Initial thoughts
Posted: Mon Jul 27, 2020 7:56 pm
I have started a new sub-forum for this project. I have talked about it several times before in various areas already and I know people are not overly enthusiastic about this project.
To summarise, the H4/H5 series was planned to be easily upgradable "STF" machine. Which of course we have done. But we should remember that the "H series" is still actually a development platform. Where developing add-ons will keep us all busy for some time yet. Basically to the point that I cannot actually envision at this point there being a future H-series board.
As discussed before, the next logical step would be to bolt-on various upgrades rather than having them as a plug-in add-on. That in itself has its own pros and cons and a topic in its own right. But the problem is at this point with the H5, we are just simply out of room now. Making further changes to this board is basically getting very messy.
With original boards failing all the time, I want to produce a series of boards to replace originals which I am calling the legacy board. Where I would take the H4 design and strip it back to a bare minimum STF. In part to keep costs low to make them as affordable as possible, but also to start building a new motherboard layout which could potentially be built upon into its own series, which could ultimately take over from the H series of boards. I suppose the way of looking at it is more of a "Fixed Alpha" board but with more SMT.
This does not mean the legacy board will be without upgrades. I do plan to bolt on items such as Flashyclock, basically giving it a flash ROM and alt-ram. With this being a legacy machine, being able to flash any version of operating system should really be a must in this case.
As for adding future upgrades like accelerators, because the intention of this is a legacy board, I'm really leaning towards not even having any expansion sockets on it at all. Because bolting a accelerator onto a legacy board basically defeats the object of having a legacy board. People who want all the bells and whistles really would still have to get a H5 type board. the future of the H series board may well ultimately have a built on 030 CPU for example. but of course, every upgrade and change which is made, is taking us away from the original ST architecture.
For example as people know we are working on FPGA MMU, GLUE, SHIFTER etc, while these will remain highly compatible, and firmware updated at a later time, It is never going to be 100% compatible IMO. The only way to have that full compatibility is to use the original chipsets. With the H4/5 series heading towards faster CPUs and higher resolutions and better sound systems, you're basically left behind the original machines at that point which is where I think the legacy board will ultimately fill that gap.
It will also allow me to totally reorganise the layout of the motherboard. The current layout was to allow the main chips to be in close proximity to each other to allow a FPGA board to plug "over" the entire area replacing those chips. Unfortunately the routing in doing that was probably the worst it could ever be.
Ironically, the "silly style" board layout is probably more like how a optimise layout should be. Unfortunately those silly style boards ended up being the worst style of motherboards. Even so the 70789 types have always been a reasonably good layout. Mostly it is just the relocation of the GLUE. because it drives a lot of chip enables, it should really be on the right-hand side of the board so that multiple routes don't have to run right across the whole CPU bus.
As mentioned already, not having any 68K sockets will literally allow the board to be half the size.. Again this will be cheaper and very small to the point we could design a half size case similar to the STM to house it. With shipping getting more and more expensive as time goes on, I think trying to scale down the motherboard and the case is a good idea anyway. I *might* fit a single 68K socket vertically to at least allow some expansion capability, but again this is not really the design goal for the legacy board to be upgradable.
I don't honestly think this board will be more of much interest to people here. Basically because it is in many ways a huge step backwards from the H4. But with that series heading off into its own direction, it will essentially leave original machines "dead" in times soon to come with no legacy alternative.
It is becoming more and more apparent over time that original machines are becoming ever problematical and will also become incredibly expensive and time-consuming to fix. But if there was a legacy board available which was a straight swap and only need a few chips changing over to get it working, its keeping the spirit of the original machines alive for many more years to come.
EDIT:
I wonder also if it could be pushed onto a 2 layer board
To summarise, the H4/H5 series was planned to be easily upgradable "STF" machine. Which of course we have done. But we should remember that the "H series" is still actually a development platform. Where developing add-ons will keep us all busy for some time yet. Basically to the point that I cannot actually envision at this point there being a future H-series board.
As discussed before, the next logical step would be to bolt-on various upgrades rather than having them as a plug-in add-on. That in itself has its own pros and cons and a topic in its own right. But the problem is at this point with the H5, we are just simply out of room now. Making further changes to this board is basically getting very messy.
With original boards failing all the time, I want to produce a series of boards to replace originals which I am calling the legacy board. Where I would take the H4 design and strip it back to a bare minimum STF. In part to keep costs low to make them as affordable as possible, but also to start building a new motherboard layout which could potentially be built upon into its own series, which could ultimately take over from the H series of boards. I suppose the way of looking at it is more of a "Fixed Alpha" board but with more SMT.
This does not mean the legacy board will be without upgrades. I do plan to bolt on items such as Flashyclock, basically giving it a flash ROM and alt-ram. With this being a legacy machine, being able to flash any version of operating system should really be a must in this case.
As for adding future upgrades like accelerators, because the intention of this is a legacy board, I'm really leaning towards not even having any expansion sockets on it at all. Because bolting a accelerator onto a legacy board basically defeats the object of having a legacy board. People who want all the bells and whistles really would still have to get a H5 type board. the future of the H series board may well ultimately have a built on 030 CPU for example. but of course, every upgrade and change which is made, is taking us away from the original ST architecture.
For example as people know we are working on FPGA MMU, GLUE, SHIFTER etc, while these will remain highly compatible, and firmware updated at a later time, It is never going to be 100% compatible IMO. The only way to have that full compatibility is to use the original chipsets. With the H4/5 series heading towards faster CPUs and higher resolutions and better sound systems, you're basically left behind the original machines at that point which is where I think the legacy board will ultimately fill that gap.
It will also allow me to totally reorganise the layout of the motherboard. The current layout was to allow the main chips to be in close proximity to each other to allow a FPGA board to plug "over" the entire area replacing those chips. Unfortunately the routing in doing that was probably the worst it could ever be.
Ironically, the "silly style" board layout is probably more like how a optimise layout should be. Unfortunately those silly style boards ended up being the worst style of motherboards. Even so the 70789 types have always been a reasonably good layout. Mostly it is just the relocation of the GLUE. because it drives a lot of chip enables, it should really be on the right-hand side of the board so that multiple routes don't have to run right across the whole CPU bus.
As mentioned already, not having any 68K sockets will literally allow the board to be half the size.. Again this will be cheaper and very small to the point we could design a half size case similar to the STM to house it. With shipping getting more and more expensive as time goes on, I think trying to scale down the motherboard and the case is a good idea anyway. I *might* fit a single 68K socket vertically to at least allow some expansion capability, but again this is not really the design goal for the legacy board to be upgradable.
I don't honestly think this board will be more of much interest to people here. Basically because it is in many ways a huge step backwards from the H4. But with that series heading off into its own direction, it will essentially leave original machines "dead" in times soon to come with no legacy alternative.
It is becoming more and more apparent over time that original machines are becoming ever problematical and will also become incredibly expensive and time-consuming to fix. But if there was a legacy board available which was a straight swap and only need a few chips changing over to get it working, its keeping the spirit of the original machines alive for many more years to come.
EDIT:
I wonder also if it could be pushed onto a 2 layer board