You're right; Adrian's patience is remarkable to say the least - had it been me doing that, i'd probably end up smashing something cheap to atoms in frustration.
Hey Adrian, I'm actually just dealing with the exact same issue as you're having (black screen and sound). ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-tPDP4izVqrk.html I was also hoping it was the ROMs, so I'm waiting for some semi-compatible EPROMs I can use, although after having seen your video, it's making me think it could be something else as well. Hopefully we'll manage to solve our problems! Let me know if you want to compare signals or do any kind of tests. Good luck!
Ah -- I solved the issue on this board in part 3. It'll be out next week although these issues really can be caused by just about anything that I talked about in this video. There won't be a common smoking gun unfortunately.
@@adriansdigitalbasement Good to hear that! I'll be looking forward to it. How much in advance do you make your videos BTW? I'm horrible about finishing them up just about the same day I release them 😜
You put in so much time for these machines and then the setup time and then the editing time and research time etc. Very much appreciated. Outstanding videos from outstanding effort.
Back in the day when I was repairing micros, one useful technique I used on bus issues was to use the “component test” facility found on Hameg oscilloscopes of the era. Basically the scope injects a low voltage AC signal to the circuit and then traces a voltage current shape onto the screen. It’s simply a case of stepping along the bus connections and compare the “signature” (voltage current/VI) shapes. Anything significantly different between bus lines and I’d start to pull devices and see if anything changes. It was possible to test the pins on the removed chip in the same way. All of this is done with no power applied. I later bought a system that can functionally test many 74 and 4xxx CMOS devices in circuit and also display the VI signatures concurrently for each pin on the chip, up to 64 pins at time. It was even possible to record the signatures from a good device and run an automatic comparison with the device being tested!
Great idea! I used the "Huntron Tracker 2000" for troubleshooting when I worked in IBM Argentina un the 80's. It was a very interesting tool! m.ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-W7LWnY4oOgw.html Greetings from Buenos Aires Daniel
Da ny The Huntron is similar to one of the modules in my system. I have an ABI System 8 and one of the modules is called an “AICT” which could do the more advanced VI curve tracing as seen on the Huntron. It could also functionally test analogue devices like op-amps and transistors in circuit! I think curve tracing is an underrated troubleshooting tool and perhaps not as well known as it should be :) Interesting to hear IBM was using Huntrons!
My second computer after the Timex Sinclair. The plug-in speech synthesizer was absolutely amazing at the time. Time spent troubleshooting is always time spent learning, so it's never time wasted!
I feel like I get more practical knowledge from watching your troubleshooting videos than I did from four years of engineering school. Thanks for the content, keep up the great work, and good luck with the repairs!
The 99/4a was my first computer, I'm still quite fond of the little guy, for all the quirks he brings. For what it is worth, the 256-byte "scratchpad" RAM attached to the CPU is because the TMS9900 was designed to be paired with fast SRAM in TI's minicomputers. The processor's registers actually exist in SRAM, so there has to be SRAM on the processor bus. And since the VDP used DRAM, they couldn't just directly share the same pool of RAM like a Commodore or an Atari computer. My understanding is TI was going to use a variant part with an 8-bit bus(like the Intel 8088) and an integral SRAM, but that part never actually went into production. And the 99/4 project was already late, so they ate the expense of installing a full TMS9900 on the board alongside a small SRAM and a multiplexer to interface it to an 8-bit system design rather than do a more significant reworking of the system to bring the full 16-bit bus anywhere(it would've been so nice if the expansion port hadn't been multiplexed).
I was just explaining to my daughter this morning that in the "old days" we would have to wait for shows to come on TV during certain times and certain days. Even though I can watch it anytime I want, I have found that ADB has become my Sunday TV show that I look forward to.
I love these challenging retro computer repairs. Funny, I was just going to suggest that you team-up with *Noel's* *Retro* *Lab* as he seems to be in a similar situation as you, but I see that you have already connected. I have been learning a lot about the TI99/4A in this series (as well as Noel's series). Great work and I eagerly await Part 3.
Because some companies were paranoid as hell about their intellectual property, they wouldn't even give their own licensed repair services specs for the machines they were trying to fix. IBM made a lot of waves when they released the technical manual for everyone, but they were in the middle of a series of consent decrees and antitrust suits based on how they ran their mainframe business, which is why the 5150 was built with off the shelf components.
