DRAM

[exxos]

I have been somewhat puzzling over memory solutions for a long time. But for this legacy board I'm leaning towards just simply soldering on the same RAM as I use in my 4MB STFM kits.

I have somewhat been reluctant to directly solder RAM onto the PCB, after knowing how much trouble RAM has always been on the original machines, I really wanted this to be a removable solution, so memory could easily be swapped, hence using a simm.

Part of the problem is sourcing RAM. Mouser used to sell the chips at about £10 each (2 needed) it seems that is discontinued now so the only source is again China Though I think I can build a H4 RAM tester board to test out the RAM before soldering it onto the motherboards, as there would be nothing worse than soldering the RAM down only to find it is faulty.

As to why the RAM fails originally, is somewhat unknown, but from what I have seen over the years, and indeed in Falcons, is that faulty power supplies spiking up to several volts like I have seen in the past could may well be the cause why the RAM is failing. Of course if people use a new power supply such as the ones I sell, then it should not be a problem, but if people are reusing original power supplies, it could may well kill the on-board RAM.

To get around this problem, at least in part, I could place a transorb on the motherboard which should help clamp down on the spikes and protect the RAM ( and indeed all the other chips as well) . Other options would be to literally crowbar protect the motherboard, in that if spikes above six volts are seen it will short out and cause the power supply to shut down. Or possibly some ultrafast voltage supervisor which will latch on a error LED to indicate the power supply is bad. But thinking about it, this is the sort of thing which should probably be included on the H4 boards as pretty much any power supply could be bolted onto the thing and potentially cause damage.

I was also thinking originally of doing a small RAM circuit board similar to the Acorn ones. So that if a RAM module goes faulty it can be easily changed.But the problem is this turns into another project, it would need more soldering, more connectors, I would have to physically solder the boards to test them out in the first place, so I don't think in the case of the legacy board it is a good idea.

[stephen_usher]

Is there any reason not to use current SRAM chips + voltage level conversion even on the recreation?

If on a daughter board, then the components, as they are SMD, should be able to be installed on the PCBs buy the fabricator rather than you having to solder them, leaving you to add whatever plug/socket you'd like to use to connect to the motherboard. Using current chips, which are easy to come by, should save a lot of hassle, especially if you got the PCB manufacturer to source and install them. i.e.less soldering for you to do.

Potentially you could use the same RAM system on the development board system too.

What do you think?

[Tomswork]

I remember the were dram or sram that used a socket rectangular plcc style. 16 mb.

Edit
found one 64by 4 plcc 18

Also we use these on some of are stuff

MT4C16270

Tom

[exxos]

Buying dram isn't a problem as such, I just need to test it. But because they are obsolete, they can't be done at factory. Even so jlc only do generic parts.

3.3v stuff is more plentiful, but adding buffers will ramp up costs due to all the buffers, and if I end up soldering them., well if rather not

In terms of assembly, I've got the same problem with the video DAC. But I have plans to speed up assembly of such stuff.

[Tomswork]

Screenshot_20200728-170148_Drive.jpg (167.04 KiB) Viewed 7333 times

This stuff may work I part is a 520st

[stephen_usher]

3.3v stuff is more plentiful, but adding buffers will ramp up costs due to all the buffers, and if I end up soldering them., well if rather not

Surely if all the chips are current SMD components then the fabricator would be able to assemble everything and you'd only have to deal with a through-the-hole sockets?

As for the cost of the chips, given the total cost per board, surely they'll be a relatively marginal cost of a few pounds at most?

Actually, I've just had a thought. How feasible would it be for the RAM board to use an edge connector and have the socket on the motherboard, with screw posts to secure the board? That should make it fully SMD and able to be totally built by the fabricator, meaning no soldering for you.

Alternatively, as you're designing this system from the ground up you can define the (S)RAM card interface, you could put all the buffers and voltage leveling stuff on the main board and make the RAM card even more simple. You can reuse the same interface on your towering inferno development boards, which you were saying will possibly be partly 3.3v anyway.

Anyway, it just feels wrong to base a new implementation on already obsolete and increasingly rare RAM chips when in-production ones are available at relatively low prices.

[exxos]

Like I said, jlc only do generic stuff mostly ,so what chips are used doesn't matter as they just won't get done.

As for obsolete ram, everything on the board is pretty much obsolete anyway

I have talked about plug in modules, but its more work, more bits, there's £100+ in generic bits, multiply that by 50 and costs soon ramp up. Adding buffers just adds more chips, more expense, more work. I would have to design and test a plugin module which I don't have time to do either. I want to get parts count down and reduce costs as much as possible with this board. I'm going to be pushed for space as well.

Likely I will try and fit the simm on the thing as a backup option and solder the 2 drams to the board.

[stephen_usher]

Fair enough. I didn’t realise that your fabricator only did generic components. I was mostly thinking how to decrease the amount of soldering you’d have to do to save you from getting hand pain.