TWC Today Forums
Classics - The Weather Channel from 1982-1999 => Local Forecast => Topic started by: techknight on May 02, 2020, 07:51:55 PM
-
Some of you may have been following this project of mine, but some of you may not have.
Through some crazy reverse engineering efforts, and then software development right behind that, I managed to get the STAR 4000 hardware running again since its demise in 2014.
I present to you, the first graphical Local Forecast since that fateful day.
https://www.youtube.com/watch?v=HCcgaJrjaYs (https://www.youtube.com/watch?v=HCcgaJrjaYs)
I have a raspberry pi as the video source into the unit, and the graphics card has a hard time with the signal coming out of the Pi. its kinda noisy, but, the 4000 was originally designed and tuned to the composite output from a VideoCipher module which is slightly different than a non-encoded NTSC signal as VC generated the horizontal front/back porch and sync pulses internally during descrambling.
-
Is that how the real 4000's radar looked towards the end? I've never seen that before.
-
Thats my own personal touch. I thought adding a regional radar would ad my own little spin on it.
Same basemap ratio so when I switch over to local radar, it'll look right. Keeps the math simple, and keeps the memory usage low as its very limited. I only have 2MB to work with for assets.
Also, this is an enhanced radar so it shows the snow and mix precip types as well. The radar the 4000 never had, and should have. :-)
But the unprojected basemap works with the NWS CONUS as well. so I have options, and can go either way.
-
Heres what it looked like when that QLCS moved through awhile back:
-
and I cant forget this tidbit:
Captured during the 3 supercells that trained in south carolina a couple weeks ago.
Note: the video source I am using is a Pi, and it distorts the framebuffer a bit which I need to fix. I have to retune the graphics card to match that of the Pi's composite signal since the card was originally designed to lock onto a VC descrambled signal.
-
Oh and one final thing:
one of my 4000s was heavily degraded and I had to fix this mess:
you can kinda see my raspberry pi video source superimposed into the signal. That obviously wasnt normal. But luckily I fixed all that mess. :-)
-
And boy, Aint that a pretty sight?
Both my plasma and CRT.
-
Now THAT is really interesting. Never before would I see a 4000 functioning in some form that isn't simulated before.
-
Now that's impressive. :thumbsup:
-
Impressive. Can't get enough of it.
-
Oh and one final thing:
one of my 4000s was heavily degraded and I had to fix this mess:
you can kinda see my raspberry pi video source superimposed into the signal. That obviously wasnt normal. But luckily I fixed all that mess. :-)
That's pretty cool that you got it to work again. As for the degradation issues, how did that happen? Was the circuit or video boards dirty or dusty?
-
Heh, I wish it were that simple. But, nope.
there are ICs that fail on the graphics board which cause that mess. In the case of this particular one, one of the 2 SRAMs had failed, and the RAMDAC had failed.
Then after fixing that, the framebuffer had bad jitters which was the result of the 3.58Mhz color burst oscillator crystal going bad. I had to change that too. These things have had a hard life.
-
That's amazing. How did you set this up for your area? And how did you customize the radar?
-
Well, its a collaboration of a few guys including myself. This STAR and its data is remotely managed just as if TWC would have done it, so I have it setup as a "set it and forget it" style. So now it operates as a self-contained unit. I put it on the same system that we manage our WeatherSTAR Jrs on. So they all inter-operate together.
Anyways, it took me almost a year to get to this point. After I had reverse engineered and got the WeatherSTAR Jr running again, I took knowledge from that and brought it over to the 4000 project.
Since the 4000 has no program or assets until its downloaded from the satellite, I had to basically write all new code from scratch. Only thing the unit had was the RTOS Kernel in the ROM, Which I scrapped as well.
I ended up writing my own parallel tasking kernel thats much simpler, and loaded into RAM instead, so I can always patch it and recompile the C code.
Since I wrote all of the code, I simply added a new product descriptor into the presentation lineup and rendering it. I still need to add local radar, but I am waiting on my new Data Cards to come in which has a few circuit changes to add a secondary RX Channel back from the 68K CPU, so I can get current buffers back into the presentation scheduler. (So things draw in the right slots, and off-screen).
