Mike's PCjr Page

Trixter wrote:

GHiero wrote:I wish J. Alex posted PCB layouts so other people could have boards made from OSH Park.

He shared the design; I believe this is the link: https://oshpark.com/shared_projects/9MPsJlrJ
It is on my to-do list to build once I build some smaller projects first to get my skills up, something I will be doing tomorrow afternoon.

Was there any particular reason given why the green is off on the bit-c128 board?

In my discussions with the designer, he designed his board by measuring the post-conversion voltage values of his actual Commodore 1084S-D monitor. It's my theory that his monitor was faulty/damaged/non-calibrated when he did his measurements.

I am not intentionally trying to paint him in a negative light, only reporting my experience. I tested two boards, including one certified by him as 100% functional before he sent it to me, and got identical wrong green levels with multiple capture devices and monitors. I even took the entire setup, Gonbes and all, to a vintage computer gathering so that its output could be compared to a real C128 hooked up to a 1084S-D, and no amount of Gonbes POT adjustment could make it look right.

Thanks for the schematic link. It would make sense that his monitor is off in some way. The logical function of the circuit is to make the green gun fire at half the strength of the red gun when blue and intensity is off and red and green are on. He really should not base his design on the idiosyncratic behavior of his own monitor, but as I know only too well, it's all about him. Not improved.

I feel like Alice.. rabbit holes everywhere... I really shouldn't get distracted from other things including my 9 to 5, however this kept eating at me. So I'm looking for a relative sanity check to see if I'm on track (I know I'm not sane - "relative")

I've been thinking about a few configurations:
1) No output, USB capture only
2) Simple VGA output - been done and doesn't address the scan rate issue
3) VGA conversion with on-board vertical scaling. Could eliminate the need for a gonbes whatever...
4) Why stop at #3, conversion to HDMI w/ vertical scaling. There is a new gonbes that does this already

I hate the idea of going in and out of the analog domain to process a signal that starts off purely digital. In the hardware I am considering, I've added a clock synth and VCXO that should allow perfect capturing of any digital RGBI signal. For cost, I'm still leaning towards option #1. It is also the simplest. Just the active video area and only packed pixels will be sent across USB. Once it's in the PC, you can do a tremendous amount of stuff, like:

a) Unlimited screen captures
b) Record the raw stream
c) Apply any color correction you want w/o mucking with trim pots
d) Apply any nifty scalers like HQnX or nxBG
e) Display your host machine on whatever video technology you like

Having a simple dongle that allows my laptop or Android enabled tablet to become a MDA, HGA, CGA, or EGA monitor is pretty sweet thought to me. So I took a few hours and roughed out what the app would look like with some sample static screens. Just imagine the 'Demo' menu screen shots would be moving captures. The screen-shot demos are actual native resolution captures from the real machines. The download (Win32) is very large thanks to the Qt 5.7 framework - apologies, but can be downloaded here (temporary link):

https://www.retrotronics.org/pgc/rgbi2usb_test1.zip

I would add missing things like custom palette selection, the magic "Auto" detect scan button most VGA monitors have, ability to send timing profiles to the cloud, record streams to files, and many more features. All open source where the community can add to it.

On the hardware side, I was also thinking a programmable keyboard and mouse output (PS/2, XT, AT, bus, serial, etc) could be added to allow anything you type in the window to be set to the connected vintage machine. Maybe an audio input jack to record sound in lock-step to video as well. I'm trying not to let my mind feature creep. Just curious about general feedback before I go deeper down the rabbit hole on the USB only KVM concept.

It would be next year sometime before I would have enough time for a working prototype.

Personally, I would like to see option 3, maybe with option 4 on the same device, with true colors on one board. A simple plug and play to convert RGBI to modern formats. If this could be done for under $100, I think you could sell a lot of them.

With a small proto-board and a Gonbes converter, you can make #3 or #4 with ease today. I'm sure a target price for a scaling converter I made could be less than $100 - maybe even for HDMI. But I can't come close to the DIY price today. And the Gonbes board can be used for much more like analog input from JAMMA systems and such.

At least with the equipment I purchased, it took a lot of adjusting to get the colors close to right. The light blue still looks a bit purple to me. I was thinking a one-piece solution that didn't need adjusting.

It was noted earlier that the C128 board had the green level off, is there a simple change that would correct this? Like just changing a resister?

My $0.02: I have long had a need to produce video captures of IBM CGA TTL RGB. How I get there, I don't care, as long as the result is accurate. Currently my setup is CGA -> bitc128 -> startech usb3hdcap -> Adobe Premiere so I can fix the bad green level coming out of the bitc128. This is not ideal, but it works. My next step is learning basic soldering and electronics so that I can build a few J. Alex boards (see prior post).

I agree that a digital-to-digital conversion would be cleaner and make more sense, but I also recognize that the greater public would probably want something that converts to VGA so that they can hook up VGA monitors on all their old PCs/PCjrs/Tandys/Amstrads/etc. So while it is against my personal preferences, I think the most sensible product to design is an all-in-one that properly converts TTL RGB to VGA.

I have no doubt alan can do it. My only real concern is sensible scaling options. Meaning, in addition to a nice smooth aspect-ratio-correct-bicubic-whatever scaling for monitors, also allow something like simple line doubling or line tripling so that individual pixels can be preserved as sharply as possible.

KenG wrote:It was noted earlier that the C128 board had the green level off, is there a simple change that would correct this? Like just changing a resistor?

Probably, but they are surface-mount, and I don't have the experience to know what to change anyway. I think building the J. Alex board, which uses through-hole components, is the better option. I have exchanged captures with someone who owns a J. alex board and I saw for myself they produce correct R/G/B output levels.

I've decided my main focus will be my immediate needs/wants. That is just a simple RGBI2USB converter. I might add the keyboard/mouse/audio to that if time permits for a true virtual/retro KVM.

I thought going into this that the software app would be the most difficult part. That's why I initially said UVC, Pi camera input, etc, until I researched the limitations. But after a half days work, I already have nearly the entire app done shy of the hardware driver - which was one of the clearest pieces of the puzzle in my mind. The above demo app already supports no scaling, aspect ratio correction only scaling, and 4:3 scaling to the window resolution with automatic pillar/letter-boxing. It supports bi-linear only scaling as well as a soft low-pass filter. It would be trivial to add a simple line double/triple scaling mode. The app already supports sending the raw frame to the clip-board as well as saving under a half dozen or more image formats. I'll add a screenshot hot-key. I planned to also add clipboard copy and file save for the scaled output as well. First pass of recording will likely be a packed raw file with a simple header, a reference palette, time code before each frame, and raw 2 bpp, 4 bpp, or 6 bpp frames. An offline converter can then take it to whatever.

This answers 100% of my needs. 110% if the Android port isn't difficult. If I ever get to a v2, I'll add additional features. I'll keep you posted on the progress.

This solution can still do #3 and #4 above, but it requires an entire host computer to do it. While not as small or cheap as a converter board + GRS8xxx, it still solves the problem while allowing crystal clear capture. It also allows extending the way that computer up-scales to its native output via software support anyone can do. I suspect even CGA composite artifacts could be simulated during scaling.