Saturday, March 30, 2024

Refurb weekend: Data General/One (and the worst LCD in the world)

I mentioned earlier that while I prefer specializing in non-x86 laptops, that doesn't mean I don't collect interesting or unusual x86 laptops, like the Brother GeoBook NB-60 (I finally tracked down a mostly working NB-80C, the top of the line model, which will be the subject of a future restoration). However, this one is a unit I've had since about 1998 when they were getting rid of it at the University I worked for, and in many ways a landmark PC laptop since it was the first battery-capable system with a full 80x25 or CGA 640x200 LCD — though with a notorious deficiency: more on that shortly — and also came with a terminal and text editor in ROM. This is the 1984 Data General/One.
I suspect it was primarily used to dial into the campus network using the built-in modem and ROM terminal emulator, at least until better alternatives became available, and then ended up forgotten somewhere until they were cleaning things out and asked me if I wanted it. (Yes, I was retro even before retro.) It came with its bag, power supply and modem cable, and of course I said yes. It wasn't pristine, though: the floppy disk eject buttons wouldn't stay on anymore, it couldn't remember its hardware configuration, and the main drive had issues with nearly full disks. And then there was that infamous LCD.

Ah, what the heck. I finally got a round tuit. That makes it time for ... another Refurb Weekend.

Early luggables, of course, existed well before the DG-1, such as the CP/M-based Kaypros and Osbornes. After the emergence of the 5150 IBM PC and XT, the Compaq Portable shocked the market in 1982 with its legally unassailable clean-room BIOS that induced IBM to produce the 5155 IBM Portable in 1984. (Naturally my favourite luggable of that era would have to be the Commodore SX-64, the first colour portable.) All these systems used small CRTs for screens, so there was no realistic way to run them off a battery smaller than the one in your car. Though the 1982 GRiD Compass 1101 had a full-sized flat electroluminescent display (ELD), ELDs were nearly as power-hungry, and thus the machine could not run untethered either.

Instead, the 1983 Gavilan SC is usually regarded as the first true DOS PC laptop. Its 66x8 (400x64) LCD was detailed enough to support a rudimentary GUI using its touch pad as well as MS-DOS 2.11, it came with an internal 3.5" floppy (the GRiD did not), and the LCD and battery-backed RAM storage were light enough to finally run the entire system on batteries. Still, it was quite expensive and its small display size limited its utility, a problem not fully addressed by other systems like the 1983 Visual Commuter, which had a full 80 column LCD but in its first iteration only 16 lines of text. Both machines, catering to the traveling professional, included 300 baud modems and ROM-based terminal software.

Meanwhile, the era of the personal computer generally spelled trouble for minicomputer makers not named IBM, and Data General was no exception. While the new 32-bit MV series (the subject of the classic The Soul of a New Machine) was their first unqualified success in years, the expense of its development combined with cheaper microcomputers eroding their mainstay 16-bit sales made it impossible to launch another large technical undertaking. To preserve cashflow, smaller projects thus became the order of the day, and one of these was to develop a microcomputer of DG's own that could nevertheless still provide market support for their larger systems. Mindful of Digital's own disastrous attempts like the DEC Rainbow in 1982, DG decided to take on the portable segment where there were relatively few competitors and they could stand out more easily.

Designed cooperatively by DG and its Japanese subsidiary Nippon Data General, the DG-1 hit most of its marks. Not least of its innovations was its size and weight (less than 14" by 12" or about 35cm by 30cm, three inches thick and around ten pounds or five kilos) as well as the 720K 3.5" floppy drives and custom full-size monochrome LCD panel, both manufactured by Epson. To solve the refresh issue on a panel this size, NDG split the screen into five segments multiplexed in parallel, though the result was lower contrast than on a smaller display. The built-in modem option and internal communications program made it a portable option for existing customers who needed a Dasher terminal on the road (and emphasized that Data General hardware was preferred: the option was specifically for remote workers using DG's Comprehensive Electronic Office suite on AOS) and it was at least as DOS-compatible as most other non-IBM PCs were at the time (provided you stuck to BIOS calls, that is).
DG also intended it to have a constellation of expansion peripherals. The rear docking port could allow connection to an external chassis for a hard disk, an external monitor or other expansion options, depicted here from a marketing photo. With a 4MHz 80C88 CPU, the base 128K of memory and a single floppy, expandable to two drives and 512K, it had an announced MSRP of $2895 (about $8660 in 2024 dollars) and hit the market in November 1984. An optional battery-powered thermal printer sold for an additional $500 ($1500).

While early users raved about its portability and power (Peter Norton himself wrote approvingly in PC Magazine that one could "use it as your one and only PC," and Academy Award-winning documentarian Mike Hoover used one on the road with a solar panel, dialing direct to news agencies), even its boosters found the LCD's legibility notoriously poor, especially as it wasn't backlit and the earliest models did not have an adjustable screen tilt. A potentially greater concern was that most DOS software still came on 5.25" floppies at the time, for which the DG-1 required a separate wall-powered unit connected to the expansion port. Data General addressed the first problem in less than six months with an updated display that could lock in multiple angles using a higher-contrast "reverse" polarizer and removing the lens. Customers agreed it was an improvement — as you'll see it could hardly have been worse — and it was sold new for the same MSRP, but existing customers had to pay $350 for the upgrade despite Data General's claims the upgrade was sold at a loss. As less than 10,000 had been sold to that point, the market damage was lessened. This updated first generation system is what we'll be working with here.

In those days PCs weren't yet (entirely) commodities, and Data General was determined to make the DG-1 a distinctive upmarket product. The DG-1 official carrying case was a sharp plush job designed by Pierre Cardin with an embroidered Data General logo and internal compartments for the power supplies (yes, plural) and wall and phone cords. Even the open-cell foam was higher quality, not the rubbishy stuff you find today sloughing off into tarry filth over other items of the era. Unified branding covered the packaging, the display lid, the bag tag, ...
... and even the official Data General/One license plate, of similar concept to the more famous Digital UNIX one. (I have this one plus the later Compaq UNIX plate after the DEC buyout.)
The styling extended to the side panels and the twin floppy drives (there was never an internal hard disk option for this particular iteration).
The keyboard is also custom. All PC keys of the era are present except for SysRq and Pause (Break was moved to the Escape key), but some have been moved to other places to save space, such as Page Up/Down and Home/End using SHIFT plus arrow keys, and some function differently such as Ctrl-Num Lock to pause scrolling despite there being an actual Scroll Lock key. For function key jockeys, DG thoughtfully included a removable, writeable cheat card whose tabs fit neatly into slots above the keyboard.

Three keys are unique to the DG-1: a Cmd (Command) key, used for specific key combinations by the ROM, plus a Spcl (Special) key and a blank (!) key that are unexplained by the manual or the tutorial and don't seem to do anything in the U.S. layout. I suspect these key positions are used by the keyboard layouts for other languages and later iterations of the DG/One employed Spcl as another type of Alt key. In another nice touch, an outcropping on the display bezel will automatically flip the power switch to off when you close the lid.

Curiously, the battery was an option and not standard. A large Ni-Cad block fits here with a connector, adding a couple extra pounds but also around eight hours of runtime. The paper clip is used to push down a small tab to slide off the top of the compartment.

I mentioned power supplies (plural): in early models, the charger for the battery was a separate unit and connected to a separate port. Even in this updated model where the power supply for battery and main power is now one unit, it still requires a separate cable to the battery charge port. I don't have the battery, so I don't have the original sidecar supply or the cable, but I also don't have leaky NiCd guts all over the inside of the machine either.

Those power ports, and all the others, are on the rear. The ports are protected by a sliding cover which also acts as a laptop stand when the cover is open, another nice design touch. Unfortunately, none of these ports is fully standard. The 5-pin DIN at the far left is the phone connector for the internal modem option, here the later 1200 baud card (early models with the internal modem option were limited to 300bps), which terminates in a regular RJ-11 phone jack that DG also included a T-coupler for. Next to it is the RS-232 port, a seemingly regular DB-25 except if pins 13 and 25 are grounded to make it an RS-422 interface instead (so your connector must not touch those pins). In a like fashion, the printer port next to it is not a parallel port: it's a serial port too, just without the modem lines, intended for DG's battery-powered thermal printer. This printer expects data at 9600bps 8N1 and is basically a teletype without the "type." The two plugs next to it are a barrel jack for the battery charging circuit and a two-pin power port for the system's main power, both 6.4V at either 2A or 3A.

Below that is the expansion port, but I don't have anything that connects to it. While I know the external 5.25" drive exists (model 2233), I've never seen the 2235 expansion chassis, though I have seen a smaller 2234 "dock expander" that also connects here. This expander is thumbscrewed onto the back over the rear ports using the two standoffs near the bottom. Although it replicates the printer and power ports, it blocks the RS-232 and modem ports in lieu of a 37-pin floppy disk connector and two external monitor ports (DE-9 and BNC, presumably CGA and composite). Since the whole concept with the DG-1 was portable communications, this particular option didn't seem especially well thought-out.

On the underside of the unit are the labels. These upgraded units have a hastily pasted-on model sticker over the original model designation which varied based on the factory options. This machine has, near as I can tell, the maximum available internal configuration: 512K of RAM (128K base and three additional 128K RAM cards), a second floppy disk drive and an internal 1200bps modem. I always found the handwritten serial number on this machine's lower label to be interesting. On modem-equipped machines an additional FCC sticker was supposed to be in the lower right well of the embossed plastic label, but this unit doesn't have one for some reason.
Oh yeah: the screen. Let me preface this paragraph by saying that our machine's LCD here has degenerated further and we're going to compensate, but despite the significant engineering resources put into its fabrication, even at its best and even at the time everyone agreed it was bad. InfoWorld in 1985 called it "godawful" and "a better mirror than [a] display," and even Peter Norton admitted its readability "isn't too good." PC Magazine 7/86 retrospectively theorized that achieving sufficient contrast and visibility wasn't feasible at the time with such a large panel, adding, "The exchange 'Why don't you turn it on? / It is on' is no joke. It happened in our offices." There isn't an ideal tilt position for the display owing to its limited viewing angles, but you can get a sense of the original contrast level with the first couple lines of text.
At this point in photography the floppy eject buttons chose to come off again, so we'll start the refurb there. The buttons attach with two prongs, a long one and short one, and it looks like the long ones (on the left in this picture) just plain snapped off. We need to make sure that whatever fix we implement lets us still remove the eject buttons to get into the case, so I chose to go with Velcro here: I cut some Velcro loop material to the proper size and punched holes in it where the button prongs would have fit, then made a little roll of Velcro hooks to go next to the intact short prong to act as a replacement long prong.
Pressed in firmly, this is enough to let the small prong engage with its receptacle and then hold it there with the Velcro, but the eject button can still be popped off if necessary. As a nice consequence, every press on the eject button will keep the grip maintained. Having done, let's take them off again for the floppy drive swap and to replace the CMOS battery, since we'll need to disassemble the machine to do those tasks.
We turn the unit over, slip off the back cover by freeing its tabs through the holes in its track, and then remove the exposed screws. Despite the appearance of only two types of screws, there are actually five: six of one type on the left which hold in the floppy drives, three in the front of another type that holds down the keyboard, one each of a thicker screw used to hold the main assembly in, and then a smaller and sharper one.
Now turn the unit back over and remove the side panel covering the floppy drive(s) and the corresponding panel on the left side of the machine with a spudger (they're held in with tabs). To make a rigid "sandwich," Data General decided to hold the unit together with metal snaps, three on each side. These pop out by wedging a small blade or metal spatula into the depressions next to them. Be careful, they're in tight! When putting them back on at the end, make sure not to get one in the floppy drive like I did or you'll have to get the drive back out again to sweep it free.
As with all laptops, there are invariably ribbon cables connecting the two halves. While the DG-1 has two ribbon cable connectors, only one of them actually has a ribbon cable in it which goes to the display. Open the battery compartment with your paper clip, disconnect and remove the battery (duh) if present, and then through the hole at the bottom gently lift the top of the occupied connector and slip out the display ribbon cable.
The top half then simply lifts up and to the side (the battery connector will slip out free on its own, but mind the connector for the power switch, which can be disconnected if you need to completely separate both halves). You may need to do a little wiggling around to get it out of its moorings; it's probably easiest to start from the back. Unlike many laptops where you need to remove the keyboard first, you should not do that here, at least not to get into the case.

With the top off we can see the A: drive (closest to the keyboard) and next to it towards the back a stack of two other cards we'll have to remove to get to the B: drive. These are all connected with Molex ribbon cables. The power board and DC-DC converter is on the backside perpendicular to the mainboard with the power switch and battery connectors wired to it. The other main landmark is a silver box marked with "UPGRADE MEMORY CARDS SHOULD BE ADDED IN ORDER FROM THE TOP DOWN."

That silver box has three little connectors for the RAM expansion cards, 128K each (model 2252). They are completely populated in this maximally expanded machine and the box can be pulled off the mainboard (we won't do that this time around) if they need to be swapped. Under the box and not shown here is a Mitsubishi M5M82C37, a CMOS second-source version of the Intel 8237 four-channel DMA controller. The base 128K of RAM is on the underside of the mainboard, so with the three cards we have our 512K.
Next to the memory expansion box are two chips, the 32K 27C256 system EPROM (here marked copyright 1984, 1985 NDG, i.e., Nippon Data General, and the code 5699-02) and the CPU itself, an Oki Electronic 80C88. On-screen this ROM revision identifies itself as version 1.21; I have not seen a more recent version of this system ROM for the original DG/One. The 80C88 is a CMOS version of the Intel 8088 and can operate from 0 (completely stopped and static) to 8MHz, though here the nominal clock speed is 4MHz. That's slower than the 4.77MHz IBM PC/XT but the underclocking achieves additional power savings.

Both chips are socketed and of course the EPROM could be erased and rewritten, which yields potential for fun upgrades like hacking the built-in software. I did think about replacing the 80C88 with the more performant pin-compatible NEC V20, but I'm not sure how well the chip would run downclocked to "just" 4MHz (the datasheet gives the lowest speed as 5MHz) nor on a battery, and even if it doesn't utterly blow the machine up we may need to make some adjustments to the ROM for timing-sensitive routines. A project for another day.

The keyboard is not attached by anything at this point other than its ribbon cable, so we can swing it up and out of the way to expose two more main ICs. The Mitsubishi chip on the right is an M5M82C59AP, another CMOS second source version, this time of the venerable Intel 8259 programmable interrupt controller.

The custom DG chip in the middle is marked 4658-01 (the 8452E7 looks like a date code based on the copyright year). I don't have the systems guide for this machine so I'm not sure what this chip does, but best guess is it handles the video and possibly the keyboard. Other versions of the board I've seen have a DG sticker covering up a UV window, suggesting this is a microcontroller of some sort with onboard ROM; nearby Intel markings on that chip hint at possibly an Intel 8751, a microcontroller from the Intel MCS-51 family with 4K of EPROM, and that would be the right number of pins. We last encountered the 8051 in the Universal Lab Interface, the spiritual descendant of the AtariLab. Early prototypes apparently used two chips here, one directly handling the LCD and another acting as a functional clone of the Motorola MC6845 as used in CGA. It is likely that they were combined into this single chip after later design refinements.

The little metal boxes here sheathe crystals. Next to the 4658 is a 3.579545MHz oscillator, likely for composite NTSC colourburst (another reason to theorize the 4658 contains the video circuit), and beside the 82C59A are 11.9808MHz and 10MHz oscillators. The 10MHz is most likely divided down to generate the CPU and system clocks, while the 11.9808MHz crystal is probably the clock for the 4658, assuming it is indeed an MCS-51 (most instructions take about twelve clock cycles).

The lifted-up keyboard also shows the internal speaker.

We now turn our attention to that stack of cards in the upper right on top of the B: drive. The top one is held in by two screws and a pin header to the card under it.

This card is the optional model 2750 1200bps modem card, another 8051-based device (an Oki 80C51 on a 7.3754MHz crystal). The 5-pin DIN connector is electrically a regular telephone modular jack and the same wires are present. The modem responds to the standard Hayes AT command set of the time and has a license for the Hayes escape sequence. If you send it an ATI0 command, it will return the PCB type and revision, in this case 120.
Under that is the I/O board, our first stop. The CMOS settings battery is here, a CR2025, and as expected it is completely dead. We'll replace that when we put this back together. The other main chips on top are an NEC μPD765AC (D765AC), which is the floppy drive controller, and an Epson SED9420C variable-frequency oscillator (VFO) data separator with MFM sync field detection, loop filter switching and timer functions. Interestingly, its datasheet only mentions use with 8" and 5.25" drives, but Epson chose to use it here as well. Along the back are the two serial ports (modem and printer), protected by three Tokin D-05N data line filters.
But what drives those serial ports? We'll have to get the card out to know. We remove the standoffs on the rear ports securing the card to the case (the standoffs on the card can be left alone) and wiggle the ports through their holes, then pull the card off the pin headers.
Turned over at top left, the card has two Oki M82C51A USARTs, CMOS versions of the Intel 8251 that is not register-compatible with the more common 8250 UART. This was a source of some incompatibility with communications software that expected 8250s to be present and tried to drive them directly (programs that used slower BIOS calls through int 14h usually worked). They are each accompanied by an Oki M82C55A (Intel 8255) for interfacing.

Our drives are now finally exposed and we've already removed their screws, so now we can just free them and lift them out.

Both drives are Epson SMD-100s. Their interface and power connectors are also ribbons and can be pushed off with a nylon spudger.
Unfortunately, these disk drives are not commonly found anymore, which is why we're just going to switch them for the time being. The SMD-110 is a close relative that should directly substitute and it may be possible to replace them with even later SMD drives, but you'd also have to account for the eject buttons and disk slots with the later slimline versions.
Lifting out the other drive. A small outcropping in both bays acts as an alignment pin and fits into a hole on the underside.
Before reinstalling the disk drives we must rejumper them first (the floppy cable doesn't set primary and secondary for you). The primary in front should have the jumper across the DSO pins; the secondary in rear should have the jumper across the pins labeled "1" (one). Reinstall the drives, plug back in the I/O card, reattach the port standoffs, plug back in the modem card, screw it in place, make sure the keyboard is back down, and shimmy the top case back on (watch out for the power switch and battery cables). I do not recommend putting in the screws or reattaching the display cable until the snaps are all on because if one decides to break free, you may need to disassemble the machine again to go find it. Once all six snaps are on and secured, then put the screws back in, reconnect the display cable, and put back on the side panels, floppy eject buttons and the battery compartment door. Whew!

Now, that screen. While never good I think it's obvious this is worse than it ought to be, and the mottled contrast suggested a degenerating LCD polarizer. Most twisted nematic LCDs have a 45-degree polarizer and this is widely available as cling film but we can't assume that here because of the custom Epson display. I ended up buying a couple Chinamazon sheets with a 0-degree line to test it. Both were smaller than the LCD, which is roughly 8.5" by 6.75", but it was an easy way to find out what I was actually dealing with. Also, by trying the film, we shouldn't be doing anything to the screen we can't undo later. Before opening the case I tried it out:

It's still pretty bad, but it's definitely better, and it's definitely also at 45 degrees. Since we have the film, I went ahead and taped the sheets together to do a quick install.
Doing cuts on the paper cutter.
Checking the fit against the bezel.
Now properly cut to size, I removed the backing and stuck it on the screen, then peeled off the front.
Here it is without the bezel. The bezel is actually only held on by a lip at the bottom and two snaps at the top. It came out a little more bubblicious than I'd like because this particular screen had a bit of a hard life with little bits embedded in it I could not clean off (and rubbing it made alarming shimmers in the LCD, so I didn't feel much like going at it with something slightly more abrasive). As a result, even with a couple tries and squeegeeing all the way, the little embedded bits leave bubbles and those aren't going away. It's like trying to apply a PDA screen protector but five times bigger and ten times worse.
But with the machine fired up the display is now much more legible (and it's possible to adjust the contrast a bit in software, read on). Still, it could obviously be a lot better if we got rid of the join line and the bubbles.

I thought about some alternatives. The most obvious is to apply a larger sheet of film and/or buy a 45 degree polarized sheet of sufficient dimension since we've established this is otherwise a TN panel. On the other hand, while that should eliminate the join line, any adhesive film will still run into the same problems with the surface that this one did.

The other thought is to get a new clear acrylic plate cut to size and put the film on that, and then put the plate in front of the screen. (There are polarized acrylic sheets also, but the quotes I was getting were substantially more expensive than just regular acrylic and polarizing film together.) The ~4mm bevel depth gives space for such a plate to sink into, well within typical stock thicknesses here in the United States, and unless I really mess up there should be no bubbles. However, since the edges of the screen are tapered, something would have to keep it in place or it will slip down (and out). Since we're still working with polymers, I suppose I could try transparent aluminum later. All that said, none of these alternatives would yield significantly better contrast — if at all — so for the purposes of this article let's go with what we have.

Let's take a tour!

The internal 32K ROM contains a simple terminal program, text editor, setup utility and basic diagnostics. It is entered either by turning on the machine with no disks to boot from, or pressing Ctrl-Alt-Cmd in DOS. It is not a built-in TSR or some such: the only way back to DOS is through a restart.
The Notebook is the text editor. The manual says it can hold up to 500 lines of 80-column text, though lines may not have more than 80 columns and joining long lines will cause them to be automatically broken at the 80 column margin. You can do basic editing, search and replace, and printing. The text in the buffer can also be transmitted in the Terminal, and you can also add data from the Terminal.

How does it write to disk without DOS? Simple: it doesn't. If you turn off the machine, it's gone. Transmit it, print it or lose it.

The Terminal program can emulate either a Data General Dasher D2 ("DGC Terminal") or a Lear-Siegler ADM3A ("Std Terminal"). We had multiple ADM3A clones on campus at the time, so I'm quite sure it was used in that capacity. It can either use the internal modem or an external modem or null modem at up to 9600bps. There is no hardware flow control, but XON/XOFF is supported with the "OFC Mode" (Output Flow of Control). You can send a document in the Notebook buffer, or you can receive data into the Notebook, to which it is appended. You can switch into and out of the Notebook and Terminal without dropping the connection.
The diagnostics mode is not too useful because it only tests memory and attached disk drives, without any other confidence testing, and all the tests are destructive (you'll lose contents of memory, contents of any scratch disks used for testing, etc.). If there is bad memory, the test will report a value x:yyy:z where x is (0) onboard RAM or (1-3) one of the expansion cards, yyy is the chip number in decimal, and z is whether it was (0) a read/write error or (1) an addressing error.
Last but not least is setup. Now that we have the battery replaced, we can actually save our settings. The defaults are generally appropriate but notice that the number of disk drives is not automatically detected: you need to tell the machine you have two or three drives, as appropriate. This allows you to have only one drive but connect the external 5.25" drive, for example, which is also a valid two-drive configuration. We'll select two disk drives, commit the settings, and boot from the DOS floppy.
I don't have the original Data General disks, just copies. The first disk contains Data General's customized version of MS-DOS 2.11 on a single bootable 720K floppy. Notice that we start off with only 464K of RAM despite 512K in the machine because 48K is used for vampire video (so a basic 128K system would only have 80K available). The use of MS-DOS 2.11 is particularly interesting because support for 720K 3.5" floppies didn't officially exist in MS-DOS until the OEM release of 3.2.
Besides the general tools of the period — but notice no on-disk BASIC — there are a few DG-specific utilities. The PRN_SET*.EXE utilities set the printer character set in use and VDISK.COM provides a RAM disk driver (DEVICE=VDISK.COM *xxx where xxx is the number of kilobytes, minimum 16K). This drive appears as D: and you can have multiple ones (E:, F:, etc.), but I found it much less useful with a two-drive system. I'm not sure what GRAPHICS.COM does; it only has a DG string in it and a small amount of code, and is not documented in the user's guide.
The second disk contains the tutorial. We'll boot up from this disk and spend a little time there.
Starting up. It also has a bootable MS-DOS 2.11.
The title screen. Despite some reports claiming 640x256 resolution, the hi-res screen is 640x200, just like regular CGA hi-res.
The main menu. The program alternates between graphics and text mode as appropriate. Although there are some inconsistencies suggesting work by separate authors at various times, it's easy to follow, and if you don't have the manual it provides at least the basics.
It also assumes relatively little of the user, though this baseline knowledge requirement does wander a bit.
I mentioned there is software contrast adjustment, though I would call it optimistic at best. The Cmd-PageDown combination lightens the screen and Cmd-PageUp darkens it (the "Try it now !" is blinking, so just ignore it and focus on the rest of the text). The text does indeed darken with higher contrast settings, but the rest of the screen darkens somewhat too, so the actual incremental improvement is small. I've chosen to go a little light here so the bubbles in the polarizer film are less in my face.
Part of the introductory text.
Communications is strongly emphasized. After all, this was the machine that would make you love to dial into your big Data General AOS/VS host again. Right? Right??
I'm not sure if the disk pack got updated for the 1200bps modem, but this answer is definitely wrong.
The demo disk culminates in a simulation of how you might connect to the Dow Jones News/Retrieval® service. To save your eyes, here is the complete session as we log on to find out how Data General's stock is doing tonight (spoiler alert: it's super and you should buy now). Notice that in those days this went through the national Telenet X.25 network, the same way you'd log onto The Source and many other online services. Boldface is what you would have typed. Your password is shown as @s. I'm not sure if the spacing shown here was accurate for the stock ticker, though.

CONNECT

TELENET
617 18C

Terminal=D1

@C 60942

609 42 CONNECTED
WHAT SERVICE PLEASE????
DJNS

ENTER PASSWORD
@@@@@@@@

DOW JONES NEWS/RETRIEVAL
COPYRIGHT (C) 1984
DOW JONES & COMPANY, INC.
ALL RIGHTS RESERVED


ENTER QUERY
//CQ
ENTER QUERY
DGN

STOCK    BID    ASKED
        CLOSE  OPEN     HIGH    LOW    LAST      VOL(100's)
DGN       40     48½       49½    47½     48½       967

DISC

Recall that Telenet addresses more or less resembled North American Numbering Plan area codes, so the local concentrator answering here was a node in area code 617, then the eastern half of Data General's home state of Massachusetts. Terminal code D1 is a dumb terminal equivalent. The prompt WHAT SERVICE PLEASE???? was not an exaggeration: others have documented this string on several nodes in area code 609, which was then southern New Jersey. These were apparently large timesharing minis that handled multiple services but I don't know who operated them. //CQ, DGN and DISC were Dow Jones commands for "current quote," Data General's stock ticker symbol, and to disconnect.

MS-DOS 2.11 basically works and particularly for Ones with small RAM ceilings may well be the best choice, but it lacks the creature comforts of later DOSes. With sufficient memory it should be possible to run at least MS-DOS 6.x and possibly even 7.x in a basic sense, but at the time I first played with it Caldera had just made DR-DOS 7.02 free for non-commercial use and I ended up using it instead. DR-DOS booted pretty well for the simple tasks I used it for until the front drive started complaining about files on the extreme periphery of the disk and at that point it went back in the Pierre Cardin bag. Let's get out the USB floppy and see what other alternatives we have.

The most obvious choice today would be FreeDOS, but aside from it wanting a hard disk to install to, that didn't end up working out so well for other reasons. Although over the years FreeDOS has intermittently come in unofficial one-disk variations, most of them unmaintained, nearly none of these could fit on a 720K floppy. However, one notable exception was ODIN, which specifically advertised a 720K floppy version that packed a fair bit of the disk with UPX. ODIN has long since disappeared from the Web but you can still get the last 0.6 release from the Wayback Machine in a version that was advertised as compatible with the 8086. Maybe it was, but not with the DG-1:
It would boot and start the kernel, and then the command prompt would appear to start, but as soon as you tried to press any key you'd get Invalid Opcode at 996C F000 F246 00CF 0070 0894 001E 00CF 0001 0004 0080 01C1 0070 and lock up so hard the machine had to be powered off at the switch. The "mainstream" 720K packed floppy didn't even get that far.

More recently was Svarog86, the precursor to today's SvarDOS. I would have used that except I didn't find an obvious 720K one-disk version whereas Svarog86 did indeed offer one. These use newer FreeDOS kernels than ODIN and both Svarog86 and SvarDOS prominently advertise their 8086 compatibility, which should work just dandy on the 8088 as well. Unfortunately, it blew up even more explosively when you touched a key.

Not only did we get a invalid opcode crash, we even got a divide-by-zero error: Invalid Opcode at 996C F000 F246 00D8 0070 03B6 0626 00D8 0006 0019 0100 0528 0070 followed by Interrupt divide by zero stack: 0528 0070 FOO6 0A0C 00D8 084D 0800 000A 0000 4353 0000 0000 0000. The invalid opcode trace in particular seems similar to the other fault in ODIN modulo the different kernel version, but I'm not aware of the 80C88 not supporting any opcodes the 8088 or 8086 do, so I don't know exactly what's wrong. Most likely there's just something weird about the hardware in the DG-1 that FreeDOS simply doesn't like, perhaps the absence of a conventional keyboard.