Wow, this Ti99/4A is fighting you until the END!! It really doesn't seem happy about living another life! "Just let me die in peace!!!" :) AWESOME TROUBLESHOOTING!! So looking forward to part 3!!!
Ahh the days of an easy to follow troubleshooting chart. I remember getting tunnels of doom for this system in the early 80's. Lots of good times on this system.
Noan YoBiseniss I remember tunnels of doom; needed both a cartridge and a cassette?; I heard that there was supposed to be more than one cassette for it; do you remember how many cassettes you had?
I'd suggest looking carefully at the data bus signals with a scope; the traces have resistance and impedance, so the voltages will be lower close to a chip that's dragging them down or higher close to a chip that's dragging them up. I'd use two channels on the scope to compare very closely so it's more apparent where they differ quantitively but not qualitatively.
skonkfactory Excellent advice! I think Adrian has used an analogous technique to locate shorted tantalum capacitors on a board by looking for decreases in resistance along positive traces relative to ground. Using a carefully reference calibrated dual trace scope could definitely yield visible results.
Thank you so much for your videos! I got me a TI/99-4A recently (EU version with component video, I wish it was a composite US version...) and had an issue with glitchy video output. I was able to fix it by detecting that the voltage on the board was 4.4V instead of 5V. Fixing PSU made it work like a charm! Your video helped me better understand the documentation and the internals of this glorious machine. The quality of your videos is awesome, you really are putting a lot of effort into these. It's so easy to comprehend & follow on what you're doing. Thanks for all the detailed explanations & visuals! I still cannot get a color picture, but maybe it's due to my shitty self-made YPbPr cable, or maybe a faulty Chinese YPbPr->HDMI transcoder that I bought specifically for my TI. But hey, at least I can boot and get the grayscale visual now! EDIT: Any tips on how to test if my YPbPr cable is OK without owning an actual functioning YPbPr receiver? Do wire lengths matter? (ring & sleeve wires of my Pb/Pr are like 1 inch longer than my luma cable) Man, I wish I had a composite version. YPbPr is so rare these days!
I'm loving the troubleshooting. As much as I'm hoping you can get the thing working, seeing all the effort you are putting into this is very entertaining.
Looking forward to the continuation! I have wasted countless hours trying to repair a vtech Laser 200 that exhibits many of the same symptoms on the oscilloscope. It's really frustrating when you just get a bunch of noise on the data and address lines, this video has inspired me to try again at least :)
Hi, nice to see someone working on a TI99. About half the performance of a Amiga if you give it some ram (piggy back on the ROM use a miniMEM to program) less than 1/3 the performance of a C64 if you leave it standard. Xtal can be upped to 16Mhz without a problem, only the video timing will change. This raises effective clock speed from 3,3Mhz to 4Mhz. A common point of failure are the G-ROM chips and those are not easy to find. These chips are used out of specs, driving more than they should. The processor is a interesting but strange beast. For example: - IO and memory don't live in the same address space. - There are "long distance fetch" instructions to access the memory of another 9900. - And there is no stack, instead the whole RAM might be used as registers for the processor. As a result it excels in interrupt response time and is very good at multitasking. Without modifications the TI99 is a 240 Byte computer with 16K video RAM. Code from the video memory has to be copied to the 240Byte RAM before it can be executed. One of many ways TI tried to slow down the TI99 and make sure this cheap computer would not compete with the companies $$$$$$ mainframe systems. Have fun.
I know you've probably already noticed this and found a solution, but at 25:10 when you're going through the workflow, you actually start reading it the wrong way. YES is down and NO is right (as per the previous page), so the next troubleshooting step (NO) was to "Check P12 0607 for High Level", not replace the Groms.
YOU ARE A VERY HARD WORKER YOUR VERY KEEN IN TRYING TO FIX ANY COMPUTER YOU ALWAYS COME UP WITH GOOD REMEDIES IN SWAPING EPROMS AND MAKING NEW ONES WORK I THINK YOU WILL FIX THIS TI BOARD WHEN YOU GET A NEW CPU AND SRAM CHIPS MAYBE JUST CPU LOOKING AT SCOPE.. GOOD WORK ADRIAN BYE
The troubleshooting instructions supposed you have a regular analog oscilloscope. The waveforms they show are what you get after the processor is in some idle loop. But if your oscilloscope has memory you would get a lot more information looking at the first few cycles after reset. Depending on what went wrong it would be normal for it to get stuck in a loop where some of the address or data lines are always low or always high (or even switching from high impedance to always low). The weird waveform you saw didn't look normal, however.
