Games that look better with Composite vs RGB colors

[samiam]

Here's a shot from the TG16 schematics showing where to tap Y, R-Y and B-Y. This would be easy to get started with. You'll want to take off resistors 130, 131 and 132 anyway, and this gives you open pads to solder to instead of tiny IC legs.

With a cheap old oscilloscope, the 240p test suite SMPTE color bar pattern, and this handy document that explains precisely what each YUV signal should look like in that pattern, you could make sure you had a very accurate amplifier design. An enterprising individual might even make up a board to sell.

I've got all the tools for this, just not the time...

[turboxray]

I plan on buying a core or engine to do some tests with (y,r-y,b-y, etc). I just didn't want to do anything with my SGX (I lost my pc-engine and duo in the move). I also want a non-A revision of the VCE too (just to see if I can find any differences)

samiam: I have Y of the VCE mapped out and it matches the entries in the 1990 US patent (the first 10ish and the last 10ish). I just need to figure out B-Y and R-Y. I think I can do that by cutting the trace for both and just enabling one one at a time (and do a complete capture set of the RGB colors). It's that, or build a demodulator for AM signals which I rather not do haha.

There's a 1995 US patent that shows a different 5bit YUV table and how to derive it (and it's the full table). Unclear where that's used, but it doesn't appear to be the A version of the VCE and most likely isn't thing for PCE systems.

[samiam]

EDIT: Hang on, I think I screwed something up in the experiment I was trying to post about...

[samiam]

Aaaaand now I get it.

First, I'd been wrongly assuming all along that the output of pins 26 (B-Y) and 33 (R-Y) weren't modulated up to 3.58MHz yet. They are, as should have been obvious just by looking at the amplifier circuit for composite video. This means that unless you create a demodulator for each of them (as turboxray mentioned and I now fully understand) you're not going to get YPbPr out of the PCE.

(Does anyone know what the pins on the HuC6260 for R-YC, R-Y+, R-Y-, B-YC, B-Y+, and B-Y- do?)

I managed to take a tired old white PCE I had laying around and do some experiments on it. Removing the resistors after Y/B-Y/R-Y and measuring on an oscilloscope, the most interesting thing I saw was that Y is whopping 3Vp-p on top of a 2V DC bias, with the negative sync tip about 1.2V below the black level, making the whole signal about 4.2Vp-p. I didn't have time to test what happens when you put this under load, however.

I see that a lot of people have been doing S-Video mods like this, or even simpler. I notice that there doesn't seem to be any mention of disconnecting pin 40 (Y) from the rest of the on-board circuitry, which looks to me like you're asking for R-Y and B-Y to contaminate it. Maybe the high amount of resistance in between keeps you safe, I don't know. I'm also a little alarmed about the idea of filtering chroma out of the composite signal. Why not just take out the resistor by pin 40 and not have anything to filter in the first place?

Is there a "high-quality" S-Video mod out there that people who've been around longer than me know about?

If you do get chroma from composite video, filtered or otherwise, note that the composite video inside white PCEs doesn't have proper 75 ohm output impedance. Neither does Y in that mod I posted. This would be another nice thing to pin down.

[turboxray]

Dec 26, 2019 11:03:08 GMT samiam said:

(Does anyone know what the pins on the HuC6260 for R-YC, R-Y+, R-Y-, B-YC, B-Y+, and B-Y- do?)

Ohh.. I don't remember seeing those! I'll check them out on the scope, if you haven't already.

Yeah, that svideo circuit has chroma contamination on the Y line (as well as color burst which I'm pretty sure should be with C, but I can't remember haha).

Demodulating R-Y and B-Y shouldn't too difficult. At that point, you can do a nice Svideo, YPbPr, and YPbPr->RGB output circuit.

[samiam]

Ohh.. I don't remember seeing those! I'll check them out on the scope, if you haven't already.

I can certainly take a look. If not today, then probably tomorrow.

The thing I'd like to try next is an S-video mod, with luma fully isolated and chroma simply taken from the internal output (although on a white PCE, this probably needs to go through another stage). What the best amps are, however, I'm not sure. I think I could get away with all sorts of configurations, but something that's efficient, accurate, and electrically sound as far as transmission goes is not so easy, at least for me.

Do you know your way around transistor-amplifiers? My grasp on the subject is tenuous at best. Here are some possible luma amps along with links to a circuit simulator. These include the 75 ohm termination at the TV end and pin 40's luma output as the 40Hz source, which I tuned to match what I measured. I'm sorry to say that other than the first one, I'm not actually sure which one has the closest to 75 ohm output impedance:

1:

(Link to simulator)


2:

(Link to simulator)


3:

(Link to simulator)
(More power-efficient version that draws a bit more from pin 40)

4:

(Link to simulator)


Thoughts?

[samiam]

Does anyone know what the video output circuit is like for PCE models with composite output built in? Is it the RF-only model's circuit plus the additions in the Turbo Booster, or is it something else?

