Mega Drive / Genesis 2 Rainbow banding/fringing

[Syntax]

Ive recently acquired one of those random VA0 ASIA MD 2 units and it seems to have bad banding after a crystal switch.

These units come pal so I changed the main crystal out for one from an NTSC MD and wired up the 60hz jumpers.
(The donor MD got the pal crystal and a 4.433MHz crystal on the video encoder. So not perfect CPU clock but great composite still)

Composite image has terrible rainbow bands, as can be seen in this similar picture.




I kind of wanted to blame the Fujitsu MB3514 video encoder they used on this model but I swear before I modded it to use the 53,6931MHZ NTSC crystal I tested it on an LCD and there was no banding, I would of noticed it I'm sure.

I figured that using the 53,6931MHZ crystal along with a 60hz mod would make the Pal machine a true NTSC with proper CPU clocks.

53,6931MHZ - NTSC 68k speed: 7.67MHz
53,203MHZ - PAL 68k speed: 7.60MHz

So I'm starting to think that the NTSC subcarrier signal in these MD2s is shit and causing the banding.
I don't have an oscillator on hand yet but as soon as I do ill cut the feed from the original crystal/chips that is supplying the 3.579MHz NSTC color subcarrier to the Fujitsu MB3514 and replace it with a 4.433MHz PAL one.

I have this funny feeling that PAL60 will fix the crappy picture. Ive never seen it before on a PAL console.
Maybe some Americans can chime in.
Not that any of this matter to me, its what happens when I'm waiting for my RGB cable to arrive lol.

**EDIT**
Looking at the Fujitsu MB3514 video encoder it seems its designed for PAL. There's no reference for NTSC Burst Phase or Burst Level Ratio, only PAL
http://www.smspower.org/uploads/Develop ... MB3514.pdf

[xAzurexEonx]

Did you ever make any progress on this?

I've only ever known to fix this was by changing caps, using RGB over the composite encoder or use 32x.

You may be correct that maybe the signal is just bad for this conversion.

[Syntax]

I have a 4 leg oscillator here but it's in a genesis.
Been ripping apart dead tv's but only found 2 leg crystals and making the oscillator yourself is dodgy.

Already proven that you can run 60hz with pal subcarrier.
The genesis now has the megadrives main oscillator with a seperate pal subcarrier oscillator feeding the amp working in 50 and 60 hz mode.
It is PAL60 also. Colors are nicer.

It really makes me wonder why some people install the wrong oscillator and put a switch to it..

Retrosales are really fast at posting stuff to me so I'll update the second I do it.

It could be caps, this unit I have is BRAND NEW so the caps may of died in storage. But I swear the original pal picture was rainbow band free.
Ntsc subcarrier is also know to introduce artifacts.

Looking over the amp pdf I really get the feeling the amps made on pal but "compatible with ntsc. Just a feeling..

[Hoagtech]

Sonic team used composite video processing to create certain effects on their genesis Sonic titles.

The popular effects are the shimmering semi-transparent shield which is checker-boarded on RGB, and the rainbow effect is argued to be intentional as well.

It looks like mist is falling from the waterfall and creating a rainbow whereas the RGB output looks like blue pinstripes.

I actually prefer those effects and miss them when I play through RGB.

https://www.reddit.com/r/emulation/comm ... ncy_trick/

[Syntax]

Yeah, If only every single Mega Drive game had waterfalls in the background then you'd be set....
Its not intentional, it has occurred as a result of my NTSC mod I believe and I intend to fix it.

[Blair]

we definitely had the rainbow banding problem in the US, I know my Genesis 2 system from the 90s had it pretty bad. my Sega CD-X is much less noticeable. (it might not have any at all, I haven't used composite in a while)

as I recall the video encoder on the 32X produces no rainbow banding. (here's a meh comparison screenshot)



so if you need composite video out of a Genesis it's probably best to get something like that, however, why not use RGB or S-video instead? (the Genesis and 32X natively output RGB, and I've heard that an s-video modification is fairly trivial for somebody with basic electronics skills) unless of course you have a specific composite only monitor in mind that you wish to use.