But DR-DOS isn't bothered by any of this, probably because it sticks to BIOS calls and doesn't assume anything about the hardware. So I decided to tune up the disk I had by removing a few unnecessary things like HIMEM.SYS (doesn't work on the DG) and SERNO.EXE (irrelevant) and TASKMGR.INI (hahaha). I had already added a simple editor (the Manx port of vi as used in Aztec C) and my own basic paging utility MOO.EXE that I had written a couple years earlier in Turbo Pascal, and with the extra space I added GWBASICA.EXE from TK Chia's rebuild of GW-BASIC using Microsoft's MIT-licensed source.

Indeed, DR-DOS runs very well on the Data General/One and pretty much everything works like it should.
And, as proof of CGA compatibility, here's a brief little test of SCREEN 2 in GWBASICA.

If you're looking for bootable floppies for your own classic Data General/One, you can download the original MS-DOS 2.11 floppies (both the boot disk and the tutorial disk) from the Internet Archive as 720K images. As for my enriched DR-DOS floppy, I think I'm on okay ground offering it for download given the various pieces' prevailing licensing but under these conditions:

If you accept these terms, you can grab it from the Gopher server.

While I think about what I'll do for the screen next time this comes out of the bag, let's finish our story, as we always must. While this updated release mitigated the display's issues, it did not end complaints about it, and more advanced (and more readable) LCD panels of similar dimensions started appearing on competitors' systems. By July 1986 PC Magazine had dropped it from its regular roundup of DOS laptops because "[t]he combination of low contrast and high price proved all but fatal to the DG/One and it's no longer a factor in today's market."

Data General apparently took feedback like this to heart. In October 1985 it slashed prices by hundreds of dollars while developing the next iteration. Waiting until IBM had fired its next bolt with the 5140 IBM PC Convertible in April 1986, DG introduced the Data General/One Model Two in May with a higher contrast LCD and two new SKUs using an ELD instead, one of which even had a 10MB hard disk and a full 640K of RAM. However, the ELD required so much power it reduced runtime to just one or two hours and the CPU was still the same 4MHz 80C88.

In May 1987 Data General upgraded the Model Two into the Model 2T. The CPU was bumped to the full XT speed of 4.77MHz (or could run turbo at 7.16MHz if you wanted), two expansion slots were in-chassis, and it came in both backlit supertwist LCD and ELD variants each also available with the 10MB hard disk option. The ELD version could not run on battery power, but the LCD version could, with advanced power management features that even could control current to individual laptop components. Despite complaints about the persistently non-standard keyboard, reviewers nevertheless heartily approved. "Finally," raved PC Magazine in July, "a modern laptop out of Data General!" In October 1987 InfoWorld lauded its performance and EMS capability and called the 2T "well-thought-out" and "a very good value." The Model 2T was adopted internally at Data General as the "Pharaoh" testing and diagnostics rig using a large attached sidecar module, and it was also rebadged as the Allen-Bradley T45 Portable Programming Terminal. In February 1988 DG added a new backlit blue LCD and a 2400bps modem option.

These bright spots notwithstanding, the size and design of the DG/One even in its more advanced iterations were no longer popular with the market, and it was quietly discontinued around 1991. In fact, its replacement had already arrived two years earlier: the five-pound 1989 Data General Walkabout, which had its own built-in modem and ROM-resident terminal and text editor, plus a phone book, dialer and alarm scheduler and ran on AA batteries or Ni-Cad. Unfortunately, buyers weren't interested in a portable terminal either, and an attempt to expand it into a 386 laptop with the Walkabout/SX (also rebadged as the Allen-Bradley T47) and successors failed to sell due to weight, price and ergonomics. Data General cancelled the line in 1993, and with its only major remaining success in storage systems, DG was bought out by competitor EMC in 1999 and quickly dismantled — in the process selling its venerable dg.com domain name to Dollar General. DG's remaining IP now rests with Dell, who bought EMC as a subsidiary in 2016.

3 comments:

  1. You mentioned SvarDOS, what about trying a latest 720K dist from SvarDOS on this?

    ReplyDelete
    Replies
    1. There's no single-floppy version of SvarDOS I'm aware of, and a single install disk isn't much use to me. (I also suspect that whatever the kernel doesn't like about the DG-1 hasn't changed.)

      Delete
  2. Hi. I have acquired one these on the just recently the DG One, and the expansion unit pictured above in the black and white picture with all the other peripherals.

    I am assuming there are not too many of these left, going to get it serviced and connected reasonably soon and will pop it into a few retro groups. Just a great piece of history to have.

    ReplyDelete

Comments are subject to moderation. Be nice.