I solder 3 oscillators back, but if the problem is in PAL, there is nothing i can do right?
It may be possible to infer correct logic input and output of the PAL by monitoring the memory read /write results, and watching with a logic probe attached, but it certainly wouldnt be fun.
Electronically, a PAL is exactly what the name implies, a programmable logic chip. It contains a program that tells it what combination of output logic bits to raise high based on a specific combination of input logic bits. They are generally faster than other kinds of programmable chip.
There are tools to 'dump' the program from working PALs using inexpensive microcontrollers/SBCs, but a broken one will give an invalid read.
It might still be a starting point though.
If you manage to get the logic table for that chip, it can be replaced with and equivalent logic array or suitable replacement programmable logic chip.
It's getting that table for a mystery chip that's the trick.
That's assuming the PAL is bad though.
butjer1010 wrote on 2025-07-20, 17:49:First 4 switches gives me picture (with Hsync broken as You stated) in combination of : 0-0-0-0, 0-0-0-1, 0-1-0-0, 1-1-1-0, 1-0-0-1 and 1-1-0-0. In other combination, there is a blank screen.
Oh, that's interesting. With an XT BIOS, the 6 combinations for "EGA primary" should result in a non-blank scereen. I don't see that those combinations perfectly fit the pattern, but assuming that 0 is binary 0 and 1 is binary 1, and the most significant bit is on the right, your "working combinations" all end up on valid EGA "combination numbers", in fact they are 0, 8, 2, 7, 9 and 3. Ordered: 0,2,3,7,8,9. This mostly makes sense, as all these numbers are numbers for "color monitor connected to the EGA card". There are two further combinations that also indicate "color monitor connected to the EGA card", which are 1-0-0-0 and 0-1-1-0. I suppose they would output a working picture as well, and you missed those combination for some reason (maybe bad contacts in the switches broke them?). The combinations for "monochrome monitor connected to EGA card" should be far more out of sync (as the EGA card generates an 18kHz image, but you monitor is designed for either 15kHz or 21kHz), and they look "blank", because what would be gray, amber or green on an MDA monitor will be displayed as a very faint green on EGA monitors. There is no point (except for curiosity) in trying those combinations again, especially as they generate a timing that's not even supposed to work on your monitor.
Am I correct in assuming that 1-1-0-0 and 1-0-0-1 look more out-of-sync than the other combinations? These two switch settings select 350-line 21 kHz text modes, while the other 6 settings select 200-line 15.6kHz text modes. If that is the case, the first four switches are decoded, and you give them in exactly the reverse sequence as I wrote them down in Re: EGA Graphics card beeps .
butjer1010 wrote on 2025-07-20, 18:15:Need to desolder oscillator 16.257, and maybe 25.000 also, well see.... This is the trace from pin 8 of ega connector. It goes to resistor 330 first leg, first leg have resistance of 330 with 2nd leg of this resistor array, then it goes to this green capacitor, and trace from positive leg of capacitor goes under oscillator....
EDIT: Of course it went under all 3 oscillators, and it ends on upper 3rd from right leg on PAL16L8!
Hmm, I don't like that. And it's kind of surprising. I wouldn't expect a PAL to be strong enough to drive a long monitor cable. Most EGA cards use some kind of buffer chip (like the 74LS240, 74LS244 or similar) to drive the monitor signals, especially more timing critical signals like HSYNC. OTOH, I never built an EGA card myself, and the manufacturers of those card did, so maybe driving HSYNC directly from a PAL is fine.
butjer1010 wrote on 2025-07-20, 18:59:I solder 3 oscillators back, but if the problem is in PAL, there is nothing i can do right?
I wonder what the purpose of that PAL is. I looked at the schematics of the original IBM EGA. That card had a dedicated latch for the "miscellaneous output register" that contains 8 operation control bits, two of them setting the sync polarity. The IBM EGA then had two dedicated XOR gates to invert the H-SYNC and the V-SYNC signal if the latched control bit in the miscellaneous output register required to do so. A PAL is a good way to integrate like 3 or 4 seperate 74-series chip into one custom logic chip, and the XOR gates for the SYNC signals would be a good candidate to get integrated into a PAL. On the other hand, while the IBM EGA had dedicated componets for the miscellaneous output register, the CRT, the attribute controller, the graphics controller (split into two halves!) and the memory timing sequencer, the PEGA2A chip integrates all those components, so there is no reason the PEGA2A does not already include the XOR gates as well.
