Extended expansion connector...

[Peter]

Just a personal thought... it seems to me that writing, reading, considering, changing and re-reading those specs takes quite some time. I have to admit that my concentration is beginning to fall.

Wouldn't it be more effective to build a working mainboard prototype first? I see three advantages:

1. Ideas which appear during practical realization can be taken into account.
2. The audience knows there is a realistic chance for series production, and therefore looks into the connector specs much closer and with full attention, resulting in better feedback.
3. Issues which became irrelevant by lack of development time can simply be dropped.

The connector pinout could still be changed in the re-design after the first prototype.

I'm not at all trying to say the connector pinout is irrelevant here in the forum. I just think the mainboard itself deserves higher priority, and it might be more effective to do it first.

All the best, Peter

[Nasta]

That's a very good point and in fact one of the reasons for this entire thread.

QL people are becoming few and far between, so we've made an attempt to dredge out some old data from times when there was much more involvement in such projects.

Why specifically this discussion?
Actually if has a lot to with decisions about what to put onto the motherboard and what not - and where on the motherboard to put things that end up being put on it.
The 'problem' is that the motherboard sees everything, and in fact needs nearly any kind of hardware - on one hand, some sort of boot ROM, basic peripherals etc are all 8-bit, even today. Simpler to route buses, source one component instead of 2 or 4, and speed is generally not an issue. On the other hand, every design is a hole of sorts that the designer digs himself into - so let's do some thinking an at least provide a rope ladder to get out of the hole, if not an elevator, should the need arise. Some semblance of future-proofing can be considered, especially in the 'thinking about it' stage of design, and it's all the more logical if it can be had with minimal extra effort. Hence the wide bus mode ideas, which are based on the existing bus, simply because one does not need then to go through all those decisions again, and, one does not need to support them immediately or indeed in full.

However, since everything can't necessarily be designed 'in one go', or at least not supported from the get-go, some dose of compatibility with the old should be there, so that some semblance of a working system can be built, even though not optimal - at least at first. So, even the working prototype needs a J1 at least at first. One other thing to consider in this light is that certain old style peripherals apparently still have a market, but it's really not that good an idea to just clone more of the old versions, with 10 or 20 years of hindsight into problems and issues they have in use. These are under quick development (or rather re-development, which makes design much faster) and it makes sense for them to have some forward compatibility built-in (or at least forethought, given that there is a prospect for a new motherboard). These are likely to be used in the development of the new motherboard.

The basic idea is to think up something that will satisfy the above goals. Also, there are things that will definitely NOT go onto a formal expansion connector - instead they are connected to the local CPU bus. If that's through some connector or not, remains the business of the motherboard designer.

So, this discussion is actually based on a general idea of a motherboard, which has the basics: CPU, RAM, I/O (Note that something like video actually looks like RAM in this context, though there could be associated I/O). The idea behind this topic is to have a good discussion where to draw lines between various domains - local to the CPU or not, before the _peripheral_ expansion connector or after, etc. So far the discussion has actually been extremely helpful, at least I see it that way

[Dave]

Before reading your post, I had a good resolve that just dropping physical compatibility would be a good way to go. However, you make some really good arguments.

Is there anything that we lose by retaining backwards compatibility?

I think the one thing we gain by having a new layout is we lose dependence to supply specific power to the expansion connector. We could then have a rational grounding scheme, and provision of regulated 12v, 5v and 3v3. Not needing regulator space on expansions cuts costs and board space.

I personally find the DIN41612 connector very bulky. It's not electrically ideal to have signals spread so widely. I would love to see a smaller more dense connector in a different orientation, like the PCIe connectors on ATX PCBs, but with a QL-based pinout. I see people using one of these inside a black box QL, unexpanded, but if you want to add anything after it already comes with almost every expansion known to man, well.... Maybe a PC case is for you Also, 164 pins, high density. Finally, the Q40 and Q60 did alright with ISA slots. And you can get 3 or 4 of them on a very short bus. No more backplanes.

So, yes, mixed feelings here. I see both paths have a lot of points to recommend them. However, if familiarity is the best we can offer, are we really thinking about the next 30 years?

What do you think? This turns into a broader philosophical discussion, so non-techheads should feel welcome to post in this thread too - it is you guys that we're trying to get this right for!

[Peter]

I'd keep things as simple as possible, so the major goal can be reached at all.

A mainboard with incompatible or less than optimal expansion slot seems better to me than a perfectly defined expansion slot with an unfinished mainboard.

E.g. when defining the expansion connector for the Q68, I first tried to get enough pins so most of the QL bus could be attached (via external buffers). But when it became clear I'd have to use a much larger FPGA package, breaking the form factor of the design and adding many work hours, I simply accepted the limitations. (Heavy shortage of address lines in this case.)