The schematic for the Turbo Booster shows that there isn't much else on there for video. I'm tempted to just play it safe and recreate the exact amp from the system itself twice for Y and C. However, having modeled it in the circuit simulator, I'm surprised to see that it's very power-hungry. At peak brightness, one of these amps consumes ~66mA all by itself. The yet-unproven power-efficient amp I modeled above uses less than half that. The two that don't seem to aim for impedance matching use more like a third.

This is the white PCE/TG16, however. The other models may have done things differently. 

EDIT: turboxray, I think I found what those +/-/C pins are for. They're on the other side of the chip in that service manual schematic. 

[keithcourage]

This is what most of us who do mods have been using for a while now for s-video. I should note that for Black DUO systems the ground needs to be tapped from the center pin of the AV jack. Otherwise you get blurry image. 

[samiam]

I'm sure that the results of that mod look way better than composite. It could also be that quality-wise, there won't be any discernible difference between that and a more electronically-kosher mod. 

But there is definitely room for a more electronically-kosher mod. Take a look at the TG16 schematic.




The mod you posted allows luma and chroma to contaminate each other. If you just tap pin 40 for luma without somehow severing the connection to pins 33 and 26, then some amount of chroma is going to leak back through that 4.7k resistor into your luma line. Also, when you tap composite video for chroma, you need that 1000pf cap in order to filter chroma out of luma; chroma would almost certainly be better off not being mixed with luma in the first place. 

Both of these issues can be easily dealt with by removing that 4.7k resistor - chroma stays out of luma, and luma stays out of chroma. You could also forgo the 1000pf capacitor in the chroma line if you did this.

Additionally, using a 220 ohm resistor on the output of that transistor creates an impedance mismatch with the 75 ohm termination resistor that's inside everyone's TV. As a practical matter, this isn't going affect many people, but I guarantee that if you were using a 10 meter S-Video cable to connect your system to your TV, you'd get visible signal reflecting. 

Although it will require a few more parts, we really ought to have a mod design with 75 ohm output impedance. This is a bit tricky since the HuC6260's Y output is so unusual, and it also apparently can't supply much current, but hopefully it's still possible with a single transistor and a few resistors.

EDIT: Using two transistors, we can make something like this. It has extremely high input impedance for pin 40, which is good, and it can drive a 1Vp-p signal with no DC bias to the TV. I left out the AC coupling capacitor because it was distorting the signal, which ain't kosher, at least as far as being industry-compliant goes. However, with no DC bias and no way to short power directly to ground, it should be fine. I don't think there are any TVs out there with a live chassis that have S-Video inputs. 





(Sorry to go so far down this tangent. I'll make a new thread the next time I want to post about an updated S-Video mod.)

[turboxray]

Dec 30, 2019 0:55:21 GMT samiam said:

Both of these issues can be easily dealt with by removing that 4.7k resistor - chroma stays out of luma, and luma stays out of chroma.

This! Y isn't filtered coming out of the PCE and it's incredibly sharp haha.

[samiam]

Dec 26, 2019 14:19:22 GMT turboxray said:

Ohh.. I don't remember seeing those! I'll check them out on the scope, if you haven't already.

OK, I had a chance to scope a few of these properly. I'm not sure exactly what their function is, but nothing much interesting is happening on them. There is certainly no unmodulated R-Y or B-Y for us to exploit.

Over the new year, I built a couple of test-circuits and discovered some interesting stuff, including another issue with the old S-Video mod that's much bigger than anything brought up so far. Any guesses as to what it is? It involves pin 40 and the transistor.

I'll put the answer in the new thread I'm making. 

[turboxray]

I dunno haha. Some of the analog stuff is out of my experience.

[samiam]

Jan 4, 2020 5:40:12 GMT turboxray said:

I dunno haha. Some of the analog stuff is out of my experience.

It certainly explains why NEC's amp uses 3+ transistors...

[Arkhan]

Dec 16, 2019 6:11:15 GMT elmer said:

Back-in-the-day most Western artists would often check their work by looking at it on a TV, and then tweak things if necessary.

The difference between what the artist was seeing on their computer, compared to the actual composite-TV output of different consoles was a well-known issue.

I did that with blue in Insanity but gave up anyways because drawing is hard.

[samiam]

I got luma to appear on a TV, with all the proper buffering and scaling of the waveform.


One pattern from the PCE's 240p test suite, which I've been using, looks like this:



Converted directly to greyscale on my PC, it looks like this:



If you were to convert the first image to YUV without any of the PCE's rounding errors, the second image is what Y (luma) would look like, right?

What I saw on my TV screen last night was unmistakably different. I guess I need to re-calibrate the TV's brightness and contrast before I say too much, but blue especially was very different. Its two brightest grey-shades were the same, and the next two after that were also the same. Interestingly, when I connected chroma (which currently has problems but still kinda works) there were four distinctly different shades visible.

This isn't anything too different from what turboxray was explaining, but I was excited to see the difference come out nonetheless.