Q&A with Dr. Dave Bradley
Start of the exchange
At 04/05/2025 Dr. Bradley Wrote
I would be happy to answer any questions I can about the Datamaster. But it's been nearly 50 years and I have no documentation. I was there at the beginning and can tell you why certain things were done. I was involved completely with the BIOS and reviewed large amounts of the BASIC and Diagnostics for the machine, but I don't think I can answer anything very specific.
For instance, there is a 3-key sequence in Datamaster that will invoke a simple debugger. I don't know whether that made it into the final production, or even which three keys did it. But that is the origin of Ctl-Atl-Del on the IBM PC. In fact, the answer to almost any question of "why did the IBM PC do it this way?" can be answered "Datamaster."
About component relabeling
I have some questions about this machine. Since we got at home our first System/23 I have been studying the motherboard and the keyboard, and therefore I have some questions nobody but the developers can answer. The Datamaster is a very misterious machine, but this obscurity has played against it in the long run and for this reason not so many units remain, from those even less are working. And so comes my first question: Do you know why this computer's components were relabelled? The truth is that it goes against the doctrine of using the standard parts they used. I have seen also relabelled parts in an AS/400 though.
At 21/05/2025 Dr. Bradley answered
I don't know directly why the parts are relabeled, but I have a guess. IBM certified their vendors, and this was true of common components. IBM certified the processes by which they were made, and probably asked that components made on this verified process be labeled as such -- hence, relabeled parts. This was probably done for all parts destined for IBM. I think.
About the ROM memories corruption
Then there is one of the main reasons a Datamaster can fail, which are the ROM memories. They tend to corrupt very easily and it is very easy to see Datamasters with broken romsets. For example, if you looked at the Wikipedia, the main image of the computer shows a 5322 with the 12h error on the POD screen. Those memories were sourced from Motorola and Mostek. Wouldn't you know by any chance why those memories were selected from them instead of other ROM makers?
At 21/05/2025 Dr. Bradley answered
Motorola and Mostek were probably selected from a bidding process, with their components verified. I doubt that a 50 year lifetime was part of the validation process.
About components from the Datamaster present in the PC prototype
Other than the Datamaster, I would also like to talk about the PC prototype (not the final product) and its relation with the System/23. I have two questions about it. I saw the power supply connector was maintained from the Datamaster to the prototype, but was replaced in the final product, although the signals are the very same. Do you know what was the main motivation to replace it? Then, there is the DRAM matter. I saw in the prototype that the memory used in the prototype was the 4332 (also used by the Apple III), pin-compatible with the 4132 that the Datamaster employs (I was able to determine it when studying the 64KB memory card). However the final product used its predecessor, the 4116. Does this reversion have to do with the release of the 4164 as a way to close distance between both chips trace layout?
At 21/05/2025 Dr. Bradley answered
the prototype was put together with Datamaster leftovers and whatever new components were necessary (8088, obviously). The PS and connector was one of those. Why it changed I don't know, although it may have been a space issue. Narrow connector along the side of the board rather than the fat one elsewhere? Don't really know. As for RAM, the prototype board has two banks of 32kb, while the final board has 4 banks of 16kb -- so we could sell a 16kb stripped version for an advertisable low price. That's why you have 4332s on the proto and 4116 on the final -- 32kbits vs 16kbits.
The key combination you stated was the precursor to Ctrl-Alt-Del wouldn't be by any chance Command-Test-Erase? This is a well known combination in the Datamaster but haven't noticed before how similar both combinations are.
At 21/05/2025 Dr. Bradley answered
I don't recognize any of those keys on the s/23, but it's been 50 years, and I've spent all of that time looking at the IBM PC keyboard. And C-A-D.
About the Datamaster clock frequency
I would like to inform you about the ridiculous situation I am involved in, and one of the reasons of why I have contacted you. It all started when we (I collect computers with my father) acquired our first unit, a 5322-124. It didn't work, so I had to conduct research about the computer, and the culprit of the issue were some capacitors in the 64KB DRAM board. That card was fixed and later studied. And there goes the myth that the Datamaster has redundant memory, which I found it was false when I determined the type of RAM IC it was carrying was the piggy-backed TMS4132. Then I wrote to all sites that were spreading the rumour in order to correct it. This has happened a few times during my interactions with the Datamaster hardware. One of them is an error that originated at the English Wikipedia, then spread to the other Wikipedias, then spread to third sites and finally Wikipedia used those sites as references. I am talking about the microprocessor's clock, which according to Wikipedia was 4.77MHz and should be 6.14MHz / 3.07MHz effective.
