NeXT
Veteran Member
That hacked up power supply is the most horrific and yet most rewarding thing I've seen!
- VCF East - April 12 - 14, Infoage Museum, Wall NJ
- VCF South West - June 14 - 16, Davidson-Gundy Alumni Center at University of Texas at Dallas
- VCF West - Aug 2 - 3, Computer History Museum, Mountain View, CA
- VCF Midwest - Sept 7 - 8 2024, Schaumburg, IL
- VCF SoCal - Mid February 2025, Location TBD, Sothern CA
-
Please review our updated Terms and Rules here
Restoring a Sun 386i: What a mess, anyone have a running one?
- Thread starter czunit
- Start date
Yeah it's pretty sad. I found the supply on Ebay in a not working condition and bought it for the enclosure. The board itself had burned in some places and had some pery bad soldering attempts done on it, so I cut it apart for the edge connector. The important part was figuring out all the voltages (esp -5 and -12) as well as realizing that the last pin was essentially a "DC OK" signal of 5 volts once the supply was up.
With that knowledge and given the system's timeframe I realized that the 386i supply was really just a PC-AT power supply and I could wire in all of the voltages and run with it. Sure enough it works and given that most of these supplies are toast nowadays it would be worth it to build a proper substitute.
One possibility is I could make a replica of the power supply board with nothing on it, but that would fit in the existing case, with standoff holes for an AT supply and copper connectors from the Sun card edge to stakes for the ATX supply connectors. That way I could just mount the ATX supply on the card, have space for a small fan, plug in the wires, and be done with it.
Any idea where one can custom make PCBs these days?
With that knowledge and given the system's timeframe I realized that the 386i supply was really just a PC-AT power supply and I could wire in all of the voltages and run with it. Sure enough it works and given that most of these supplies are toast nowadays it would be worth it to build a proper substitute.
One possibility is I could make a replica of the power supply board with nothing on it, but that would fit in the existing case, with standoff holes for an AT supply and copper connectors from the Sun card edge to stakes for the ATX supply connectors. That way I could just mount the ATX supply on the card, have space for a small fan, plug in the wires, and be done with it.
Any idea where one can custom make PCBs these days?
NeXT
Veteran Member
If it works, it works! 
Nothing wrong with restuffing a power supply if the result isn't explosions.
Nothing wrong with restuffing a power supply if the result isn't explosions.
Well, the one big question is cooling: This supply is probably more efficient than the 386i one but it was designed to have a fan pull air over the (smaller) heat sinks. Problem is those sinks are mounted so they are perpindicular to the airflow through the supply, not sure how hot they are getting. Plus the 386i fans blow air into the case and out vents on the side and although there is a baffle to get air into the supply box it might not be enough.
I'll see if I can measure the temps with an IR gun, worst case I might have to fit a small fan in there to pull air though and blow it out the exhaust vent. And I'll still try to figure out what is wrong with the other two supplies, the main FETs are not shorted so that is odd. There has to be a small "snooper" circuit that powers up the PS controller board so it can gate the real +5 and +12. We'll see.
I'll see if I can measure the temps with an IR gun, worst case I might have to fit a small fan in there to pull air though and blow it out the exhaust vent. And I'll still try to figure out what is wrong with the other two supplies, the main FETs are not shorted so that is odd. There has to be a small "snooper" circuit that powers up the PS controller board so it can gate the real +5 and +12. We'll see.
Meantime I dragged out the dremel, cut the lines from the battery to the timekeeper, soldered on new wire-wrap wires to the metal stubs, then hooked it up to another 2 aaa battery holder and mounted it under one of the metal card supports. Put the board back in the second 386i and we're back up and running with system #2.
Which is a bit tricky as it's configured to be a yp slave to the first system. And I still have only one working power supply.
