Man best of luck with this thing. I've never seen 60C on anything of that era unless it had an internal short. That might also explain the power consumption. Check those two hot chips.
Those hot chips could also be a symptom of something different and not directly a failure of the chips themselves. If something is loading their output circuits down or driving the outputs from somewhere else, they will source or sink current they're not supposed to. If they carry high frequency signals, bad capacitors could very well be a cause of excessive load on that circuit.
While MDA has only one mode of operation - the text mode, where the character clock is the crystal clock devided by 9, the Hercules card has two modes of operations. In graphics mode, the character clock is the crystal clock *divided by 16*. And 16.257 MHz / 16 is 1.016 MHz. This means your card has switched to graphics mode because of some signals it received from the bus. The fact that you can measure 1.8MHz on the bench and 1.0MHz in the computer actually is a *good* sign for the clocking circuit - it operates well in both text and graphics mode - but something on the card is misinterpreting the bus signals - the IBM PC/XT BIOS is unable to switch a working Hercules compatible card into graphics mode.
@@MessalineApghar The IBM PC is set to MDA using its dip switches. I haven't found the manual of the Tseng Labs UltraPak but the output should be limited to monochrome MDA (default) and hercules. There are 2 jumpers on the card but tried all combinations. But will keep on investigating. Will start with the 2 hot chips.
@@RetroSpector78 there were usually software drivers included with the boards as well to change modes. And not all mda graphics cards were hercules standard. Some did cga emulation instead or in additio .
I would suspect those two 74LS chips, if they are somewhere near a control line, they could be pulling some pin of the CRT controller low which causes the controller to switch to Hercules mode instead of MDA mode.
I tried to reverse engineer the card just from a chip list taken from your pictures on imgur. Most likely the 4 74LS258 chips are generating the address lines for the video RAM. The video RAM on your board is comprised of 8 RAM chips with 65536 entries of 1 bit each (64k x 1). This means the RAM chips need 16 address bits in total, split into 8 column and 8 row bits. Your card has two sources of addresses for the video RAM: Either the address is taken from the CRTC (for screen refresh) or it is taken from the ISA bus (for processor access). This means that you have 32 address input signals (16 CRTC, 16 ISA), and 8 address output signals (RAM address pins A0..A7). Each address pin for the RAM has 4 different possible sources: It might be the row address from the CRTC, the column address from the CRTC, the row address from the ISA bus or the column address from the ISA bus. If my intuition is right, this selection is performed by the 74LS258 chips, which are "quadruple 2:1 multiplexers with tristate output". As these chips contain just 4 multiplexers, but the card needs 8 output bits, each two of them are grouped as 8-bit multiplexers. The selection process works in either of the following two ways: a) two multiplexers select the RAM row address by receiving 8 address bits from the CRTC/graphics card logic and 8 address bits from the ISA bus, and the other two multiplexers select the RAM column address by receiving the respective other 8 bits. The select input in the multiplexes select whether you want the CRTC address or the ISA address to reach the RAM chip. The output enable input is either active for the row address multiplexers or the column address multiplexers, depending on which address half is required at the RAM chips. b) two multiplexers deal with the ISA address and select either the row or the column part of it, the other two multiplexers deal with the CRTC address and chose the row or column part of that address. The select input input of the multiplexers is connected to a signal the determines whether a row or column address is required, and the output enable either activates the ISA address multiplexers or the CRTC address multiplexers. In either way, only two of the multiplexers must ever be enabled at the same time, as the second level of the 4:1 demultiplexing is done by only ever enabling one of the two groups of 8 chips. As you have two very hot 74LS258 chips, it is quite likely that those two chips are driving the same address lines, and must not be enabled at the same time, because otherwise they drive against each other, causing excessive current consumption and heat generation. I suggest that if doing the simple test suggested by other people, namely testing the output pins of the hot chips for shorts against ground (and +5V) doesn't result in any findings, you probe the output enable lines (pin 15) of all four chips. You should find two different signals (each at two of those chips), which are *never* low at the same time (output enable is active low). If the output enable waveforms are correct, it is quite likely that one of the two hot chips is driving its output even though it is requested to be in high impedance state (aka Z or tristate) and needs replacement. If the output enable waveforms are incorrect (both of the hot chips are indeed enabled at the same time), you would need to trace down where the output enable waveforms come from.
I now know that you are very knowledgeable and I understand even less about computers than I thought I did. Just fascinating. I now want to learn more so thank you for that.
Hey there, been watching your series from some time now so I thought I'd give you a little help with my insight. To me, the main steps of diagnosing over consumption are as follows : 1- Checking the power rails integrity : with a simple ohmmeter, pick your usual +5v/-5v/+12v/-12v on the main connector (as you did) and check that each (and I mean all of them) chips VCC are connected. Do the same with each GND. 2- Power it with a bench supply, preferably not more than 500 mA on the +5v rail. If current gets limited, check what is consuming that power. You can feel the heat from the components with the back of your hand (or even your face because it's even more sensitive... silly, I know, but it works). 3- Now you've identified something. Those two 74ALS258N sitting in the middle that are overheating seems to be linked to the excessive idle power consumption (2 Amps at idle is a lot). These are data bus multiplexers, and their outputs might be tied down to ground by what's after them, causing them to sink all the power through them. 4- With the power off, check again the outputs of the two chips and the GND. With an ohmmeter, then with the diode tester mode (you could find bad transistor junctions). 5- Then proceed to remove all socketed chips you have on the board. The idea is not only to find the culprit, also being able to reproduce the fault. With (almost) no circuits, you will get the big picture of whether the chips OR the board is at fault. Dirty traces under sockets can cause high impedance MOS inputs to go low or high. (By the way, once you know the board's PCB is in good health (clean, no bad traces), soldering sockets for the chips is helpful) My personal advice would be to remove all RAM chips, they can go bad for nothing much. (Don't focus too much on the "low" 1 Mhz clock signal, it might be related to low VCC) Once you've identified something wrong and eliminated it, repeat from step 2. Then, once we have every chip present and not causing problems, we willl look for the proper signals (clocks). I'd advice not plugging it in the PC until then. Good luck with your fix, it takes time and patience but it is worth the happiness once you get it done :) so please don't give up.
Hehe .. sure looks like it :) but already tried a lot and it can get complex.... without schematics and a lack of some fundamentals and experience it can be tricky. Fixing a PC is 1 thing, drilling down to the chip or even logic gate level is another :) but as long as it stays fun I will continue. If all else fails I’ll run over it with my car and make a video on that :)
@@RetroSpector78 You're welcome. Don't be discouraged by the apparent complexity, just check one thing at a time so you will be more efficient. Read some datasheets (datasheetcatalog.com is nice for this era), and you will get a general idea of the inner workings. Experience comes this way. I can still guide you through what you will find when troubleshooting. But please... no car...
The PC power bus is probably dropping lower than the bench power supply and the clock is being affected by that. Those chips at 60degC are DEFINITELY too hot and should be replaced.
DAC chips will get that hot. Which is normal-ish. 10 watts isn't outlandish but is higher than I would expect. My genuine herc card used 8.8 watts IIRC. Clock won't be effected by vdroop much, or shouldn't be. It is being pulled into a mode that is not usable for post.
@@wishusknight3009 There are no DAC chips on that card, the chips that were hot appear to be 74 series logic chips next to the CRT controller. The closest thing to a spec that I could find says 4.5A is the limit of the ISA bus. I can't believe the actual connector is rated for that. I'll have to check my Hercules card, it would be surprised to see 8.8w.
That power consumption is a clue, I don't think that the video card should take ~10w. You probably have some shorted parts on there and a "fingertip test" should be done to find any warm chips.
First things first, the card isn't going to run on 4.1V even if it was working perfectly fine (although I've seen a Williams Joust run on 4.2V, but very unreliably) Clock: There's a good chance that this is a multi standard card where the clock frequency depends on what video standard it emulates. It has 64k of RAM, so it's quite thinkable it can do MDA, Hercules and CGA, so that depends on how it's initialized which depends on what's on that ROM. The 6845 is really flexible. Go check HSync and VSync. Hot Chips: If it's a 74LSxxx it shouldn't get hot. Some get warm. There are exceptions (like the 188). If it's a PAL or a GAL, these usually run hot. At least early ones like on this card. Battery Acid: It's rather common to see chips that suffered battery corrosion run fine for a few hours or days and then fail. Logic Levels: Check the voltages on the pins on the ICs to the left of the ROM. If there's any signal where low is > 0.3V or high < 3V, that chip is bad. On the right side of the ROM are lots of tri-state devices where these levels are actually OK.
