Atari RGB to Composite Video Converter

RGB TO COMPOSITE VIDEO CONVERTER DOCUMENTATION

December 27, 1986

version 1.0

The following document is placed in the public domain. You may make as many copies of it as you like and transmit it in any form you want provided it is not sold commercially, nor any product derived from it is sold commercially. The author is not responsi- ble for any damage, physical, mental or otherwise caused by fol- lowing the instructions given below. Please mail corrections and/or suggestions for improvement to the address given below.

Anees Munshi
58 York Road
Weston, Ontario
M9R 3E6
Canada.
(416) 246-0670

Introduction

The schematic described in this document converts the analog RGB video signal that is output by the ST to an NTSC composite signal which can be displayed on a colour or monochrome composite monitor, or on a TV set by adding a modulator. Please be careful when building the circuit. Should anything in the schematics or this documentation seem suspicious, use your better judgement. A certain amount of experience at building electronic circuits will be very helpful. Also, a good oscilloscope and knowledge of the theory of RGB to composite conversion may be necessary in order to debug the circuit (should you have to). I have tried to pro- vide some background in this article.

An Overview

The MC1377 RGB to composite video converter IC used does most of the work in converting the red, green and blue signals to composite video. The red, green, blue, horizontal sync and vert- ical sync signals are taken from the ST and fed into the 1377. A colour burst carrier (3.579545 Mhz) is fed in from a separate oscillator [1]. The 1377 generates the R-Y, B-Y and luminance signals by passing the red, green and blue signals through a matrix. Then, the B-Y signal is modulated using the colour-burst frequency carrier, and the R-Y signal is modulated using a 90 deg. phase shifted carrier. This results in the I and Q (in- phase and quadrature) components of the chroma signal. The two components of the chroma signal are added and amplified and made available on pin 13 of the chip. This allows the chroma to be band-pass filtered [2] externally and then fed back into pin 10. The band-pass filter should be centered at the colour-burst fre- quency and should have a bandwidth of about 1.6 Mhz. Not having a very good colour TV to experment with, I chose to do an el-cheapo filter since it wouldn't make any difference on my set anyway. (Besides, I plan to use my board on an old green-screen monitor to run an occasional colour-only program). A simple second order LC filter may be used. Set the resonant frequency of the tank to the colour-burst frequency, and choose an appropriate R so that you get the 1.6 Mhz bandwidth [3]. Introducing a second (or higher order) BPF to do chroma-filtering will probably result in a visible delay on the chroma-signal. So, to make sure the colours are not offset from the B&W image (like in sloppily coloured comic books), the luminance signal must be delayed an equal amount so that the luminance information is not ahead of the chrominance information. To allow this, the luminance signal is looped out from pin-6 to pin-8. An approprate delay line inserted between pin-6 and pin-8 will create the required delay [4]. The luminance signal does not need any filtering.

Interfacing

The horizontal and vertical synchs from the Atari must be combined and fed into the composite synch input of the 1377 (pin-2). The HSYNC and VSYNC are taken from the ST's monitor out- put, AND-gated and fed into the comp-sync input. This works fine since the syncs are active low, TTL level signals and satisfy the (-0.6V, 0.9V) active and (1.7V, 8.2V) inactive threshold levels.

The colour signals, red, green and blue must be capacitively coupled through 22uF capacitors and attenuated through 2-4 Kohm resistors (in series with the input) so as to not interfere with the chip's bias and satisfy the 1Vp-p signal requirement respec- tively. All the three colour signals output by the ST have a 1.8Vp-p range with a 1.2V DC bias. 1377 inputs: Pin 3: red input; Pin 4: green input; pin 5: blue input.

The colour-burst carrier signal is generated as shown in the attached schematic and coupled to pin-17 through a 2.2Kohm resis- tor and a 0.1uF capacitor in series.

The colour-burst is added after every sync pulse, (burst is not suppressed after VSYNC) approximately 5.5us (micro-seconds) after the sync's leading edge and it lasts for approximately 3us or 10.7 cycles of the carrier. This timing is done by an R-C timer in a fashion similar to an LM555 operating in astable mode. A 0.001uF capacitor is connected between pin 1 and ground, and a 51Kohm resistor is connected between pin 1 and an 8.2V reference (available at pin 16). When a sync occurs, the capacitor is unclamped from ground and begins charging through the 51Kohm resistor from 8.2V DC. When the capacitor voltage reaches 1.0V (approx 5.5us after the sync), the colour-burst carrier is gated on. When the capacitor voltage reaches 1.3V, the colour-burst is gated off. The capacitor continues to charge until the voltage reaches 5.0V. At this point, the capacitor is discharged to 0V. Clearly, by changing the RC values (and hence the time-constant), the burst can be made shorter or longer (8 cycles of colour-burst is the NTSC spec, but I don't think a slightly longer period will hurt (gives the PLLs in the receiver plenty of time to lock)). If the time constant (Time const. = R*C) is increased the burst will occur later and last for a longer time. The converse is true if the time constant is decreased. If the time-constant is made too long, the ramp may not reach 5.0V before the next HSYNC, which will result in the some missing bursts (not too good). If you feel like it, put a 50Kohm potentiometer instead of the 51Kohm resistor. You may be able to get some cheap special effects while calibrating the pot. :-)

