The humongous Sandy Electronics thread...

[Dave]

Hi all,

Wow, it's been a crazy couple of weeks. I am so grateful to QUANTA for the advert, and to so many of you for ordering or pre-ordering - your faith and confidence will be rewarded.

Firstly, I have had to make a new batch of microdrive felt pads. I sold most of the first batch, and sent a lot more on to Rich at RWAP for those who would prefer to buy from a more established seller. I also sent him some QL on a Stick, new editions! I'm sure he'll have those listed soon after he gets them.

I have changed how I package the felt pads, and it will soon allow me to reduce the cost even further. I now attach them to a piece of tape, then mount them on the instruction sheet. I tri-fold the sheet, and it protects the pads well enough - even if they get completely crushed, they fully recover in about 20 minutes. You should see the lower pricing once the current batch finishes selling. Oooer, I hope that doesn't kill my sales while you wait for the new batch!

The battery adapters have finally been released from customs and should be here tomorrow! I am so excited to see them coming and hope to get the first orders shipped out this week.

I have almost finished the Minerva MK IIs, and am just waiting for two components to arrive.

UltimIDE is currently a box of components. I have almost finished the prototype wiring and have started the PCB design in earned. Currently, it supports two traditional IDE devices (more with an expander) plus two CF cards and a 1.8" or 2.5" SSD on-board. It's almost entirely SMD. It will have an option for 512K of on-board RAM, which will be selectable, in case you want an unexpanded QL for compatibility or (S)GC reasons.

UltraQ is taking shape quite nicely. It currently is specced for floppy, parallel, QIMI mouse, IDE, plus its own CPU and 4MB of RAM. We are hoping it will run most QLs at 15 MHz.It will not be quite as fast as a Gold Card, but it will have twice the memory. It will also use some GC-style tricks like shadowing video RAM. It will come with a custom version of Minerva and, hopefully, SMSQ/E on CF card.

Lets talk about ethernet for a moment. We would like to include ethernet: the hardware is easy. Fitting it on the card is a bit more difficult, but once we concede it'll be a double-deck interface, a lot of room opens up. Maybe if we include the hardware the driver will come. This presents us with a quandry. Including ethernet adds approximately £20 cost to the interface - we have discussed if it should be on every card, or an optional extra. If it is on every card, adoption will be wider and people will find uses for it. However, if it's not on every card, some cards will be cheaper, but the cards that do have it will cost a little more. If you, humble reader, could have a sensible discussion of the merits of both options, we'll read closely and take your thoughts into consideration.

Which leads to... USB. We have found a way to reasonably implement USB *hardware* but again, there are no drivers. USB takes up less space than ethernet, but it is complimentary. I personally am deeply interested in USB for a number of reasons; especially of the "build it and they will come" variety. Secondly, the hardware has the potential to solve a difficult problem for us, whilst also giving us more options with hardware. Let me explain:

We talked with Tony Firshman and Lau Reeves about open-licensing superHermes, and they have agreed in principle to do so. However, superHermes is a complex piece of kit: it isn't cheap to incorporate both in terms of component cost, and production cycle. What it brings us is both relatively great and relatively little... It allows us to replace the inbuilt keyboard with a PS/2 keyboard and it gives us better working serial. However, by modern standards, it still only gets us to where we should have been initially anyway. It corrected a fault and an in-hindsight omission. As an alternative, if we could get to where there is an HID driver for USB and a USB component on the QL, most HID (Human Interface Devices) - mice and keyboards plus all sorts of exotic input devices - could be made to work using default profiles. Imagine being able to use modern keyboards, cordless laser mice, etc. It fits well with the picoATX cased QL concept I am trying to achieve. It also means the hardware is open for people to write the other two main classes of generic USB driver: storage devices and "the other one" (I forget. It's late!)

The point is, if we can get a basic HID USB driver functional in the next few months, it can go out on the QL2, and it might even make it interesting for the UltraQ. With the ROM image in flash, the capabilities could be extended by a community effort to cooperate and develop the drivers in a way that will really free up the QL for all sorts of expansion.

I am well aware of the fine work Adrian Ives did with the USBWiz. We'd like to use that as a springboard, or at least an inspiration. That is why I am more than willing to sink resources into making prototype hardware for both ethernet and USB and get it into the hands of anyone who has an ability to develop for them. Remember, I'd like all of this work to be open-source and free. This is the only way we can inspire others to develop, and to secure the future of the platform.

We're also still scratching our heads at 4am on QL2. Literally. We know it'll incorporate the majority of the above, but faster, on wider buses, and with huge amounts of memory attached to a very fast CPU. It's just too early to say, but it looks like it's going to be insanely great.

I am looking for little side projects that I can use to raise more funds for QL2. For example, I was shocked at the price of good quality prototyping boards that are anything even remotely large or specialized. 20x30cm prototyping boards can cost well over £200! I welcome feedback and suggestions for little things like that, which I can do without too great an investment and that won't be inclined to soak up all my free time.

​Finally, thank you once again to everyone for your support both in encouragement here, on the QL-Users mailing list and in emails, and by buying from my store. I hope this can grow into a real and meaningful resource for the community.
​
​Cheers!​

[1024MAK]

Hi Dave

As a retro community member where cost is the second thing I look at, not the first, I like the idea of both ethernet and USB being a standard built in feature. I would be very interested in buying QL interfaces / system boards that include such features

Of course, for people where money is tight, paying for something that may or may not be useful to them would not go down so well...

I think the answer is to try to include a wide range of features as standard, while not letting the price go out of control.

Including ethernet and USB will help to "future-proof" the design at least a little

superHermes is great (I have one). However, it is a "fixer" board. Until we get USB keyboard support, we do need some kind of keyboard interface. We also need a sound generator. So maybe the superHermes part should just do these functions. We can then move the RS232 serial ports to either modern UART chip(s) or to a custom programmed chip (whatever type is best for the job).

