new hybrid mode found

Página 2/2
1 |

Por Jorito

Mr. Ambassadors (1782)

Imagen del Jorito

08-03-2012, 19:08

But what are the boundries for the wiki?

As far as I'm concerned there are no real boundaries to the wiki. And as I'm crew, I hope to speak for the entire MRC team Tongue

Treat it like Wikipedia; try to be as factual as possible. If you don't get it complete or fully right the first time, there's always the community to help and polish the wiki. So don't be afraid to take some iterations before you're fully done.

Por hit9918

Prophet (2923)

Imagen del hit9918

20-03-2012, 21:51

@hap, did I get this right, the 9118 still got mask registers for pattern and color, but the character mask bits were turned to color mask bits? Does it still have the sprite bug? The question arises which kind of mask registers the 9938 got.

I had funny ideas like detecting TMS by 5th sprite of the mirror sprites. But now another VDP variant seems to have popped up which got only a part of the mask features. So the game got to have a manual setup for every detail used, "character masking yes/no". To "F18A" etc makers I advice to implement the character masking, nice performance potential in there.

Thoughts about how history turns out:

If 9918A would have no sprite bug, the feature would be used, and no clone VDP without the feature exist. Oh what pity. On the other hand, who knows, without that bug MSX 1 maybe would have only 4k vram! Maybe not all models, but you would have it as the lowest common denominator.

Because the sprite bug tends to enforce that a 9918 machine got to have 16k vram:
NES got 2k or so RAM. While coleco got 16k vram.
TI99 got ZERO k RAM. Just 256 bytes scratchpad. But 16k vram.

Por PingPong

Prophet (3898)

Imagen del PingPong

20-03-2012, 22:25

@hit9918: can you explain in detail what is this new hybrid mode? is really a new undocumented one?
I've read your post, but i cannot understand what is this new video mode...

Can someone create a video demo?


Por hap

Paragon (2040)

Imagen del hap

20-03-2012, 23:30


the 9118 still got mask registers for pattern and color, but the character mask bits were turned to color mask bits? Does it still have the sprite bug?

yes, and yes.

Detecting type of TMS chip automatically by undocumented features like sprite cloning is not very reliable.
The sprite clone glitch has a peculiar 'feature', where the number of clones changes depending on the temperature of the VDP chip. More clones in a fridge, less clones with a hot blowdryer to the chip. Tongue
Another example is the 9918A addressline scrambling when setting it to 4K mode (I'm sure you know of this glitch).. but, it gets more awkward that this also disconnects part of RAM from refreshing, causing it to slowly degrade to its initial state.

Por hit9918

Prophet (2923)

Imagen del hit9918

21-03-2012, 00:42

@PingPong, the mask runs on the "ASCII" values, allowing doublebuffering without writing the nametable,
see right at the start of the thread, there also is URL for the TI users video.

@hap, tested on 9929:

It is like with pattern mask bits pulled low, there is too little fan-out for the sprite circuit to receive the correct Y value. The top mask bit affects "Y=128", the next one "Y=64", they pull to 0 the bits of the Y value the sprite engine gets. The beam is down the screen, but the sprite engine thinks it is more up.

The color mask does add to this effect, but colormask alone is not enough to cause the bug. The TI99 video got similar wording.

So when using both masks, the bug is permanent, while when only using pattern mask, I saw the bug fade from "fully present" to "gone" as the chip got hotter! The transition happens in a funky "sparcle" style, every frame and every scanline and there can be bug yes/no.

The character mask bits got no effects on sprites, and this is compliant with the "mask registers hurt Y value fan-out" theory: character masking does not work dependant on the Y counter, but on the byte read from the nametable. So it got no business with the Y value and does not draw too much current in that area.

Back then, one could have explored whether one can remove the bug completely by running the VDP extra hot Big smile Today better use software hacks.

Por Bit Addict

Resident (34)

Imagen del Bit Addict

15-03-2015, 21:59

I agree that a WIKI topic covering the undocumented or hybrid modes of the (different) VDPs would be a good addiction, as some mentioned in this post. I am curious if it is actually done.

Por Grauw

Ascended (10605)

Imagen del Grauw

15-03-2015, 22:43

If anyone wants to write an article for the MSX Assembly Page, submissions are welcome Wink.

Página 2/2
1 |