Video options for the HEC computer
Vložit
- čas přidán 27. 10. 2022
- I'm sorry for the duplicate video! (and a video that doesn't match its thumbnail image)
This video is about what video options I have for the HEC. Video is arguably the most important aspect of the 'personality' of a retro computer and the range of selection is WILD. This video only covers about half of the options.
I have been working on quite a few projects off-camera recently (for example, the 'Frankenstein' AST Premium 286 has been fully fixed). As a result, I don't upload videos as frequently as usual. Things have been moving incredibly fast. I now have a high tendency to use the V9958 as the VDP since the supply has mysteriously gone up, and the compatibility with TMS9918 makes it a good choice for emulating/virtualizing several famous platforms. - Věda a technologie
Thank you for choosing easy-to-find chips for your computer. You put a lot of thought into your project! I hope I will buy an electronics design that YOU invented! Cheers!
Daydreaming about a super console or home computer is one thing; practicality is another. I kind of learned that in the hard way (like, most things I learned). The first priority is getting the HEC to the market as soon as possible and as cheaply as possible. (talking like an evil businessman haha) Because unlike the 8-bit guy I don't have three years to wait or 100k dollars to spend.
It's great to see someone building a homebrew Z80 machine that has graphics at all; so much the action in the Zilog realm is CP/M running with terminal-style I/O.
The problem I guess I have personally with FPGA solutions like VERA is, honestly, it kind of feels like you might as well just do the whole computer inside the FPGA and cut out the middle man. But as you handily point out, if you want colorful accelerated graphics with an old-school 8-bit CPU it's pretty much a choice between that and the various options which all suffer from potential availability issues. The V9958 seems like about as good a choice as any in that category, and definitely optimal if you want to be able to emulate MSX. It would certainly be an interesting challenge to try to virtualize the *interface* of a TMS9918 and translate the operations for it into, say, VERA function calls, but the overhead would probably be immense. (Assuming there's complete enough functionality overlap to do it at all while retaining meaningful hardware acceleration.)
For my own perpetually morphing homebrew computer ambitions I've opted to embrace a *very* retro aesthetic and am building my own memory-mapped video hardware; the current design substitutes an Atmega8 CPU for a CRTC chip like the 6845 for sync and address generation so I don't rely on extinct parts, but otherwise *could* be implemented with straight TTL. (I also use GALs to cut the part count down a lot, though. Which does mean I'm technically using programmable logic... but only barely.) It can do both linear bitmap and "tile" modes (IE, character mode with user defined fonts), but no sprites, no color, and a maximum resolution around 512x384 (interlaced) monochrome. Basically something you could have built for real in the late 1970's.
But, yeah, my goal is specifically to emulate stuff from the 1970's, not the late 80's and onward, so I'd fully agree that taking a discrete approach doesn't seem like an option here. Tacking on sprites/deep color/etc. would rapidly turn my approach into a multi-board set in a card cage unless compressed into CPLDs/FPGAs. And doing without those things and settling for large dumb framebuffers on a VGA chip likewise seems unlikely to scale. For a Z80 there's a huge difference between shoving around 12-24K of framebuffer verses hundreds of K.
You have three other vintage options not mentioned - PD7220, Intel D72020, and the EF9366. The last one can be implemented with the least logic. You could also cheat and use a Parallax Propeller VGA or ESP32 VGA solution, but I like that your system uses vintage parts. Keep up the interesting videos!
Wait for the next episode where I mention them! These are called 'smart framebuffers', yet they also have their own problems: mainly due to the fact that they don't provide very 'useful' hardware drawing for games (more suitable for CAD software, GUI, etc.) 'Cheatgraphics' is always an option, but I kind of dismissed that idea when I say I don't want to use FPGA. I just don't want a modern chip used in the first 'baseline' implementation. However, HEC being HEC, you can always add them and run existing software on them, as long as you provide the driver!
I noticed that V9958 chips are being sold in kits for the RC2014 single board computer.
I bet they are! The RC2014 is becoming the new S-100 with everything you can think of put on it!
Andy !. I've been meaning to comment. Please don't make the VGA mistake. Try for progressive 224/240p RGB if at all possible. But so long as it's not VGA the computer will be ok even if composite. Also how's the table tennis, nice table I saw hiding there.
I know VGA is a mistake, but VGA is the only analog format that will give you crisp image without expensive hardware (like an RGB monitor). My concern for composite is that it is bad at high-resolution image (like 640x400 or 640x480), which I consider a 'computer' would have. However, I finally chose the V9958, which outputs 256x192 because it just has a lot of advantages. HEC being HEC, VGA with powerful blitter or with stupidly fast Z80 is always on the table for future project, though. About the table tennis, finally someone pointed it out! I play table tennis once in a while, but currently it is mostly used as a workbench. It has the same color-scheme as the Amiga workbench!
I'd been thinking about adding [8bit] ISA slots to my z80 computer... Then I could start with CGA, then move up to VGA, and possibly even VGA with simple onboard graphics-processing (surely early VGA cards eventually had things like sprites?).
That said, I guess your "most advanced" z80 design could go in many directions, as you've detailed here. Do you want to stick with parts available at the time to "show-off" what could've been, or do you want to show off the things it could've enabled a lowly z80 to do,, allowing it a longer life as a viable CPU in later machined, given developments in other fields (such as GPUs)?
...
You're in a difficult spot, here, because the very clever idea of virtualization will inherently slow things down, turning each single I/O read/write into a many-instruction endeavor, even if the attached hardware is identical. But, even slower if you intend, say, to emulate one piece of hardware on another. (e.g. emulating sprites in a framebuffer).
Personally... I think your board-design would best show the capabilities of virtualization if you *don't* limit it to specific hardware, like a specific video chipset. That's the whole point, isn't it? Make a CGA "driver", attach a CGA card. Make a VGA "driver", attach a VGA card...
Make a simple square-wave oscillator for a speaker, and its driver. Then a SID, then maybe a soundblaster with a wavetable...
Even before all that, maybe start smaller and just have a serial "dumb" terminal attached. Most CP/M programs I've seen aren't graphical at all.
You could demonstrate virtualization actually functioning long before implementing graphics! I wanna see that!
I have considered ISA-based solutions SERIOUSLY. (and ISA is currently the forward-running bus solution candidate for my Dragonfly and the 16/32 bit CISC processor project) However, when you look at the actual chips you will be disappointed by the fact that most 8-bit ISA graphics chips/cards don't have hardware graphics processing! The best you can have is TVGA8900D or RTG3105iEH which are very dumb framebuffer chips. VGA with graphics processing didn't come until VL and even PCI era.
In terms of slowing things down, I have said in many occasions that I plan to overclock the Z80 to 24MHz, mainly to compete with the X16 in raw performance. it will provide a 6-fold performance boost over traditional z80-based machines. As long as your program is not filled with I/O operations, it will run decently fast. (But emulation is another story! Heavy optimization will be needed!)
Another thing to notice is that graphics calls doesn't need to be individual reads and writes. The call can be 'draw a line', 'move a sprite' or 'fill the screen with a certain color', which the OS can perform with a series of reads and writes. This will speed thing up significantly. (however, if your graphics solution is an add-in card, the 'triple blind' system may eat away this benefit, I will talk about that in a later video)
Demonstration without graphics or even keyboard is definitely possible. In fact the graphics and keyboard circuit wasn't even in the first design -the 8251 USART is. My original plan call for a terminal-only computer just to demonstrate the virtualization, but then I realized I need something to show to my viewers, so I added the TMS9918 chip and an SN76489 and a keyboard chift register.
@@andyhu9542 I guess the question I have is what is your main goal with this?
I came across your z80 project from the Hackaday Blog, which linked your video about virtualization, as my introduction to your project. I think many others found out about your project similarly.
It's a great idea which in and of itself could be the basis of an entire, albeit short, build series.
But you're also, it seems, trying to implement an entire computer, based on it, from scratch, without yet having even demonstrated a proof-of-concept or prototype of the underlying concept.
As you've said, it seems no one has done it before... Don't you think it's a good idea to try it before building an entire system based on it?
Besides, if it's as unique an idea as it seems, then don't you want to highlight that part of the accomplishment , before burying it so deep within a complex system?
If I understand correctly, basically all you need is the NMI circuitry of your design, a z80 , a ROM, a UART, and a tiny bit of software to prove the concept... and all those things you'll want in the bigger system, later anyhow.
It's just a thought...
@@andyhu9542 Looks like we cross-posted, heh!
And, apologies... this is YOUR project, please don't let the likes of me tell you how to go about it. If you do, you'll likely find, as I have numerous times, conflicting advice from countless armchair directors.
My own interest, frankly, is probably obscure in comparison to yours and most others', as I have no nostalgia for vintage z80 systems. I look at the z80 as a simple-enough processor that I can actually understand what it does at low levels, as a way of slowly building my understanding about newer concepts like memory-paging and virtualization. So, for me, compatibility with MSX, CP/M, or games from different systems is merely a way of demonstrating those concepts, as opposed to a motivation. But, I understand that's probably not the same for many others, who otherwise might want to play an MSX game on a TRS-80.
@@ericwazhung We may share more in common then you think. I look at the Z80 (and the 6502) exactly as you do. I consider this project as a 'next step' to 8-bit computers like the Commander X16 (not to say its any better, but more like 'the next thing you may want to learn after you have studied the X16 inside out'). This project was a no-video, no-sound, no-keyboard, serial-terminal-only single board computer. But I also need views (and I'm not going to avoid saying that). Most people on CZcams who watch contents about the Z80 because of nostalgia, and I deliver it. I'm not motivated by it, either. In some sense i'm motivated by the view number. I am probably more superficial than you, but I'm still amazed by how close we are when it comes to learning the Z80. I guess this is the magic of CZcams, isn't it? You can get to know your mirrored image (in one aspect) who lives maybe thousands of kilometers away!