As I have full control over the software I wrote, I can pretty much do what I want, and when I want with the exception of the limitations of the hardware which some of it was fun to work around. Just as TWC had to work around it. Doing something simple as the Travel Cities Forecast takes an orchestra of delicate IRQ/code execution timing bouncing back and forth between 2 CPUs on 2 different cards. All while being executed during the VBI. Nightmare. :cry3:
All of the graphics and text assets are rendered in code just as it was originally with the exception of the bitmaps which are recreations, However the Font Assets are from the WSJr. the Jr had 8 fonts, you only ever saw 3 or 4 onscreen.
Thats why it looks so realistic, Because, well, it is. B)
I don't want to hog the entire spotlight because I do really have to thank the group of guys I worked with and continue to work with that helped me bring this into fruition, they shall remain nameless unless I have their permission to mention them, or they mention themselves :-)
-
That's awesome!
-
This looks sick! I wonder if it'd be possible to reverse engineer the other STARs.
-
which ones, exactly?
I already have the WeatherSTAR Jr and the 3000 working. I did that beforehand.
-
I was always a sucker for the XL. I love the graphics it came with, especially in V3 of it.
-
Oh, Yea I suppose they probably work too, but I dont own any of those. So I cant personally verify.
I am an analog STAR guy, I grew up in the 4000 era.
-
I don't want to hog the entire spotlight because I do really have to thank the group of guys I worked with and continue to work with that helped me bring this into fruition, they shall remain nameless unless I have their permission to mention them, or they mention themselves :-)
You've got my permission :rofl:
Other STARs do, in fact, work. It just takes a bit of computer knowledge, patience, and some very long nights to grasp exactly what is going on with them.
I already have the WeatherSTAR Jr and the 3000 working. I did that beforehand.
Figured I'll drop some pictures on here of the III running on my end... some people may like to see it.
(https://i.imgur.com/gthDPopl.png)
(https://imgur.com/DJ1bqqUl.png)
There's two other pictures on Imgur: https://imgur.com/a/oaCnmlq (https://imgur.com/a/oaCnmlq)
-
That's how it would look on a real TV set like that one for example.
-
Parse Error Reading Sentence....
I am going to assume you mean that This is the way its supposed to look on a TV? yeah as its a real STAR on a real CRT TV. Nothing photoshopped here... No emulation either.
So I am uncertain of your statement above.
-
Here is a video I did of Flavor J on my Unit, I just completed the flavor script earlier today.
Enjoy:
https://youtu.be/Ac91sE46pTI
-
Did you program all the weather data parts? I've never seen "Scd T'strm" or VAR6 on any real STAR. And apparently there isn't going to be any sun light in Boston or Cleveland tomorrow. :blink:
-
Did you program all the weather data parts? I've never seen "Scd T'strm" or VAR6 on any real STAR. And apparently there isn't going to be any sun light in Boston or Cleveland tomorrow. :blink:
The data comes from an API so it is whatever the API gives me. I dont have any preprocessing code except for the basics I need to provide to the teletext frames for the presentation engine. Its not like I have access to the satellite feed anymore, or have the exact data that TWC sent back when these were in service. I have to handle the cards that I have been dealt.
Now there is a chance I have a bug. it took me several months to write this engine and debug it to this point. It was not easy as I had to work within the same limitations that they did.
Anyways, I noticed you like to nitpick at this stuff. Any particular reason as to why?
-
The thing to keep in mind with these older WeatherSTAR units is they are essentially "dumb". They are just character generators that can be addressed over satellite... so, if memory serves, a Digital PDP-11 mainframe at TWC would create slightly modified UK Teletext packets to send to the STARs, addressing a specific unit that is hooked up to the satellite feed.