Speaking of supplies, man they are weird. They don't seem to have the usual snooper circuit to power the PS electronics directly, in fact it looks like they have a complete second DC-DC converter to make the -12, -5, and +5 ready signals on the sidecard board. As for the main board I still can't figure out what switches what: One of the smaller transistors was shorted on one of the supplies but is fine on the other. Likewise one of the small signal resistors is burned on one supply but I can't tell if it's a 4.7 ohm resistor or a 470k ohm. Color codes are not too clear.
Oi. I may just build 2 adapter boards on PCB stock, then wire in more AT type supply boards. Sad, but without a schematic this is a really weird power supply design. Small wonder it blew up.
Which is a bit tricky as it's configured to be a yp slave to the first system. And I still have only one working power supply.
Speaking of supplies, man they are weird. They don't seem to have the usual snooper circuit to power the PS electronics directly, in fact it looks like they have a complete second DC-DC converter to make the -12, -5, and +5 ready signals on the sidecard board. As for the main board I still can't figure out what switches what: One of the smaller transistors was shorted on one of the supplies but is fine on the other. Likewise one of the small signal resistors is burned on one supply but I can't tell if it's a 4.7 ohm resistor or a 470k ohm. Color codes are not too clear.
Oi. I may just build 2 adapter boards on PCB stock, then wire in more AT type supply boards. Sad, but without a schematic this is a really weird power supply design. Small wonder it blew up.
Ok, breakthrough (of sorts).
At first I was tackling the problem by trying to trace things out from the +320 DC supply through the FETs, then the transformer to the secondary rectifiers for +5 and +12. I can see the gates on the two main high voltage switching transistors are both low, and both are controlled by a small transformer that's configured with one primary (signal input) and two secondaries that are 180 out of phase with each other. The secondaries are wired to the gates, the whole thing is an isolation system, and if you put a signal into the transformer one way it activates one FET on and one off, and the other way activates one fet off and one on.
Built in crossover protection. Ok, good. That seems to be powered by ANOTHER set of transistors in a push pull arrangement at which point I thought:
"Screw this, what is calling the shots?"
On the accessory motherboard there is a single chip that has no visible markings. However with the right light I can see it's a MB3759 buck controller that probably gates stuff that gates stuff that eventually hits the FETs. Ok, no biggie. However the important thing is it shows me that ground is on pin 7, VCC is pin 12. NOW I have something.
So I traced the pins around the accessory board I ripped off the bad supply (helpful) and finally figured it out. That's powered by a transformer on the accessory board that has 4 windings. One set goes to a diode then a 7812 regulator for -12, one goes into the vcc and ground of the 3759 through a small diode, and the other two I have no clue where they go. One probably goes to ANOTHER switching supply that is sourced by either mains AC or the +320v raw DC I don't know but something, and the other goes heck knows where.
But. Now that I know where the power is, and that the transformer isn't working, what if I backfed the 3759 circuitry with another power supply? So I hooked up my trusty vicor to ground and the circuit side of the diode, set the current to 0 and the voltage to 15 (recommended voltage) and fired it up.
Voltage 0, current limited. Brought the current up and at 200ma the voltage stabilized. Good! Now we can do something here. With the clips on carefully and the vicor off, I plugged in the naked board, set up a camera to record the fireball, and turned on the main switch.
No output on the +5 and +12. Now turn on the Vicor and wait for the sparks....
No sparks. More interestingly nothing exploded. I turned down the current to zero on the vicor and noted the voltage stayed at 10v. Which meant something was now powering the damn MB3759. Could it be????
Checked the +5 volt rail on the main power supply. Perfect 5v. Likewise the +12 has a perfect 12 volts. -12 is down, but -5 and +5 DC ok were both fine.
Which means the main supply circuitry, the FETs, the drivers, and the control circuitry are all good. The problem is in the snooper circuitry or whatever powers that little transformer. But it does mean that SOMETHING still works on the board, and the big supplies (+5 and +12) are good. At least on this board. It also means that once the board is up the snooper circuit is no longer used, which means something else is also sourcing power from the board, but it's probably not the transformer as the -12 is down.