Not an electrical engineer, but I would definitely check if those two 60C chips are working the way they should. There can be short in those two (closest experience I have is burned Arduino nano - atmega chip was very hot and never worked again after I accidentally shorted it).
Yeah will check those. Only decided to checked the temperature hours before editing the video :) sometimes you get tunnel-vision when debugging those things for too long.
Let me say I like this sort of videos very much. I like your normal videos too, but it's a plesure to see you working as a detective, understanding the most elementary electronic part. Congrats.
Thx appreciate it ! Try to add some variety but most of the time its just whatever I fancy doing at a certain moment in time :) I try to complete these series in a timely fashion but also want to add some other videos in the mix. Cause I spent several days spending countless hours on this thing :) might not look like that in a 20min video.
18:16 maybe it can be handy: i was fixing bad amd radeon hd 6870. The card was on the bench - and I've checked resistances on each power rail, then I've "injected" proper/correct voltage from lab power supply on each power rail step by step to observe power consumption. And then I've decided to supply all voltages to the card, which was on bench. So I've put 12V on psu connectors, 12v on pci-e pin, and using step-down voltage converter I've put 3.3v on pci-e pin (made from 12v) So all voltages was present but I still not get high state on enable pin of main phase buck controller, so it doesn't start to produce 1v for gpu core and 1,59v for video ram. But when I've disconnected everything and I've inserted video card into computer pci-e slot, and start it up, then I've got high state on enable pin and then buck controller start working and it provides 1v for gpu and 1,59v for ram. So the conclusion, which can help you, is that - there must be a signal from computer (via socket pin) to wake up video card. I understand that here you have retro architecture and power saving was different in that time, but maybe this is the case. Maybe you get 1,8MHz on the bench, because the card is NOT fully on. But as soon as you put it into computer, the video card receives a signal to start rest of the components (which all wasn't supplied on the bench) and maybe some of that components is bad and that bad chip is taking down the 1,8MHz clock signal. Second thing. 16:55 I see 1,78 Amps on that videocard and it is too much i think. And 60 C degrees on any old/retro chip is abnormal. That hot chips explains amperage. If you can't touch it (because is too hot), then it is the clear clue, that it is short, or something around him is overloading the rail which that chip provides.Please check that hot chips datasheets, and measure resistance on all pins to ground and if (in the datasheet) that chips provide any output voltage, then lift that leg out from the board, or take out that chip and inject voltage from lab psu on that pad on motherboard side. To check if something on that rail is shorted or if short in on the chip side only (i suggest to remove the chip or lift the output pin before injecting voltage from psu, because sometimes reverse voltage from the other side to the output side, can burn the chip or inside diode) Third thing: When you be able to discover the reasons of the hot chips and fix it, and if then the videocard will still produce black screen, then maybe you should check the video bios. is any eeprom chip there? Sorry for my english, I could use wrong words, but I hope you've understood me. greetings from poland, and thank you for part two :)
Thanks a lot for the comments.... I will check the 2 hot chips and see if I can replace them. Running it off the test bench was just more convenient to probe around, but I can imagine it is missing some vital signals at that point on the ISA bus. But it is a little bit more difficult to probe around with the video card in the PC. Should have some kind of ISA expansion board + cable to get it out of the PC.
You shouldn't use "I've" before the verbs, just "I". For example it should be "I checked resistance", not "I've checked resistance". "I have"/"I've" is used is other situations. Other than that your English is pretty good :)
@@WTFBOOMDOOM Thank you. But wait a minute - I remember, form the high-school, that the was "I've,you've..." required in some situations in past time. Could you tell me, in simple way, when should I use it?
I'm the only one who didn't understand most of what's going on and still got hooked up? I really want to see this card working again and now I'm interested on learning a bit of basic electronics
Even though this is way beyond my skill level, it is fascinating. I hope you get the help you need so that this piece of technology doesn't go to waste. Cheers.
interesting! was the crytal still oscilating at 18MHz when you plugged it in? I think maybe some of the +12 -12 -5V is disturbing either your mass or your 5V signal. Try using standard PSU on the bench
both crystals were oscillating fine in the PC. I think the 18Mhz was used for the serial ports (this is one of those multi-functional cards including serial ports, a battery, and a header for a memory expansion.
it looks like the hot ICs you pointed at are some 74LSxxx ICs. Those do not usually run at 60°C. Last time (a long time ago) when I fixed one of those cards they just took over 1A and not almost 2A like that one. But it was a bit more similar the other working card you showed.
Ditto on the overheating TTL. Run the board on the bench with all 3 supply voltages and measure the current draw on all the rails with the 2 overheating chips in and out of the board. A current limiting power supply module is useful in this type of situation. A frequency counter is useful to measure the clock divider chain. There may be several faults here. Have fun!
HGC info from this link: en.wikipedia.org/wiki/Hercules_Graphics_Card "The monitor is connected to the Hercules card through a DE9F connector using 5V TTL electrical signaling; the same as for an MDA card.[5][6] Nominally, the Hercules card provides a horizontal scanning frequency of 18.425 ±0.500 kHz, and 50 Hz vertical.[7] However, the Hercules actually ran at two slightly different frequencies depending on whether it was running in Text or Graphics mode. This is a side effect of the CRT controller using 9-pixel, 14-line character cells when impersonating the MDA, but having to use 8-by-4 pixel blocks to turn memory bytes into 1-bit pixels in graphics mode, with neither the total pixel count in a horizontal period nor total number of scanlines matching up completely. As the card's onboard clock crystal was also slightly different, at 16.000 MHz vs the MDA's (and later, EGA's) actual 16.257 MHz, neither mode was an exact match for the original sync structure, with the exact figures being 18.141 kHz and 49.84 Hz in text mode, plus 18.519 kHz and 50.32 Hz in graphics. However both are close enough to MDA's own 18.432 kHz and 49.82 Hz timings to fit within the sync tolerance of and work reliably on a typical IBM monochrome monitor or compatible. This tolerance was taken advantage of by the various Hercules clone cards, which themselves often exhibited less than accurate timings versus the original.[8]"
I have and I will .... didn’t know it could do logic chips :) something for part 3 (all the socketed chips have been replaced and brand new). Will check the “hot” chips also.
@@RetroSpector78 Yes the common MiniPro TL866/TL866II EPROM programmers you can get from ebay, etc have the ability to test many 74 series logic chips and even static RAM chips. Sadly they aren't up to testing TMS4164 or TMS41256 DRAM chips. I'd also suggest dumping the EPROM(s) on the board as this will test them out if nothing else.
60C on some TTL Chips (?). That would seem shorted to me. Maybe that's also explaining your high Amps and low voltage on the card, which in turn might be affecting your crystal?
Yeah that will be the next step replacing those. On the working ibm mda card I also get 1.2amps power consumption. Dunno what is considered “normal” for these cards. And low voltage is the headers / connectors I think. Cause in the pc it is getting a clean 5v.
Any chance you can check those larger specialized ICs? I had the honor of repairing an Atari ATW800 a while back which had video issues, while I suspected a number of TTL chips on the video board it was the SRAM (for the color lookup table) inside one of the DACs in the end. Getting a TTL tester and checking the EPROM contents should be the first priority at this point tho. EDIT: Another thing worth checking ... is the card a 4 layer one? Might be worth checking that there are no shorts in the inner layers - that's the sad fate an Amiga 3000T from a friend is currently suffering due to battery corrosion.
60°C of heat on a standard TTL logic IC is not normal at all. Probably those 2 chips are internally shorted and cause a high power draw when you put the board in the ISA slot, this could affect the correct working of the entire card. I would start by replacing them.
I checked that these pins were electorally connected and as such it didn’t matter what pin I used ? In that case is it possible that one chip is on pin29 and would not affect pin 3 ?
@@RetroSpector78 Ignore that part of the ISA slot of my comment, I actually deleted it after re-watching your video, where you connected pin 1 and 3 to the power supply and measured the voltage on pin 29 and 31. This confirms that the two +5V lines are connected so powering the board from pin 3 or pin 29 shouldn't make any difference. I suspect that those two ICs are internally shorted to ground. I remember when one chip died on my Roland Jupiter 6 analog synthesizer, I was playing some music on it and suddenly the synth went out. No magic smoke, strange noises and such, it just stopped producing any audio and all the lights on the panel went off. No response from the keys nor the controls on the panel and a power cycle didn't fix that (of course...). I opened it and by touching the components with my finger I identified one analog IC on one voice board that was really hot. I replaced it and the synth came back to life - that chip suddenly shorted to ground. Chips can short at any time and when you less expect it even without any bad power event, so after that experience I'm not suprised to see issues like yours, especially on such old boards. This is why I suggest to start to concentrate around those 2 ICs, like others adviced in the comments BTW.
