Simple DIY composite video mod

[Keatah]

RF loves to jump (or couple) across parallel wiring.

[stlouisrod]

That construction technique is going to be prone to problems. I would start by looking at how close the component legs are to other solder joints —that transistor looks to be very close to the pads below, for example. That proximity may be producing an interaction that you don’t want.

Ideally, you want to keep components well isolated and any wires as short as possible. That cable is unshielded and may be acting as an antenna, or a path for unwanted electrical interaction.

 

It could be that the problem is unrelated to the additional components, too. Perhaps an existing connection was compromised in some way — a chip not making correct contact to all points in the socket, a socket having a bad solder joint, a damaged pad or track... Unfortunately, a lot of things can be wrong.

 

If you can put up a photo of the output on the display that shows the problem that you are getting, or a short video clip if that gives a better example (e.g. colours shifting over time), then it may help figure out the cause.

Thanks for your continuing assistance!

 

The transistor is riding at least a centimeter above the PCB, so little chance of interference there.

I also don't think it's interference, as the color issues are VERY consistent and repeatable. They don't waiver or fade like RF interference causes. This appears to be a definite rounding error or color-clock translation issue, which likely a diode or transistor issue not resistor or RF issue.

It's not a chip issue, as I swapped all 3 chips from my unmodded 4 switch unit that had good display.

Colors don't "shift over time". They are present exactly in same games, same conditions, from power on to power off

I've added more pics of the exact issue below. Please ignore the weird reflection in upper right of each pic.. my "Game Room" neon sign was on behind me.

Edited September 20, 2018 by stlouisrod

[boggis the cat]

Thanks for your continuing assistance!

 

The transistor is riding at least a centimeter above the PCB, so little chance of interference there.

I also don't think it's interference, as the color issues are VERY consistent and repeatable. They don't waiver or fade like RF interference causes. This appears to be a definite rounding error or color-clock translation issue, which likely a diode or transistor issue not resistor or RF issue.

It's not a chip issue, as I swapped all 3 chips from my unmodded 4 switch unit that had good display.

Colors don't "shift over time". They are present exactly in same games, same conditions, from power on to power off

I've added more pics of the exact issue below. Please ignore the weird reflection in upper right of each pic.. my "Game Room" neon sign was on behind me.

My guess is that the signal from the composite is out of spec. If you have an oscilloscope available then you can check this.

 

The mod I used (from the top of this thread) uses two transistors to amplify the signal sufficiently. A single transistor mod may not have enough amplification. Transistors vary from part to part, so you may just have some that are lower in amplification.

[stlouisrod]

My guess is that the signal from the composite is out of spec. If you have an oscilloscope available then you can check this.

 

The mod I used (from the top of this thread) uses two transistors to amplify the signal sufficiently. A single transistor mod may not have enough amplification. Transistors vary from part to part, so you may just have some that are lower in amplification.

Hmmm... I will review your process, and I DO have an oscilloscope (digital), but I'm not sure which wires/contacts to hook it to, nor what to expect in the scope readout. Any details you can offer would be much appreciated .

[Osgeld]

just hook the clip to ground and the probe to the video output while connected to a tv, it should be 1v peak to peak (it will be a stepped pattern, sync is from 0 to 0.3 volts, video signal is from 0.3 to 1)

 

from your screenshots it looks like its a bit strong, but 6 switch units IMO always look like crap even in RF ... its probably been there but now you can see it better

[boggis the cat]

just hook the clip to ground and the probe to the video output while connected to a tv, it should be 1v peak to peak (it will be a stepped pattern, sync is from 0 to 0.3 volts, video signal is from 0.3 to 1)

 

from your screenshots it looks like its a bit strong, but 6 switch units IMO always look like crap even in RF ... its probably been there but now you can see it better

 

Good advice.

 

Sometimes you don’t want to connect the ground on the probe. Phantom mains power is sometimes present on the ‘ground’ of TV connectors (no power there, but the signal will screw up the oscilloscope display). This confused me when I wasn’t getting an expected TV pattern, until I realised it was swamped out by mains coupled through the composite ground. Removed the ground and the signal was fine (TV and oscilloscope power from the same power strip).

 

With a digital ‘scope you can capture some scanlines and check for problems. My thinking is that the signal amplitude may be a bit low, but it could be overdriven as well.

 

The Atari doesn’t produce a fully compliant signal, so a ‘picky’ TV may not produce a good image — or any image, if you’re unlucky. I have a fairly new Sony LCD that does an OK job of displaying both NTSC and PAL 2600 output, so the claim sometimes made that no new TVs support the old systems isn’t correct.

[stlouisrod]

Good advice.