[dilwyn]

I'd keep things as simple as possible, so the major goal can be reached at all.

A mainboard with incompatible or less than optimal expansion slot seems better to me than a perfectly defined expansion slot with an unfinished mainboard.

E.g. when defining the expansion connector for the Q68, I first tried to get enough pins so most of the QL bus could be attached (via external buffers). But when it became clear I'd have to use a much larger FPGA package, breaking the form factor of the design and adding many work hours, I simply accepted the limitations. (Heavy shortage of address lines in this case.)

I tend to agree with the first sentence. While it is likely there will only be one chance at this motherboard, we have to accept that once done it cannot be undone.
I'm not a "hardware" person so potential problems with this or that might not be obvious to me. For example, if retaining compatibility increases complexity and cost we have to think twice.

So if the motherboard includes everything - loosely defined - I include expanded ram, floppy interface, sd card and/or qubide, vga, possibly Ethernet) I can't see plugging in older cards would make much difference unless you intend to plug in such obscure hardware as a Q+4, Maurice Computers sound card, Q-Control boad etc etc - compatibility would only be for (a) plugging in obscure devices (b) plugging in old interfaces to test them. QEPIII Eprom programmer would probably be the only valid old expansion device.

[1024MAK]

I think the best solution depends on a clear objective.
In the end, it comes down to physically fitting a new CPU card into a case (be that a PC type case or a QL case) and cost.
So if new fast/wide bus expansion cards are likely to be produced, it is logical to design a new expansion system with new pin-out and maybe a new (to the QL world) connector.
If compatibility with existing expansion cards is part of the objective, I think it is better to have a legacy expansion connector that is QL compatable. This does not prevent currently unused/rarely used signals from being redefined for use with cards that can make use of new features. You may want this so that some new cards can be used on existing QL hardware, as well as with the new main board where they can sense that new features are present, and so use these to operate in a faster/wide bus mode.

If possible, a new main board will have at least two new standard expansion connectors plus a QL compatable expansion connector (as described above).

Of course, if either space or cost are a factor, something must be dropped...

The other consideration, how big would the market be for expansion cards?

Any that use features not present on a normal QL, could only be sold to owners of a new QL main board. Not every owner will buy one, so this could be a small market.

Mark

[Dave]

My proposal[1], simply put, is to drop the traditional J1 DIN connector on the left edge of the board.

Then, imagining an ATX motherboard with the back closest to us, and the PCIe connectors on the front right side.

I see a row of three connectors - PCIe x16 physical connectors, but wired differently. They have 164 positions. My idea is to implement both pre-decoded and full expansion signals to each of the connectors. An interface, when inserted, can select to use either the pre-decoded level for simplicity, or the full level for high speed.

This way, every 4th pin can be a ground pin, and every card will be able to draw enough power by spacing GND, 12v, 5v and 3v3 liberally along the connector.

Other advantages: PCIe cards are standard templates in most PCB design programs, and with 164 pins it's certainly expandable. They're cheap, and don't require an expensive DIN connector. $10 cheaper per board.

Disadvantage: I'd need to make new case bottoms because the connectors would require a physical size 12mm wider than the current case bottom. A simple vacuum formed black bottom would be easy to do, and laser cutting the connector holes would be accurate, cheap and easy, plus highly customisable.

[1] I say proposal, but it's more a devil's advocate statement to gauge peoples' reactions. The responses are informative. We have one chance to get this right. It's not like we're looking to make a range of computers - I'm constantly being warned not to overestimate demand.

[Mr_Navigator]

Dave, for my two cents worth, I would say go for new design, my main reason for not trying to keep compatibility is this.

The BBQL

A majority of expansion peripherals are for memory and disk interface
In order to be fully compatible you would need to check all available peripherals, just because the connector is specced, doesn't mean it was followed exactly. You only need one person to say your interface doesn't work with my addon to cause some doubt. Where as a new design there is no doubt.

Your New Thing
Will already have sufficient memory, easily replaced memory, expanded memory
Will already have a disk interface
Will be designed to work as is

Those who purchase your new device will make those peripherals redundant to them, but not others and they could recoup some of their hard earned cash selling said peripherals on SellMyRetro or eBay.

[Peter]

I wouldn't do anything which makes use inside the QL case difficult, like the proposed 164-pin connector. Fitting the QL case seems the main point for this mainboard.

(Outside the QL case, the board would be compared to Q60 and Q68.)

[Dave]

If PCIe connectors are fitted, the QL case can still accommodate. However, they don't have to be fitted - anyone who can wield a soldering iron can install them.

However, it does get the whole backplane issue under control.

A design goal is to have it fit inside a QL case if someone should want to do so, but that's just a board dimensions restriction. It's not a content restriction as long as that content can fit inside the envelope.

Maybe it's time for a poll?