I havent seen your last video and have just started this one, but I thought I would mention that if you have a TL88 eprom programmer, you can use them to test the logic chips on this board. They test pretty much everything I can see on your card.
great video and great analysis!!! if you want to test logic chips, do yourself a favor and pick up a tl-866 pro 2, it will test all sorts of logic chips without you having to go through the trouble of figuring out what each does and finding a way to test. should save you a ton of time!
As soon as you brought that bread board into the picture, I was thinking oh, he's watching Ben Eater too. Hah! :) Sorry the card is giving you headaches, but this is great content. I enjoy the detective story! Sadly I think this is above my head so I don't think I can help much. I'd probably start swapping out all the chips, if you can get replacements that is. Hoping to see a working card down the road, good luck with it!
For the IBM MDA the clock rate is changed with HRES pin 2 on the LS153 on sheet 3 and that one is in turn coming from a d-type flip flop on sheet 7 which is in turn written to by an I/O write to 0x3b8 bit 0 . The 1.016MHz clock is probably the "TEST CLOCK" signal but I can not find where it is coming from.
Verify that the voltage when the card is in the pc is 5 volts. If it's not, the crystal won't vibrate at the correct frequency. If that's right, try replacing the chips that are going up to 60C. Those chips may be shorted out internally
Checked the voltage on the video card headers and it had a clean 5v. Low voltage on the test bench is the result of the jumper wires / connectors / headers. (Initially tried it with croc clips and there the voltage drop was even worse). So voltage is ok. Will look at those 2 chips getting hot. Thx.
@@RetroSpector78 I know this is going to sound tedious, and it is, but you really should check the voltage at every chip. Any one of the chips could pull the voltage down and you wouldn't know it checking at a single point. Or the chips could be introducing noise into the system and the frequency divider doesn't see all of the information it needs to see because it's masked by noise. I personally would check the voltage at those hot chips and go from there
TTL chips voltage of the clock circuit is ok. And on the output of the oscillator circuit I am getting a clean 16.257mhz so on the source everything seems to be fine
Hey, I might not be an electronics engineer, but for start i would suck out those hot chips out, and also, go around the board with scope to test that all the 74 logic chips levels are correct. If some of the chips inputs are bad, they could be driving down the logic rails to strange levels, and then strange things can happen.
Will look into the hot chips, verify them and replace them if needed. Regarding the logic levels. That would involve probing all inputs / outputs and seeing if they are in range (valid logic low and logic high signals) ?
Yeah, probe logic pins with scope and verify it is oscillating between logic low and logic high (i don't know if it is TTL, or CMOS, or what ever are normal logic levels for that card) On some types of logic/dram chips if logic levels are between high and low, very strange things happen. And it can happen if, for example, one out goes to multiple chips, and one of those chips has a bad input that is pulling it down from 5V (eg, TTL) to something around 3V (or even lower). Then all inputs on other chips are going to be at very strange state, and some oscillation could happen. Also, ISA bus provides +5V, -5V, +12V and -12V power rails. Some older DRAM chips, and some chips in general need multiple rails to run (eg OpAMPs need positive and negative to run). Check all rails, current draw on all rails while supplied with everything that ISA bus provides
Also, don't just go replacing chips. Those hot chips might just be fine. Things start to get very strange at some times. You can have chips warming up crazy, but they could still be perfectly working, the problem might be further down the line with some other chip, or component pulling a data bus low, and causing extra current draw. This is why checking if you have some kind of digital activity and valid logic levels makes sense as well. Check chip type and do best guess where would you like to see activity
My guess would be to replace the two hot chips and retest on bench to see if temp is the same. If this doesn't resolve the issue then I would say it's something to do with the PC's power supply. Try using another PSU.
The difference between the bench and pc could be that the voltage of the pc bus is dropping due to high current caused by a short in the card and could be what is causing the overheat. Try checking the voltage supply in the pc case and see if it is dropping. If so, you might need to check those heating chips and their surrounding.
The timings on a video card are highly dependent on the video mode. The mode will determine what video (dot clock) will be generated along with the hsync and vsync timings. The MDA card will have timings that are different than a CGA, EGA, or VGA card. It is very possible that the TSENG timings are different than the MDA timings for a number of reasons. Specifically, the mode defined when the board is powered on should mode 3, however that is a color mode. Therefore, a number of other modes may be defined that support 80 x 25 mono text. These might be mode 7 or mode F. Take a look at this link for the some details. cs.lmu.edu/~ray/notes/pcvideomodes/ At this link there are some details about the ultrapak board. It seem that it had MDA and HGC modes. en.wikipedia.org/wiki/Tseng_Labs It was capable of displaying graphics (640 x 200 and 320 x 200) in monochrome, so the timings must be different to support this. It also supported a 132 column mode. Check the hsync and vsync timings and compare them to the Hercules graphics controller timings. You may find that the timings match the HGC. search for tseng ultrapak for more information.
What is the minimal signals from the pc needed to get this working. could you make a circuit to emulate the pc to help diagnose the card fault.Or add something between the pc and the card that would allow you to break the connections one by one.
@@RetroSpector78 I was looking at it from the ISA connection. You said it drops down from 18mhz to 1mhz in the pc. if you disconnected the signals until it goes up to 18mhz it might give you a clue . I am just giving wild ideas and no expert.
couple of side notes - too thin wires from bench psu, to long inductance loop when extending osc probe leads like that, I'd suggest backing up character eprom in the first place (also checking if it is even alive). cmos chips are designed for low power dissipation, if they'r hot - something is wrong. Commands to switch modes are directed to CRTC by the CPU, that's why you are seeing different measures when connected to system bus, and flat on the bench. I would poke around with logic analyzer trying to understand what's going on (on the addres/databus of the system in the first place, that would give a clue - where, and probably why processor stops)
4 ปีที่แล้ว
I guess the voltage drop you mentioned near to 16:50 is because of the thin wires connecting the power supply and the video card. At that current it will show quite a resistance (thus voltage drop), I am not sure if it's the connector's fault you proposed (or at least not only) ... But anyway it would be interesting to measure voltage drop between the power supply and the connector on the wire itself ;)
Thanks for the videos. Also, FWIW, thank you for acknowledging the covid situation. Many other youtubers are just going on like nothing is happening and I think it's weird.
From page 23 of the IBM MDA Schematics 'Monochrome Display Adapter Sheet 1 of 10' the card needs +5V(B03 + B29) and +12V (B09). You only seem to supply +5V to one pin and one GND pin(your pins look wrong) for some reason? You need to connect to all the +5V pins and all(three) GND pins too (B01, B10, B31). Plus Reset(B02) needs to be held high in order to enable the video. The card appears to disable the video by turning off the logic that generates the video frequency. Apparently you can talk to the card via the parrallel printer port and read memory in the card to determine the status of the card ( I/0 Register Address '3BA ' )
Any warm circuit needs to be replaced. I've heard about a siberian way of repair : make the card powered up, let it warm, use a cooling spray, the temperature shock will destroy faulty components. Replace'em. Also, no TTL circuit could work reliably with 4,11 volts. make a better connection and check you get at least 4,75 volts on the furthest circuit from the power. Take and old dead mobo, get yourself an ISA connector, solder 1,5mm² wires to all power pins, and use it to power your card, you will at least get the correct voltages and amps.
I'm glad you are keeping going with this investigation how ever frustrating it is. The card just uses jellybean logic so it should respond to reason...eventually!
Whenever troubleshooting these vintage card's, it may not actually be the card's fault, it may be the PC's CMOS itself. Have you tried using this card in another system?
@@RetroSpector78thanks so very much for the timely response. I'm curious could the uvprom data be corrupted. For some odd reason I believe I've seen this type of fault in operation unless you've already singled that out and rewritten the eeprom.
You should try to install a replacement battery. I've had several PC components that had a battery that would just not run correctly if a good one wasn't present. Some circuitry only powered by it, or memory that will not keep its contents at all without the battery.