Another possibly interesting clue is the feature connector. Contrary to my belief, the EGA feature connector and the VGA feature connector work quite differently. The 8 pins on the EGA feature connector contain the 8 signals sent to the monitor (6 color bits, H-SYNC, V-SYNC) as they are produced by the EGA card. The feature connector also has 8 output pins that allow the feature connector to control the signals that are sent to the monitor. In case there is no hardware connected to the feature connector that modifies the signals, the feature connector can be bypassed. On the original IBM EGA, this is implemented using the 74LS244 U36 that forwards the signals from the EGA video generation directly to the monitor. There is a control pin to turn off the forwarding in case you want the feature connector to inject a different signal. That signal is also available on the feature connector, so hardware connected to the feature connector knows whether it should output its own monitor signal or the EGA operates in "feature connector bypass mode". While I don't know why it would make sense to do so, the purpose of the PAL on your card might be just the controlled bypassing of the feature connector. The feature connector is the blue 32-pin connector next to that PAL.
The pin of that PAL that is connected to the HSYNC output is pin 13. This makes sense, because pins 12..19 are the 8 tristate-capable output pins of a PAL16L8, so that actually is (might be) an output pin. I expect that pin is also connected to the HSYNC output of the feature connector (that is the signal that is sent from the feature connector to the monitor), which is pin 30. On the other hand, the HSYNC signal from the EGA core that is sent to the feature connector is present on 24. If the purpose of the PAL is a controlled bypass of the feature connector, both pin 30 and pin 24 are likely connected to the PAL. As pin 24 is an input to the PAL, I expect it to be on one of the dedicated input pins, that is pins 1 to 9 or pin 11. You don't have to desolder anything to verify that if you have a continuity tester, for example as part of a multimeter.
As a shot in the dark, you might try just connecting pin 24 of the feature connector to pin 30 of the feature connector, and see whether this resolves your issue, as this removes the dependency on the feature connector bypassing logic.
mkarcher wrote on 2025-07-20, 20:41:Oh, that's interesting. With an XT BIOS, the 6 combinations for "EGA primary" should result in a non-blank scereen. I don't see […]
butjer1010 wrote on 2025-07-20, 17:49:First 4 switches gives me picture (with Hsync broken as You stated) in combination of : 0-0-0-0, 0-0-0-1, 0-1-0-0, 1-1-1-0, 1-0-0-1 and 1-1-0-0. In other combination, there is a blank screen.
Oh, that's interesting. With an XT BIOS, the 6 combinations for "EGA primary" should result in a non-blank scereen. I don't see that those combinations perfectly fit the pattern, but assuming that 0 is binary 0 and 1 is binary 1, and the most significant bit is on the right, your "working combinations" all end up on valid EGA "combination numbers", in fact they are 0, 8, 2, 7, 9 and 3. Ordered: 0,2,3,7,8,9. This mostly makes sense, as all these numbers are numbers for "color monitor connected to the EGA card". There are two further combinations that also indicate "color monitor connected to the EGA card", which are 1-0-0-0 and 0-1-1-0. I suppose they would output a working picture as well, and you missed those combination for some reason (maybe bad contacts in the switches broke them?). The combinations for "monochrome monitor connected to EGA card" should be far more out of sync (as the EGA card generates an 18kHz image, but you monitor is designed for either 15kHz or 21kHz), and they look "blank", because what would be gray, amber or green on an MDA monitor will be displayed as a very faint green on EGA monitors. There is no point (except for curiosity) in trying those combinations again, especially as they generate a timing that's not even supposed to work on your monitor.