 

Sometimes you don’t want to connect the ground on the probe. Phantom mains power is sometimes present on the ‘ground’ of TV connectors (no power there, but the signal will screw up the oscilloscope display). This confused me when I wasn’t getting an expected TV pattern, until I realised it was swamped out by mains coupled through the composite ground. Removed the ground and the signal was fine (TV and oscilloscope power from the same power strip).

 

With a digital ‘scope you can capture some scanlines and check for problems. My thinking is that the signal amplitude may be a bit low, but it could be overdriven as well.

 

The Atari doesn’t produce a fully compliant signal, so a ‘picky’ TV may not produce a good image — or any image, if you’re unlucky. I have a fairly new Sony LCD that does an OK job of displaying both NTSC and PAL 2600 output, so the claim sometimes made that no new TVs support the old systems isn’t correct.

I have little experience using my oscilliscope, but i hooked up the composite plug collar to negative and pole to positive, then turned on the atari (so not hooked to TV). The voltage kept switching from .811 mV to 1.3mV, fwiw. So on your design, I see 2 transistors. Since I kept the original one after removing it from the PCB, can i just add it inline to my existing one to increase the amplification up to 1v spec? How did you wire yours? You just showed a picture of the top, not the bottom solder points. Are they in series or parallel? The transistor provided for my hack is 2N3904. Thanks.

Edited September 22, 2018 by stlouisrod

[boggis the cat]

I have little experience using my oscilliscope, but i hooked up the composite plug collar to negative and pole to positive, then turned on the atari (so not hooked to TV). The voltage kept switching from .811 mV to 1.3mV, fwiw. So on your design, I see 2 transistors. Since I kept the original one after removing it from the PCB, can i just add it inline to my existing one to increase the amplification up to 1v spec? How did you wire yours? You just showed a picture of the top, not the bottom solder points. Are they in series or parallel? The transistor provided for my hack is 2N3904. Thanks.

Simon.Plata is the designer of this mod (see the first post). I slightly modified it to work well in my PAL machine.

 

Check post #13 for a good circuit diagram and board layout. You would have to adapt the circuit to use the connection points you have now, or remove the mod entirely and start over with Simons mod.

 

If you have a reasonably modern oscilloscope it should have a setting for TV signals. You set that then have to adjust the triggering to get an appropriate signal. To get a proper signal you are best to hook up to your TV as normal, then pick off the signal from the composite output it is likely easier to connect to the wiring on the board, just take care not to short anything. Your polarity (positive and negative) should be correct, but your values should be around 1 Vpp maximum volts, not millivolts (probably a probe setting issue, as you wouldnt get any picture at mV levels). 1.3 Vpp is way too high an output, so if that 1.3 mV is actually Vpp then that may be the issue but you really need to terminate the cable into the TV to get an accurate measurement (or connect a 75 ohm load across the output).

 

Unfortunately, this is just something that you have to figure out for your particular modifications and the test gear you have. There is a lot of trial and error in this. An alternative would be to buy something like the UAV mod which is pre-built and much smarter than these basic signal level amplification hacks.

Edited September 24, 2018 by boggis the cat

[simon.plata]

Thanks for your continuing assistance!

 

The transistor is riding at least a centimeter above the PCB, so little chance of interference there.

I also don't think it's interference, as the color issues are VERY consistent and repeatable. They don't waiver or fade like RF interference causes. This appears to be a definite rounding error or color-clock translation issue, which likely a diode or transistor issue not resistor or RF issue.

It's not a chip issue, as I swapped all 3 chips from my unmodded 4 switch unit that had good display.

Colors don't "shift over time". They are present exactly in same games, same conditions, from power on to power off

I've added more pics of the exact issue below. Please ignore the weird reflection in upper right of each pic.. my "Game Room" neon sign was on behind me.

 

Hi,

 

Can you share the schematic and component values you're using? I Will give a look and see if I found something.

 

Looks like a bandwidth problem, it could make a color shift or distortion.

 

Best regards,

[stlouisrod]

 

Hi,

 

Can you share the schematic and component values you're using? I Will give a look and see if I found something.

 

Looks like a bandwidth problem, it could make a color shift or distortion.

 

Best regards,

Thanks a ton for offering to take a look. Here are the pertinent pages from the install manual of my mod:

[simon.plata]

Thanks a ton for offering to take a look. Here are the pertinent pages from the install manual of my mod:

 

The mod you're installing is not the same as the one stage (one transistor) posted in this tread. From the information you send it's very difficult and time consuming to reconstruct the schematic in order to analyze the circuit and verify the configuration.

 

 

However it looks like a bandwidth issue or a unwanted low pass filtering. This could happen because:

* The 2N3904 transistor is out of specs or in the lower limit for gain. I suggest to test it for a minimum gain of 200 or replace it.

* May some of the minimum removal components are still in place. These components needs to be removed from the VCS board, otherwise they Will attenuate or filter the signal to unusable levels for this mod.

 

If neither of the above Works for you, I suggest to implement the mod posted in this tread, then we can offer better support or advice.