Crystal was fine .. had to replace it in the end cause I broke one leg off and was difficult to solder back on. But it is outputting the correct signal. Need to find out why the division by 9 is not working in the pc
@@RetroSpector78 Yeah the Chinese croc clips are notorious for being terribly thin or making bad contact even inside the clip, really got to measure them and trash the bad ones
I would start with socketing every IC if you're really wanting to get it working. but I would start by swapping out the 2 chips that are getting hot. that seems a bit high of a temp to me. I am not an expert but I would start there. Also in my time fixing junk I have come across components that only short when the circuit is live. so it can be a pain in the butt to find the issue. this could be what's happening to your CRT clock. just a guess. have you thought about bypassing the clocks and injecting your breadboard version to see if that changes the results?
Maybe one of the memory chips on the card could need replacing, check the flow of the wiring on the card, using a 'small amount of heat' may help. Edit: clock signal is derived by some component, so if that "said" component reports incorrect clocks, then it seems logical to replace it. Though 'what_chip'? allows this.
You happy that the voltage on the motherboard is stable and ok? Whilst you may have tried another card and it's good. The current pull from this card may be something that the M/B cannot handle. You checked the power lines into the GPU on the PC whilst operational and is it in line with the specs?
Isn't it an option to just swap the two chips that are getting too hot to touch and see what happens before spending ages on tracing lost Mhzs? Or are they not available anymore?
They are available (3 EUR a piece !!) , but didn’t really notice them getting hot until pretty late in the video footage. Will remove them and see if I can get some replacements. There were a bunch of other logic chips broken also, so in the heat of the moment you encounter lots of issues and try to fix / debug what you can. I’m sure if I ever get this working I can create a 3min video explaining people how I fixed it. But not there yet :)
Sir maybe there's a bad connection on the card edge connector and the clock signal of the pc interferes with the one on the videocard. It's a wild guess and I don't even know if it altogether possibile. Take care and thanks for all your work :)
I think some of this issues causing the excessive power consumption when just powered on the bench with just the 5V power supply is because there are other chips on there that are expecting the 12V, -5, and -12V rails to be present. This can cause latch up in the chips that are not getting correct power and can damage them. If you have TL866 or equivalent programmer they usually can be used to test the logic chips and some ram chips as well. I would check the pins on the chips that are getting hot with the scope to see if they are being properly driven to 0V and 5V and not some where in between. Have you verified the video card bios is ok?
Don't you have an identical working board? Something I guess it would help is to measure the frequencies and other parameters in a working board to compare.
At part 3 you probably will be able to build the same graphics card from scratch faster than repairing this one. ~2 Amps is surely not OK for such a device. I think that the connector itself might be even not designed to withstand such power. If you ever need to quickly find if something is overheating, Louis Rossmann once used an interesting trick. He just spilled alcohol on the board (best done immediately after disconnecting power), and overheated parts became dry much faster. And please don't get discouraged even if it doesn't work in the end. You've probably already gained a significant amount of knowledge that can be a big help in any future repairs.
Fun fact, the Dialog4 CNC controller on my FP2-NC uses the same Motorola 6845 graphics chip, along with 68k CPU. www.scnct.de/shop/images/product_images/popup_images/34_0.jpg
i think there is an burned part, just send 6v power input and see where temperature go up... inside of computer there are weak power rails so you dont see expected frequencies (and power on actual ics probably fall below 4.7v)
hot chips + high power consumption, seems very much like bad chips to me... and it could well be that, say, on lab bench, you test its power state, but it wouldn't get any signals, wouldn't show anything, but in a computer the computer tries to send data, which interacts with the bad chips, which mess it up?
for a second there I thought you meant things would turn back to normal in some months for the video card :P - what will get fixed first this video card or coronavirus?? that is the question :P
I am not much help on the troubleshooting side, but this might help you see what is going on a little more clearly: github.com/Upcycle-Electronics/ChipLabels I have most of the chips from this board already premade. The GIMP folder has the .xcf files with the labels premade as layers. There is also a folder with hundreds of premade labels in PDF. They are setup to print on 8.5 x 11in paper. Just set the scaling to this in your computer's print settings. Most labels have multiple versions for each part number. I also made the logic chips with 2 different widths. If you have the narrow 1/2 width end pins, I have specific labels for those too. I find it much easier to understand the circuit when I'm not trying to translate the schematic to decipher the pinout and then counting pins. I have the Motorola 6845 label aready made too :-) -Jake
If all the troubleshooting gets you nowhere I would say try replacing ALL the glue logic. At this point if you have a high current consumption and hot chips then you know it must be the board. That or the EEPROM is fried. Not sure how to test / fix that.
Maybe send to Eli over at Eli Tech. He's the only You Tuber i've seen that has replaced memory IC's, GPU's and traces on Video/GFX cards and actually fix them.
At the end you mention how its weird you cant get the same measurements on the character clock when your connected to the pc. Now I'm no expert but sounds like it could be a problem within the PC. Or perhaps the pc isn't feeding the card enough power because it sees that something is wrong with the card? But once again, I'm not a expert and this is just purely a guess. lol
Man best of luck with this thing.
I've never seen 60C on anything of that era unless it had an internal short. That might also explain the power consumption. Check those two hot chips.
Will do ! See you in part 3 ;)
Those two chips would also be my best guess. What's their type?
Those hot chips could also be a symptom of something different and not directly a failure of the chips themselves. If something is loading their output circuits down or driving the outputs from somewhere else, they will source or sink current they're not supposed to.
If they carry high frequency signals, bad capacitors could very well be a cause of excessive load on that circuit.
@@Stoney3K Admittedly, that could very well be the case. Therefore the type(s) of the ICs and the circuitry around would be interesting.
Stoney3K yep - but either way, I’d bet they’re involved in the problem, so that’s where I’d start.
While MDA has only one mode of operation - the text mode, where the character clock is the crystal clock devided by 9, the Hercules card has two modes of operations. In graphics mode, the character clock is the crystal clock *divided by 16*. And 16.257 MHz / 16 is 1.016 MHz. This means your card has switched to graphics mode because of some signals it received from the bus.
The fact that you can measure 1.8MHz on the bench and 1.0MHz in the computer actually is a *good* sign for the clocking circuit - it operates well in both text and graphics mode - but something on the card is misinterpreting the bus signals - the IBM PC/XT BIOS is unable to switch a working Hercules compatible card into graphics mode.
Yes..it had to be switched by a dos driver command . but default should be MDA mode..unless jumpered to HGA.
@@MessalineApghar The IBM PC is set to MDA using its dip switches. I haven't found the manual of the Tseng Labs UltraPak but the output should be limited to monochrome MDA (default) and hercules. There are 2 jumpers on the card but tried all combinations. But will keep on investigating. Will start with the 2 hot chips.
@@RetroSpector78 there were usually software drivers included with the boards as well to change modes. And not all mda graphics cards were hercules standard. Some did cga emulation instead or in additio .
@@RetroSpector78 beautiful work on the refurb btw.
I would suspect those two 74LS chips, if they are somewhere near a control line, they could be pulling some pin of the CRT controller low which causes the controller to switch to Hercules mode instead of MDA mode.
PLEASE! DONT GIVE UP!
DONT THROW THE CARD AWAY!
Have faith!
I tried to reverse engineer the card just from a chip list taken from your pictures on imgur. Most likely the 4 74LS258 chips are generating the address lines for the video RAM. The video RAM on your board is comprised of 8 RAM chips with 65536 entries of 1 bit each (64k x 1). This means the RAM chips need 16 address bits in total, split into 8 column and 8 row bits.
Your card has two sources of addresses for the video RAM: Either the address is taken from the CRTC (for screen refresh) or it is taken from the ISA bus (for processor access). This means that you have 32 address input signals (16 CRTC, 16 ISA), and 8 address output signals (RAM address pins A0..A7). Each address pin for the RAM has 4 different possible sources: It might be the row address from the CRTC, the column address from the CRTC, the row address from the ISA bus or the column address from the ISA bus. If my intuition is right, this selection is performed by the 74LS258 chips, which are "quadruple 2:1 multiplexers with tristate output". As these chips contain just 4 multiplexers, but the card needs 8 output bits, each two of them are grouped as 8-bit multiplexers. The selection process works in either of the following two ways:
a) two multiplexers select the RAM row address by receiving 8 address bits from the CRTC/graphics card logic and 8 address bits from the ISA bus, and the other two multiplexers select the RAM column address by receiving the respective other 8 bits. The select input in the multiplexes select whether you want the CRTC address or the ISA address to reach the RAM chip. The output enable input is either active for the row address multiplexers or the column address multiplexers, depending on which address half is required at the RAM chips.
b) two multiplexers deal with the ISA address and select either the row or the column part of it, the other two multiplexers deal with the CRTC address and chose the row or column part of that address. The select input input of the multiplexers is connected to a signal the determines whether a row or column address is required, and the output enable either activates the ISA address multiplexers or the CRTC address multiplexers.