A chroma band-pass filter must be introduced between pins 13 and 10 as mentioned before. You may use the simple bandpass filter shown in the schematic, or to get better results, use one of the many shown in the Application Notes or design one your- self. As mentioned before, if you design your own, the filter should have a bandwidth of about 1.6-2.0 Mhz and a center fre- quency of 3.58 Mhz. Since most colour TV's have a BW around 3 Mhz, the BP filter will help reduce cross-talk between the lumi- nance and chroma components (on some TVs, you would never see the cross-talk in all the noise!). If your TV has a comb filter, you won't need a fancy BP filter (and the corresponding delay line); in this case, the simple bandpass filter sketched in the schematic will do.

Again, as mentioned before, if you do insert a fancy band- pass filter and see a noticeable shift between the outlines of objects and their colour fill, you will need to put a delay line between pin 6 and 8. The Motorola Application Notes should help in this department.

Pin 20 must be grounded to select NTSC operation [5].

Pin 19 provides the reference voltage for the voltage- controlled phase shifter (needed for I-Q phase shifting). It must be capacitively de-coupled to ground through a 0.01uF capacitor to provide a stable voltage reference at the pin. By pulling the pin up through a resistor to 8.2V, or by pulling it down through a resistor to ground, the axes can be tilted to get some colour adjustment (about 7 degrees, for a slight effect, according the application notes).

A 12V power supply is to be connected to pin 14. The power supply need not be regulated.

The 8.2V DC reference voltage appears at pin 16. A 0.1uF cap between pin 16 and ground will provide adequate filtering for this reference voltage.

Pin 15 is the ground connection.

Pins 11 and 12 are coupled through 0.1uF capacitors to ground. By sourcing or sinking some current through these pins, whites may be made whiter and the blacks blacker (sounds like a detergent commercial) by compensating for the balanced modulator feedthrough thus.

Pin 7 is coupled to ground via a 0.01uF capacitor.

Pin 9 is the composite video output. It has an output impedance of about 50ohms. To drive a 75ohm monitor input termi- nal, put a 25 ohm resistor in series with pin 9 and a 75 ohm co- axial cable. Connect the other end of the co-axial cable to the monitor.

Construction

You will need to buy a colour monitor cable from Atari [6]. Start by cutting the cable in half. Then, strip about 2 inches of insulation off the end of the cable. Now carefully remove the shielding without cutting the wires underneath. This should expose 7 wires. Four of the wires are shielded themselves; these are the red, green, blue and audio out wires. Strip the shielding off these wires as well. The following is a colour-code chart to help you find the wires: 
                  colour of wire                  signal                  pin#

                  white                           RED output              7
                  red                             GREEN                   6
                  black                           BLUE                    10
                  yellow                          HSYNC                   9
                  blue                            VSYNC                   12
                  green                           AUDIO out               1
                  brown                           ground                  13
The pins are labelled as follows (looking at the monitor output plug from outside). Please test the cable with an ohm-meter to see if my colour code is applicable to your cable. 
                  pin 4 -> 0 0 0 0 <- pin 1
                           0 0 0 0 <- pin 5
                           0 0 0 0 <- pin 9
                              0 <---- pin 13
Solder the cable onto the board you will be using, then solder or wire-wrap the circuit as shown in the schematic. Try to keep the construction as clean as possible. Keep as much of the shielded cables shielded as you can. Try to RF shield the whole enclosure if you can. All the best.

Bibliography

[1]

The separate oscillator is not really necessary since 1377 contains a common-collector Colpitts oscillator which can be used to generate the colour-burst on-chip. However, I found it easier to generate the signal off-chip to make sure the thing is indeed oscillating.

[2]

If the chroma signal is not band-pass filtered, the low frequency components it contains (those components having frequency less than 2Mhz or so) will interfere with the luminance signal since it is very hard to put the chroma specral lines exactly in between the luminance spectral lines without any interference between the two.

[3]

Note Filter bandwidth = Wo/Q, where Wo is the resonant frequency in radians/second (Wo=2*PI*3.15 Mhz) and Q is the quality factor required. Q=R*sqrt(C/L). A few filters are sketched in the Motorola Application Notes if you don't want to design one. A reference to these notes is in the appendix.

[4]

The Motorola Application notes show how to hook up a TDK delay line if you need one.

[5]

If pin-20 is connected to the power supply, PAL operation is selected. For all you PAL hackers: If you replace the 3.58 Mhz crystal with 4.43 Mhz crystal, and if you have a properly serrated vsync, you should be able to use this circuit to get PAL composite. If you do hack and get it working, please post. (PAL is the colour system used in England, India and some other countries.)

[6]

If this is not available, you may kludge one up as mentioned in the Abacus Internals book.)