The TEA 2000 integrated circuit is responsible for PAL enconding the RGB signal from the ZX Spectrum 128K/+2/+3 ULA. This IC is rather sensitive to electrostatic discharges, and hence, it can suffer damage. Besides, it can become very hot.

I came across a Spectrum +2A, with an issue number I haven't seen before, issue 4. This Spectrum had a number of faults that could easily be solved, but one remained: although the RGB output worked fine, the composite signal obtained at the input of the RF modulator didn't carry color. A faulty TEA200 was the culprit.

Becoming a rare IC, I studied the possibility of replcing it with a modern IC that could carry the same task: get a RGB signal plus sync, and give a composite PAL signal. The Analog Devices AD722 came to help.

While the TEA2000 is designed to accept a 2-bit per color digital signal, that is, 6 bits of color data, the AD722 and nearly every modern RGB-to-PAL encoder accepts analog RGB signals. The Spectrum +2A/+3 generate an analog RGB signal from the 6 bit output using diodes and resistors to add the bright component to each red, green and blue component, but this is done outside the TEA2000 circuit.

I'm interested in a drop-in replacement, so I have to deal with this 6-bit color data. Luckily, the arragment for these 6 lines is similar to the arragment used to generate the analog output: if we look at the inputs to the TEA2000, we have one input directly from one color component (R1, G1 and B1), and another input from the bright signal, with a diode in series (D28, D29 and D30). Note that these diodes are connected the same way as D33, D34 and D35, so the anodes of these diodes carry the same signal as the anodes of D28, D29 and D30, which go to R0, G0 and B0.

So, to regenerate the analog output needed for the AD722, we will use the same resistors that have been used in the analog circuit: six 150 ohm resistors, two for each color component, which will add the information from the primary color and the bright signal. The complete schematic diagram for the TEA2000 replacement is as follows:

It does not use the clock signal from the Spectrum, as this would need a frequency divider to get 4.43 MHz from the given 8.8 MHz. I designed an alternative circuit in which this is done, but didn't work very well. The version with a 4,43 MHz crystal gives a decent picture, as we will see.

The only thing that have not been included in this design is adding the modulated audio carrier onto the video signal. Besides, the output from this circuit may not be adequate for the RF modulator. As it's my goal to get a composite signal, the audio subcarrier and the RF modulator don't bother me.

Here we have the PCB for the above schematic. I've made it so it's one side only. There are three wire jumpers that have to be soldered in the component side. All but the crystal and the pin header for the TEA2000 socket are soldered directly on the solder side.

The PCB's for the two prototipes designed, ready to be solded. The left one is the design with the 4,43 crystal. The right one uses a 74HCT74 to convert the 8.8 MHz clock input for the original TEA2000 to a 4,43 MHz signal, suitable for operating the AD722 in FSC mode.

The PCB's, fully populated. Both are single side designs.

The component side of the PCB reveals the crystal, the wire jumpers and the pin head for the TEA2000 socket in the Spectrum.

The replacement, compared to a TEA2000 IC.

Installed into a Spectrum +2A (errr... a +2E, as this Spectrum has a ZXMMC inteface with two slots.) The RCA socket is located to the left, near the audio socket. A shielded cable goes under the Spectrum PCB just to the pin 6 of the TEA2000 socket, where the composite signal should go.

And a little test, using the Shock Megademo. Actual picture looks better.

This picture was taken at the same distance (more or less) showing the same Spectrum outputting video through RGB, then outputting composite video using the first design (the one featured in this article), and last, using the second design (the one that get the clock signal from the 8.8MHz clock using a divider). The noise in the last example is prominent, and barely noticeable in the middle part.

And for the last, the Eagle files for this project, for those interested.