How many frigging power supplies does this thing have in it?
Anyway I'm going to walk away for awhile and think about this. Something in the primary snooper supply circuitry is bad, I'm not sure where of course, but it's possible I could just put in a pair of small 120/240 to +12 supplies and bypass the whole primary system anyway with connections to the -12 regulator and the 3759 subsystem. Put a blocking diode in to ensure the main supply doesn't somehow toast the custom built snooper and Bob's your uncle....
Progress. Well on one of the supplies. I haven't tried this with the other one yet, it could still explode in a new and exciting way. But it is interesting.
So I powered
At first I was tackling the problem by trying to trace things out from the +320 DC supply through the FETs, then the transformer to the secondary rectifiers for +5 and +12. I can see the gates on the two main high voltage switching transistors are both low, and both are controlled by a small transformer that's configured with one primary (signal input) and two secondaries that are 180 out of phase with each other. The secondaries are wired to the gates, the whole thing is an isolation system, and if you put a signal into the transformer one way it activates one FET on and one off, and the other way activates one fet off and one on.
Built in crossover protection. Ok, good. That seems to be powered by ANOTHER set of transistors in a push pull arrangement at which point I thought:
"Screw this, what is calling the shots?"
On the accessory motherboard there is a single chip that has no visible markings. However with the right light I can see it's a MB3759 buck controller that probably gates stuff that gates stuff that eventually hits the FETs. Ok, no biggie. However the important thing is it shows me that ground is on pin 7, VCC is pin 12. NOW I have something.
So I traced the pins around the accessory board I ripped off the bad supply (helpful) and finally figured it out. That's powered by a transformer on the accessory board that has 4 windings. One set goes to a diode then a 7812 regulator for -12, one goes into the vcc and ground of the 3759 through a small diode, and the other two I have no clue where they go. One probably goes to ANOTHER switching supply that is sourced by either mains AC or the +320v raw DC I don't know but something, and the other goes heck knows where.
But. Now that I know where the power is, and that the transformer isn't working, what if I backfed the 3759 circuitry with another power supply? So I hooked up my trusty vicor to ground and the circuit side of the diode, set the current to 0 and the voltage to 15 (recommended voltage) and fired it up.
Voltage 0, current limited. Brought the current up and at 200ma the voltage stabilized. Good! Now we can do something here. With the clips on carefully and the vicor off, I plugged in the naked board, set up a camera to record the fireball, and turned on the main switch.
No output on the +5 and +12. Now turn on the Vicor and wait for the sparks....
No sparks. More interestingly nothing exploded. I turned down the current to zero on the vicor and noted the voltage stayed at 10v. Which meant something was now powering the damn MB3759. Could it be????
Checked the +5 volt rail on the main power supply. Perfect 5v. Likewise the +12 has a perfect 12 volts. -12 is down, but -5 and +5 DC ok were both fine.
Which means the main supply circuitry, the FETs, the drivers, and the control circuitry are all good. The problem is in the snooper circuitry or whatever powers that little transformer. But it does mean that SOMETHING still works on the board, and the big supplies (+5 and +12) are good. At least on this board. It also means that once the board is up the snooper circuit is no longer used, which means something else is also sourcing power from the board, but it's probably not the transformer as the -12 is down.
How many frigging power supplies does this thing have in it?
Anyway I'm going to walk away for awhile and think about this. Something in the primary snooper supply circuitry is bad, I'm not sure where of course, but it's possible I could just put in a pair of small 120/240 to +12 supplies and bypass the whole primary system anyway with connections to the -12 regulator and the 3759 subsystem. Put a blocking diode in to ensure the main supply doesn't somehow toast the custom built snooper and Bob's your uncle....
Progress. Well on one of the supplies. I haven't tried this with the other one yet, it could still explode in a new and exciting way. But it is interesting.
