BadGoldEagle
Well-known member
Ladies and Gentlemen,
Today, I would like to talk about the rapidly decreasing population of working Quadra 9x0 (900 and 950) specimens. If no action is taken within the next 5 or 10 years, the entire specie will soon disappear. Although the exact cause of death is still unknown, most researchers have concluded based on the symptoms that failing PSUs play a major role in the evolution of this disease. Experiments have proved that transplanting PSUs from younger compatible individuals saved almost all affected Quadra 9x0s. At this point in time, only one did not survive the transplant, with all other attemps being successful. The goal of today’s discussion is to specify a new effective PSU transplant procedure to save the remaining individuals from extinction.
Joke aside, my objectives are the following:
1/ To provide an easy solution/kit to replace the dying original PSUs. I’m pretty sure ALL are bad at this point and the lack of schematics makes repairing them uneconomical and replacing caps before failure proved to be not enough in my case. I will be using @GeekDot’s brilliant write up as a base for my mod.
2/ Make it so that there won’t be any drawbacks (which means creating a daughterboard to add back the square molex connectors as well as the monitor pass through feature)
3/ Make it silent.
4/ Meet or exceed the original PSU’s specs.
But first, some additional background: The Delta PSU in the Q9x0 computers is a very strange animal. Because motherboard and CPU design changed so much in the last 29 years since the Quadra was first introduced, traditional modern PSUs don’t actually meet the old PSU’s specifications in some areas. While the original could happily deliver (back in its heyday) up to 33A on the 5V rail, modern high power PSUs struggle to output more than 25A. Some industrial-grade ATX PSUs on the other hand have to be compatible with old school systems, and that means that there are modern PSUs out there that are as powerful as the old Delta unit was. I plan on using a Bicker BEA-630 because it’s a modern, efficient, quiet (has a temperature controlled fan controller and operates most of the time without any fan at all) relatively small factor (ATX) PSU designed for 24/7 use (MTBF: about 20 years). I’ve found it for as low as 50 euros (which is really not bad for a brand new ATX PSU).
There are unfortunately a few problems that need addressing: (ELECTRICAL ENGINEERING HELP NEEDED)
- The 'minimum output current‘ problem. Industrial grade ATX PSUs usually have a 'minimum output current‘. This is a common problem for ATX bench PSU modders. Basically (if I understood it correctly), the rail won’t be able to maintain the proper voltage unless there’s a minimum load. In the case of the Bicker, the minimum output current for both 5V and 12V rails is 0.5A. Some people suggest taking one wire of each rail and wiring it to ground through a 5 or 10 ohm, 10 watt resistor. If my math is correct, I’d need a 10 Ohm resistor for the 5V rail (5V = 10 Ohm x 0.5A), and a 24 Ohm resistor for the 12V rail (12V = 24Ohm x 0.5A). Is this correct? If so the additional power consumption would be about 8.5W, right?
- The 'there’s not enough pins' problem. ATX has 5x Ground and 5x 5V pins. The original PSU had 10 of each. Geekdot suggests to only populate every 2nd slot in the Q9x0 plug. In his excellent post, he proved that this arrangement worked and he was able to power up the Quadra. But with only half the pins, it probably halves the maximum current the board can take (‚probably‘ because we don’t know what the power repartition between the pins really is, so let’s assume it’s the same for each pin). Apparently ATX PSUs are designed so that each individual wire can deliver up to 6 Amps. So if I split one wire/post in two outputs, I can get about 3A max for each pin (assuming full load and without burning the PSU). Now the theoretical max. deliverable 5V amperage to the logicboard would be balanced across all pins and equal to 3A x 10 (# of 5V pins) = 30A which is almost the max. rated amperage (33A) of the original PSU for the whole computer. Of course, on the original Delta unit, not all 33A can go into the motherboard. So using one 5V pin from the Becker’s ATX 24 pin connector for two quadra 5V pins should be safe and will match or exceed the original specification. Any comments? I am no electrical engineer, so what is above might be completely wrong…
- There are other small problems but they’re easily solvable so I won’t bother you with the details right now. I’ll have to learn how to use Kicad properly and will probably end up taking lessons. But hey, that might turn out beneficial for me professionally so I think this is a cool and easy enough first project.
