Atari video circuits experiment

[Bryan]

I've been experimenting with the Atari's video circuits (using the SuperVideo articles and others as a guide), and I've come up with the following observations:

 

First, some basic stuff:

 

The Atari 8-bit uses a Hue-Luma (H-L) color model, with a fixed Saturation (S) level (which, as we'll see, doesn't actually translate to a fixed level).

 

(Hue is the base color, Luma is basically brightness, and Saturation represents a point between a pure color and gray/white, like when you turn the color knob down on your TV, you reduce Saturation)

 

Luma is generated as a voltage level that varies across a line. When a bright pixel is being displayed, the level jumps up, and back down accordingly. This is your B&W base picture.

 

Chroma is generated as a wave at a fixed frequency (3.58MHz for NTSC, 4.43MHz for PAL). The rest of this discussion relates to NTSC Atari's since I don't have any experience with PAL (yet!).

 

The wave has 2 properties that determine how the B&W image will be modified, phase (hue) and amplitude (saturation).

 

If the wave is moved to the left or right over an area, the color changes to a new hue. If the wave drops out, you get gray. If the wave is taller, you get more saturated (pure) colors, and if the level drops, you get less saturated (paler) color.

 

Back to the NTSC Atari...

 

The Saturation value (amplitude) is normally varied on dark and light images of the same color (less saturation is needed for color on darker images). But there is only one level produced by the Atari, so darker colors are more saturated than bright ones. This is why the more you jack-up the LUM value, the paler the color becomes.

 

Over saturated dark colors receive an artificial boost in brightness as well. This is why Setting LUM to 0 does not produce black when color is used. Even though there's no Luminance being generated (should be black), the high saturation value produces a color that "blooms" a bit, producing an ill-defined (from lack of LUM) dark color.

 

(to be fair, few systems of this era could produce such a range of colors at all)

 

This is what causes color "shadows" on the 8-bit. If a bright pixel of one color is followed by a dark pixel of another color, a dark gap of the true LUM value will appear between them. This is because it takes the color circuits a short while to respond to the shift in the color signal and boost the pixel.

 

(to see the shadows, let a BASIC blue screen go into Attract mode, and watch the color transitions along the edges as different colors and shades are introduced)

 

The elaborate 1200XL circuits sought to increase the satutation of bright colors by boosting the Chroma signal somewhat. This does increase color purity on bright colors, but also helps create shadows on darker colors.

 

The solution?

 

A Chroma signal that varies with the Luma signal. Such a circuit would produce the sharpest, best color signal at all intensities without the shadows.

 

The down side is that it would change the relative brightness of colors to gray shades (especially at darker shades, where oversaturated colors receive the artificial luminance boost).

 

The compromize I'm considering is a circuit that leaves lower shades alone (so they look as they've always looked), but boosts saturation as the brightness increases. This would lead to much more vibrant display, with shadows at their current level (although minimized by cleaner video circuitry).

 

Anyway, I've come up with video discoveries as well, I just wanted to pass along this info.

 

-Bry

 

P.S.

 

The Antic color modes have a resolution of 160 because the NTSC 3.58MHz clock is used to clock the pixels as well, so the color carrier goes through one complete cycle every pixel (on PAL the relationship between pixels and color is different). On the 320 modes, GTIA still technically generates color at 160, but uses a different LUM value for each half of the color cycle.

 

The blurry color circuits in the XL (and other models to a lesser degree) are 'almost' right in most respects, but suffered from what I think is the relegation of analog engineering duties to digital engineers who don't understand all the math involved with amplifier and filter circuits.

[Shawn Jefferson]

I've done the Supervideo upgrade on two 800XLs and one 1200XL. I noticed that the two 800XLs vary significantly in the quality of their output and I don't know if that is something to do with my work, or something else.

 

Have you checked out Bob Woolley's recent work with the 1200XL:

 

http://www.retrobits.net/clearpic.html

 

I haven't done it, but it is supposedly better than the SV upgrade.

[Bryan]

I've done the Supervideo upgrade on two 800XLs and one 1200XL.  I noticed that the two 800XLs vary significantly in the quality of their output and I don't know if that is something to do with my work, or something else.

 

I've done several machines too. There shouldn't be any noticable difference. Somthing may not be hooked up right.

Have you checked out Bob Woolley's recent work with the 1200XL:

 

http://www.retrobits.net/clearpic.html

 

I haven't done it, but it is supposedly better than the SV upgrade.

 

I've looked at it, but he removes the saturation boost circuit entirely, and bypasses the input buffer (clipper) transistor. This leads to a messier looking chroma signal.

 

The chroma amp is made up of 3 transistors in the XL (including the 1200). The 1st is used to remove the DC offset from the chroma signal (it isn't the pretty wave it should be straight out of GTIA) and to make it look nice and smooth. The 2nd is a gain amp (common emitter amp), to boost the level and the 3rd is the buffer amp (emitter follower).

 

I've tried removing components but it usually leads to a degraded signal.

