Is the composite video of an NTSC Genesis really this bad?

You can talk about almost anything that you want to on this board.

Moderator: Moderators

User avatar
TmEE
Posts: 960
Joined: Wed Feb 13, 2008 9:10 am
Location: Norway (50 and 60Hz compatible :P)
Contact:

Re: Is the composite video of an NTSC Genesis really this bad?

Post by TmEE »

Those rainbows are present in many NTSC models, some worse than others. The mechanism of why is explained and it stems from very simple(aka cheap) filters on CXA1145 external circuitry which converts the RGB output of VDP into CVBS for AV out and RF modulator. I remember some later models use CXA1465 which is much improved over 1145, but MD2s use a variety of encoder chips, some worse than others and with suboptimal filter bits on the external paths on chips that do need it.

Strangely, MD clones with chips that are pin compatible copies (aka the superclones) often have much better image quality than original Sega hardware, and mostly because they deviate from what is done at CXA1145 parts. (Sound is often better too lol)
User avatar
tokumaru
Posts: 12427
Joined: Sat Feb 12, 2005 9:43 pm
Location: Rio de Janeiro - Brazil

Re: Is the composite video of an NTSC Genesis really this bad?

Post by tokumaru »

Fisher wrote: Sun Feb 12, 2023 6:47 pmWhen transcoded, an external crystal is used, eliminating the problem.
Ohhh, I looked inside the console and the transcoding was indeed done using an extra crystal connected directly to the video encoder chip! Do you think this is why there are no rainbows?

So, do you think that if I replace that PAL-M crystal that's directly connected to the video encoder with an NTSC one, and configure the chip for NTSC encoding rather than PAL, would that result in proper NTSC output but without the rainbows?
User avatar
TmEE
Posts: 960
Joined: Wed Feb 13, 2008 9:10 am
Location: Norway (50 and 60Hz compatible :P)
Contact:

Re: Is the composite video of an NTSC Genesis really this bad?

Post by TmEE »

In theory yes, but there will be moving dot-crawl/jailbars interference pattern as the clock isn't locked to video signal coming out the VDP. What freq is the master clock oscillator in the machine ? It should be 53.203424MHz for true PAL machines or 53.693175MHz for NTSC ones, in which case you could use the subcarrier VDP itself produces, then you can trade the moving interference for static "jailbars".
PAL-M freq is close enough to NTSC so that filter adjustment shouldn't be necessary.
User avatar
Fisher
Posts: 1249
Joined: Sat Jul 04, 2015 9:58 am
Location: -29.794229 -55.795374

Re: Is the composite video of an NTSC Genesis really this bad?

Post by Fisher »

tokumaru wrote: Mon Feb 13, 2023 3:08 am
So, do you think that if I replace that PAL-M crystal that's directly connected to the video encoder with an NTSC one, and configure the chip for NTSC encoding rather than PAL, would that result in proper NTSC output but without the rainbows?
Most probaly yes.
I think you also need to undo some circuit alterations around th CXA chip, most probably pin 7 should be pulled high. I'm not sure if any other modifications needs to be undone, but would be nice to see all alterations properly documented. :wink:
User avatar
tokumaru
Posts: 12427
Joined: Sat Feb 12, 2005 9:43 pm
Location: Rio de Janeiro - Brazil

Re: Is the composite video of an NTSC Genesis really this bad?

Post by tokumaru »

TmEE wrote: Mon Feb 13, 2023 5:05 amIn theory yes, but there will be moving dot-crawl/jailbars interference pattern as the clock isn't locked to video signal coming out the VDP.
Sounds better than the rainbows!
What freq is the master clock oscillator in the machine ? It should be 53.203424MHz for true PAL machines or 53.693175MHz for NTSC ones
It's whatever the original NTSC oscillator on a US Genesis is, that wasn't changed. As far as I can tell, the only changes are a PAL-M crystal directly connected to the video encoder, and it being set to PAL instead of NTSC.
in which case you could use the subcarrier VDP itself produces, then you can trade the moving interference for static "jailbars".
Isn't that how things normally are?
PAL-M freq is close enough to NTSC so that filter adjustment shouldn't be necessary.
I'm still very confused about what to do. What I have is a US Genesis converted to PAL-M with a dedicated crystal and one input of the video encoder changed. I considered restoring it to its original state, but probably won't do it if that will give me those dreadful rainbows.

From what you're saying, it looks like it's possible to end up with an NTSC unit that doesn't have the rainbows, but I'm not sure I fully understand the logic behind that (not that I fully understand why the rainbows are there in the first place), and to be honest I'm not sure I'm okay with replacing a mod with another mod, so I may just leave that console as is...