Am I correct in assuming that 1-1-0-0 and 1-0-0-1 look more out-of-sync than the other combinations? These two switch settings select 350-line 21 kHz text modes, while the other 6 settings select 200-line 15.6kHz text modes. If that is the case, the first four switches are decoded, and you give them in exactly the reverse sequence as I wrote them down in Re: EGA Graphics card beeps .
butjer1010 wrote on 2025-07-20, 18:15:Need to desolder oscillator 16.257, and maybe 25.000 also, well see.... This is the trace from pin 8 of ega connector. It goes to resistor 330 first leg, first leg have resistance of 330 with 2nd leg of this resistor array, then it goes to this green capacitor, and trace from positive leg of capacitor goes under oscillator....
EDIT: Of course it went under all 3 oscillators, and it ends on upper 3rd from right leg on PAL16L8!Hmm, I don't like that. And it's kind of surprising. I wouldn't expect a PAL to be strong enough to drive a long monitor cable. Most EGA cards use some kind of buffer chip (like the 74LS240, 74LS244 or similar) to drive the monitor signals, especially more timing critical signals like HSYNC. OTOH, I never built an EGA card myself, and the manufacturers of those card did, so maybe driving HSYNC directly from a PAL is fine.
butjer1010 wrote on 2025-07-20, 18:59:I solder 3 oscillators back, but if the problem is in PAL, there is nothing i can do right?
I wonder what the purpose of that PAL is. I looked at the schematics of the original IBM EGA. That card had a dedicated latch for the "miscellaneous output register" that contains 8 operation control bits, two of them setting the sync polarity. The IBM EGA then had two dedicated XOR gates to invert the H-SYNC and the V-SYNC signal if the latched control bit in the miscellaneous output register required to do so. A PAL is a good way to integrate like 3 or 4 seperate 74-series chip into one custom logic chip, and the XOR gates for the SYNC signals would be a good candidate to get integrated into a PAL. On the other hand, while the IBM EGA had dedicated componets for the miscellaneous output register, the CRT, the attribute controller, the graphics controller (split into two halves!) and the memory timing sequencer, the PEGA2A chip integrates all those components, so there is no reason the PEGA2A does not already include the XOR gates as well.
Another possibly interesting clue is the feature connector. Contrary to my belief, the EGA feature connector and the VGA feature connector work quite differently. The 8 pins on the EGA feature connector contain the 8 signals sent to the monitor (6 color bits, H-SYNC, V-SYNC) as they are produced by the EGA card. The feature connector also has 8 output pins that allow the feature connector to control the signals that are sent to the monitor. In case there is no hardware connected to the feature connector that modifies the signals, the feature connector can be bypassed. On the original IBM EGA, this is implemented using the 74LS244 U36 that forwards the signals from the EGA video generation directly to the monitor. There is a control pin to turn off the forwarding in case you want the feature connector to inject a different signal. That signal is also available on the feature connector, so hardware connected to the feature connector knows whether it should output its own monitor signal or the EGA operates in "feature connector bypass mode". While I don't know why it would make sense to do so, the purpose of the PAL on your card might be just the controlled bypassing of the feature connector. The feature connector is the blue 32-pin connector next to that PAL.
The pin of that PAL that is connected to the HSYNC output is pin 13. This makes sense, because pins 12..19 are the 8 tristate-capable output pins of a PAL16L8, so that actually is (might be) an output pin. I expect that pin is also connected to the HSYNC output of the feature connector (that is the signal that is sent from the feature connector to the monitor), which is pin 30. On the other hand, the HSYNC signal from the EGA core that is sent to the feature connector is present on 24. If the purpose of the PAL is a controlled bypass of the feature connector, both pin 30 and pin 24 are likely connected to the PAL. As pin 24 is an input to the PAL, I expect it to be on one of the dedicated input pins, that is pins 1 to 9 or pin 11. You don't have to desolder anything to verify that if you have a continuity tester, for example as part of a multimeter.
As a shot in the dark, you might try just connecting pin 24 of the feature connector to pin 30 of the feature connector, and see whether this resolves your issue, as this removes the dependency on the feature connector bypassing logic.
Yes, I did opposite from You, because I marked 1 as ON, and 0 as OFF on the DIP SW 😀 So what do You think, 1-0-0-1 (if 0 is ON than 0-1-1-0 as You wrote earlier) is EGA card on EGA monitor?
Also yes, that pin from PAL goes to feature connector (see picture) pin 30. I will solder wire from 24 to 30 and let You know.