As an example, imagine you have a WeatherSTAR III unit who's address is 00011111 in Denver, and a WeatherSTAR Jr with address 00022222 in Minneapolis. Both units are hooked up and connected to a VideoCipher II that has been authorized by TWC. The mainframe at TWC would gather all of the necessary weather data, 36 hour text forecasts, etc, and build pages to push to the units. The mainframe would fetch the information for Denver, and create a page header packet specifying how many lines should be on that page with some other information, then packets would follow it with what text should be displayed on what line. These packets would be addressed to unit 00011111, and sent over the satellite. Then it would do the same thing for Minneapolis. Gather the necessary information for Minneapolis, and create packets to address unit 00022222. The mainframe defines exactly what will be seen on-air. "Conditions at Denver" could be changed to "Your weather for Denver", or whatever else in the current conditions product... if they made that change at the mainframe.
I hope that makes sense. It's not an in-depth explanation, and there are a lot of technicalities -- but that's just a dumbed down example of whats going on behind the scenes. The 4000 operates a bit differently, but it follows the same premise. It's addressed by the mainframe, and just displays whatever is sent to it. This is why during sun outages, you would see random characters appear on these older STARs... because the data packets are getting screwed up.
Information overload. But that's the basic run down. I might do a little write up here of my WeatherSTAR III and the road taken to make it run again with the combined efforts of Techknight and a few other forum members on here... that way I don't clog up the 4000 topic with III/Jr stuff ;)
(https://i.imgur.com/vzkaWgdm.png)
Edit: Here's an example off of my YouTube channel. I'm sending these lines to the unit... totally custom page I built. Only example of this that you'd see in the field was back in the early 80s when they'd advertise upcoming programming, or show that eclipse warning from like 1984 or something around there.
-
Anyways, I noticed you like to nitpick at this stuff. Any particular reason as to why?
I'm a programmer, so I find this stuff really interesting. I've asked Mike (the guy who programmed the 4000 for TWC) a lot of questions too.
-
Well maybe I am interpreting it incorrectly, it just felt like a debbie-downer type response. Anyways.. I digress....
I don't know any of the original guys unfortunately, and ive tried getting in touch with them but nobody is willing to respond. Maybe they all just want to forget TWC entirely, Which I cant blame, it was just a job to those guys.
Mike was the only one who did respond, but he was the graphics designer and didnt really know too much about the under the hood operations unfortunately.
If I could get the 3 wishes type questions answered from the original under the hood programmers, I have some open-ended questions and issues I have with framebuffer operations/hardware that I would love to have answered. but ya know, doubt itll happen.
Anyways, so I had to go it alone and totally reverse engineer the 4000 back down to a schematic diagram for each board, just so I could understand what was going on. Plus, I needed to reverse engineer and dump the PAL logic. I even tried cracking the FPGA bitstream but that got me nowhere since the reverse engineering script for the XC2018 was based off of the XC2064 and has bugs. Doesnt work properly as the XACT output was wrong.
-
The thing to keep in mind with these older WeatherSTAR units is they are essentially "dumb". They are just character generators that can be addressed over satellite... I hope that makes sense. It's not an in-depth explanation, and there are a lot of technicalities -- but that's just a dumbed down example of whats going on behind the scenes.
Yes, that does help explain things. I assumed that if you were able to get one working, it would work pretty much the same as it did when it was at TWC. I didn't know you could customize it that much. It also explains what Mike told me about how whenever they wanted to update the graphics, they had to either ship new hardware or have the cable operators download the update from the satellite. Which explains this somewhat cryptic message that aired in early '92:
https://twcclassics.com/view.php?file=4000-change92
-
Yea, the 4000 was entirely upgradeable over satellite, it was also volatile. so if the power was cycled, Poof everything was gone. All of it. So it had to redownload it all from the satellite.
The 3000, and Jr had the software in ROM, which unfortunately was non-upgradable unless they send new ROMs or new boards.
Anyways, since the 4000 runs at 8Mhz, I had to crank the GCC optimizer up all the way to 3 just to squeeze any ounce of performance out of it. But hey, it works.
-
Is there any way you can get other local forecast areas such as Montana or Utah to appear? So crazy you got this to work