Component modding CRT TV - "China TV"

[Zebidee]

   please.

I wasn't aware 100Ω resistors were commonly found on RGB and YUV lines... I also maintained 1μF electrolytic capacitors in series and now realize ceramic disk capacitors with lower values are better to maintain the needed AC coupling, but improve frequency response a bit.

A modular approach works best.

I suggest you build a breakout PCB with connectors for your Y Pb Pr inputs. This will allow you to quickly and easily swap the whole unit out, to test new arrangements, avoiding strain or risk to TV components.

Start with this (below), which I already posted on the thread above. Use the 100R inline resistors + 100nf (104) ceramics for Pb Pr, and you can also use it for Y (luma). Don't use 1uF for Pb Pr, signal is too high frequency. Y is lower, so a higher value cap is acceptable, but the main effect will be a slight fuzziness as some of the signal is delayed slightly. If the TV has problems syncing, you could try a 1uF (105) or even 10uF for Y. If using electrolytics ensure polarity is correct, negative to GND.

You should use 75R termination resistors because the video signal is impedance matched for 75R and expects 75R termination. The 100nF cap passes high frequency video signal. Any remaining low frequency noise prefers to go to GND via the 75R resistor.

This base design is a good place to start from. So I recommend you try it first, then you can modify from there.


     

I've done RGB mods before and haven't had any such issues with the signal level, and I read where some ICs may have current limiting built in. But
...seems something with the Y clamp circuit in the chip failed, maybe I burned it out doing this mod incorrectly.

In your case, I'd recommend you add some 12v zener diodes for overcurrent protection, as shown in the circuit below. The zeners give some protection from high voltage spikes on the video inputs. I have learned about this the hard way on a TCL TV mod, click on the picture for a link to the thread.

I replaced the chip since I added a socket, then I encountered another issue where it would power on for about a second after I plugged it in then turn off and not respond. It would do the same thing regardless of what chip I put in it

I also recommend that you look at backing up your EEPROM. This is typically an 8-pin DIP or SOP/SOIC-8 chip near the jungle. Jump online and grab a "CH341a", they are cheap, and you can use this to copy/backup/write/hack the EEPROM. You could even socket it, like you did for the jungle itself. I have another thread on this here.

I also wanted to ask about the 391 capacitors on the cathode amplifiers, these are used for filtering according to the schematic Adrian Black on YouTube scanned in from his China kit board. He bought his in 2023, and he doesn't seem to have any blue streaking issues. I checked and the blue capacitor on mine is indeed a higher value so I'll be replacing it. I'm not sure if I'm seeing any symptom of that though. There is some noticeable smearing of the reds and blues when using s-video but that might just be chroma delay seen on these chips.

What are these "cathode amplifiers"? Are you talking about caps on transistors in the colour drive circuit, on the neckboard? Who is Adrian Black?

  More pics please. What works for me is to pretend that everyone here has either dyslexia or attention span issues (maybe not far from truth).

EDIT: Make sure you have removed any GND clamping capacitors on the Y Pb Pr inputs to the jungle. Often a 103 cap. Like this:

In this pic we can see that the trace connects 19 to 21, and then to C804, a 103 (10nF) ceramic capacitor. This 103 cap clamps the Pb, Pr signal pins to ground, to prevent unused inputs from "floating ". It needs to be removed. The first pic below is untouched and shows which pins are the signal pins; the second pic (from an identical but different TV) has had the trace scraped off between Pr and Pb, and C804 has been removed.


 

[titan-91]

Here's how I had things wired thus far. Most of the Reddit post links in that Pastebin document contain full annotated Imgur album links with many more photos as I went along.

https://pastebin.com/ppVWTe7H





I assume the standard method you found will give a brighter picture with more dynamic range, but thanks for the breadboard suggestions. What I think happened is I allowed too much current to sink into the chip and it fried the Y clamp circuit. My assumption was the inputs were high impedance, so I used a higher impedance termination. Guess that wasn't the case.

I've already taken a detour with the EEPROMs, I had a FlashCatUSB Classic on hand so wiring that up with 10kΩ pull-up resistors on the clock and data lines was simple. I uploaded several EEPROM dumps along with board details and photos to my Archive.org account:

https://archive.org/details/@titan91?sort=-addeddate

At first I was concerned the EEPROM contained the remote code data. I've now learned that is not the case and I also have an old Android phone with an IR blaster I use with a Remotify app from 2024 some guy made. 6 of the 8 emulated remotes work fine with these 8895 chips of mine.

