Sidequest: VGA scan doubler

[Smonson]

Hi Smonson,

Cool project. How are you doing?

Hi @ijor, long time no see. Thank you! I'm doing quite well, just a few health concerns these days. My warranty ran out in 2020 it seems. How are you?

Don't despair, it is very likely that these chips will be available next year, but almost certainly not before.

I understand Intel may need to prioritise their military customers and do what they can to avoid SLA penalties for their industry contracts. It's annoying, but understandable, that we must just wait and see.

You can infer HSYNC from DE and make this a solderless plug-in upgrade. It might be difficult in those few cases that alter SYNC. But these programs that alter SYNC will fail to display correctly in many, perhaps most, modern displays anyway.

It still needs an outbound connection to HSYNC to inject the doubled HSYNC into the video-out circuitry though - so no matter what, it will need at least one soldered wire for HSYNC, and another for BLANK (also an output). I also need one to detect VSYNC in order to generate the vertical part of the blanking. Considering that soldering these wires is already necessary, the cut track and the extra wire for HSYNC (inbound) is only a small amount of extra work.

As HSYNC from the GLUE is - presumably - always TTL logic (you'd know better than me), it may have been possible to leave the HSYNC track uncut on the PCB and have both the GLUE and this mod drive the line at alternate times. But I didn't want to take the risk of damaging someone's GLUE. A cut track is easily repaired, not so much with the GLUE.

I agree with you that most people would like a plug-and-play mod, if that were possible.

[ijor]

Hi @ijor, long time no see. Thank you! I'm doing quite well, just a few health concerns these days. My warranty ran out in 2020 it seems. How are you?

I'm fine, thanks. Hope you are better now.

It still needs an outbound connection to HSYNC to inject the doubled HSYNC into the video-out circuitry though -
...
As HSYNC from the GLUE is - presumably - always TTL logic (you'd know better than me), it may have been possible to leave the HSYNC track uncut on the PCB and have both the GLUE and this mod drive the line at alternate times.

Ah. I missed that you wanted to use the original ST video connector. I thought you were still working on using a separate VGA output.

GLUE is CMOS, not TTL. But IMHO, requiring cutting a trace on the motherboard is not a good idea. You have escalated the complication of installing this upgrade at least one order of magnitude. I think very few people would dare to cut a trace on the motherboard. Of course that advanced users, like most people that actively post in this forum would have no problem. But this is not a representation of the typical user.

I agree with you that most people would like a plug-and-play mod, if that were possible.

Well, a solderless upgrade it is definitely possible if you use a separate VGA connector. May be you should consider leaving the original video connector output as an option for advanced users. This way you can get a solderless upgrade with a separate VGA connector, but still optionally inject the video through the original ST video circuit and connector.

[Smonson]

Ah. I missed that you wanted to use the original ST video connector. I thought you were still working on using a separate VGA output.

My first prototypes used a new connector, but it's difficult to physically mount a VGA socket on the ST case without cutting, so I changed plans. Many people use an ST-to-VGA cable already, so it's a convenient setup. It also makes the mod about half as complicated and therefore twice as fast to solder.

GLUE is CMOS, not TTL

It's good that I didn't make an assumption there.

Well, a solderless upgrade it is definitely possible if you use a separate VGA connector. May be you should consider leaving the original video connector output as an option for advanced users. This way you can get a solderless upgrade with a separate VGA connector, but still optionally inject the video through the original ST video circuit and connector.

It's a reasonable approach, and having a new connector raises the possibility of a better picture quality (the ST's video circuitry is not the cleanest). It may be better to have 2 versions of the mod rather than one expensive one that does both: the primary consideration is whether it can all be made to fit into the metal box around the shifter without making it expensive. I'm experimenting with 4-layer PBCs soon, which might overcome it, we'll see.

I'm not going to be too worried if a lot of people decide not to install this due to the difficulty of installing it. People who aren't confident to cut that track probably shouldn't be installing the flying wires or removing resistors either. It's a trade-off, but there is also the option of using a more expensive external scandoubler like OSSC, and now Lotharek's Medusa.

[ijor]

It's a reasonable approach, and having a new connector raises the possibility of a better picture quality (the ST's video circuitry is not the cleanest). It may be better to have 2 versions of the mod rather than one expensive one that does both: the primary consideration is whether it can all be made to fit into the metal box around the shifter without making it expensive. I'm experimenting with 4-layer PBCs soon, which might overcome it, we'll see.

Sound like a good plan. Good luck!

[Smonson]

STILL no MAX-Vs available. Intel is obviously not prioritising these parts and I've got to move on. It has been three years! I have ordered a Lattice FPGA dev board and have started looking into how the open-source tools work. Being FPGAs, they cost more and have more pins to solder, but that is offset by the fact that I can leave the two SRAMs off.

[troed]

It's ...

...


...


... ALIVE!!!

[Icky]

STILL no MAX-Vs available. Intel is obviously not prioritising these parts and I've got to move on. It has been three years! I have ordered a Lattice FPGA dev board and have started looking into how the open-source tools work. Being FPGAs, they cost more and have more pins to solder, but that is offset by the fact that I can leave the two SRAMs off.

I feel your frustration. Most of my FPGA work has faltered. I have been lucky to grab some QMTech Cyclone V boards to continue developing in some guise for a while. However building boards is impossible due to chip availability.

Will be good to see progress here though.

[exxos]

STILL no MAX-Vs available. Intel is obviously not prioritising these parts and I've got to move on. It has been three years! I have ordered a Lattice FPGA dev board and have started looking into how the open-source tools work. Being FPGAs, they cost more and have more pins to solder, but that is offset by the fact that I can leave the two SRAMs off.

Glad to hear the project has not been abandoned :thumbup:

What lattice chips are you planning on using then ?

[Smonson]

It's ...
... ALIVE!!!

* no guarantees

I feel your frustration. Most of my FPGA work has faltered. I have been lucky to grab some QMTech Cyclone V boards to continue developing in some guise for a while. However building boards is impossible due to chip availability.

It's been a rough three years for the hobby, that's for sure! I've been holding onto my last few Max-Vs like they were made of gold. I wonder if the shortage will actually end, or Intel will just go the way of Zilog. All empires fall eventually - and I've read some fairly pessimistic news articles about the company's internal structure.

Glad to hear the project has not been abandoned :thumbup:
What lattice chips are you planning on using then ?

Hello Exxos! Well, I tend to dive into these things the way a blind man goes swimming, so I bought an iCE40HX1K dev board after about 2 minutes of googling (from Olimex in Bulgaria, hopefully it doesn't take too long to get here). I've also been playing around with the verilog compiler as I mentioned and it looks like my viable options are:

• iCE40HX1K (cheapest option, on a VQ100 package)
• iCE40UP5K (smallest hand-solderable package, an SG48 package)
• iCE40LP1K (QN84 package - soldering it might be a crapshoot)

The scandoubler was made for a CPLD (I used a 240-LE sized Max V) so it will easily fit into even the smallest modern FPGA. In the open-source compiler I've been playing around with, it only uses about 350 logic cells out of 1,200 on the iCE40 range. That's with picking pin assignments at random. The onboard SRAMs are enough to fit the scanline buffers (32kbits needed, I will have 64kbits) so will actually have enough room to add more features if I can think of any to add! You can use the RAM as ROM, so a small boot-up image wouldn't be out of the question.

The software is all CLI-driven, which suits me. Better than the crazy 12-gigabyte Altera IDE. The downside is that being open-source, it's basic on features.

If this ends up being a colossal failure, at least I'll have learned stuff.

[exxos]

All empires fall eventually - and I've read some fairly pessimistic news articles about the company's internal structure.

That is my continual concern as well. A lot of us get near the finishing line just to have the rug pulled out from under us as we cannot get the chips anymore :cry:

Hello Exxos! Well, I tend to dive into these things the way a blind man goes swimming, so I bought an iCE40HX1K dev board after about 2 minutes of googling (from Olimex in Bulgaria, hopefully it doesn't take too long to get here). I've also been playing around with the verilog compiler as I mentioned and it looks like my viable options are:
....
If this ends up being a colossal failure, at least I'll have learned stuff.

I think terriblefire mentioned using the ice chips at one point I think. Please keep us updated on how you get on with all the stuff. I think a lot of people definitely need alternative options for FPGA stuff these days.