Talk:TMS9918

"RGB" output and RAM

The "2" versions do not output Red, Green, Blue video. They output Component_Video of the type where color (chroma) information for only the Red and Blue is output, and the brightness Y (luminance) signal is taken from the Green signal. Green color information is derived from the R and B color, and the Y brightness.

See YPbPr. This is sometimes seen as R-y, B-y, Y.

Computers using the component video version can be connected to a component video monitor for color, or the Y signal can be connected to a composite video monitor for greyscale.

The article also is unclear about the 16k of RAM, making it read as though the RAM is internal to the VDP. The RAM is on an external bus.

As used in the TI-99/4 and TI-99/4A computers, the 16k RAM was the VDP RAM, while the 9900 CPU had only 256 BYTES of RAM to work with. The design was such that the powerful (for the time) CPU was reduced to being a data shuffler for programs running on the VDP. Only when running pure assembly language programs, which required the 32k expansion, was the CPU brought into full use.

Dates

When (what year) was the chip released? Is it still in production and if not when was it discountinued? RJFJR (talk) 03:16, 10 December 2007 (UTC)

• As far as I know, the TMS-9918, as used in the TI-99/4 computer, was released in 1979. When the TI-99/4 became the TI-99/4A in 1981, the chip was replaced, as a result, by the TMS-9918A. It was this version of the chip that had been well-known among early computer and videogame fanatics. The TMS-9928A (NTSC) and TMS-9929A (PAL) chips, which are the ones that address the R-Y and B-Y color signals as well as luminance, were released in 1982 (the ColecoVision and its computer counterpart, the Adam, both employed the TMS-992xA series, among others). WikiPro1981X (talk) 09:08, 30 August 2009 (UTC)

Amateur's Tool

This section needs a serious revamp, or at least someone to proofread it. Salgat (talk) 01:23, 29 May 2008 (UTC)

Fixed colour palette...?

I guess that's what the (unlisted) "additional modes and registers" of the Sega SMS/GG/MD were for? To allow a wider master and per-sprite/tile colour palette vs the Coco etc? IE 16 for the background alone, another 16 usable by both sprites and BG, taken from a gamut of either 64 (SMS, MD) or 4096 (GG) colours? Would it be relevant for me to add these details in? Or at least a link to those systems' articles' spec sheet sections? 193.63.174.211 (talk) 09:12, 4 August 2014 (UTC)

VDP defn?

The Acronym VDP is used, but never defined. What does VDP stand for? — Preceding unsigned comment added by 2601:1C0:8381:2C20:65C4:240C:4D00:AAC8 (talk) 05:56, 7 June 2022 (UTC)

Colors, gamma and CRT display

I split the color section into two. One for values sourced from the datasheet, and another for colors as displayed on a CRTs.

This second section (CRT display) is interesting and yes gamma must be converted, but all those calculations might be considered Original Research and challenged. It needs simplification and references. 4throck (talk) 12:58, 25 May 2021 (UTC)

In principle you're right. But if you intend to document how the colors of the TMS9918 series chips originally looked like, you need to explain why only referring to the datasheet would nowadays result in a wrong impression. TI994A (talk) 15:06, 25 May 2021 (UTC)

Agreed. Your contributions are great! But the article is not the place for all those calculations. We only need to show the final table with the converted hex values.

Doing the mathematical demonstration is original research and outside the scope of the article. Perhaps you can move the calculations here, to the talk page? Seems a better place.4throck (talk) 17:42, 25 May 2021 (UTC)

In my opinion it's not pure mathematics, but rather describing the physics of the chip. Especially you can see it at the remarkable example of the color 9 "light red". On the other hand I admit that this differentiation is somewhat pedantic.

Are you able to provide an example within another Wikipedia article, where the main text had "outsourced" some parts into the talk page, so that I can get an impression about best pracice how to do that? TI994A (talk) 19:30, 25 May 2021 (UTC)

Hey, your new amendments are very well done! Greatly appreciated! TI994A (talk) 14:35, 26 May 2021 (UTC)

Some problems with your math there, first you're using the wrong formula for Y, R-Y and B-Y to RGB, the formula linked is the HD TV spec, not the same as SD colour space, and assumes 0.7Vpp chroma signals, not the 1Vpp values coming out of the chip / used in the datasheet (and even assuming standard SD colour space was the intended result is dubious). Regarding gamma flat panels do have gamma, sRGB curve (approximately 2.4 average gamma but not flat) is generally assumed whenever hex values are used, not linear. Also CRT TVs were typically well above 1.6 gamma, BT.470 spec for system M regions (aka the places that used NTSC) specifies a 2.2 target. jamvanderloeff 15:15, 4 October 2021 (UTC) — Preceding unsigned comment added by Jamvanderloeff (talk • contribs)

First consumer device using DRAM; DRAM too expensive for a frame buffer

The Karl Guttag source appears to be nonsense in its claims that, "Back in the late 1970s, only very high-end CAD computers could afford a frame buffer. The TMS9918 was the first consumer device to use DRAM...." The Apple II, which was released two years earlier (June 1977), used both DRAM and a bit-mapped frame buffer for its high-res graphics. Further, sprites are not what saves memory for graphics but instead using "characters" is what saves the memory if your set of character glyphs is small enough that you cannot fill every character slot with a different character glyph. (This is obviously not true of Mode 2.) Because of these obvious issues, I've removed from the article the statement that "The TMS9918 was the first consumer device to use DRAM. Because enough DRAM to store a complete frame buffer was too expensive at the time, the TMS9918 uses sprites." Cjs (talk) 18:11, 11 August 2022 (UTC)