MMC1 MAPPER impossible to do with discrete parts

Discuss hardware-related topics, such as development cartridges, CopyNES, PowerPak, EPROMs, or whatever.

Moderator: Moderators

darleiv
Posts: 69
Joined: Wed Feb 23, 2022 10:55 pm

Re: MMC1 MAPPER impossible to do with discrete parts

Post by darleiv »

Bregalad wrote: Tue May 17, 2022 5:25 am
Ben Boldt wrote: Sat May 14, 2022 6:40 pm We should have a challenge. Design 100% compatible MMC1, only use 74xx chips (any, including shift registers, rare ones etc), minimum # of chips wins.

How many do you think it takes? What’s realistic?
The problem is that doing so is completely useless. Many chips technically existing in the 74xx series are rare or even very rare. You could play and make a schematic with them but you'd never be able to even try if this works. And even then, to get the exact replica of an MMC1, you'd need at least 5 chips to get just the shift register to behave the same. The major problem is the shift register who is 5 bits, and internal registers are 5 bits, when most chips have 4 of the same element. This thing alone increase the # of chips needed to get it done.

On the other hand, it's possible can use PAL22V10 or GAL22V10 chips which are common and still available, and only a 5 of them can replicate a full MMC1 who is identical (3 are enough if you use CHR-RAM). But even when suggesting doing that, I was critiquized this way crazy and that using a single CPLD chip is cheaper and easier.

I already saw your project. actually the mmc1 register is the difference of it not being so simple. but I believe that to run a game only CHR PRG rom. Be practical using TTL , the SXrom mappers are quite peculiar .
I think there is room for everything. each project sums + the global retro console community
your code is very well improved beautiful.
mmc3table2.png
lidnariq
Posts: 11432
Joined: Sun Apr 13, 2008 11:12 am

Re: MMC1 MAPPER impossible to do with discrete parts

Post by lidnariq »

The other problem with GALs is that they're very thirsty for current - using 5 could easily consume 200mA.
User avatar
Fisher
Posts: 1249
Joined: Sat Jul 04, 2015 9:58 am
Location: -29.794229 -55.795374

Re: MMC1 MAPPER impossible to do with discrete parts

Post by Fisher »

Wouldn't be easier to hack most of the games to use a mapper like mapper 78?
lidnariq
Posts: 11432
Joined: Sun Apr 13, 2008 11:12 am

Re: MMC1 MAPPER impossible to do with discrete parts

Post by lidnariq »

INL historically asserted that every existing MMC1 game only uses a subset of the functionality, so any one given game can fit into a single 36-macrocell Xilinx CPLD.

(The full amount of state needed for the MMC1 is only 24 bits, but the other macrocells are needed for the glue logic)

The problem is that the only Discrete Logic mapper with 4+4 banking is NINA-001. It's a lovely mapper for that reason by itself, but it means that there's no ready-to-use mapper for, say, mirroring control and 4+4 CHR banking. I could have sworn the wiki used to describe some de-serialized MMC1 mapper variant but I can't find it now.
User avatar
aquasnake
Posts: 515
Joined: Fri Sep 13, 2019 11:22 pm

Re: MMC1 MAPPER impossible to do with discrete parts

Post by aquasnake »

Mmc1 series is a huge family. Some sub-mappers do not have a superset-subset relationship, so the implementation consumes logic resources and has to be isolated by state machine, which leads to the discretization of mapper. One or two sub mappers can be realized by 74 series logic gates, but it is impossible to emulate all of family

Why still insist on using 74? I think this is a meaningful topic. Building circuits with 74 logic gates has less learning cost
darleiv
Posts: 69
Joined: Wed Feb 23, 2022 10:55 pm

Re: MMC1 MAPPER impossible to do with discrete parts

Post by darleiv »

aquasnake wrote: Fri May 20, 2022 7:38 pm Mmc1 series is a huge family. Some sub-mappers do not have a superset-subset relationship, so the implementation consumes logic resources and has to be isolated by state machine, which leads to the discretization of mapper. One or two sub mappers can be realized by 74 series logic gates, but it is impossible to emulate all of family

Why still insist on using 74? I think this is a meaningful topic. Building circuits with 74 logic gates has less learning cost


a xilinx would be ideal because the ES ports accept simultaneous 3v/5v without level shifters
there's a mapper made by the Latin of the consoles for siempre that they say it works
Post Reply