So I powered
Put the second power supply on the bench. Hooked up the variable supply to the chip inputs and noticed that it was pulling more current than the other board. At 7 volts I was pulling 1.5 amps which was *very* wrong. Got out the IR scope to take a look and see which component was getting hot and spotted the rectifier diode roasting. Pulled the diode, unit is pulling .2 amps at 12 volts and when powered up with main power outputs a smooth +5, +12, and -5 just like the other supply.
So the issue is in whatever chopper is putting power into the snooper circuits. I'll see if I can back trace, this is a lot less real estate to figure out but still odd. I might just say screw it and install 2 very small 120/240 to 12v dc-dc converters for the chip and for the -12 supply. Problem fixed, and I bypass what appears to be a bad design issue with modern stuff.
One thing I did hear: When I cranked the voltage above 12.2 volts on the driver circuitry I started to hear a whine. I think it's the isolation transformer that gates the fets and could be a warning not to bring the voltage above 12v. 10v works fine, I might make one of the DC-DC's a 10v supply to lower the stress on things. Below 9 volts it shuts down the supply.
So the issue is in whatever chopper is putting power into the snooper circuits. I'll see if I can back trace, this is a lot less real estate to figure out but still odd. I might just say screw it and install 2 very small 120/240 to 12v dc-dc converters for the chip and for the -12 supply. Problem fixed, and I bypass what appears to be a bad design issue with modern stuff.
One thing I did hear: When I cranked the voltage above 12.2 volts on the driver circuitry I started to hear a whine. I think it's the isolation transformer that gates the fets and could be a warning not to bring the voltage above 12v. 10v works fine, I might make one of the DC-DC's a 10v supply to lower the stress on things. Below 9 volts it shuts down the supply.
NeXT
Veteran Member
I've had cap issues in the past cause the MOSFET PWM driver to lock up and the avalanche blew up or sent a bunch of parts into the marginal range.
Possible. Another odd thing I noticed: That thermal switch is open, yet it seems to tie into the power inverter that is supplying that little transformer. I wonder if that's how they shut down the system if the heat sink temp gets too hot. Also I wonder if that switch is normally open and closes with heat or if it's supposed to be normally closed and is just bad....
Off to look up what it is....
Off to look up what it is....
Probably not the issue. While it is open on both systems the part is a Tokin OHD3-80M. Which means it's a 3 amp temp sensitive switch, 80c op temperature, and makes contact. The specs seem to indicate that it closes at 80c. So it's probably not the problem.
Back to the drawing board. Trying to sort this thing out is exceptionally difficult, but it looks like on the board there is a circuit to shut down power (such as it is, coming from the 300v - side of the supply and what looks like the + side via a small resistor) using a pair of transistors if that thermal switch is closed. On one board one of the transistors blew shorted (replaced) and one of the resistors going back to that 300v supply is showing 314k resistance while on the other board it only shows about 4.9 ohms.
Since that resistor is the only thing connecting this pile of stuff to the positive of the 300v supply line (DC though, why again?) I wonder if on at least one board the blowing of the transistor took out that resistor as some sort of fuse. I don't know, but here is a picture of the resistors in question on both boards. I think they're 4.7 ohm ones.
What's making me wonder is this then goes up into the board where it pretty much enters the transformer. If it was straight 300 volts then the transformer would see a dead short. That can't be. Now the second winding that goes back to the main board could be something that interrupts the first transistor turning off the DC voltage and making it kind of a somewhat regulated-ish chopper. Then the transformer would see a positive voltage, then as soon as the voltage spike hit the first secondary it would turn off, then when the voltage dissapated the transistor would turn "on" again keeping this disaster a running. The diode and power resistor on the input side could serve as a load to limit current or give the circuit something else other than the transformer to "bite" on. Hm.
