cwilkson

Yeah, that was the excuse. If it ain't broke don't fix it! Weather wasn't so bad...things that were exposed were typically in weatherproof enclosures. And sometimes potted in rubber or even epoxy. Water wasn't getting into those. But some stuff like in the signal shacks by crossings was just normal metal cabinet enclosures. Which is fine. Until the roof leaks. Or when it floods.

Having worked in the railroad industry for a short time, I can tell you that NOTHING changes quickly. Around 2010 when I was in the industry, they were still using relay logic systems in parts of the freight world infrastructure. Yes. THOSE relays. The electromechanical clicky ones. Used to perform control logic. They had considered using vacuum tubes but they were not reliable enough. And those new transistor things were still too new and unknown....after 60 years in mass production! It was insane! To be fair other parts of the system were fully modern with microprocessors, cell radios, etc. But not all of them. Fun fact: Train detection track circuits use the rails as wires and the signal is something like a 1 kHz sinewave....at 10 amps peak! A train is detected in a section of track when the bogeys (wheels + axles) create short circuit between the 2 rails, passing the signal from one rail to the other. The train itself is the switch.

I have a DMP132? I think it's a 132. It was a GREAT printer! It had both parallel and serial ports. I used it with my Coco 2 in high school and briefly in college. But I haven't used it for 30 years. I need to just sell it. I remember having a program to do graphics screen dumps. It took forever to dump a PMODE 4 image. But it was really cool. I wrote my own basic program that was much faster. IIRC it could dump the screen to paper in about 1 minute. I always wanted to go back and see if ASM could speed it up. But I think the print head was the limitation. Or maybe it was the baud rate. I think it was limited to 1200 or 2400 baud over serial.

That thing is cool. I was looking for an EAI machine for a long time but gave up finally. Would love to have one.

2nd data point. I just purchased a Harmony Encore with this issue. (Shipped 9/6/23, received 9/11/23) I downloaded the latest software (HarmonyCart-1.3-win32.exe) and I beat my head against this issue for a couple of days before I found this thread. I can confirm the updated 2k4k.arm file (424 bytes) fixed the issue.

Ok. I know you've been to hell and back on this thing. But I didn't see where you had confirmed measuring a buffered clock (PHI_x). It's mainly a check box/sanity check kind off thing and I also wanted to document for posterity. I'm glad you did that that check. One other thing I thought of if you ever come back to it. Maybe the board is dirty? Not standard dirt, but something special. Like carbon scoring or chemical contamination. (even soda could do it). You mentioned the original TIA was damaged. Or at least is was suspect. If there was an "event" it could have left tiny conductive paths that are stealing energy from the oscillator.

One sanity note. "Never" measure a crystal oscillator's output directly. Others have alluded to potential loading effects. In many crystal circuits, adding a scope probe's capacitance in the wrong spot can completely kill the oscillations. Best case, it can slow the circuit down a bit. You should measure the frequency of PHI_0 on TIA-pin4 instead. The ideal frequency here is (105/88)*1MHz. But you should divide your crystal's marked frequency by 3 to get your "correct" value for the CPU clock. To get the best resolution from most scopes, put a single period of your waveform on the screen. And turn on averaging if your scope supports it. Use a big number for the averaging. And measure both frequency and period. Period may be more accurate on your scope. I really don't know...I've never used Siglent gear. Compare your PHI_0 frequency from the problem machine with another one instead of the using OSC frequency.

Got it. That's very interesting. I know that the '723 was pretty popular. It's quite a versatile IC. It can support linear voltage, linear current regulation, and even temperature regulation. And I'm pretty sure I used it to control a switcher back in the 90's, based on a Radio Shack app note. But it seems kind of wasteful in a simple linear regulator vs. a 3-pin regulator. BTW, with a dropout voltage of 3000mV I would never call the '723 a Low Drop Out (LDO) regulator. No matter what TI marketing wants you to believe.

Thanks for the catalog reference! The 26-210 looks like it might work with an MPI. But maybe not. I was looking for one a few years ago but didn't even know what to call it. Or that is was designed for the Tandy 1000-EX. Now I have a few new search terms to work with! Has anyone here actually ever *seen* a 26-210? LOL

Weird sockets on that supply board. What ICs are used? LM723? Or some type of quad opamp?

What is the revision status of the NTSC CleanComp? (the website says v1.0?) I'd like to buy one to compare/review, but it clearly has some pretty serious issues remaining. If the next revision with fixes for these issues is almost ready, then I'll wait. Otherwise I'll get the current revision and live with (or fix) any problems that exist.

What? Really? Still? I thought they supported horizontal video for a year or two. I admit I'm not into TikTok, but I could swear they support it. Damn.

Why are you using a SMPS? I don't see a need and it will only cause you extra grief as you've already pointed out. But maybe there's something special in your circuit that requires a weird voltage? For transistors...are you using BJT or FETs? What type? BJTs are probably faster at the cheap end, but storage delay will kill you if you don't compensate for it correctly.

Real quick cinematography advice. Turn your phone 90 degrees!!! And tell all your friends to do it too! To take advantage of basic human vision wetware. And every computer monitor everywhere. You'll have a bigger viewable image without wasted pixels at the top and bottom and you won't need to pan and scan. You can then use a cheap tripod to avoid shaky cam effects. Even propping the phone against a stack of books will work. I said cheap!

Nice update. There is quite a bit of noise and edge fuzz in the screen captures. I would like to see photos vs. captures to see how much of that is caused by sampling. Your composite waveforms look surprisingly nice. Good band filtering where there's color. Some over/undershoot on the luma stairsteps but it's minimal. Good job! It's great that you're using a scope to actually *see* what's happening vs. guessing and hacking. Makes it *much* easier to improve things. Keep working on the timing between signals. Those spikes are pretty severe and might be visible on a good monitor. Especially if they happen at a bad time. Need to hammer them flat! Be sure to check the waveforms with color active at both edges of the blank pulse. That's one of those bad times. Looking forward to the deep dive for S-video.