At 21/05/2025 Dr. Bradley answered
I can't verify the S23 clock speed, but what you have sounds reasonable. It's not 4.77MHz because that's the PC, which is 14.31818 MHz divided by 3. 14.31818 was chosen to generate the color burst frequency of 3.58MHz for use by the CGA with a home TV.
About the DRAM cards of the Datamaster
I would like to know what motivated the design team to structure the RAM in cards instead of placing them in the motherboard, please. I am aware that while the banking mechanism allows for a theoretical 256KB of DRAM, the detection and identification mechanism of such cards prevents the inclusion of more than 128KB of DRAM in cards. I am also aware that are other ways to place RAM in the Datamaster, but will leave those for a later time. My question about the mechanism is if you know why an identification for 128KB signal like the one used for the 64KB RAM board wasn't created to profit from memory over the 128KB boundary.
At 21/05/2025 Dr. Bradley answered
I got no idea
About the Light pen test
Then, when writing the emulator I found test 05h (CRTC + DMA). It has sparkled some other questions to me. The first curious thing about this test is the video interrupt signal. The trace between the interrupt RST5.5 and the 8275 /IRQ is cut and the interrupt is checked using polling. What is the motivation behind that behaviour? Then, in the same test, I found the light pen test. It was a pain to trace and describe in the MAME source, but I am proud of having found it. I imagine this test was done to check if the DMA device was working and delivering data to the CRTC; is this correct? I also found the test signal to be gated with a pin in the video connector. Such pin is not documented in the service manuals of the machine. Does it means there was an initial requirement for a light pen that was eventually dropped?
At 21/05/2025 Dr. Bradley answered
I didn't write the diagnostics, and only proof-read/optimized the ones in ROM. I do know that the Customer Engineer types who wrote the code did everything they could to test something in two different ways to verify everything they could. There may have been a light pen design early on, but I'm pretty sure it went away. Same thing happened with the PC --we made allowances for a light pen and they never caught on. I'll suggest that a pointing device like a mouse removed the need for a light pen, and mice appeared the late 80s.
About the Datamaster and MDA
This one is a question about the MDA video subsystem and its relationship with the Datamaster video subsystem. To be fair, I have had my dose of research on the issue and found the two subsystems, while they serve the same purpose, are designed and act in ways completely different. The Datamaster video subsystem is based around the i8275 and the i8257. The i8275 does not generate video addresses, so it relegates the task to the DMA device. MDA on the other hand is based on the MC6845, which does generate video addresses. This fact changes the design of the video subsystem radically. Also, the i8275 has internal queues of 7 bits each, element not present in the 6845, and which limits the characters to 128 per bank (there are seven banks of characters plus one empty). MDA was designed to bypass this specific limitation, at the same time some of the hardwired characters of the i8275 were transfered to the MDA character generator ROM. The pixel clock also differs between both subsystems (with the Datamaster having two different pixel clocks).
In short, we have two very different implementations of the video subsystem, however it is claimed that the MDA is based on the Datamaster video subsystem. Tracing the source of the claim was one of your interviews. Please, could you clarify what it is meant that MDA is based on the other video system even when there is nearly no common point?
On 28/06/2025 Dr. Bradley answered
If i said the mda came from dm, then i was wrong. Your analysis of the dm video seems right on as i recall it from 50 years ago. The idea of 80x25 mono video does translate from dm to mda. I can't recall why we didn't like the intel chips , probably the limitations you indicated. Pc was intended to be more open and accessible while the dm was closed. Or, using 6845 on both mda and cga was a better solution...made them more alike . Intel chips wouldn'thave done the graphics.
About the Datamaster word-processing card
There is also a doubt about the expansion cards. There is the word-processing card, which contains a second i8275 and possibly a 8048 as well as 32KB of extra memory. However, in most cases there is a second board which is identical to the word-processing card minus some components (for instance, the i8275 is removed). With no direct connection with the word-processing card and no connectors other than the pre-ISA bus, the purpose of that card remains unknown. Please, if you know what that card is, could you state what is its function?
On 28/06/2025 Dr. Bradley answered
I know nothing about the word processing card. Might have been a different bunch from ibm building a special pirpose thingie.