Nice video and care for the TI99/4A, it brought me good memories. I had one of these when young, my father bought is super discount (like 1983 for $100). I remember learning to program basic on that machine. I think it is still at my parents house.
I find these extensive rabbithole type videos the most interesting due to the educational aspect. Sometimes I'm rooting for failure just so that I get to pick your brain. :D
I know you could do it!😁 Allthough it was fine if it wasnt, I like seeing the steps you take trying to fix unknown hardware. Great find/workaround on verifying these eproms on your minipro. Cant wait for part 3!!
Thank you for your work and video!!! The TI99/4A (silver and black) was my first computer. I loved it. Sadly, I had to retire it after it became unreliable.
Excellent. I really like these detailed, component-/schematic-level bug hunts. The PROM dump resuling in all FFs made me immediately suspicious though. That didn't match the issue you had already identified with some data lines always sticking to low. But, well, things are always obvious once you know the solution... Looking forward to part #3.
One of the original CMOS processors, and they did not have the capability to integrate the clocks into the chip, so you had external IC's to give you the 4 clock phases needed to run the internal state machine in it, and 12V levels because NMOS needed the high voltages to get high speed. Later on you got processors that did all the voltage conversion inside the die, so all that was needed was a single phase clock and a single 5V supply, and now once again you get processors with internal voltage converters, so you can generate the internal sub volt rails they need, while still having 3V logic interfaces.Power has gone up though, with some needing over 100A on die to power it, so half of the pins on the package are power and ground.
@@SeanBZA interesting stuff, indeed. But it's the whole system architecture that is so baffling to me, with only the video processor having access to the main RAM and that GPL interpreting a BASIC interpreter, interpreting a program shenanigans. Really odd stuff.
That's what I was thinking too. I haven't looked at one of these since I played with BASIC on it in K*Mart as a tween, but the fact that it has a 16-bit CPU/bus and 32 hardware sprites is really surprising to me. Too bad they had to skimp on the sound and RAM for cost reasons - otherwise it could have been a real contender.
I know little to nothing about the technical side of these things, but I always wondered what the 99/4A would have looked like had they been able to get it 16bit through and through and what that may have led to.
...and not nearly as elegant as the Rube-Goldberg shenanigans going on in the Apple IIGS, because Woz _had_ to do his crazy stuff to retain backward compatibility. This thing was more-or-less first-of-its kind, so it was all unforced errors man. :-)
Back in the later 70's-early 80s, I built an EPROM programmer for those 27xx chips and the programming voltage was the hardest part to get right and I do remember how slow the process was.
Oh man, we had one of these when I was a kid, and it stopped working. Now that I'm in my 40s and would be more than comfortable to poke around inside, I really wish I still had it.
Your falt is the same as on mine, going to use this in fixing mine up, currently on the back burner as it is quite an involved repair. Not something simple like a C64 issue.
This video reminds me that my rusty found-outdoors Sega Genesis VA3 board needs some love and I have a schematic / flowchart I could be going through to diagnose it.
I haven't finished the series yet, but I swear it looks like pins 5 & 6 are shorted together on that one chip. Looking forward to the thrilling conclusion.
Given the frustration I'm having in repairing (or trying to) a Sony 21" TV, I think I would have given up hope on something like this, you've a lot more patience than me... :D
These videos are great, but hopefully you're able to document what you learn and leave something in writing for the next guy who needs to repair one of these. Maybe a PDF or Word document fleshing out the gaps in the existing older documentation? Best wishes.
When you have exactly the same signal on two adjescent bus signals you might be dealing with a short between two bus traces, example D5 and D6. The wide signal, changing up an down into a small spike can be caused by a floating select signal, causing an intermittent bus signal, so as a whole you might be dealing with a bus conflict. I would take a look at the glue logic. Could be one select signal stuck to active. You might also want to take a look at busses connected to the outside world. They might have suffered some kind of discharge and stuck in a permanent bus hold state.