Bests,

[boggis the cat]

* May some of the minimum removal components are still in place. These components needs to be removed from the VCS board, otherwise they Will attenuate or filter the signal to unusable levels for this mod.

Just a note that you can remove one leg of a resistor or capacitor and leave it attached by the other. This makes putting it back in place easier if you have to do so. Transistors are best to remove entirely, or leave attached by one leg.

[stlouisrod]

 

 

The mod you're installing is not the same as the one stage (one transistor) posted in this tread. From the information you send it's very difficult and time consuming to reconstruct the schematic in order to analyze the circuit and verify the configuration.

 

 

However it looks like a bandwidth issue or a unwanted low pass filtering. This could happen because:

* The 2N3904 transistor is out of specs or in the lower limit for gain. I suggest to test it for a minimum gain of 200 or replace it.

* May some of the minimum removal components are still in place. These components needs to be removed from the VCS board, otherwise they Will attenuate or filter the signal to unusable levels for this mod.

 

If neither of the above Works for you, I suggest to implement the mod posted in this tread, then we can offer better support or advice.

Bests,

 

I will try to remove the extra components you highlighted. I believe I found the 820p and 47p caps on the right, but I can't seem to find the RF modulator related 150p cap or 15k resistor (did you mean 1.5k perhaps?). Since this is a heavy 6er, the RF modulator is on the separate switch board along with c105 (4.7uF capacitor) on it, but no resistors. Of course, I removed the RF box long ago as part of my mod instructions, in case you were curious.

Edited September 25, 2018 by stlouisrod

[stlouisrod]

Update: I removed the 820p and 47p caps in the right red section of your diagram, with no change to the picture. I also tested the 2N3904: emitter to ground before trim pot = 1.1v; after trimpot adjusted for best saturation of picture = .95v. So it seems well within tolerance. Even if i adjust the trimpot to lowest level (highest resistance), the picture gets dim but the pixel coloring anomaly remains. So it never appeared to be a low current issue, and now I've proven that electrically as well.

[simon.plata]

Update: I removed the 820p and 47p caps in the right red section of your diagram, with no change to the picture. I also tested the 2N3904: emitter to ground before trim pot = 1.1v; after trimpot adjusted for best saturation of picture = .95v. So it seems well within tolerance. Even if i adjust the trimpot to lowest level (highest resistance), the picture gets dim but the pixel coloring anomaly remains. So it never appeared to be a low current issue, and now I've proven that electrically as well.

 

Despite the mod is just a simple video amplifier, it exists several reasons to not work properly: One of them could be amplifier saturation to any or both rails.

 

The real problem can't be found easily without an schematic or a signal analysis with an oscilloscope.

 

The alternative is to revert the mod you did and try with the mod proposed in this tread which is proven and has worked for several people including myself.

 

Best regards,

[stlouisrod]

This looks fantastic.

 

I have been using an A/V mod that I purchase off eBay https://www.ebay.com/itm/Atari-2600-7800-A-V-Composite-RCA-Audio-Video-Mod-Kit-NEW-IMPROVED-KIT-1G-/311705759237?_trksid=p5731.m3795

 

I have been pleased with it so far, but I am always looking for something potentially better. I am planning to install your design in a light 6'er as soon as I can get all the components together.

 

If possible I would like to get a copy of your PDF Eagle file.

 

Thanks.

Quick shout out for this recommendation. I bought this for $13.00 on Ebay and installed in my 4-switch woody. Installation only involved removing the RF box, one resistor and one transistor. Only 4 new wires to solder to the board. The picture results are very good, with accurate colors, detail, and sound. The only mild distortion I could find is the blue suit on Keystone Kapers. It kind of jumbles up against the green background. But I suspect this is likely true over standard RF signal as well. The Activision logo on all games reads nicely, so no need for adding distortion "smoothing". All testing was done on a 32" CRT TV and 19" LCD HDTV. My only mild complaint is that while the mod PCB has a resistor to attenuate the audio, it is still quite loud. But it is always clear. Even with my CRT volume on only 20 out of 100 it provides playable volume.

Edited September 29, 2018 by stlouisrod

[Osgeld]

4 switch models dont buffer the video signals though the 4051 though which is the entire source of ghosting bleeding and shadowing on the 6 switch models

[stlouisrod]

4 switch models dont buffer the video signals though the 4051 though which is the entire source of ghosting bleeding and shadowing on the 6 switch models

Interesting... I can definitely confirm my heavy 6er exhibited noticeable ghosting through the original RF video, and while diminished quite a bit with my Best Electronics composite mod, some ghosting is still visible.