The cathode amplifiers are referring to the transistors on the neck board, yes. I don't have a picture of mine at the moment. But I can tell the blue capacitor is definitely not the same capacitor as the other 2 channels. Very strange how that is.

And yes, before I first did the mod the Y pin was floating on the board (not soldered) and the Pb/Pr pins were soldered directly to ground. No capacitors. Not great to isolate from potential DC voltage transients on the ground plane, and leaving Y floating also exposes the chip to electrostatic discharge through that line. When I have Pb and Pr floating I see red and blue stripes across the screen, so there's definitely some coupling that happens there by default.

[Zebidee]

Here's how I had things wired thus far. Most of the Reddit post links in that Pastebin document contain full annotated Imgur album links with many more photos as I went along.

Excellent! It is good discipline to keep detailed logs.

As you've already worked out, you can link to those archived images from here, to quickly pull together a concise story.

I assume the standard method you found will give a brighter picture with more dynamic range, but thanks for the breadboard suggestions. What I think happened is I allowed to much current to sink into the chip and it fried the Y clamp circuit.

The 75R termination pulls the input voltage to the correct level, which for Y is up to 1.0vp_p, though most ICs will accept up to 1.2v. For Pb Pr the input voltage is a little less. Anyway, without that 75R termination the Y input signal can be up to 2v, which the jungle won't like for sure. So it is probably less about current and more about voltage, though obviously you can't have one without the other.

You have done OK for a first time, but honestly some connections there are pretty dodgy-looking. So you also run the risk of accidentally shorting the input signal with something else and perhaps momentarily creating a voltage spike, which could also damage the IC video channels.

I've done these mods many times now. In rare cases the PCB has the vacant spots already there for the extra components, but otherwise a small breakout PCB is the way to go for safety and convenience. It will save you hassle in the long run, and is:
- Structurally much more sound
- Reduces stress to jungle IC and fragile connections
- Easy to connect/disconnect
- Allows you to implement experimental changes easily
- If you use the circuit I suggested above, especially with the zener diodes, it will help protect your circuit from interference and exposing the jungle to excess current/voltage.
- the connectors are cheaply available from your favourite online retailers. You can even buy them with the wires already connected.

I've already taken a detour with the EEPROMs, I had a FlashCatUSB Classic on hand so wiring that up with 10kΩ pull-up resistors on the clock and data lines was simple. I uploaded several EEPROM dumps along with board details and photos to my Archive.org account:

https://archive.org/details/@titan91?sort=-addeddate

FYI  These China TV boards often have a white 4-pin connector (5v, data, clock, GND) allowing you to read/program the EEPROM directly with a customised cable. If there is no connector, there might still be a footprint on the board in which case you can pop one in! I cover all that in the EEPROM thread linked above.

At first I was concerned the EEPROM contained the remote code data. I've now learned that is not the case and I also have an old Android phone with an IR blaster I use with a Remotify app from 2024 some guy made. 6 of the 8 emulated remotes work fine with these 8895 chips of mine.

Cool! What a great idea.

The cathode amplifiers are referring to the transistors on the neck board, yes. I don't have a picture of mine at the moment. But I can tell the blue capacitor is definitely not the same capacitor as the other 2 channels. Very strange how that is.

I once thought that too. After looking at many CRTs I've found it is actually very common, at least in Asian brand sunset-era CRT TVs, but I cannot tell you exactly why the blue cap is different. I've tried swapping it (for the same value used by red/green) in most of my mods, and in many cases the image seems crisper and the blue definitely less smudged (to the right). Occasionally it looks better to leave it in. It might be because a certain amount of smear on blue makes AV/composite/RF actually look better?  It isn't really noticeable until you use a component or svideo input.

[titan-91]

Yeah those photos are pretty early in the process, I've cleaned things up since then. My board didn't appear to have those spots. I'll see if I have any applicable zener doides as well. Your schematics have been very helpful, so thanks for those. The unloaded voltage was probably just too high as well, so I'm otherwise not sure why the actual signal level was so dark before. Lack of current limiting is the only other thing I can think of, as others have done YUV mods to these boards without that issue. I do also recall the original capacitor I decided to re-use from one of the audio output lines early on was causing sync issues, I believe it was 10μF. Before I replaced it with a 1μF electrolytic the picture would lose vertical sync between bright and dark transitions and sometimes bounce a little vertically. That no longer happens but I'm still swapping those with the ceramic 104 caps as suggested.

I looked carefully and don't seem to have that I˛C header on my board. Those traces don't go anywhere else except between the EEPROM and main microcontroller.