I had the same issue reading a mask rom from my compaq portable, where all I would get are FF. Turns out that while most of the important stuff that the motherboard cared about were the same on an eprom and the mask rom, the mask rom actually had an OE pin or two where an eprom might have NC or a program pin. When I jumpered the OE pin high and the TL866 was able to read it just fine. According to the data sheets, the state of the OE pin was configurable by the customer.
Adrian, do a search for "99-4A_Console_Peripheral_Expansion_System_Technical_Data.pdf" This has a fairly readable 'silk screen' type of layout diagram, and very readable schematic. Also found "TI_Manual_Project.zip" which has a bunch of good schematics and other documents. Both hosted by 99er.net. They have archives of message boards and links to a lot of other resources. I also found nouspikel.com and ninerpedia.org what has links to a few good resources/sites.
Hi Adrian, at about 43 minutes when your showing your adapters, where did you get those pin sockets? do you have a pn? I've been looking and cant find those.
If you have trouble to see low quality images, why don't you try to upscale them with Topaz Gigapixel? I used it for... upscaling a menu before ordering. Works good and might help you
Hi Adrian, love you work, one thing I noticed to me it sounds like a sound chip problem, with the phasing that you are experiencing that could be the course of your problem.
Hey Adrian! Good work over there! (At least..A LOT of! ;) Ideas for the next video: Find the TMS9900 opcode NOP and make a NOP generator! So as to check TMS is running and the Address Bus lines! Regards from Argentina! - = Diego = -.
@39:32 weird that your programmer couldn't read the ROMS... when I insert mine I give them a horizontal wiggle as I tighten the ZIF socket to try and get rid of crud on the legs... FF suggests no power to the chips?
Could you make an video about video/vga to hdmi upscalers to connect old machines to modern monitors... i serach an vga upscaler wich is able to upscale 320x200 (mode 13h)
When one chip is pulling down the line, what you can often do is check at which pin the signal is the lowest, thats often the best candidate that is actually pulling it down
For the cursor issue during video capture, it looks like you've got a HW accelerated cursor plane (those are typically 64x64 pixels) and the colors are messed up somehow. I'd guess the recording SW thinks it's in one RGB mode but it's actually in another.
Often to figure out which chip is U31 (or whatever) you'll need to look at the board layout. If the company is kind, they'll provide an easy to read version that just lays out which chips are where.
I was curious about the tms9901 interface and it turns out: it really does not have a data bus connection but the address bus is not carrying the data it is only for addresses -> the chip gets data in and out via the "CRU" shift register interface to the cpu
I have one that many years ago I left it on all night and the next morning the screen was blank just like that one. I still have it in hopes of fixing it one day.
The TI9900 series connected all I/O chips via a single bit serial bus. The I/O was actually BIT addressed using this CRU interface. (It worked sorta like a single channel PCIE bus).
Hey Adrian, I feel your pain!! I have a Texas Instruments DS990 Model 1 in my collection that I have been trying to get booting for... 3 years now. Geez its been that long. It is a TMS9900 system like the Ti99/4A. Apparently, TI used the DS990 Model 1 to design the original Ti99/4a, so I'd like to think they are similar in many ways. Would it help if I told you that the TMS9900 doesn't contain internal registers? It uses part of the system memory to store its registers. I'm... not too familiar with very low level system logic like that so I don't know if that matters, but I can presume that if those registers are not working, the rest of the system would be dead too. I vaguely recall you saying there was a 120x16bit RAM space..? Good luck!!
I had the exact same problem on my ti-99/4a and it was the mcm6810p ram chips and these can be easily ordered via eBay 12$ for 5 pcs after replacing these it worked just fine and they tend to fail.
The groms maintain a pointer, which the cpu can set, on a read the at the location of the pointer is placed on the bus and the point is increased, each subsequent read returns the next byte. it is designed for fas access, but also means eproms wont work.
On your oscilloscope, some of your weird looking readings could be related to an averaging or "high resolution" mode. Turning it off will look more noisy but stuff like address or data lines will usually look like nonsense with those modes on.
Another place the data bus comes out is the expansion/game port .... also a great place for the curious child to cram something...also there is a couple of pullup resistor packs ....