I'm not as hardware savvy as many of you guys, so every time I do any work on one of my consoles I nearly shit my pants thinking I'm gonna ruin the thing. Most of the time it works, but when it doesn't, I'm not knowledgeable enough to debug the problem and fix it. The only console I ever actually ruined was a perfectly good NES, trying to revert it back to NTSC. I have no idea what went wrong, and even undoing everything I'd done wouldn't bring it back to life.
User avatar
tokumaru
Posts: 12427
Joined: Sat Feb 12, 2005 9:43 pm
Location: Rio de Janeiro - Brazil

Re: Is the composite video of an NTSC Genesis really this bad?

Post by tokumaru »

Okay, so I went and tested all my consoles, and I can't believe I had never noticed some of this stuff until you guys pointed them out to me.

gen-waterfall.jpg
gen-palmtree.gif
gen-palmtree.gif (808.9 KiB) Viewed 1791 times

First, The PAL-M transcoded Genesis. There are no rainbows whatsoever, but I can see a dot crawl effect on the palmtrees. Can't believe I never noticed that before! The dot crawl is not consistent across time either, it slows down, speeds up, stays still for a while...

md-waterfall.jpg
md-palmtree.gif
md-palmtree.gif (783.33 KiB) Viewed 1791 times

Above is the NTSC Japanese Mega Drive. The rainbows are very noticeable, and there's no dot crawl on the palmtree.

md3-waterfall.jpg

Finally, the PAL-M Tec Toy Mega Drive 3. The first thing I see is that the image quality in significantly worse - there's more noise and the colors are way too saturated. The rainbows are there, but are less prominent. There's also no dot crawl (no image because of the attachment limit and it doesn't look much different from the Japanese MD anyway).

So I guess I ultimately have to chose between rainbows and dot crawl effects... The dot crawl is probably more subtle, and makes the image look less blocky, but it's still pretty weird. Maybe it's better to just keep the composite signal as authentic as possible, and use S-video or RGB when the it feels like the composite artifacts are too much.
lidnariq
Posts: 11432
Joined: Sun Apr 13, 2008 11:12 am

Re: Is the composite video of an NTSC Genesis really this bad?

Post by lidnariq »

A Megadrive should use a 15·colorburst (NTSC,PAL-M) / 12·colorburst (PAL-B) clock driver. In raw numbers, those should be:

NTSC: 53.693182...MHz
PAL-B: 53.203425MHz
PAL-M: 53.363418...MHz
(PAL-N: 54.106584...MHz ?)

A PAL-M console will generate the 1/4 dot crawl that you see, regardless of whether it's emitting NTSC or PAL-M chroma. (3420 clocks is 228 chroma cycles, and for PAL-M that's 15682.5Hz hsync, which is 228.251... NTSC chroma periods, or .75 too many PAL-M chroma periods) Hence the drifting every 4 scanlines and also every vsync - there's an excess of .716 chroma periods per vsync.

Tepples mentioned that in the Mega LD PAC the master clock has already been replaced with the "correct" one to generate 227.5 chroma periods per scanline, with a crystal that's 53.811188MHz. This will result in "perfect" alternation of chroma phase every scanline, and no drift from one vsync to the next.

(Or maybe one that's 3420·262÷56905.5·colorburst = 53.810731MHz? Which will result in "perfect" alternation of chroma every vsync, and very close to perfect alternation (off by a degree of hue angle) from one scanline to the next)

To reiterate, all of these require separate clocks for the chroma modulator vs the VDP
Pokun
Posts: 2681
Joined: Tue May 28, 2013 5:49 am
Location: Hokkaido, Japan

Re: Is the composite video of an NTSC Genesis really this bad?

Post by Pokun »

Although the rainbows are clearly visible on my Japanese MD, they don't look as bad as they do on Tokumaru's pictures. My composite picture is much more desaturated however, and Sonic looks kinda greyish blue. Might be the TV's setting.
I'm using an early Japanese model 1 Mega Drive with the circle-R on the Sega logo. The board is IC BDM5 VA4.

Waterfall stand
Waterfall stand
Waterfall jump
Waterfall jump
The real thing looks better than these photos, but the rainbows still don't poke you in the eye as much as in Tokumaru's pictures.

Waterfall RGB
Waterfall RGB
RGB looks much more vibrant, but the waterfall transparency effect is lost.
Oziphantom
Posts: 1565
Joined: Tue Feb 07, 2017 2:03 am

Re: Is the composite video of an NTSC Genesis really this bad?

Post by Oziphantom »

given https://twitter.com/GameSack/status/162 ... 81312?s=20 maybe its a "it was mostly fine on the CRTs they had, but is a lot worse on LCD screens"
93143
Posts: 1718
Joined: Fri Jul 04, 2014 9:31 pm

Re: Is the composite video of an NTSC Genesis really this bad?

Post by 93143 »

I just saw a video today in which I learned there's a piece of concept art for a cut Sonic CD level that has rainbow waterfalls:

SCDconcept-R2RainbowWaterfalls.jpg

Hmm...
Post Reply