My 100nF capacitors are on the way, still waiting on the resistor kit I ordered to ship.

[Zebidee]

Cool-o   

Yeah essentially the 100nF (or 1uF for Y) gives AC coupling (blocks DC), the 100R gives current limiting, the 100R + 75R is a voltage divider, and the 75R sinks DC noise. What you said

For the zeners, I selected 12v as they are there to prevent voltage spikes (like ESD or momentary short) from damaging the jungle. 1.2v zeners seem an obvious choice, but can be very "leaky" and allow a fair bit of current through anyway (quantum tunneling) at lower levels, which would be bad for a video signal. So that choice isn't written in stone, you might be able to go as low as 6.8v or 5.6v and probably be OK, but if going that way probably want *another* diode (e.g. 1N4148) in addition to prevent quantum tunneling leakiness. Just using 12v zener is simpler while doing mostly that same thing. I looked far and wide on the matter, you can read more about my reasoning at the time on my TCL TV mod thread.

Otherwise, there is a lot of stuff in the service menu to optimise like tint and colour systems, RGB brightness levels etc. and honestly I've half-forgotten most of it already. There are like 20 pages of settings. I think I attached an English version of a complete service manual PDF earlier on this thread, with some notes.

[titan-91]

I spent about an hour reworking all the input connections, I removed the 165Ω termination on all 5 video lines and replaced them with 75Ω. Strangely one of the China kit board schematics shows 82Ω termination. I don't see the point of these little tweaks to the design, same with the 391 capacitor replacement on the neck board. I then added the 100Ω series resistors on the input side followed by the 104 ceramic capacitors and the 391 cathode amp capacitor on the neck board. Thanks for the diode suggestion, I'll keep that in mind if I come across any 12V zeners.

Here are the results:

https://www.imgur.com/a/0bdB5ny

Man, for the first time since I started this project in the spring the picture actually looks amazing! Before the luma/gamma levels were too dark and the colors were oversaturated. Had no idea I was just blasting the chip with really hot video and the luma clamp was turning it down, at the same time limiting the contrast. Before the corrections I was also getting that really bad buzzing in the audio. No buzzing now at all! And I was actually able to get the pictures with the lights on.

Thank you so much for your suggestions. Now I need to open up my other 2 RGB modded TVs and fix them as well so I don't do any further damage.

[Zebidee]

That awesome! You seem to be across it and your pics look good too.

Yeah I've pored over dozens of schematics and actual TVs and they often had these funny variations that didn't seem to make sense, although maybe they do if you consider that most customers were only looking for a cheap TV to watch soap operas and shopping channels. So low cost & quality video was the game, blurry was sometimes better. They were optimised to make a composite/RF input look good, rather than high quality YPbPr or RGB signals.

The input schema/circuit above attempts to provide a baseline input arrangement for 90's/2000's YUV-capable TVs with similar jungles, and the same principles work as a starting point for RGB mods.

Now what you need is a GreenAntz transcoder (see link in my sig below) so you can convert the component TV into an RGB monitor for retrogaming & videos 

[titan-91]

I believe I've reached "cheap retro gaming TV nirvana." I did some more dialing in and the results via component video are just incredible with this chassis! I have a Sony PVM, a couple of larger TVs I've RGB moddded, and another one I've done an s-video mod on. None of the tubes (even the small PVM) look as good as this one. It's a Philips CRT, so I wasn't sure how good the color purity would actually be. But I do have quite the unfair advantage due to the tube being practically mint, pegging the needles on an emissions test. This TV was just dumped in someone's front yard!

I appreciate you giving people another good option for converting from RGB to component, but in my case I won't be needing it since I have my Wii for some emulators and fat PS2 for my PSX/PS2 games. S-video is an option too but I haven't been a fan of the color smearing and offset the NTSC decoder in these Toshiba 88XX series chips have.

The pictures I got didn't do it justice at first, so I boosted the color saturation in the these photos to best reflect how the games look to the eye. I also used to help out with a family arcade operator for many years growing up and after the mods, this thing looks every bit as good as a brand new arcade monitor running on RGB.

https://www.imgur.com/a/NUgpvlB

[Zebidee]

I believe I've reached "cheap retro gaming TV nirvana."

Heh, welcome 

Even if the TV is busted, try to keep the tube!

None of the tubes (even the small PVM) look as good as this one. It's a Philips CRT...