[xAzurexEonx]

we definitely had the rainbow banding problem in the US, I know my Genesis 2 system from the 90s had it pretty bad. my Sega CD-X is much less noticeable. (it might not have any at all, I haven't used composite in a while)

as I recall the video encoder on the 32X produces no rainbow banding. (here's a meh comparison screenshot)



so if you need composite video out of a Genesis it's probably best to get something like that, however, why not use RGB or S-video instead? (the Genesis and 32X natively output RGB, and I've heard that an s-video modification is fairly trivial for somebody with basic electronics skills) unless of course you have a specific composite only monitor in mind that you wish to use.

now that you mention S-video. I can't recall the banding because when I got a new Gen 2 around the time SNES mini came out both were like 50 bucks at Walmart, it came with only a S-video cable which at the time I hated because none of our TVs had this input (barely composite too). So it had to be plugged into a VCR instead.

[Syntax]

so if you need composite video out of a Genesis it's probably best to get something like that, however, why not use RGB or S-video instead? (the Genesis and 32X natively output RGB, and I've heard that an s-video modification is fairly trivial for somebody with basic electronics skills) unless of course you have a specific composite only monitor in mind that you wish to use.

Ill be using RGB into a CRT but later on in life I may want to use something like a framemeister which can have issues with syncing to a moded genesis.
This is one of 2 reasons I switched the main megadrive oscillator for a genesis one( the other reason being correct cpu clock speed)
At this point because I did not have a megadrive 2 rgb cable yet I was testing via composite and noticed the rainbow bars instantly.

The only thing that changed to the video encoder is the subcarrier signal is now NTSC, which is know to introduce artifacts.

If I change the incoming signal for the video encoder to a PAL one and it fixes the rainbow bands then it puts the debate about it to bed.

Just a bit of fun for me while I wait on amps to come in really.

[Syntax]

Spent very many hours tonight playing around with oscillators, here are my findings
osc = Oscillator

Connected Pal osc and it fixed the rainbow banding, but introduced heavy ghosting and color bleeding and interlace like noise.
Attempted reducing the osc voltage,signal grounding and resistance on 10k pots and different coupling caps too.
Nothing fixed the heavy ghosting, some things I did made the color bleed either run left or right.

Connected NTSC osc and it reduced the banding by about 30% but introduced interlace like noise. No ghosting.
Connected a switch between the original subcarrier NTSC signal and a new NTSC one to check differences on the fly.

Noticed instantly the noise on the new NTSC osc smoothing the picture out.

Decided to externally power the osc and vary its voltage to see if over 5v helped.

Found out very fast that the noise on the screen is infact directly related to the voltage supplied to these oscillators.
I had no noise at 6.7v on the pal one and the NTSC one operated nicely at 3.4v and 5.3.

After tuning the noise from the NTSC osc as best I could it still has a noticeable interlaced/buzzy thing going on.
Flicking between the original signal and the "tuned" NTSC osc one the new one almost seems nicer, like a poor mans composite low pass filter, if only it didn't have the buzz.

Still the original signal is so much pixel for pixel clearer and noise free.

After all this I wont be using the new osc, ill keep using the original signal as I use composite video for sync on a few screens here and the original signal was so much more stable.

Only thing left to do is hook up a megadrives subcarrier signal to my genesis and see if the banding is gone and ghosting gone also, then ill scope the megadrives subcarrier vs the new PAL osc. Which I'm sure is probably impossible for me to recreate, and a mostly pointless exercise.

Till then here's some pics or it didn't happen


NEW NTSC OSC


NEW PAL OSC


What I had to drive the PAL osc to get a stable signal


NTSC TUNED TO 5.SOMETHING VOLTS (over 5v that's all that matters)


Original Subcarrier

[Syntax]

So running an original megadrive subcarrier signal clears the bleeding up some but the colors are gross and still has ghosting.