So the positions 5-8 on dip sw are irrelevant for now? I don't need to change them in any way, leave them on OFF?
butjer1010 wrote on Yesterday, 05:50:Yes, I did opposite from You, because I marked 1 as ON, and 0 as OFF on the DIP SW 😀 So what do You think, 1-0-0-1 (if 0 is ON than 0-1-1-0 as You wrote earlier) is EGA card on EGA monitor?
Also yes, that pin from PAL goes to feature connector (see picture) pin 30. I will solder wire from 24 to 30 and let You know.
So the positions 5-8 on dip sw are irrelevant for now? I don't need to change them in any way, leave them on OFF?
I didn't mean you swapped 0 and 1, so this card assigns ON and OFF the other way. I meant you wrote the in the opposite order, so in your list, 0-1-0-0 is the combination I labelled 0-0-1-0. Indeed, I wouldn't start worrying about DIP switches 5 to 8 until you get a picture that is in sync.
Having a connection from pin 13 (see footnote) of the PAL to pin 30 of the feature connector confirms that one of the duties of the PAL is likely to forward the hsync signal from the EGA core to the monitor if the feature connector does not supply a modified hsync signal, so I still have hope that the wire can work around the issue, if the PAL stopped working correctly.
1: you start counting in the lower left corner. That's pin 1, then you count along the lower edge until 10, which is at the lower right corner. Then you continue at the upper right corner with the pin directly opposite to pin 10 and call it pin 11. You go on counting along the other edge to the left until pin 20, which is opposite to pin 1.
mkarcher wrote on Yesterday, 06:18:I didn't mean you swapped 0 and 1, so this card assigns ON and OFF the other way. I meant you wrote the in the opposite order, s […]
butjer1010 wrote on Yesterday, 05:50:Yes, I did opposite from You, because I marked 1 as ON, and 0 as OFF on the DIP SW 😀 So what do You think, 1-0-0-1 (if 0 is ON than 0-1-1-0 as You wrote earlier) is EGA card on EGA monitor?
Also yes, that pin from PAL goes to feature connector (see picture) pin 30. I will solder wire from 24 to 30 and let You know.
So the positions 5-8 on dip sw are irrelevant for now? I don't need to change them in any way, leave them on OFF?I didn't mean you swapped 0 and 1, so this card assigns ON and OFF the other way. I meant you wrote the in the opposite order, so in your list, 0-1-0-0 is the combination I labelled 0-0-1-0. Indeed, I wouldn't start worrying about DIP switches 5 to 8 until you get a picture that is in sync.
Having a connection from pin 13 (see footnote) of the PAL to pin 30 of the feature connector confirms that one of the duties of the PAL is likely to forward the hsync signal from the EGA core to the monitor if the feature connector does not supply a modified hsync signal, so I still have hope that the wire can work around the issue, if the PAL stopped working correctly.
1: you start counting in the lower left corner. That's pin 1, then you count along the lower edge until 10, which is at the lower right corner. Then you continue at the upper right corner with the pin directly opposite to pin 10 and call it pin 11. You go on counting along the other edge to the left until pin 20, which is opposite to pin 1.
Ok, got that with 0-0-1-0 😀
Got that with counting pins on PAL also, pin13 is indeed this pin that is connected with pin 8 on ega connector. Feature connector has labeled numbers, so it is not difficult to "guess" pin 24 😀
Need to fix the EGA connector also, because when i did hold the card in my hand, i broke brass pins from pin 6 to pin 9, and i need to fix that first 🙁
I tried to connect 24th and 30th pin of feature connector, but no change, still is out of Hsync. Best picture i got is when only pin2 is on, and 1,3 and 4 are off. I can see all the letters, flying letters are a lot slower than on other combinations.
I didn't try to remove U32 and 34 as You suggest some posts ago (You thought that the Hsync is connected to them). Even though all the LS's past the test on my T48, this one at U17 was definitely faulty. Card stopped beeping and it gave me the picture. Broken, but still i have a picture now. If there are another suspects, please let me know, i will try to exchange them, if i have a spare one (full box of LS's).