In either way, only two of the multiplexers must ever be enabled at the same time, as the second level of the 4:1 demultiplexing is done by only ever enabling one of the two groups of 8 chips. As you have two very hot 74LS258 chips, it is quite likely that those two chips are driving the same address lines, and must not be enabled at the same time, because otherwise they drive against each other, causing excessive current consumption and heat generation.
I suggest that if doing the simple test suggested by other people, namely testing the output pins of the hot chips for shorts against ground (and +5V) doesn't result in any findings, you probe the output enable lines (pin 15) of all four chips. You should find two different signals (each at two of those chips), which are *never* low at the same time (output enable is active low). If the output enable waveforms are correct, it is quite likely that one of the two hot chips is driving its output even though it is requested to be in high impedance state (aka Z or tristate) and needs replacement. If the output enable waveforms are incorrect (both of the hot chips are indeed enabled at the same time), you would need to trace down where the output enable waveforms come from.
someone push that response to the top! Awesome analysis!
I now know that you are very knowledgeable and I understand even less about computers than I thought I did.
Just fascinating. I now want to learn more so thank you for that.
Hey there, been watching your series from some time now so I thought I'd give you a little help with my insight.
To me, the main steps of diagnosing over consumption are as follows :
1- Checking the power rails integrity : with a simple ohmmeter, pick your usual +5v/-5v/+12v/-12v on the main connector (as you did) and check that each (and I mean all of them) chips VCC are connected. Do the same with each GND.
2- Power it with a bench supply, preferably not more than 500 mA on the +5v rail. If current gets limited, check what is consuming that power. You can feel the heat from the components with the back of your hand (or even your face because it's even more sensitive... silly, I know, but it works).
3- Now you've identified something. Those two 74ALS258N sitting in the middle that are overheating seems to be linked to the excessive idle power consumption (2 Amps at idle is a lot). These are data bus multiplexers, and their outputs might be tied down to ground by what's after them, causing them to sink all the power through them.
4- With the power off, check again the outputs of the two chips and the GND. With an ohmmeter, then with the diode tester mode (you could find bad transistor junctions).
5- Then proceed to remove all socketed chips you have on the board. The idea is not only to find the culprit, also being able to reproduce the fault. With (almost) no circuits, you will get the big picture of whether the chips OR the board is at fault. Dirty traces under sockets can cause high impedance MOS inputs to go low or high.
(By the way, once you know the board's PCB is in good health (clean, no bad traces), soldering sockets for the chips is helpful)
My personal advice would be to remove all RAM chips, they can go bad for nothing much.
(Don't focus too much on the "low" 1 Mhz clock signal, it might be related to low VCC)
Once you've identified something wrong and eliminated it, repeat from step 2.
Then, once we have every chip present and not causing problems, we willl look for the proper signals (clocks).
I'd advice not plugging it in the PC until then.
Good luck with your fix, it takes time and patience but it is worth the happiness once you get it done :) so please don't give up.
Thx a lot for this elaborate comment ... will definitely try out your suggestions, provide feedback and hopefully see you in part 3 :)
Damn dude this is like an entire wikihow article with a step by step guide on how to fix this card!
Hehe .. sure looks like it :) but already tried a lot and it can get complex.... without schematics and a lack of some fundamentals and experience it can be tricky. Fixing a PC is 1 thing, drilling down to the chip or even logic gate level is another :) but as long as it stays fun I will continue. If all else fails I’ll run over it with my car and make a video on that :)
@@RetroSpector78 No you won't...
@@RetroSpector78 You're welcome.
Don't be discouraged by the apparent complexity, just check one thing at a time so you will be more efficient.
Read some datasheets (datasheetcatalog.com is nice for this era), and you will get a general idea of the inner workings. Experience comes this way.
I can still guide you through what you will find when troubleshooting.
But please... no car...
Is there a part 3 yet? I couldn't find it.
3:30 so why use a jumper wire? Rather than pulling the hook off the probe... Still this is beyond my skill set.
chazzy2501
Exactly!
The PC power bus is probably dropping lower than the bench power supply and the clock is being affected by that. Those chips at 60degC are DEFINITELY too hot and should be replaced.
Agreed
Yup
DAC chips will get that hot. Which is normal-ish. 10 watts isn't outlandish but is higher than I would expect. My genuine herc card used 8.8 watts IIRC. Clock won't be effected by vdroop much, or shouldn't be. It is being pulled into a mode that is not usable for post.
@@wishusknight3009 There are no DAC chips on that card, the chips that were hot appear to be 74 series logic chips next to the CRT controller. The closest thing to a spec that I could find says 4.5A is the limit of the ISA bus. I can't believe the actual connector is rated for that. I'll have to check my Hercules card, it would be surprised to see 8.8w.
@@eformance Makes sense. Shows what I know lol. I see some other people here have given some good explanations to the heat spots.
That power consumption is a clue, I don't think that the video card should take ~10w. You probably have some shorted parts on there and a "fingertip test" should be done to find any warm chips.
EFormance Engineering probably leaking gates to a rail
Totally agree. These cards take about 4-6W.
Just spray the card with alcohol. If it catches fire then there was a chip that was getting too hot.
First things first, the card isn't going to run on 4.1V even if it was working perfectly fine (although I've seen a Williams Joust run on 4.2V, but very unreliably)
Clock: There's a good chance that this is a multi standard card where the clock frequency depends on what video standard it emulates. It has 64k of RAM, so it's quite thinkable it can do MDA, Hercules and CGA, so that depends on how it's initialized which depends on what's on that ROM. The 6845 is really flexible. Go check HSync and VSync.
Hot Chips: If it's a 74LSxxx it shouldn't get hot. Some get warm. There are exceptions (like the 188). If it's a PAL or a GAL, these usually run hot. At least early ones like on this card.
Battery Acid: It's rather common to see chips that suffered battery corrosion run fine for a few hours or days and then fail.
Logic Levels: Check the voltages on the pins on the ICs to the left of the ROM. If there's any signal where low is > 0.3V or high < 3V, that chip is bad. On the right side of the ROM are lots of tri-state devices where these levels are actually OK.
Thx a lot for all the tips. Will look at all of them and I hope to see you in part 3
I have to admire your persistence, don't give up now!
Not an electrical engineer, but I would definitely check if those two 60C chips are working the way they should. There can be short in those two (closest experience I have is burned Arduino nano - atmega chip was very hot and never worked again after I accidentally shorted it).
Yeah will check those. Only decided to checked the temperature hours before editing the video :) sometimes you get tunnel-vision when debugging those things for too long.
It's incredible how many ICs have been literally crammed onto the board!
Let me say I like this sort of videos very much. I like your normal videos too, but it's a plesure to see you working as a detective, understanding the most elementary electronic part. Congrats.
Thx appreciate it ! Try to add some variety but most of the time its just whatever I fancy doing at a certain moment in time :) I try to complete these series in a timely fashion but also want to add some other videos in the mix. Cause I spent several days spending countless hours on this thing :) might not look like that in a 20min video.
18:16 maybe it can be handy: i was fixing bad amd radeon hd 6870. The card was on the bench - and I've checked resistances on each power rail, then I've "injected" proper/correct voltage from lab power supply on each power rail step by step to observe power consumption. And then I've decided to supply all voltages to the card, which was on bench. So I've put 12V on psu connectors, 12v on pci-e pin, and using step-down voltage converter I've put 3.3v on pci-e pin (made from 12v) So all voltages was present but I still not get high state on enable pin of main phase buck controller, so it doesn't start to produce 1v for gpu core and 1,59v for video ram. But when I've disconnected everything and I've inserted video card into computer pci-e slot, and start it up, then I've got high state on enable pin and then buck controller start working and it provides 1v for gpu and 1,59v for ram.
So the conclusion, which can help you, is that - there must be a signal from computer (via socket pin) to wake up video card. I understand that here you have retro architecture and power saving was different in that time, but maybe this is the case. Maybe you get 1,8MHz on the bench, because the card is NOT fully on. But as soon as you put it into computer, the video card receives a signal to start rest of the components (which all wasn't supplied on the bench) and maybe some of that components is bad and that bad chip is taking down the 1,8MHz clock signal.
Second thing. 16:55 I see 1,78 Amps on that videocard and it is too much i think. And 60 C degrees on any old/retro chip is abnormal. That hot chips explains amperage. If you can't touch it (because is too hot), then it is the clear clue, that it is short, or something around him is overloading the rail which that chip provides.Please check that hot chips datasheets, and measure resistance on all pins to ground and if (in the datasheet) that chips provide any output voltage, then lift that leg out from the board, or take out that chip and inject voltage from lab psu on that pad on motherboard side. To check if something on that rail is shorted or if short in on the chip side only (i suggest to remove the chip or lift the output pin before injecting voltage from psu, because sometimes reverse voltage from the other side to the output side, can burn the chip or inside diode)
Third thing: When you be able to discover the reasons of the hot chips and fix it, and if then the videocard will still produce black screen, then maybe you should check the video bios. is any eeprom chip there?
Sorry for my english, I could use wrong words, but I hope you've understood me.
greetings from poland, and thank you for part two :)
Thanks a lot for the comments.... I will check the 2 hot chips and see if I can replace them. Running it off the test bench was just more convenient to probe around, but I can imagine it is missing some vital signals at that point on the ISA bus. But it is a little bit more difficult to probe around with the video card in the PC. Should have some kind of ISA expansion board + cable to get it out of the PC.
I agree, chips running very hot mean they probably are shorted.
You shouldn't use "I've" before the verbs, just "I". For example it should be "I checked resistance", not "I've checked resistance". "I have"/"I've" is used is other situations. Other than that your English is pretty good :)
@@WTFBOOMDOOM Thank you. But wait a minute - I remember, form the high-school, that the was "I've,you've..." required in some situations in past time. Could you tell me, in simple way, when should I use it?
@@Radek__ That would be verb tense "past perfect". Google explains it much better than I could...
I'm the only one who didn't understand most of what's going on and still got hooked up? I really want to see this card working again and now I'm interested on learning a bit of basic electronics
That’s great to hear .... there was a lot to unpack, and I am far from a professional :) glad you liked it. Also a learning experience for me.
Well, I learn a LOT on those videos! I wish I knew more.
me too :)
Amazing, please keep this up! Excellent level of details without omitting even the basic stuff
Totally agree. Bens channel is great. Very well done.
I really love these kinds of repairs! This is what keeps me going further into depth when fixing these issues. Love it! 😍
Also keep up the great work!
Thanks ... I'm going to need it :)
Even though this is way beyond my skill level, it is fascinating. I hope you get the help you need so that this piece of technology doesn't go to waste. Cheers.
I was waiting a long time for this part 😭
Spoiler alert : there will be a third :)
Finally!
@@RetroSpector78 admit it, you got it working now :)
great video explanation of the clock divider circuit!
interesting! was the crytal still oscilating at 18MHz when you plugged it in?
I think maybe some of the +12 -12 -5V is disturbing either your mass or your 5V signal.
Try using standard PSU on the bench
both crystals were oscillating fine in the PC. I think the 18Mhz was used for the serial ports (this is one of those multi-functional cards including serial ports, a battery, and a header for a memory expansion.
good luck, seems like Cookie made some uberawesome suggestions!
I wouldn't trust the frequency measurement on your DMM. I'd probe it with the scope in both cases.
Watching your fault tracing of this card is quite inspiring! =) Good luck and stay safe in these times.
Hehe ... thx ... you too !
it looks like the hot ICs you pointed at are some 74LSxxx ICs. Those do not usually run at 60°C. Last time (a long time ago) when I fixed one of those cards they just took over 1A and not almost 2A like that one. But it was a bit more similar the other working card you showed.
any chance for a part 3? love the work you're doing
Ditto on the overheating TTL. Run the board on the bench with all 3 supply voltages and measure the current draw on all the rails with the 2 overheating chips in and out of the board. A current limiting power supply module is useful in this type of situation. A frequency counter is useful to measure the clock divider chain. There may be several faults here. Have fun!
Great Video! Thank you and hopefully you will be able to repair this nice card. ☺️. BR, peter
HGC info from this link:
en.wikipedia.org/wiki/Hercules_Graphics_Card
"The monitor is connected to the Hercules card through a DE9F connector using 5V TTL electrical signaling; the same as for an MDA card.[5][6] Nominally, the Hercules card provides a horizontal scanning frequency of 18.425 ±0.500 kHz, and 50 Hz vertical.[7]
However, the Hercules actually ran at two slightly different frequencies depending on whether it was running in Text or Graphics mode. This is a side effect of the CRT controller using 9-pixel, 14-line character cells when impersonating the MDA, but having to use 8-by-4 pixel blocks to turn memory bytes into 1-bit pixels in graphics mode, with neither the total pixel count in a horizontal period nor total number of scanlines matching up completely. As the card's onboard clock crystal was also slightly different, at 16.000 MHz vs the MDA's (and later, EGA's) actual 16.257 MHz, neither mode was an exact match for the original sync structure, with the exact figures being 18.141 kHz and 49.84 Hz in text mode, plus 18.519 kHz and 50.32 Hz in graphics. However both are close enough to MDA's own 18.432 kHz and 49.82 Hz timings to fit within the sync tolerance of and work reliably on a typical IBM monochrome monitor or compatible. This tolerance was taken advantage of by the various Hercules clone cards, which themselves often exhibited less than accurate timings versus the original.[8]"
have you got a eprom programer as you can plug in the chips and use a program to test them
I have and I will .... didn’t know it could do logic chips :) something for part 3 (all the socketed chips have been replaced and brand new). Will check the “hot” chips also.
@@RetroSpector78 Yes the common MiniPro TL866/TL866II EPROM programmers you can get from ebay, etc have the ability to test many 74 series logic chips and even static RAM chips. Sadly they aren't up to testing TMS4164 or TMS41256 DRAM chips. I'd also suggest dumping the EPROM(s) on the board as this will test them out if nothing else.
@@davefiddes You can build you own small microcontroller based DRAM testers.
60C on some TTL Chips (?). That would seem shorted to me. Maybe that's also explaining your high Amps and low voltage on the card, which in turn might be affecting your crystal?
Yeah that will be the next step replacing those. On the working ibm mda card I also get 1.2amps power consumption. Dunno what is considered “normal” for these cards. And low voltage is the headers / connectors I think. Cause in the pc it is getting a clean 5v.
Any chance you can check those larger specialized ICs? I had the honor of repairing an Atari ATW800 a while back which had video issues, while I suspected a number of TTL chips on the video board it was the SRAM (for the color lookup table) inside one of the DACs in the end.
Getting a TTL tester and checking the EPROM contents should be the first priority at this point tho.
EDIT: Another thing worth checking ... is the card a 4 layer one? Might be worth checking that there are no shorts in the inner layers - that's the sad fate an Amiga 3000T from a friend is currently suffering due to battery corrosion.
The hot chips might have shorts in them. I would test there.
60°C of heat on a standard TTL logic IC is not normal at all. Probably those 2 chips are internally shorted and cause a high power draw when you put the board in the ISA slot, this could affect the correct working of the entire card. I would start by replacing them.
I checked that these pins were electorally connected and as such it didn’t matter what pin I used ? In that case is it possible that one chip is on pin29 and would not affect pin 3 ?
@@RetroSpector78 Ignore that part of the ISA slot of my comment, I actually deleted it after re-watching your video, where you connected pin 1 and 3 to the power supply and measured the voltage on pin 29 and 31. This confirms that the two +5V lines are connected so powering the board from pin 3 or pin 29 shouldn't make any difference.
I suspect that those two ICs are internally shorted to ground. I remember when one chip died on my Roland Jupiter 6 analog synthesizer, I was playing some music on it and suddenly the synth went out. No magic smoke, strange noises and such, it just stopped producing any audio and all the lights on the panel went off. No response from the keys nor the controls on the panel and a power cycle didn't fix that (of course...). I opened it and by touching the components with my finger I identified one analog IC on one voice board that was really hot. I replaced it and the synth came back to life - that chip suddenly shorted to ground. Chips can short at any time and when you less expect it even without any bad power event, so after that experience I'm not suprised to see issues like yours, especially on such old boards. This is why I suggest to start to concentrate around those 2 ICs, like others adviced in the comments BTW.
Holy cow. About 90% of this video was over my head! Still very interesting though. I hope you are eventually able to get this card working.
I havent seen your last video and have just started this one, but I thought I would mention that if you have a TL88 eprom programmer, you can use them to test the logic chips on this board. They test pretty much everything I can see on your card.
