First, Be Useful (Home computers and Pico Projectors)

For today’s blog I am going to be setting up an analogy for what I see going on with pico projectors today.  Those that want to skip the history lesson and just learn about pico projectors may want wait for my next posting.  But just to entice you to read on, at the end of the appendix to this story, I tell of my meeting with Steve Jobs a few months before he bought Pixar.

The difference between a new electronic “necessity” and a “gimmick” comes down to being useful to the consumer.   But sometimes products start out life as “gimmicks” because the technology has not advanced enough to build a really useful product.

I lived through an example of this in the late 1970’s and early 1980’s as one of the designers of TMS9918 Video Display Processor or VDP* (see appendix for more on the 9918 and its legacy) that was used in the TI Home Computer, Japan’s Ascii-Microsoft’s MSX Computer, and Coleco’s Adam computer (the 9918 was put to better use in Colecovision).   They, along with the Radio Shack TRS-80, Sinclair ZX, Commodore-64, Amiga, and Acorn computer are all long gone (although the CPU inside the Acorn lived on as ARM when it was turned into a “core”).    I remember in the mid 1980’s the pundits were saying that failure of home computers proved there would never be a big home computer market and that they would be relegated forever to the techno-geeks.  Of course since you are reading this on what they said would be a non-existent product, things turned out a bit differently.

In reality, there was no right way to build a home computer in the early to mid 1980s as the technology to build a “real computer” was too expensive and the “use models” had not been figured out.  The prevailing theory was the “home” meant “cheap” so companies tried building these home computers on the cheap, and sometimes it looked like being cheap was succeeding with products like the Commodore-64 and the Sinclair, but the success was fleeting.   To be cheap mean having audio cassette tape writable storage (storing the data like a modem signal on audio tape — it was primitive even in its day), a low performance CPU, a very small amount of RAM, and low resolution television output.  The 64 in the Commodore-64 stood for 64K bytes of total RAM and originally sold for the “only” $595 in 1982  (about $1325 in today’s dollars!).

By this point, those that know their early home computer history probably think I left out one computer, the Apple II that managed to outlive all the others of this era by a wide margin.  It was more expensive but it supported a comparatively inexpensive floppy drive (the one-eyed man in the land of audio cassette storage) and had expansion slots (another extravagance).   With expansion slots people could plug in a more powerful CPU running CP/M using the Apple II’s CPU as essentially a keyboard scanner and with the floppy drive they could store their results.  In fact one could seriously argue that it was Steve Wozniak’s inexpensive floppy disk driver design (What Woz himself called  “the finest job I did“) that had more to do with the long term success of the Apple II than main computer.   The IBM’s PC then followed the Apple model with floppy drives (and soon thereafter with hard drives) and expansion slots.   The Apple II was more expensive than all the other home computers of the era, but it was capable of being useful and that led to it being used in both business and education.

So what does this have to do with pico projectors?   I see today’s pico projectors being like those early home computers, they just aren’t very useful.    The ones that are selling the most today are cheapest with low brightness and low resolution.

One problem is that a useful pico projector is not affordable today.  Even at any price with today’s technology one cannot meet the size, efficiency, and brightness requirements.    A second less obvious but perhaps more important problem is that the “use model” or the way the projector is being use, is all wrong in most designs.   Take it from me, I have been all over the world with pico projectors and shooting large images on wall is just a silly way to use them.  There is rarely a wall that is white or close enough to white, at the distance from the projector that makes the right size image and the room is not dark enough for a 20 lumen pico projector to make an image larger than a letter size sheet of paper.   Yes, there are some exceptions, but you can’t have a large market selling to people with exceptional cases, it has to useful to a large percentage of people on a regular basis.

This is not to say that a few million pico projectors can’t be sold a year.  In a world with billions of people, a few million can be sold.  But to get to the 10’s and hundreds of millions things are going to have to change.

So the rhetorical question is, “other than being too expensive and not that useful, what is holding back the pico projector market?”  But don’t despair, next time, I will discuss how I see the technology and use model changing for pico projectors.

. . . to be continued


For those that are interested in more about the TMS9918 (and 9928/9929/9118 and 9129) you can do a web search and find out a lot more about it but a couple of quick references are the Wikipedia Article, the Data Sheet (a good bit of which I wrote), and even a paper I wrote in 1980.