Edited September 29, 2018 by stlouisrod

[stlouisrod]

A tad off topic, but I was reading through the original Atari 2600 field services PDF, and it recommended repair shops connect TIA pins 6 and 9 through an 820 1/4 W 5% resistor to obtain a more saturated picture (aka, better). I tried this on my 4 switch, and while it was definitely more saturated, the colors were too washed out.. reds were pink, whites were gray, etc. Worse, it showed odd horizontal lines on the left edge of the screen when the brightness was turned up a bit. So I quickly undid that "improvement"

Edited September 29, 2018 by stlouisrod

[rallyrabbit]

Old thread, great idea. I just tried to make one of these little boards dual transistor composite mod. I hand built it, set the pot myself to 1477 ohm). Ran everything as close together to reduce noise, crossed +5 and video lead rater than having anything parallel leads to reduce liklihood of noise.

NTSC v13 of 4 switch board.

 

Removed these components:

C206 - 820pf

L202 - 1.8MH

L201 - pot/coil?

Q201 - Transistor

C209 - 47pf

R209 - 1.8k

R222 - 15k

C215 - 150pf

 

 

On the surface, man, this mod looks Great. Colors saturation is excellent. Picture is crisp.

 

But have a problem I cannot seem to sort. This is consistent in all games. Take a look at the pictures before and after movement in Mrs Pacman. Picture is perfect before movement. As soon as movement occurs, the left half of the screen is great. If you device the screen linearly into 4 equal columns:

 

Column 1 (left 1/4 of the screen) is good

Column 2 (left/center 1/4 of the screen) is good

Column 3 (right/center 1/4 of the screen) contains the picture for column 4

Column 4 (right 1/4 of the screen) contains the picture for column 3

 

Any suggestions?

Edited March 3, 2019 by rallyrabbit

[ChildOfCv]

4 switch models dont buffer the video signals though the 4051 though which is the entire source of ghosting bleeding and shadowing on the 6 switch models

 

... and then the 7800 went back to gating the sync and luma

 

It does seem to use a different selection of mixing resistors though.

[R.Cade]

Old thread, great idea. I just tried to make one of these little boards dual transistor composite mod. I hand built it, set the pot myself to 1477 ohm). Ran everything as close together to reduce noise, crossed +5 and video lead rater than having anything parallel leads to reduce liklihood of noise.

NTSC v13 of 4 switch board.

 

Removed these components:

C206 - 820pf

L202 - 1.8MH

L201 - pot/coil?

Q201 - Transistor

C209 - 47pf

R209 - 1.8k

R222 - 15k

C215 - 150pf

 

 

On the surface, man, this mod looks Great. Colors saturation is excellent. Picture is crisp.

 

But have a problem I cannot seem to sort. This is consistent in all games. Take a look at the pictures before and after movement in Mrs Pacman. Picture is perfect before movement. As soon as movement occurs, the left half of the screen is great. If you device the screen linearly into 4 equal columns:

 

Column 1 (left 1/4 of the screen) is good

Column 2 (left/center 1/4 of the screen) is good

Column 3 (right/center 1/4 of the screen) contains the picture for column 4

Column 4 (right 1/4 of the screen) contains the picture for column 3

 

Any suggestions?

That problem is not from a simple video mod. I would suspect you have dropped some solder or other leftover materials somewhere on the board...

[rallyrabbit]

A little update. I swapped a TIA from another unit and the swapping I described went away (v13 board has socket). The original TIA works with the RF to TV, but does not work on the single or dual transistor composite mods. Very strange, I have no explanation.

 

 

Some notes:

Another poster mentioned removing R216 (820 ohm resistor for TIA pins called out in Atari TSB. I find that with it, the colors are majorly vivid. Without it, I cannot get the proper color match on most games. So with it, the frog in frogger can be seen on the logs, without it, the frog blends in. Another example is Ms Pac Man. With it, I can get the pink/blue default setting. Without, it is at best peach on blue.

 

The original post mentioned removing C206 (cap near the audio transistor). Looking at the schematic from the first post on components to remove, C206 is the one attached to the base leg of the transistor, it should be removed. I also found that having both in seems to help the TIA. The circuit with the leftover components at that point are only feeding AUI and AUO of the TIA.

Edited March 3, 2019 by rallyrabbit

[stlouisrod]

Quick update on the simple composite mod I made to my 4 switch woody. As I said above, picture looks great on a CRT, and on most older (non-smart) LCD panels. HOWEVER, there are big image problems when sent to my 1st generation 4k LCD HDTV. While the color seem solid, there are major ghosting/double images occurring, no matter what settings I change on the TV itself. I blame the ever-worsening analog to digital chips in these newer TVs, which apparently can't quite cope with the hack's timings/signal strength.

 

[Osgeld]

I bet its still there on the crt, but its being dampened by the nature of the thing (as I have seen this my self first hand and its still there on a tube, its just really soft and blurry ... like a tube), its kind of hard to blame a 4k tv that is sold to you for showing every detail possible, then showing you the details