What happens if you supply stable voltage from outside the PC? As in, not relying on just the power from the PSU of the computer?
great video and great analysis!!! if you want to test logic chips, do yourself a favor and pick up a tl-866 pro 2, it will test all sorts of logic chips without you having to go through the trouble of figuring out what each does and finding a way to test. should save you a ton of time!
Haha ... wanna hear something funny ... used mine to check the eeprom. Never realised you could check logic chips ;) thx for the tip !
thank you for posting. i'm rooting for your eventual success w/ this card. does a part 3 [hopefully successful] now exist ? take care & stay safe,
As soon as you brought that bread board into the picture, I was thinking oh, he's watching Ben Eater too. Hah! :)
Sorry the card is giving you headaches, but this is great content. I enjoy the detective story!
Sadly I think this is above my head so I don't think I can help much. I'd probably start swapping out all the chips, if you can get replacements that is. Hoping to see a working card down the road, good luck with it!
Thanks ... see you in part 3 :)
For the IBM MDA the clock rate is changed with HRES pin 2 on the LS153 on sheet 3 and that one is in turn coming from a d-type flip flop on sheet 7 which is in turn written to by an I/O write to 0x3b8 bit 0 . The 1.016MHz clock is probably the "TEST CLOCK" signal but I can not find where it is coming from.
Verify that the voltage when the card is in the pc is 5 volts. If it's not, the crystal won't vibrate at the correct frequency. If that's right, try replacing the chips that are going up to 60C. Those chips may be shorted out internally
Checked the voltage on the video card headers and it had a clean 5v. Low voltage on the test bench is the result of the jumper wires / connectors / headers. (Initially tried it with croc clips and there the voltage drop was even worse). So voltage is ok. Will look at those 2 chips getting hot. Thx.
@@RetroSpector78 I know this is going to sound tedious, and it is, but you really should check the voltage at every chip. Any one of the chips could pull the voltage down and you wouldn't know it checking at a single point. Or the chips could be introducing noise into the system and the frequency divider doesn't see all of the information it needs to see because it's masked by noise. I personally would check the voltage at those hot chips and go from there
Appreciate every tip ! Will try it out :) thx.
The voltage drop is in the thin cable you are using and not in the connection. Also 1.7A seems too high to me.
Loving the series, I think we can figure it out! Did you verify the voltages going to the osc crystal?
TTL chips voltage of the clock circuit is ok. And on the output of the oscillator circuit I am getting a clean 16.257mhz so on the source everything seems to be fine
Hey, I might not be an electronics engineer, but for start i would suck out those hot chips out, and also, go around the board with scope to test that all the 74 logic chips levels are correct.
If some of the chips inputs are bad, they could be driving down the logic rails to strange levels, and then strange things can happen.
Will look into the hot chips, verify them and replace them if needed. Regarding the logic levels. That would involve probing all inputs / outputs and seeing if they are in range (valid logic low and logic high signals) ?
Yeah, probe logic pins with scope and verify it is oscillating between logic low and logic high (i don't know if it is TTL, or CMOS, or what ever are normal logic levels for that card)
On some types of logic/dram chips if logic levels are between high and low, very strange things happen. And it can happen if, for example, one out goes to multiple chips, and one of those chips has a bad input that is pulling it down from 5V (eg, TTL) to something around 3V (or even lower).
Then all inputs on other chips are going to be at very strange state, and some oscillation could happen.
Also, ISA bus provides +5V, -5V, +12V and -12V power rails. Some older DRAM chips, and some chips in general need multiple rails to run (eg OpAMPs need positive and negative to run). Check all rails, current draw on all rails while supplied with everything that ISA bus provides
Also, don't just go replacing chips. Those hot chips might just be fine. Things start to get very strange at some times. You can have chips warming up crazy, but they could still be perfectly working, the problem might be further down the line with some other chip, or component pulling a data bus low, and causing extra current draw.
This is why checking if you have some kind of digital activity and valid logic levels makes sense as well.
Check chip type and do best guess where would you like to see activity
My guess would be to replace the two hot chips and retest on bench to see if temp is the same. If this doesn't resolve the issue then I would say it's something to do with the PC's power supply. Try using another PSU.
Tried multiple pc’s (different XT systems) but will definitely check the 2 hot chips
The difference between the bench and pc could be that the voltage of the pc bus is dropping due to high current caused by a short in the card and could be what is causing the overheat. Try checking the voltage supply in the pc case and see if it is dropping. If so, you might need to check those heating chips and their surrounding.
The timings on a video card are highly dependent on the video mode. The mode will determine what video (dot clock) will be generated along with the hsync and vsync timings. The MDA card will have timings that are different than a CGA, EGA, or VGA card. It is very possible that the TSENG timings are different than the MDA timings for a number of reasons. Specifically, the mode defined when the board is powered on should mode 3, however that is a color mode. Therefore, a number of other modes may be defined that support 80 x 25 mono text. These might be mode 7 or mode F. Take a look at this link for the some details. cs.lmu.edu/~ray/notes/pcvideomodes/
At this link there are some details about the ultrapak board. It seem that it had MDA and HGC modes.
en.wikipedia.org/wiki/Tseng_Labs
It was capable of displaying graphics (640 x 200 and 320 x 200) in monochrome, so the timings must be different to support this. It also supported a 132 column mode.
Check the hsync and vsync timings and compare them to the Hercules graphics controller timings. You may find that the timings match the HGC.
search for tseng ultrapak for more information.
What is the minimal signals from the pc needed to get this working. could you make a circuit to emulate the pc to help diagnose the card fault.Or add something between the pc and the card that would allow you to break the connections one by one.
Trying to reverse engineer it bit by bit. But hard for a rookie like myself without schematics. Need to brush up my electrical engineering skills :)
@@RetroSpector78 I was looking at it from the ISA connection. You said it drops down from 18mhz to 1mhz in the pc. if you disconnected the signals until it goes up to 18mhz it might give you a clue . I am just giving wild ideas and no expert.
@@RetroSpector78 if you give it all the voltages present on the ISA slot does it give you any picture ?
couple of side notes - too thin wires from bench psu, to long inductance loop when extending osc probe leads like that, I'd suggest backing up character eprom in the first place (also checking if it is even alive). cmos chips are designed for low power dissipation, if they'r hot - something is wrong. Commands to switch modes are directed to CRTC by the CPU, that's why you are seeing different measures when connected to system bus, and flat on the bench. I would poke around with logic analyzer trying to understand what's going on (on the addres/databus of the system in the first place, that would give a clue - where, and probably why processor stops)
I guess the voltage drop you mentioned near to 16:50 is because of the thin wires connecting the power supply and the video card. At that current it will show quite a resistance (thus voltage drop), I am not sure if it's the connector's fault you proposed (or at least not only) ... But anyway it would be interesting to measure voltage drop between the power supply and the connector on the wire itself ;)
I cant wait for part 3
Thanks for the videos. Also, FWIW, thank you for acknowledging the covid situation. Many other youtubers are just going on like nothing is happening and I think it's weird.
I'd put my money on those extra-toasty chips.
Or something what is driving it.
From page 23 of the IBM MDA Schematics 'Monochrome Display Adapter Sheet 1 of 10' the card needs +5V(B03 + B29) and +12V (B09). You only seem to supply +5V to one pin and one GND pin(your pins look wrong) for some reason? You need to connect to all the +5V pins and all(three) GND pins too (B01, B10, B31). Plus Reset(B02) needs to be held high in order to enable the video. The card appears to disable the video by turning off the logic that generates the video frequency. Apparently you can talk to the card via the parrallel printer port and read memory in the card to determine the status of the card ( I/0 Register
Address '3BA ' )
Any warm circuit needs to be replaced. I've heard about a siberian way of repair : make the card powered up, let it warm, use a cooling spray, the temperature shock will destroy faulty components. Replace'em. Also, no TTL circuit could work reliably with 4,11 volts. make a better connection and check you get at least 4,75 volts on the furthest circuit from the power. Take and old dead mobo, get yourself an ISA connector, solder 1,5mm² wires to all power pins, and use it to power your card, you will at least get the correct voltages and amps.
I'm glad you are keeping going with this investigation how ever frustrating it is. The card just uses jellybean logic so it should respond to reason...eventually!
Whenever troubleshooting these vintage card's, it may not actually be the card's fault, it may be the PC's CMOS itself. Have you tried using this card in another system?