I have an idea. Maybe the problem is not with the card but the monitor? For example I have a green Philips BM7513 that works great for 50Hz MDA/Hercules due to long persistence. I have both Hercules clones and CGA and EGA cards that can emulate MDA/Herc. However the clock on CGA and EGA is not exactly the same as true MDA/Herc and this monitor has very narrow window in which it'll accept the HSYNC. Eventually I found settings that make stable picture with all cards, both when cold and hot (that matters too). The down side is the picture is not always nicely centered but I had to tweak all the pots to get there.
Point is, lets try forcing CGA mode at 15kHz and see if that is stable. Maybe the problem is only with 21k EGA mode. I think running "mode CO40" in DOS will force CGA text mode? Not nice looking but it should at least be stable. If not try the opposite, forcing EGA mode with "mode CO80". If at least one of these is stable then the card can be somewhat tested for correct operation of the chips. The other mode might require tweaking some pots, like my monitor does (some of which might be inside the monitor and not accesible even from the back without taking the cover off).
Deunan wrote on Yesterday, 09:55:I have an idea. Maybe the problem is not with the card but the monitor? For example I have a green Philips BM7513 that works great for 50Hz MDA/Hercules due to long persistence. I have both Hercules clones and CGA and EGA cards that can emulate MDA/Herc. However the clock on CGA and EGA is not exactly the same as true MDA/Herc and this monitor has very narrow window in which it'll accept the HSYNC. Eventually I found settings that make stable picture with all cards, both when cold and hot (that matters too). The down side is the picture is not always nicely centered but I had to tweak all the pots to get there.
Point is, lets try forcing CGA mode at 15kHz and see if that is stable. Maybe the problem is only with 21k EGA mode. I think running "mode CO40" in DOS will force CGA text mode? Not nice looking but it should at least be stable. If not try the opposite, forcing EGA mode with "mode CO80". If at least one of these is stable then the card can be somewhat tested for correct operation of the chips. The other mode might require tweaking some pots, like my monitor does (some of which might be inside the monitor and not accesible even from the back without taking the cover off).
I have another EGA (oti037) card that work normally on this monitor. Do You think those cards are different in syncs? I can change in BIOS of this 286 board to Monochrome, 40 mode, 80 mode and EGA/VGA, but i doubt this is the problem as this monitor works ok. I can try also on MDA monitor i have?
Well another EGA card should have almost identical sync pulses, there are crystal oscillators for a reason. But the width of the pulse might be somewhat different, depending on the particular chip on the card, that might play a role. Tired capacitors in the monitor might be also affecting things. It's only a theory, worth testing. Running a few commands in DOS, even if you have to type them somewhat blind, is not going to take much time. So try both mode commands. Keep the BIOS settings and jumpers as-is for now.
Sure you can try on the MDA monitor, that card should have switch settings for MDA/Hercules emulation. Just keep in mind some monitors don't like weird sync inputs and might get damaged. To be honest I've not seen a monitor (except some early original IBM ones) that would be damaged by wrong sync pulses but it's always some risk. Nothing has blown up so far though.
One thing I forgot to mention, don't assume that your switch settings are correct just because you can see something on the screen. Sometimes these can be way off, it just so happens that montitor is able to re-sync to every N-th pulse so it seems somewhat stable. It might actually be some of the fast rolling settings that are closer to what is needed - it really doesn't take much of frequency difference for H-sync to make the picture look completly wrong. This would be much easier to test with a scope or frequency counter, but I assume you have neither?
Deunan wrote on Yesterday, 10:40:Well another EGA card should have almost identical sync pulses, there are crystal oscillators for a reason. But the width of the pulse might be somewhat different, depending on the particular chip on the card, that might play a role. Tired capacitors in the monitor might be also affecting things. It's only a theory, worth testing. Running a few commands in DOS, even if you have to type them somewhat blind, is not going to take much time. So try both mode commands. Keep the BIOS settings and jumpers as-is for now.
Sure you can try on the MDA monitor, that card should have switch settings for MDA/Hercules emulation. Just keep in mind some monitors don't like weird sync inputs and might get damaged. To be honest I've not seen a monitor (except some early original IBM ones) that would be damaged by wrong sync pulses but it's always some risk. Nothing has blown up so far though.
One thing I forgot to mention, don't assume that your switch settings are correct just because you can see something on the screen. Sometimes these can be way off, it just so happens that montitor is able to re-sync to every N-th pulse so it seems somewhat stable. It might actually be some of the fast rolling settings that are closer to what is needed - it really doesn't take much of frequency difference for H-sync to make the picture look completly wrong. This would be much easier to test with a scope or frequency counter, but I assume you have neither?
No, of course, i don't have oscilloscope 🙁
I have tried all the combinations with MDA monitor also, but no picture at all, just dots on 2 combinations - 0010 and 1010. Double that with jumper on motherboard for mono and color, tried that also, same result
butjer1010 wrote on Yesterday, 10:46:I have tried all the combinations with MDA monitor also, but no picture at all, just dots on 2 combinations - 0010 and 1010. Double that with jumper on motherboard for mono and color, tried that also, same result
Well the problem with MDA is different pinout. My ATI Graphics Solution card (that's CGA/Plantronics and MDA/Hercules all-in-one) requires not only switch changes for different frequency, but also some switches change the monitor type - either CGA or MDA. Assuming your card could drive an MDA monitor I would also expect it requires different switch settings for that.
My EGA card has only switches 1-5, and works with color monitors only as far as I can tell. It can emulate MDA but on EGA/CGA monitor. With EGA/CGA timings and output pinout of course. On this card switches 1-4 select mode and 5 selects I/O range (primary or secondary EGA card). Did you check if your switches 5-8 are connected to anything? Your card has an extra oscillator (Y1: 27.256MHz), mine has only 25.0 and 16.257 ones. So clearly your card can do more than mine. Maybe that's why it has more switches.
BTW I wanted to dump a Hercules clone font ROM for you but the bloody thing is soldered on my card. So you'll have to find one on the net.
Deunan wrote on Yesterday, 11:05:Well the problem with MDA is different pinout. My ATI Graphics Solution card (that's CGA/Plantronics and MDA/Hercules all-in-one […]
butjer1010 wrote on Yesterday, 10:46:I have tried all the combinations with MDA monitor also, but no picture at all, just dots on 2 combinations - 0010 and 1010. Double that with jumper on motherboard for mono and color, tried that also, same result
Well the problem with MDA is different pinout. My ATI Graphics Solution card (that's CGA/Plantronics and MDA/Hercules all-in-one) requires not only switch changes for different frequency, but also some switches change the monitor type - either CGA or MDA. Assuming your card could drive an MDA monitor I would also expect it requires different switch settings for that.
My EGA card has only switches 1-5, and works with color monitors only as far as I can tell. It can emulate MDA but on EGA/CGA monitor. With EGA/CGA timings and output pinout of course. On this card switches 1-4 select mode and 5 selects I/O range (primary or secondary EGA card). Did you check if your switches 5-8 are connected to anything? Your card has an extra oscillator (Y1: 27.256MHz), mine has only 25.0 and 16.257 ones. So clearly your card can do more than mine. Maybe that's why it has more switches.
BTW I wanted to dump a Hercules clone font ROM for you but the bloody thing is soldered on my card. So you'll have to find one on the net.
I played with switches, assuming that first 4 are for EGA, CGA, MDA,.... and 5-8 are for something else 😀 I have found 5 combinations (in previous post i said 6, but 0000 doesn't work, i don't know why i wrote it is working) that gives me output, but with broken hsync. This 5 combinations i tried with 16 combinations of second pair of 4 switches, but that didn't gave me good results. Picture was always broken.
So, i need someone who has this card, to give me switch settings for EGA, than we can try to troubleshoot if something else is wrong, right?
Well, manual or switch settings would be preferable. Or maybe consider buying a cheap LCD "portable scope" - these are not great but usually cheap and for this kind of work it's enough.
Looking around I found some manual for similar chipset "AutoSwitch EGA". That one not only can detect the connected monitor, but also if there is another card present in the system. Switches work completly differently. But that also suggests your card could possibly work with MDA monitor in Hercules emulation, as well as 640x350 EGA 2-color mode.