[Syntax]

Did some more testing and found that

MEGA DRIVE
60hz mode requires a 4.433MHz crystal to avoid ghosting
50hz mode requires a 3.579MHz crystal to avoid ghosting

GENESIS
60hz mode requires a 3.579MHz crystal to avoid ghosting
50hz mode requires a 4.433MHz crystal to avoid ghosting
Install a 4.433MHz crystal and lift and wire pin 7 of the encoder to high low you can switch between a really shitty ghosted NTSC 60 and PAL 60
Install a 3.579MHz crystal and lift and wire pin 7 of the encoder to high low you cannot switch between PAL and NTSC 60, PAL60 has wrong colors.



These crystals all perform differently and require a trim pot to get the noise down to near 0. Even then they seem to buzz but smooth the picture overall.
The ones I used need 5v or more, and the old Mega drive only outputs 4.5v on both rails and the chip wanted exactly 5v. Genesis has 4.95v on the rails but the chip wanted around 5.3
I stole power from the powerpaks via a 10k pot for testing but that's not a solution for all those that have one of these installed. It varies too much.

So the 2 different systems require different crystals in 60hz mode, WTF??
I don't really understand that, I figured it would be 1 crystal for 60 and 1 for 50.

Another feed to the amp I missed? I thought id been pretty thorough with testing the subcarrier line and pin 7 (ntsc pal high/low)
I was fairly sure that pin 7 and 6 were the only ones that were feeding the encoder other than the rgb lines

[rama]

All oscillators / crystals will drift and wander over time, which makes it important that the video components are all driven by the same master clock. When you use an external clock for the subcarrier, then this frequency will not be syncronized with the GPU pixel rate and you get additional, moving artefacts.

A better way to improve Composite output is to replace the video encoder with a CXA1645 or even CXA2075.
Those newer encoders have most of the problematic passive components already built into them.
Also they provide an easy to use y-trap filter that nicely softens the harsh RGB pixels while reducing rainbow banding.
It's worth trying!

[Syntax]

All oscillators / crystals will drift and wander over time, which makes it important that the video components are all driven by the same master clock. When you use an external clock for the subcarrier, then this frequency will not be syncronized with the GPU pixel rate and you get additional, moving artefacts.

A better way to improve Composite output is to replace the video encoder with a CXA1645 or even CXA2075.
Those newer encoders have most of the problematic passive components already built into them.
Also they provide an easy to use y-trap filter that nicely softens the harsh RGB pixels while reducing rainbow banding.
It's worth trying!

I was considering swapping out the video encoder, might give it a go.

Thankyou for point out the synchronization issue, its something that went completely over my head. Makes total sense.

[Syntax]

Put the osccilators to the test with a scope today.

Here are the results.

PAL OSC (Dips high and low)


NTSC OSC (dips high and low)


NTSC Genesis 2 the one with rainbow banding (Dips low but not high)


PAL Mega Drive 1 (No dip)


Using a scope is new to me so if anyone has any input as to what's going on it would be great.
To me it seems the MD1 has a nicer wave.

If only I had a Genesis 1 to test the NTSC subcarrier shape and attempt to inject it into a Genesis 2 with banding to see if it clears it up.

[Blair]

Syntax, your experiments are incredibly fascinating. if you learn anything new please keep us up to date.

you also talked about wanting a FrameMeister eventually, have you looked into the OSSC (open source scan converter) project at all? since you like to tinker and experiment that might be a better fit for you. ( although the OSSC only supports RGB, Component (YUV) and VGA/RGBHV, so no composite or S-video)

[rama]

The signal shape of the subcarrier doesn't have a big effect on the final picture. If you filter it (LC or RC filters work), you might get a little less jailbars. It won't fix the rainbow banding though, I've tried

The jailbars are caused by having the MD's dot clock (pixel clock) be a multiple of the NTSC color carrier clock.
With a good video encoder, the effect would be minimized. The 1145's or their clones can't deal with that though.