Yep... multiple other pc’s. Issue is definitely somewhere on the card
@@RetroSpector78thanks so very much for the timely response.
I'm curious could the uvprom data be corrupted. For some odd reason I believe I've seen this type of fault in operation unless you've already singled that out and rewritten the eeprom.
You should try to install a replacement battery. I've had several PC components that had a battery that would just not run correctly if a good one wasn't present. Some circuitry only powered by it, or memory that will not keep its contents at all without the battery.
Tried it in the beginning but didn’t notice a difference. But would not hurt to try again. Thx !
This is so boring to me, but I'm really excited to see what happens next!
Glad to hear you are bored and excited all at once :)
is the crystal putting out the right frequency? I know its not really common, but they can go bad
Outgoing frequency from the clock circuit is fine. Both crystals (18mhz for the serial ports I think, and 16Mhz for the clock signal) seem to be fine.
Have you checked the crystal? I do remember people though very rarely having to change them other than that something has to be bad.
Crystal was fine .. had to replace it in the end cause I broke one leg off and was difficult to solder back on. But it is outputting the correct signal. Need to find out why the division by 9 is not working in the pc
I have more for you :) Have you tried Vogons or VCFED?
Sorry ... travel restrictions. Only essential travel permitted :)
@@RetroSpector78 When it's over :)
Voltage drop isn't across the connector but your super thin wires...
Also possible yes :) was even worse with my chinese croc clips :) never knew that made such a difference. And didn’t realise it would draw 1.8amps.
@@RetroSpector78 Yeah the Chinese croc clips are notorious for being terribly thin or making bad contact even inside the clip, really got to measure them and trash the bad ones
I would start with socketing every IC if you're really wanting to get it working. but I would start by swapping out the 2 chips that are getting hot. that seems a bit high of a temp to me. I am not an expert but I would start there. Also in my time fixing junk I have come across components that only short when the circuit is live. so it can be a pain in the butt to find the issue. this could be what's happening to your CRT clock. just a guess. have you thought about bypassing the clocks and injecting your breadboard version to see if that changes the results?
Maybe one of the memory chips on the card could need replacing, check the flow of the wiring on the card, using a 'small amount of heat' may help. Edit: clock signal is derived by some component, so if that "said" component reports incorrect clocks, then it seems logical to replace it. Though 'what_chip'? allows this.
Removed all of the memory chips and put them on a working mda card. (And vice versa) Worked fine so I guess memory chips are good
You happy that the voltage on the motherboard is stable and ok? Whilst you may have tried another card and it's good. The current pull from this card may be something that the M/B cannot handle. You checked the power lines into the GPU on the PC whilst operational and is it in line with the specs?
Great series, don't give up!
Thx ... appreciate it !
Isn't it an option to just swap the two chips that are getting too hot to touch and see what happens before spending ages on tracing lost Mhzs? Or are they not available anymore?
They are available (3 EUR a piece !!) , but didn’t really notice them getting hot until pretty late in the video footage. Will remove them and see if I can get some replacements. There were a bunch of other logic chips broken also, so in the heat of the moment you encounter lots of issues and try to fix / debug what you can. I’m sure if I ever get this working I can create a 3min video explaining people how I fixed it. But not there yet :)
I really hope you continue your quest in fixing this. This is awsome learning stuff. Thanks for great YT content. :-)
Don't give up! The solution will be very interesting!
Sir maybe there's a bad connection on the card edge connector and the clock signal of the pc interferes with the one on the videocard. It's a wild guess and I don't even know if it altogether possibile. Take care and thanks for all your work :)
Issue occurs on multiple pc’s. So has to be an issue with the videocard itself. You also take care and thx for the comment !
I understand and sorry for my crappy english, I was talking about the video card edge connector not the expansion slot on the various pcs.
I think some of this issues causing the excessive power consumption when just powered on the bench with just the 5V power supply is because there are other chips on there that are expecting the 12V, -5, and -12V rails to be present. This can cause latch up in the chips that are not getting correct power and can damage them. If you have TL866 or equivalent programmer they usually can be used to test the logic chips and some ram chips as well. I would check the pins on the chips that are getting hot with the scope to see if they are being properly driven to 0V and 5V and not some where in between. Have you verified the video card bios is ok?
Don't you have an identical working board? Something I guess it would help is to measure the frequencies and other parameters in a working board to compare.
Nothing to compare unfortunately :(
@@RetroSpector78 Maybe if you ask, some good soul could donate one. I would do it if I have one and if I live closer. I'm from Brazil.
At part 3 you probably will be able to build the same graphics card from scratch faster than repairing this one.
~2 Amps is surely not OK for such a device. I think that the connector itself might be even not designed to withstand such power.
If you ever need to quickly find if something is overheating, Louis Rossmann once used an interesting trick. He just spilled alcohol on the board (best done immediately after disconnecting power), and overheated parts became dry much faster.
And please don't get discouraged even if it doesn't work in the end. You've probably already gained a significant amount of knowledge that can be a big help in any future repairs.
Haha :) yeah it has been a nice learning process .... learning a lot from your comments also. I will prevail..... someday :)
In the last scope picture, the yellow trace is between 0 and 4v should be 0 and 5v. Check the power supply, can be a bad cap.
Don't give up.. solution will be amazing 👍
Thx ! Will try .... but will probably do another video in between to keep my sanity :)
Fun fact, the Dialog4 CNC controller on my FP2-NC uses the same Motorola 6845 graphics chip, along with 68k CPU.
www.scnct.de/shop/images/product_images/popup_images/34_0.jpg
Yeah pretty versatile chip ... have it on a bunch of different mda and cga cards here.
Could there be a problem with the computer expansion slot?
No have the issue on all computers I tried it on.
Part 3?
Or is it already burried in the backyard and we shan't ever mention it again?
i think there is an burned part, just send 6v power input and see where temperature go up... inside of computer there are weak power rails so you dont see expected frequencies (and power on actual ics probably fall below 4.7v)
You seem to be making good progress, I have no doubt you'll figure it out. :)
hot chips + high power consumption, seems very much like bad chips to me...
and it could well be that, say, on lab bench, you test its power state, but it wouldn't get any signals, wouldn't show anything, but in a computer the computer tries to send data, which interacts with the bad chips, which mess it up?
unreal piece of work! Was this graphics card really worth it?
never saw an oscope probe where the hook cap couldn't be pulled off to reveal a pin probe
for a second there I thought you meant things would turn back to normal in some months for the video card :P - what will get fixed first this video card or coronavirus?? that is the question :P
I am not much help on the troubleshooting side, but this might help you see what is going on a little more clearly:
github.com/Upcycle-Electronics/ChipLabels
I have most of the chips from this board already premade. The GIMP folder has the .xcf files with the labels premade as layers. There is also a folder with hundreds of premade labels in PDF. They are setup to print on 8.5 x 11in paper. Just set the scaling to this in your computer's print settings.
Most labels have multiple versions for each part number. I also made the logic chips with 2 different widths. If you have the narrow 1/2 width end pins, I have specific labels for those too.
I find it much easier to understand the circuit when I'm not trying to translate the schematic to decipher the pinout and then counting pins.
I have the Motorola 6845 label aready made too :-)
-Jake
hehe ... thanks for the link ... looks nice ... will take a look.
If all the troubleshooting gets you nowhere I would say try replacing ALL the glue logic. At this point if you have a high current consumption and hot chips then you know it must be the board. That or the EEPROM is fried. Not sure how to test / fix that.
So that was YOUR question then? It ended up on Hot Network Questions list!
What question exactly ?
Chips getting to 60 degrees C?
Is the crystal on the card still ok?
New guy here.... 👍😎
I'm ON IT!
One question; what's a video card? 😉
Thumbs up and SUBBED!
Much TechnO❤from So. Ca. 🇺🇸
Large hammer.... very satisfying. I won't tell you how I know this..
hehe ... last resort.
Had you tried turning off and on again? XD
Feed the clock directly from arudino and see that rest of the circuit is fine.
Maybe send to Eli over at Eli Tech. He's the only You Tuber i've seen that has replaced memory IC's, GPU's and traces on Video/GFX cards and actually fix them.
there is a short somewhere, thats why the chip become so hot
At the end you mention how its weird you cant get the same measurements on the character clock when your connected to the pc. Now I'm no expert but sounds like it could be a problem within the PC. Or perhaps the pc isn't feeding the card enough power because it sees that something is wrong with the card? But once again, I'm not a expert and this is just purely a guess. lol
Insane keep going and good luck