thompson-Bell
I had one of those!
Somehow spending 4 hours typing in games, then finding out you mis-spelled "peek" at line 827, before you could play them created an enormous feeling of accomplishment. This stick in a disc and off it goes malarky we get these days is for noobs I say.
Don't think I'd want to turn back the clock, but somehow proud I was there at the time.
Re: It was *garbage*
What are you talking about? It worked fine. Sounds like you took a perfectly good computer back because you made a typo somewhere. I guess it's the 80s equivalent of blaming the compiler :)