God this thing is stupid. I really think I might just replace it with a pair of 120-240 to 12v wall wart guts, hook them up to the inputs, then let that drive the -12 circuit (easy enough since the wall wart is isolated) and the weird circuit for the switcher. I could glue those circuit boards to the big capacitors and be done with it.
Or maybe I'll buy a 4.7
Back to the drawing board. Trying to sort this thing out is exceptionally difficult, but it looks like on the board there is a circuit to shut down power (such as it is, coming from the 300v - side of the supply and what looks like the + side via a small resistor) using a pair of transistors if that thermal switch is closed. On one board one of the transistors blew shorted (replaced) and one of the resistors going back to that 300v supply is showing 314k resistance while on the other board it only shows about 4.9 ohms.
Since that resistor is the only thing connecting this pile of stuff to the positive of the 300v supply line (DC though, why again?) I wonder if on at least one board the blowing of the transistor took out that resistor as some sort of fuse. I don't know, but here is a picture of the resistors in question on both boards. I think they're 4.7 ohm ones.
What's making me wonder is this then goes up into the board where it pretty much enters the transformer. If it was straight 300 volts then the transformer would see a dead short. That can't be. Now the second winding that goes back to the main board could be something that interrupts the first transistor turning off the DC voltage and making it kind of a somewhat regulated-ish chopper. Then the transformer would see a positive voltage, then as soon as the voltage spike hit the first secondary it would turn off, then when the voltage dissapated the transistor would turn "on" again keeping this disaster a running. The diode and power resistor on the input side could serve as a load to limit current or give the circuit something else other than the transformer to "bite" on. Hm.
God this thing is stupid. I really think I might just replace it with a pair of 120-240 to 12v wall wart guts, hook them up to the inputs, then let that drive the -12 circuit (easy enough since the wall wart is isolated) and the weird circuit for the switcher. I could glue those circuit boards to the big capacitors and be done with it.
Or maybe I'll buy a 4.7
Well, power supply is back together and running. The solution was to get a pair of 100-300 volt input DC-DC converters that can output 10 volts to 13.9 volts. One I set for 11 volts and wired into the line for the startup/hotel circuitry, the other I set to 13.5 volts and fed into the 7812 -12 volt regulator.
For the input I wired each one to one of the large primary capacitors that have about 160 volts DC across them. This balances the load, and ensures that the supplies will never come near the 300 volt limits.
With that... it works. I have the 5 volts set at 5.2 volts to get the 12 volt rail up to 11.8 volts which might be a tad high but I don't think it will blow anything up. The -5 is about -4.5 and the -12 is a good solid -11.9 volts. And the final voltage (the +5 "ready" line) reads a good 5 volts as well.
Supply is back together and ready to go in. So now I have one supply that uses the AT bus system, a second supply that uses the 12 volt to 120 volt adapter, and a third that is outputting -12 as well properly. I may retrofit one of these little supplies to that second one (so it can have -12 which is used for RS232 and such) just to be complete but this project is at this point, done.
Now to figure out the problem on the third 386i's motherboard. That's a different thread for a different year. Time to work on other stuff.
For the input I wired each one to one of the large primary capacitors that have about 160 volts DC across them. This balances the load, and ensures that the supplies will never come near the 300 volt limits.
With that... it works. I have the 5 volts set at 5.2 volts to get the 12 volt rail up to 11.8 volts which might be a tad high but I don't think it will blow anything up. The -5 is about -4.5 and the -12 is a good solid -11.9 volts. And the final voltage (the +5 "ready" line) reads a good 5 volts as well.
Supply is back together and ready to go in. So now I have one supply that uses the AT bus system, a second supply that uses the 12 volt to 120 volt adapter, and a third that is outputting -12 as well properly. I may retrofit one of these little supplies to that second one (so it can have -12 which is used for RS232 and such) just to be complete but this project is at this point, done.
Now to figure out the problem on the third 386i's motherboard. That's a different thread for a different year. Time to work on other stuff.