Aye, I would rate the 90's+ Philips CRTs to be perhaps the best of their breed. As cost and competition pressures bore down on the industry in its sunset era, European CRT TV manufacturers cut cost by consolidating tube manufacturing to fewer and fewer facilities, only the best and cheapest survived, and many major brands (Loewe, Blaupunkt, Grundig etc.) were using Philips CRTs produced out of France iirc.

Philips also produced CRTs with bonded yokes, all the adjustments were done to perfection in the factory and a bunch of epoxy plopped down so no-one could ruin it. They were nice, no convergence to worry about with those babies.
 
Around 15 years or so ago (in Australia) I matched some Philips 27" CRTs (from old/dead TVs) with "Sharp Image" chassis to make some arcade monitors for custom cabs I made for people. They were awesome, comparing well against even PVMs.

I appreciate you giving people another good option for converting from RGB to component, but in my case I won't be needing it since I have my Wii for some emulators and fat PS2 for my PSX/PS2 games. S-video is an option too but I haven't been a fan of the color smearing and offset the NTSC decoder in these Toshiba 88XX series chips have.

Yeah, that offset has had me spinning in circles looking for a solution. Originally I thought it was something to do with the GreenAntz, and I spent a lot of time looking into it, even experimenting with delaying the Y output (running it through a hand-made 20-pole Butterworth filter!) by a few hundred nanoseconds. But that didn't really help, just pushed stuff to the right. I also found the same issue (or usually worse) with every other type of transcoders, so it wasn't an issue with the GreenAntz either and I was just chasing my own tail. Seeing your pics shows that even your Wii does it. Eventually I realised it was baked into the TV itself.

There are some TV service mode settings that help. In particular to adjust tint balance and sharpness (low or zero), but I can't remember it all now. If an RGB colour is too bright then it blooms/smudges. There is a Y-offset setting somewhere but changing that didn't improve anything.

This picture of yours (below) shows what I am talking about, zoom right in to the "3", and the shadows before and after the letters. It is noticeable where there is a sharp change in colour and Y vs chroma intensity. I assume it is because the TV takes a little longer (few hundred nanoseconds) to demodulate the Pb/Pr signals into RGB. Difficult to see with video, "normal" TV, but visible in games. Is this what you mean?

The pictures I got didn't do it justice at first, so I boosted the color saturation in the these photos to best reflect how the games look to the eye. I also used to help out with a family arcade operator for many years growing up and after the mods, this thong looks every but as good as a brand new arcade monitor running on RGB.

I agree. This is why I gave up trying to match cheap chinese arcade chassis with old tubes and started looking at modding TVs instead. Despite small issues like above, the results are better!

[titan-91]

I can tell this tube is heavier than some other 25" TVs and arcade monitors I've moved. Like it's better quality. I checked my photos again in the early Reddit posts and this particular one is an A63AFW40X made in USA in 1994. It seems it took the CRT a while to "wake up" when I was first powering it on a few times to test. I was having to adjust the G2 voltage down a few times, like it would turn on with faint retrace lines that weren't there during my testing the previous day or two. I finally got G2 dialed in the way I want it. This chassis/series of chips doesn't have a DC restoration circuit though so on mostly dark screens the background will be a bit lit up. With a normal game or DVD, that's not noticeable as the blacks return back to their normal levels.

No the chroma delay I'm talking about is much more noticeable with NTSC decoding over composite and s-video:

https://www.reddit.com/r/crtgaming/comments/1tyxttu/new_aliexpress_board_in_rf_only_tv/

And to compare my N64 on s-video, Paper Mario's rainbow text shows the chroma offset isn't an issue with another s-video modded Funai Symphonic TV of mine:

https://www.reddit.com/r/crtgaming/comments/1ud3tq8/paper_mario_over_svideo/

[Zebidee]

I see. Thanks for the links. Well, hopefully things are looking better now that you have proper termination (75R) and those 104 caps. The way your excess luma was cut off ("clamped") would have left you with greyish blacks.

The thing I was talking about is apparent in some of your screenshots, especially in the one I copied above. I've attached a close-up to demonstrate. See the light-greenish "shadow" after the letters and the "3", and the slightly purplish tint on the leading edges? That is what I'm talking about.

Way I see it, the purplish tint at front is because luma has started drawing a letter but the Pb/Pr is still drawing the background. The greenish shadow after is because luma has changed to background, but Pb/Pr haven't caught up to that yet. The jungle takes longer to process chroma into RGB. Makes sense that it would be more obvious with composite NTSC input, which builds into why they have seemingly weird stuff like those 391 caps on blue and 10uF caps on input etc.