Mark

[Derek_Stewart]

Hi Dave,

How about an LCD interface, so that a Portable QL or QL-Laptop could be made.

Derek

[thorsinclair]

Thanks for the update Dave, this is all good news for our beloved QLs!

Personally I like the idea of Ethernet and USB on the board first for accessing the net and second for modern, may custom made or branded peripherals. Would it be possible to integrate the enhanced graphic modes of the Aurora, connection with modern LCD monitors and may even better sound? Does this make sense?

[Dave]

Sound, enhanced graphics and so forth are all options for QL2, yes. It will likely need some heavyweight custom logic, and some of the 5v tolerant devices available these days far exceed the capabilities of what was available 15-20 years ago.

That said, there are two main limiting factors: the first is an essential resource one - Nasta's time. Nasta is extremely busy working his daily life, and for him this is an 'interesting side project.' The second is financial - if the UltimIDE and SuperQ and other little assortments only raise limited funds, those funds will limit what we can do.

Fortunately for everyone, Nasta has an interest in audio. We have actually had discussions about the load that's placed on the CPU by shuttling CD quality audio around, FIFO buffers that are too short for our needs, etc.

If people have any imaginative suggestions that could ease the hardware burden we're all ears

Right now, it's shaping up that UltimIDE will be released in about 3 months. The current thinking is that every card we do should have a SQB-style riser section for a daughterboard. This way, UltraQ can be inserted either as a standalone card, or as a daughter card to the UltimIDE, or vice versa. This will double the board edge area we have outside the QL case. It's a very tight squeeze to get floppy, IDE and parallel on a single expansion card just because that's a lot of connectors for just 220mm of edge connectors. We could make the PCB slightly longer, but that impacts how we can panelize it, which increases costs for PCB manufacture, and while it increases board area (and cost) it doesn't increase edge area that much. For example, making the board 3cm longer (99x134mm becomes 99x154mm) gives us 4cm extra board edge, but 40cm extra PCB to pay for. It also is a big no-no to just have a bunch of upward-facing connectors - over time they will fill with dust and crumbs and good God I don't know what that is Dilwyn's eating!

We note that Trump Card is longer anyway, 99mm x 160mm, giving 28cm, but that board would cost about 25% more to produce.

Since it is already a known requirement for UltraQ to be a full 2-deck design, the power regulator and low-height connectors will need to be on the lower deck, and bulky connectors (ethernet, etc) will be able to go on the upper deck.

Being a major dust-trap it will need a properly designed and 3D printed case. Once we conclude it needs a cover, it makes sense to design a standard one that can work with all three cards. I think I have this covered with a 5-panel design in ABS. The bottom panel will be standard across all designs. The top, side and end panels will be two lengths depending on single or double card. The end will be a piece of black coated white acrylic, which is laser cut to fit perfectly, and laser etched with the names or symbols of the connectors. As an attractive selling point, this will look and work better than a folded piece of metal, but cost less to produce.

*stomps feet* mailman just came - no package today.

Final thought - USB and ethernet. If either of these make it onto the QL2, they will *absolutely* also be released as a standard expansion card for the QL community at large, and the plans and driver would be open, so people can tinker, improve and share. We'd need to put together a card for the beta testers to develop the drivers and software anyway, so.... why not?

[dilwyn]

It also is a big no-no to just have a bunch of upward-facing connectors - over time they will fill with dust and crumbs and good God I don't know what that is Dilwyn's eating!

Excuse me!!!

Neither do I actually...

[1024MAK]

If people have any imaginative suggestions that could ease the hardware burden we're all ears

Well, depending on memory/bus bandwidth and how many unused cycles there are from the CPU (or how many cycles you want to hold/pause the CPU), the most logical hardware device for moving data is a DMA type device. If done correctly, a flexible DMA system can move data to/from RAM to/from mass storage, or the audio system or the video system. If you make the DMA from custom logic, you could even get it to do some data processing, like the <<< blitter >>> does on Atari ST or Commodore Amiga machines. Of course, this will require more development time

Mark

[tofro]

Dave,
Second that. If you're thinking about USB in earnest, DMA is definitely the way to go, otherwise background CPU load from the driver will grow way too high. USB as such is a nuisance, but without DMA it's a nightmare.....
But it might need some entirely new concepts for OS and drivers. QDOS is definitely not built for that.

Tobias

[Dave]

DMA for the win.

We've discussed that. It seems more likely than not. For HID purposes, DMA isn't necessary and even without DMA would work infinitely better than a USB adaptor sat on the end of a 9600 baud cranky serial port. However, it would place such limitations on future development of drivers that would require more bandwidth that we think DMA would be helpful also.

That said, the way the scheduler works in QDOS is very efficient. It does a creditable job on a 7.5 MHz CPU, so you can imagine it does a much better job on a 50 MHz CPU as anyone with a QX0 can attest.

[tofro]

The main point is that nearly every USB host controller you can buy will simply assume it can use DMA. Also agree that it's definitely overkill to introduce DMA into an architecture to simply connect a keyboard - But that's PC architecture today

Most host controllers have an API that requires you to give them a chain of buffers and they use DMA to walk along that chain as they go. The OS just collects what it's interested in from the filled buffer chains and hands back empty (or re-filled) buffers for more processing. (Very much like the QDOS slave block mechanism, but with processing by two busmasters)

USB would also (ideally) require an (ideally) flash-based device/vendor database to determine device classes and driver mappings (which need to be dynamic depending on the device plugged in).

And it would also require a huge re-write of how QDOSSMSQ works today.

Tobias
(scared )