 

The 1200XL also has a 3 transistor circuit before the chroma amp, that varies the gain of the 1st stage. What it does is hold the gain low during the colorburst (off screen color & level calibration) part of the signal, then ramp it up during the visible line, for better saturation. This saturation circuit is a bit more complicated than it has to be, but it works.

 

The XE's have an entirely different (deceptively) simple circuit with only 1 transistor. The DC offset is NOT removed from the Chroma signal (which doesn't seem to offend most monitors) but it is later removed before it's mixed with Luma in the RF modulator.

 

The difference between removing the DC offset and not doing it is this:

Normal chroma signal (below) 

          __    __    __

__off__   |  |  |  |  |  |__off__ 

      |__|  |__|  |__|





      |pixel|pixel|pixel|

     	

__off__    __    __    _____off__

      |  |  |  |  |  |  

      |__|  |__|  |__|

      

GTIA uncorrected signal (above) 

 

(Suffice to say that there are only 2 digital levels from GTIA, when there need to be 3 levels, OFF (halfway), HIGH and LOW. Doing this cleanly is tricky. Thanks to jsoper for the info on using a 'code' tag.)

 

Both contain the 3.5MHz comoponents needed to derive color, but only the 1st one, when added to luma, results in an average value that is the same as the original luma value. The 2nd one will one subtract from lum, making a slightly darker picture, and bright or dark spikes may appear at the edges of the colors from the DC filters in the monitor.

 

Anyway, I'm looking for the cleanest way possible.

 

-Bry

[jsoper]

This may not completely relate, but I have s-video mod instructions for an XEGS on my website (link below). The XEGS has a similar production schematic to the XE, but different component numbers. The mod looks great on a 1702 monitor (most things do), but I've never tried it on a modern s-video tv.

BTW, you can do fixed width ascii graphics with the code tag. I'll post the production XEGS schematic as an example (since we're on the subject already):

                 

                                            

                  5v--------+-------------------,

                            |                   |

                      1.6K  |    51             |

                    ,-///-+--///--,         |

                    |  R81      R79   |         |

                    |                 |         |

         luma plus  |        ___/-----'         |               To Modulator

         sync lines |       /|/c 3904          |                   Pin 2 --------,

                ----+------(b|   ) Q3        ___/                                 |

                    |       |e/           /|/c 3904                            |

                    |           -----+----(b|   ) Q4                             |

                    |                 |     |e/    33      18                   |

                    |                 |         --///-+-///--+------|(-------+

                    |  2K       100   |             R53  |  R52   |      C55      |

                    `-///-,--///--`                  |        |               |

                       R80  |   R78                      |        |  1k     1.6k  |

                            |                            |        `-///-+-///-'

                        GND /                           |           R51  |  R93

                                                         |                |

                                                         |            GND /

                               5v---------,              |

                                          |              |

                                       ___/              |

         GTIA    5.6k    _______      /|/c 3904         |

         P21 ---///---|_______|----(b|   ) Q5          |

                 R55        L5        |e/              |

                                          ---+----|(----'    

                                              |    C44

                                    180       |

                             ,-----///------'

                             |      R54

                         GND /

[Bryan]

BTW, you can do fixed width ascii graphics with the code tag.  I'll post the production XEGS schematic as an example (since we're on the subject already):

 

Thanks... I'll go back and edit my post...

 

-Bry

[bobwoolley]

Yes, I did remove the color boost circuits in the 1200XL. The way they do it, you can just crank up your COLOR control for the same effect.

 

The big problem in the 1200XL is the poor ground returns. This leaks CHROMA into the LUMA circuits and digital noise into everything. You cannot just drive video current into the outputs without getting ground noise.

 

You get transition colors because the video level changes the LUMA at the beginning of a clock cycle before the CHROMA phase has been decoded. The later the phase change, the worse the transition error. You could fix this by storing the RGBY decoded values in the monitor until the next cycle, but most (all, that I know of...) just run the LUMA thru a short delay line. This is why the DVD component video looks so good, the CHROMA is already decoded.

 

Bob

[Bryan]

Yes, I did remove the color boost circuits in the 1200XL. The way they do it, you can just crank up your COLOR control for the same effect.

 

The big problem in the 1200XL is the poor ground returns. This leaks CHROMA into the LUMA circuits and digital noise into everything. You cannot just drive video current into the outputs without getting ground noise.

 

Hey Bob! Glad you're here.

 

Yeah, the messiest looking circuits appear to be those of the 800, though the picture quality isn't that bad. Much of it looks like it shouldn't work.

 

You get transition colors because the video level changes the LUMA at the beginning of a clock cycle before the CHROMA phase has been decoded. The later the phase change, the worse the transition error. You could fix this by storing the RGBY decoded values in the monitor until the next cycle, but most (all, that I know of...) just run the LUMA thru a short delay line. This is why the DVD component video looks so good, the CHROMA is already decoded.

 

Bob

 

What is the proper alignment for Chroma vs. Luma? I thought they were supposed to be aligned and the TV would do the delay. I noticed that the TV seems to get color alignment right when the luminance changes as well. Wouldn't all color be moved to the right?

 

I might experiment with clocking the LUM pins to slow'em down.

 

-Bry