Building PONG from scratch

[Anthony Singleton]

I am working on recreating the Pong video game based upon Dan Boris' schematics which can be found here. I have starting building it using LS TTL chips on a solderless breadboard. The original circuit uses plain 7400 TTL which may or may not be an issue.

 

This is what it looks like so far. I have arranged the chips like they were placed on the original board. Each row is labelled A (bottom) to G (top) and each column is 1 (left) to 9 (right). I have a 0.1uF capacitor across each rows power rails, again like the original PCB. There is a 220uF cap on the output of my 5V supply. It is a simple linear supply using a 7805, the current draw so far is 91mA. The measured voltage across each 0.1uF cap is about 4.9V which seems OK.

 

I have built the H-Sync, V-Sync, Video/Sync mix and Net circuits which in theory should show a dotted line down the middle of the screen. However all I get is a blank screen. When using a VCR video input, the blue screen does disappear so the sync pulses seem to be within range of proper video.

 

The CLK signal has a lot of ringing.

 

The output 256H has a lot of noise on the level parts.

 

The H_RESET when zoomed in is noisy and distorted.

 

Any help is appreciated.

[Anthony Singleton]

The circuit is mostly complete, except for the fact that it dosn't seem to work properly.

 

Here is a video of "atract" mode, which is when the game is powered up but no coin is inserted. The ball bounces up and down and looks distorted. [The red light is a reflection from the camera].

 

Here is game mode, the paddles appear and are moveable. Again, the ball only moves up and down and slightly side-to-side, something is funny with the horizontal counter circuits.

 

I have checked and re-checked my schematics against the originals and besides a few minor discrepencies I couldn't find a major problem. I suspect part of the problem is that lack of original parts. Many parts of the series of original TTL chips are becoming unavailable from places like Digikey. To accomodate the lack of original TTL, I have used LS-TTL throughout. Parts B4, A6 and B6 wiring was changed to reflect the different pinout of the LS chips. While logically identical, because of their different internal construction, parameters like clock skew and propagation delay will be different. Thus the timing of the original circuit is not properly duplicated. Comparing the 9316 logic diagram with the 74161 and 74LS161 shows different latching mechanisms used to accumulate the counters.

 

Anyway, I am waiting on some original 74161 parts which hopfully give better performance in the circuit.

[amiman99]

That's pretty good to get at least what you have.

 

I have a copy of "Video Ball Game" circuit from an Australian Electronics magazine from 1976, I was thinking to do it, but it may be over my head a little bit. It looks alot simpler then the original PONG. Only 13 IC chips.

 

Good luck with your project.

[+poobah]

Some of the signals are quite fast for a breadboard. The pong design is very sensitive to timing delays.

[+Random Terrain]

After you finish Pong, maybe you can start making new Atari 2600 clone consoles with new chips that aren't failing. You'll sell hundreds of them.

[Anthony Singleton]

Well, my supposedly original 74161 chips arrived and they are in fact regular Fairchild DM74LS161AN chips. Argh, $20 not well spent. Anyway, there was no guarantee that this would have solved the problem anyway.

 

Reading up on Dr. Hugo Holden's analysis of the Pong circuitry [pdf at bottom of page] has given me some insight. Near the end of the document he proposes a number of solutions to issues with the original design. One of the problems was the net image appearing dim due to timing differences when using newer flip-flops. Because of subtle differences in the clock skew, the signals no longer line up. To remedy this, the CLK is replaced with the 1H signal which is 1/2 the rate of CLK. This gives timing delays closer to what would have appeared with the original parts.

 

I started probing around the Horizontal Ball Counter circuit, particularly the H6 nand-gate. The signals all have to line up properly to get an output. The HVID signal appeared erratic, only intermittently changing state. On a whim, I switched CLK on the G7/H7 counters to the H1 signal. This gives a ball with proper motion but is twice as wide as it should be. I then tried using the inverse output (bar)CLK from the main clock flip-flop. This shifts the phase of the signal by 180 degrees and seems to give proper ball motion.

 

There is still work to be done as the paddle motion is limited to the top half of the screen making for rather unfair game play.

[RevEng]

Some of the signals are quite fast for a breadboard. The pong design is very sensitive to timing delays.

 

I'm by no means an expert, but I think poobah is right here. Breadboards have stray capacitance at each point (~10pf) that will result in crosstalk, and you'll get varying inductance+resistance at the points. They're generally considered unreliable at video frequencies and up.

 

Better to use something with a ground plane and soldered connections, like the perf-board that has solid copper around the holes, or dead-bug on a copper clad board.

[GroovyBee]

@Anthony Singleton: What is the bandwidth of your scope and what is the frequency of the signals you want to see on it?

 

A good rule of thumb is that you need up to around the 10th harmonic to see a "square" square wave on a scope. Anything less than that and it will look like more like a sine wave. So if you want to view a 4MHz square wave you are going to need a 40MHz bandwidth scope or higher. Fast rising edges have high frequency component signals too.

 

When you take your measurements how far away is the GND on the scope probe away from the signal of interest? For the best quality signal you need a GND close to the signal of interest.

[RevEng]

Don't know if he'll post here again, but it sounds like the original poster made it work, and designed a custom PCB.

 

The reddit thread mentions that he's planning on putting together a small run of kits. Sounds like a fun project!

[jaybird3rd]

VERY impressive! If this is ever offered as a kit, the OP should definitely put me down for one. I have an original Atari Pong coin-op that I still haven't gotten around to troubleshooting, and this new board would make a great replacement for the original.

[RevEng]

Me too. I'd love to try my hand at a bartop replica of the original.

[driph]

Awesome work! Add me to the list of folks who'd love something like this in kit form.

[jaybird3rd]

I don't believe this is related to the OP's project, but a complete re-creation of Atari's original PONG, made using discrete TTL, is now available as a DIY kit called "TTL Tennis for Two". Here is a video of it in action, posted by the designer:

 

http://www.youtube.com/watch?v=WS8JTAK34Fg

 

I've got one on order, mostly because I'm very anxious to get my original Atari PONG coin-op working again. I'll post an update after I've received and assembled it.

 

EDIT 3/9: I've received the kit, and it's exactly as I expected: a blank board and a collection of parts that you build yourself. I'm going to take some time to assemble it in a week or two.

[19rsn007]

I once created a Pong game, driven by a PIC and drawing it's power from 3 AA batteries.

It was a nice project.....must be around here somewhere.

[Osgeld]

pong on a micro is quite easy, its another to do the whole thing in what amounts to a chain reaction of delays

[jdrose]

That is what I was thinking too. It is no problem to program a Pong game. It is quite another thing to make it completely out of hardware. Neat project.