Can you make a final list of the switch combinations you tried, and what they do? It's getting hard to follow all the posts and I'm getting confused. Make sure to mention i which order you are writing it down, 1-2-3-4 or some other. Also what is 0 and 1 (up or down). I can compare it with it the notes I have about other cards, maybe something will look similar.
Deunan wrote on Yesterday, 11:41:Well, manual or switch settings would be preferable. Or maybe consider buying a cheap LCD "portable scope" - these are not great but usually cheap and for this kind of work it's enough.
Looking around I found some manual for similar chipset "AutoSwitch EGA". That one not only can detect the connected monitor, but also if there is another card present in the system. Switches work completly differently. But that also suggests your card could possibly work with MDA monitor in Hercules emulation, as well as 640x350 EGA 2-color mode.Can you make a final list of the switch combinations you tried, and what they do? It's getting hard to follow all the posts and I'm getting confused. Make sure to mention i which order you are writing it down, 1-2-3-4 or some other. Also what is 0 and 1 (up or down). I can compare it with it the notes I have about other cards, maybe something will look similar.
0 is OFF, 1 is ON.
0001, 0100, 1110, 1001 and 1100 worked with EGA monitor, but with broken sync.
These are the combinations of first 4 switches that worked (on other 11 combination there was no letters at all, either black screen or few lines or dots), and i have combined them with all 16 combinations of switch 5-8, but the result was same.
butjer1010 wrote on Yesterday, 10:46:I have tried all the combinations with MDA monitor also, but no picture at all, just dots on 2 combinations - 0010 and 1010.
"Just dots" is technically also being "out of sync", but by a way greater amount than the pictures you showed with your EGA monitor.
That's 4 and 5, which are indeed settings for "EGA is connected to an MDA monitor". The fact that you have two different monitors, the MDA monitor and the EGA monitor that both are able to sync to other cards, but not to the CGA card make it very likely that indeed the EGA card is not outputting a valid synchronization signal.
Deunan wrote on Yesterday, 11:41:Well, manual or switch settings would be preferable. Or maybe consider buying a cheap LCD "portable scope" - these are not great but usually cheap and for this kind of work it's enough.
Looking around I found some manual for similar chipset "AutoSwitch EGA". That one not only can detect the connected monitor, but also if there is another card present in the system. Switches work completly differently. But that also suggests your card could possibly work with MDA monitor in Hercules emulation, as well as 640x350 EGA 2-color mode.
I found an "AutoSwitch EGA" configuration utility (Written in Forth 83) by Paradise that seems to perfectly match what the BIOS of this card is doing. This cards seems to be hardware compatible to a "Paradise AutoSwitch EGA 480". The BIOS includes both 640x480 and 640x400 modes. On the other hand, https://www.minuszerodegrees.net/manuals/Para … %20Brochure.pdf does not include the 400-line mode. I wonder whether that one uses the same timing as the AT&T 400-line CGA modes.
The DIP switch settings at https://theretroweb.com/expansioncard/documen … ation/52480.pdf actually do match with the well-known EGA settings.
The "single monitor system" section shows configuration 11 for "single monitor monochrome", configuration 6 for "40 column", configuration 7 for "80 column", configuration 8 for "200 line" (column is a misprint) and configurion 9 for "350 line".
The "two monitor systems" section shows the same configurations again, but with SW 1/5 always in the "off" position (i.e. AutoSwitch disabled).
The "auto switch" section finally shows the original IBM EGA settings for "EGA with a color monitor is the secondary card". Interestingly, no setting on that page ever has SW1/6 set to on.
butjer1010 wrote on Yesterday, 12:08:0 is OFF, 1 is ON.
0001, 0100, 1110, 1001 and 1100 worked with EGA monitor, but with broken sync.
These are the combinations of first 4 switches that worked (on other 11 combination there was no letters at all, either black screen or few lines or dots), and i have combined them with all 16 combinations of switch 5-8, but the result was same.
As 0-0-0-0 is now removed from the set of valid configurations, my reference point that convinced me that 0 actually means 0 on this card is gone, so I will check those settings again expecting the typical "inverted logic". The settings then would decode as 7, 13, 8, 6 and 12 if read from right-to-left, or 14, 11, 1, 6 and 3 if read from left to right. Both sets don't make sense.
So, back to 0-is-0: These combinations would decode as 8, 2, 7, 9 and 3 if read from right-to-left or 1, 4, 14, 9 and 12 if read from left-to-right. The right-to-left set is the only one that has only valid EGA switch combinations in it. 8 and 9 are for "EGA monitor connected, PEGA is the primary or only card" while 2 and 3 are for "EGA monitor connected, PEGA is the secondary card". As butjer1010 has an AT system which has a different primary/seconday selection than the XT did (primary/secondary is no longer selected using the EGA switches, but the mono/color jumper on the mainboard decides which card is used for booting; also it is automatically detected if the switch designates a non-existant card as primary card), it is not that surprising that the "PEGA is secondary card" combinations work as well. What is surprising is that the 200-line-on-EGA-monitor settings 2 and 8 do work, while only one of the 200-line-on-CGA-monitor settings (0,1,6,7) works. All those settings are supposed to output the same timings in text mode.
butjer1010 wrote on Yesterday, 08:53:Best picture i got is when only pin2 is on, and 1,3 and 4 are off.
That's the combination that created Re: EGA Graphics card beeps - which, as I interpret the image is a 200 scan line 80x25 character text mode. You can recognize the 8x8 character cell not only by counting the scan lines that are clearly visible in that picture, but also be looking at letters with "descenders", like y or g. Those letters have their base line raised to fit the 8x8 cell. I don't think the monitor is nearly synchronized to a multiple of the card scan frequency, as every scan line seems to contain bright dots where the text is, and there are no repeated lines. During a whole frame, the start of the line drifts by around half a frame (262 scan line in CGA/NTSC timing), so it takes around 500 scan lines to get into the same horizontal phase again, which means the monitor is just 0.2% off the horizontal frequency of the picture. Unless the mode is just no longer in the capture range of the monitor, the monitor should have no problems capturing a sync signal that is 0.2% off the nominal frequency. Maybe a better way to phrase it is that the oscillator in the monitor is likely free-running unsynchronized just 0.2% off the real NTSC HSYNC frequency. That's impressingly close!
butjer1010 wrote on Yesterday, 08:53:I tried to connect 24th and 30th pin of feature connector, but no change, still is out of Hsync.
That's possibly good news, as this makes it more unlikely that the PAL is the cause of the issue, but the missing HSYNC is likely already missing before the PAL. Could you find a PAL pin connect to pin 24 of the feature connector? It looks like there is a fault on the HSYNC signal before the PAL, and pin 24 of the feature connector is supposed to be "hsync as generated by the card".
butjer1010 wrote on Yesterday, 08:53:I tried to connect 24th and 30th pin of feature connector, but no change, still is out of Hsync.
Wow, look at that picture closely! Unless it's an image artifact, there might be your fault! The trace from pin 24 of feature connector to the via left to it seems broken.
EDIT: I looked at an older picture of your card, and that trace is clearly OK on that picture, so I guess the dark spot on the trace is just flux residue from soldering the wire, so this is likely not the issue.
mkarcher wrote on Yesterday, 16:33:butjer1010 wrote on Yesterday, 08:53:I tried to connect 24th and 30th pin of feature connector, but no change, still is out of Hsync.
Wow, look at that picture closely! Unless it's an image artifact, there might be your fault! The trace from pin 24 of feature connector to the via left to it seems broken.
EDIT: I looked at an older picture of your card, and that trace is clearly OK on that picture, so I guess the dark spot on the trace is just flux residue from soldering the wire, so this is likely not the issue.
I was flux residue, not broken trace 😀 Didn't saw it at all
butjer1010 wrote on Yesterday, 17:27:I was flux residue, not broken trace 😀 Didn't saw it at all
OK, I have another idea. Assuming the PAL is broken and shorts pin 13 to ground. This would kill the HSYNC signal even if you bypass the PAL. With the wire between pins 30 and 24 still installed at the feature connector, remove U20 from the socket, bend pin 13 upwards, so it does not get into the socket, and re-insert the PAL without pin 13 connected. This should ensure the PAL is unable to interfere with the sync signal from the PEGA2A chip.