If you made it this far, kudos to you!
Today, I would like to talk about the rapidly decreasing population of working Quadra 9x0 (900 and 950) specimens. If no action is taken within the next 5 or 10 years, the entire specie will soon disappear. Although the exact cause of death is still unknown, most researchers have concluded based on the symptoms that failing PSUs play a major role in the evolution of this disease. Experiments have proved that transplanting PSUs from younger compatible individuals saved almost all affected Quadra 9x0s. At this point in time, only one did not survive the transplant, with all other attemps being successful. The goal of today’s discussion is to specify a new effective PSU transplant procedure to save the remaining individuals from extinction.
Joke aside, my objectives are the following:
1/ To provide an easy solution/kit to replace the dying original PSUs. I’m pretty sure ALL are bad at this point and the lack of schematics makes repairing them uneconomical and replacing caps before failure proved to be not enough in my case. I will be using @GeekDot’s brilliant write up as a base for my mod.
2/ Make it so that there won’t be any drawbacks (which means creating a daughterboard to add back the square molex connectors as well as the monitor pass through feature)
3/ Make it silent.
4/ Meet or exceed the original PSU’s specs.
But first, some additional background: The Delta PSU in the Q9x0 computers is a very strange animal. Because motherboard and CPU design changed so much in the last 29 years since the Quadra was first introduced, traditional modern PSUs don’t actually meet the old PSU’s specifications in some areas. While the original could happily deliver (back in its heyday) up to 33A on the 5V rail, modern high power PSUs struggle to output more than 25A. Some industrial-grade ATX PSUs on the other hand have to be compatible with old school systems, and that means that there are modern PSUs out there that are as powerful as the old Delta unit was. I plan on using a Bicker BEA-630 because it’s a modern, efficient, quiet (has a temperature controlled fan controller and operates most of the time without any fan at all) relatively small factor (ATX) PSU designed for 24/7 use (MTBF: about 20 years). I’ve found it for as low as 50 euros (which is really not bad for a brand new ATX PSU).
There are unfortunately a few problems that need addressing: (ELECTRICAL ENGINEERING HELP NEEDED)
- The 'minimum output current‘ problem. Industrial grade ATX PSUs usually have a 'minimum output current‘. This is a common problem for ATX bench PSU modders. Basically (if I understood it correctly), the rail won’t be able to maintain the proper voltage unless there’s a minimum load. In the case of the Bicker, the minimum output current for both 5V and 12V rails is 0.5A. Some people suggest taking one wire of each rail and wiring it to ground through a 5 or 10 ohm, 10 watt resistor. If my math is correct, I’d need a 10 Ohm resistor for the 5V rail (5V = 10 Ohm x 0.5A), and a 24 Ohm resistor for the 12V rail (12V = 24Ohm x 0.5A). Is this correct? If so the additional power consumption would be about 8.5W, right?
- The 'there’s not enough pins' problem. ATX has 5x Ground and 5x 5V pins. The original PSU had 10 of each. Geekdot suggests to only populate every 2nd slot in the Q9x0 plug. In his excellent post, he proved that this arrangement worked and he was able to power up the Quadra. But with only half the pins, it probably halves the maximum current the board can take (‚probably‘ because we don’t know what the power repartition between the pins really is, so let’s assume it’s the same for each pin). Apparently ATX PSUs are designed so that each individual wire can deliver up to 6 Amps. So if I split one wire/post in two outputs, I can get about 3A max for each pin (assuming full load and without burning the PSU). Now the theoretical max. deliverable 5V amperage to the logicboard would be balanced across all pins and equal to 3A x 10 (# of 5V pins) = 30A which is almost the max. rated amperage (33A) of the original PSU for the whole computer. Of course, on the original Delta unit, not all 33A can go into the motherboard. So using one 5V pin from the Becker’s ATX 24 pin connector for two quadra 5V pins should be safe and will match or exceed the original specification. Any comments? I am no electrical engineer, so what is above might be completely wrong…
- There are other small problems but they’re easily solvable so I won’t bother you with the details right now. I’ll have to learn how to use Kicad properly and will probably end up taking lessons. But hey, that might turn out beneficial for me professionally so I think this is a cool and easy enough first project.
If you made it this far, kudos to you!
Last edited by a moderator: