To combine the RAMs you can just put the NVRAM on the bus and select it in the right address ranges. The trouble comes in the address decoding part. That's where you're going to need some kind of PLD. It doesn't really make sense (IMO) to "redesign" an MPU and use 74 series logic for the address decoding and other basic logic functions.

The other issue is the requirement by some Bally code to have a 4-bit RAM with the higher data bits tied high. You could wrap up the 4-bit/8-bit RAM switching in PLD, but it might be cheaper/easier to use two 8-bit buffers like the Alltek MPU. Using that method you could get away with a GAL16V8 for the address decoding, making a pretty simple design.

Here's an example partially finished layout. I'm doing the address decoding, RAM compatibility, and PIA output multiplexing in one CPLD. Display interrupt generator, zero cross, and clock circuits not shown. This example is mid way through a design that I did for this MPU. In the final design I broke down and added 128 DIP switchable ROM banks. Too many projects in the queue before I get to this one though... maybe next year.

Bally_MPU_PCB_Partial_Layout.jpg

If you don't like PLD you'll really hate the fact that I used a microcontroller for the display interrupt generator and display blanking delay circuit

PLD hasn't hurt Alltek's sales, but there's obviously a market for the original design as well because exact replicas are also currently available. You can mitigate the serviceability issue by publishing the JED file for the PLD. It's not as simple as replacing a 74 series IC, but not un-fixable.

I tried to use the microcontroller to also generate the two clock speeds, but it didn't work that great because the MPU-200 clock speed is not divisible evenly with the PICs internal clock. If you only wanted to do the slower clock speed (500Khz) you could use a PIC 10F320 and provide the display interrupt with a PWM pin, clock the 6802 with a clock output pin (PICs can be configured to output the internal clock on an IO pin at 1/4 speed and has an internal clock up to 16Mhz), and still have two pins and enough CPU to do the little circuit on the display blanking. The PIC running at 8Mhz would drive it's clock out pin at 2Mhz which would then be divided by four again inside the 6802 to give the 500Khz clock which would then be present on the E pin of the CPU. That would eliminate a whole bunch of passive components and a few ICs using a single ~$0.75 part in an 8 pin DIP package, buuuuut... you have another non-off-the-shelf part in there. And you would have to write and test the microcontroller code. Sometimes it's easier to just duplicate the original.

It really depends on what you want to build. If you're only going to do -17, -35, and MPU-100 compatibility using a 6802 configured for internal RAM, it's easy enough to do it with basic logic. If you want to add MPU-200 and 6808 compatibility, it get's more complicated. I took the approach of minimizing parts for easier assembly and offering compatibility for all Bally/Stern titles. In terms of reselling for profit, the assembly time and parts count are areas where a lot of cost can be saved (IMO). Eliminating a lot of passive components is pretty compelling if you're using DIP parts and hand assembling them.

If the price is right people are not going to care if there's programmable logic on the board. Especially when it allows you to offer features which would not otherwise be available. Alltek has sold several thousand of them, I'm sure, and no one complains about the design. Being simple doesn't really factor into it. An IC with no external components is simpler by definition than a 555 timer circuit. The whole MPU is very basic. Like I said, I designed it to make assembly more efficient. About 10 less passive components to bend the legs, stick through the holes, solder, and clip the legs by adding the PIC with no additional cost. Seems like an obvious choice to me. I'll offer a warranty on the MPU, publish the .bin file and offer programmed PICs for $1. Problem solved. The time to install those passives adds up fast. Installing all of the resistors used in the MPU with DIP packages is going to be a nightmare. I knew that with this design there might be a hope that I would actually assemble a bunch and sell them. A direct dupe of the original would have been a lot easier to design, but there's no way in hell I'm assembling a dupe of the original design using DIP parts because it would take FOREVER and I would make nothing, even at $199.

Same thing with the clock and zero cross detector. They're very simple too, but there are "better", more efficient ways to do it using less parts which don't need PLD or a microcontroller. If you're planning to go into large scale production these changes will make a big impact to the overall cost because assembly is going to be a big chunk of the cost with something like an MPU. Especially if you're doing it yourself. Every minute you save is more profit in your pocket. If you're planning to have them assembled, they're going to be pretty expensive using so many DIP parts. On a small scale like this, you're going to spend an arm and a leg having them assembled offshore. The time saved by shaving off a ton of passive components and ICs is considerable. Anyone who's ever assembled one of the kits from Homepin knows what I'm talking about. Obviously, the biggest obstacle to any of that is knowing how to program PLDs and microcontrollers, but that's really worth exploring because then you have the options to do all kinds of things in pinball. The problem with rehashing what's already available and undercutting the price is that anyone can do it and pretty soon you're all competing for peanuts. I'm surprised this hasn't happened with MPUs yet. Especially Bally, but I assume the assembly effort is what stops people. It's definitely not the complexity.

I'm not trying to convince people to do it my way, but I do have good reasons for doing it the way I did which will work best for me in the long run. I could offer these things REALLY cheap if I wanted to, because the assembly will be WAY cheaper than a conventional design with DIP parts. I was also able to get the total size of the board way down. That's also going to be a big cost savings if you're not ordering the things in 1000+ quantities. All of this stuff adds up to my boards being cheaper to build than the "competition" by a considerable margin.

I started out thinking about offering a "basic" MPU as well without MPU-200 compatibility or ROM for all of the games but then I started thinking that adding ~$5 in parts could add $50 to the final sale price so I went with that option but who know's if I'll ever sell any. I'm working on other projects for the foreseeable future.

The main reason I would use the NVRAM to replace the 6810 rather than using the RAM inside the 6802 is because I've seen the RAM inside the CPU fail a lot more often than the CPU itself. If I were using pulls, I would avoid using the RAM in the 6802 for that reason.

I don't doubt that. They may have been tested before you got them. I've seen it happen a couple times where the RAM will fail in the 6802. It's not an epidemic or anything. I still trust the new RAM farther than the old and it doesn't cost me anything to configure it that way.

That's something I've been working on. Either replacing the PIAs with CPLD or the CPU and PIAs with FPGA. It's not a difficult thing to do in theory because the VHDL for the CPU and PIAs are already out there. The hard part is making it cost effective. Until the PIAs become scarce, it's still a lot cheaper to just run with the 6821s.