The term “Sprite” was first coined by the 9918 (before that they were call “player graphics” and were less capable than Sprites). I didn’t come up with the name “Sprites,” that term was coined by a manager named David Ackley, but I was one of two engineers that did the sprite architecture and the display processing and I did most of the logic design for the first Sprites.

The 9918 was also remarkable for being the first consumer device to directly interface to DRAMs and indirectly affected the SDRAMs still in use in just about every computer today.   You see,  I defined the DRAM interface for the 9918 (and later the 9118) and this later led to my work on the Multiport Video RAM (VRAM), the first form of Graphics RAM, and later to my helping define the very first commercial Synchronous DRAM or SDRAM/DDRAM/GDRAM used in just about every computer today (a story for another day).

A major legacy of the TMS9918 what that it was used the very successful Colecovision game system and one of the software developers for the system was Nintendo.  Nintendo worked with Yamaha to develop the register level clones (it had the same Sprites, only more of them) and upgrade to the 9918/9928, the V9938 and V9958 (notice the numbering), that were used by Nintendo in their highly successful video game system.

Some more “Connections” (ala Jame Burke):

Pico Projectors: My work in the 9918 (my first project at Texas Instruments) later led to my work on graphics and the Video RAM, a DRAM with a shift register to support display refresh.  Video RAMs were used as “write only memories” (although they could be read) in that the processor wrote via the random port and the memory was dumped to the display device via the serial port.  When I first got into LCOS microdisplays, my comment was, “this is like a video RAM only in reverse, the data goes in the serial port and then it appears on the display.”

Pixar and Steve Jobs Connection: The Video RAM was used by the then computer graphics hardware company call Pixar owned then by Lucas and this resulted in my meeting with them several times at Siggraph in the mid 1980’s (I presented and what on panels there in the 1980s).  To show off their hardware capability Pixar always had these great videos that they showed at Siggraph on “movie nights.”

When I met with Steve Jobs just after he founded NeXT (they will still moving into the building) he wasn’t much interested in the new TMS34010 graphics processor I was there to show him and brushed it off with the question “is that the greatest graphics you have ever seen?” to which I truthfully answered, “no, it was being built to meet a PC graphics card price point.”  Steve then pressed with, “then, so what is the best graphics you have ever seen” so based on having met with Pixar several times including a party they held at the previous Siggraph and having seen their great videos, I told him, “Pixar had some some great stuff at Siggraph” and a few months later he bought Pixar.  Ok, he probably didn’t buy Pixar on my advice and I later realized that he was probably fishing for confirmation of a deal that was in probably already in the works, but my saying “Pixar” in answer to his question probably didn’t hurt :-).

By the accounts I have read, Steve Jobs was after Pixar for its hardware not its its to make compelling animated cartoons.  Funny how things work out.

Karl Guttag
Karl Guttag
Articles: 247


  1. Karl, am I correct in assuming the 9996 chip in the TI XDS/22 I have is a is a 9995 variant? It’s 64-pin DIP as apposed to 40 like 9995, so I’m assuming it likely has either a non-multiplexed data bus OR extended addressing. Are there any instruction set differences or unique architecural features? Or am I completely off base here? I can find no reference to this chip except for a comment you made in a Delphi TI Forum conference, IIRC, a long time ago.


    • Funny enough, as you say I was asked that question in 1992 or 21 years ago. My answer back then (which I think is correct) is that it was a 9995 variant where they brought out the internal 16-bit bus. I didn’t work on that variation but I think I vaguely remember it as I was back in the U.S. working on the 99000 (which shared a lot of common logic with the 9995) when it was done. My best guess is that it was a very quick spin where they just buffered and ran out the internal 16-bit that went to the internal 16-bit wide RAM.

      From the Q&A I think your are referring to from 1992:!msg/comp.sys.ti/QlLThkrm8Po/GtnkyhtF4nMJ

      I think that might be a 16-bit bus version of the 9995. The
      9995 had a 16-bit internal bus (that when to 256-bytes of
      internal 16-bit wide RAM) but brought out an 8-bit external
      bus (ala the 8088 or 68008). It may also have been a special
      chip for supporting our hardware emmulator (or both).

Leave a Reply

%d bloggers like this: