advice on bridging the crowbar out

[Beemer Biker]

I had LW1600PG power supply go out. Symptom was having to power the motherboard on/off more and more until eventually it no longer stayed powered up., The motherboard worked fine with a small seasonic. The 1600 watt supply capacitors all looked ok and no obvious damage anywhere.

Google was unable to find data sheet for Sitronix ST9S313-DAG

By testing the chip I found that if I bridge the pins shown below, the power supply no longer turns off.

I measure 12.31 on the 12v and 5.2 on the 5v which seems OK.

Possible this will work? It will be used only to power some PCIe graphics cards running BOINC application. A smaller seasonic will handle the motherboard.
According to the only spec I could find, this chip only checks for the 3.3, 5 and 12 lines. Could the chip be bad?

The primary is fused and I only need the 12v lines.

 

[Bill_Bright]

Without seeing a schematic, shorting those two pins would be like sticking a penny in a fuse socket - a bad idea.

Edit comment: fixed minor typo.

 

[Beemer Biker]

This is the best I can come up with

ST9S313-DAG Datasheet - Supervisor IC ( PDF )

ST9S313-DAG Datasheet - Supervisor IC, ST9S313-DAG pdf, ST9S313-DAG pinout, equivalent, replacement, ST9S313-DAG schematic, ST9S313-DAG manual, data.

www.datasheetcafe.com

According to that doc, the chip only checks under / over voltage and offers no current protection.
The only reference to pins is that the FPO goes high to signal a fault.

I bridged only the IC pins which only handle TTL level voltages. (IMHO)

However, for your enjoyment, I offer this picture of a seasonic "platinum" that had a faulty crowbar.


[edit]
for what it is worth:
The bottom red arrowed pin had 4.9 v, the pin above it had 0.0
After I bridged them the voltage was 5.2 on both and the power supply seems to work ok

 

[Bill_Bright]

Yeah, that is bad but frankly, I've seen worse - where the people were lucky to get out alive, thanks to their dog waking them up.

I would not leave it unattended.

 

[Panther_Seraphin]

So from what I am seeing, Pin 2 is ground and Pin 3 is the fault detection pin which seems to sit on a 5V rail.

Your basically running a no name brand chinesium 1600watt PSU with NO protections at all now....Hows the Home/Workplace insurance?

 

[Beemer Biker]

With no diagram or pinout you are just guessing

Possibly pin 2 is supposed to see 5v but when it did not "show up" the system was told to shut down. What I was hoping for was someone who had an application diagram of the chip or at least a pinout .

This aux supply is for the coprocessors. If they lose 12v the motherboard shuts down. When the motherboard shuts down the aux power supply is switched off.

The doc I listed stated that the "Fail Power On" pin went high to signal a problem. I supplied power to a pin that was low and needed to be high. A big difference. l am trying to debug the problem and come up with a solution. I have no interest in buying any more 1600+ watt power supplies no matter where they are made. If the chip is bad it is a few $ but a lot of effort to replace.

 

[ArdWar]

It's a "standard" part, look at WT7502 for example.

Also, please, anyone with decent electronic knowledge would recognize pin 2 as GND even without looking at datasheet....

 

[Beemer Biker]

Thank you for posting that reference. l disagree with your observation that anyone would know that pin 2 is ground without RTFM. I assume you know what RTFM is.

 

[Panther_Seraphin]

Wait....so PIN 2 (Bottom one marked) was reading 5V at all time?? That is meant to be a ground pin.

Can you double check the grounds then as I suspect the Chip is not seeing a 5V difference between Ground and 5V input (Pin 6) so going into a fault state.

Can you check the voltage between PIN 7 and PIN 2 with the PSU plugged in but not powered on? Should be a 5 Volt difference.

 

[R00kie]

With no diagram or pinout you are just guessing

Possibly pin 2 is supposed to see 5v but when it did not "show up" the system was told to shut down. What I was hoping for was someone who had an application diagram of the chip or at least a pinout .

This aux supply is for the coprocessors. If they lose 12v the motherboard shuts down. When the motherboard shuts down the aux power supply is switched off.

The doc I listed stated that the "Fail Power On" pin went high to signal a problem. I supplied power to a pin that was low and needed to be high. A big difference. l am trying to debug the problem and come up with a solution. I have no interest in buying any more 1600+ watt power supplies no matter where they are made. If the chip is bad it is a few $ but a lot of effort to replace.

Found this on some Vietnamese forum, people are right, you are connecting ground to what seemingly is a pin for fault protection

 

[ArdWar]

Thank you for posting that reference. l disagree with your observation that anyone would know that pin 2 is ground without RTFM. I assume you know what RTFM is.

If non-load bearing part got copper pour on a single pin, it's GND 80% of the time.
The other 20% is the part's floating reference, which kinda the same thing as its own GND.

I have no interest in buying any more 1600+ watt power supplies no matter where they are made. If the chip is bad it is a few $ but a lot of effort to replace.

What's wrong with getting a replacement? It or its many other equivalent parts are like only a buck if you order by pieces. Or even maybe salvage one from other PSU. It's also a DIP so soldering isn't exactly hard either. At least it'll be safer than doing a bodge work. The voltages you measure might seems to be fine, but short spikes or dips won't show up on a multimeter.

 

[Shrek]

I had LW1600PG power supply go out. Symptom was having to power the motherboard on/off more and more until eventually it no longer stayed powered up., The motherboard worked fine with a small seasonic. The 1600 watt supply capacitors all looked ok and no obvious damage anywhere.

The secondary capacitors look fine? (no bulging of the top)

 

[Bill_Bright]

Also, please, anyone with decent electronic knowledge would recognize pin 2 as GND even without looking at datasheet....

Hardly the point. If anything, it makes it even more apparent shorting pin 3 to ground would create a short to ground. Ohm's law is clear. When the voltage remains constant and resistance drops, current goes up. And when current goes up, so does heat. After all, that heat is what blows (melts) the fuse, right?

So having a decent knowledge of electronics is immaterial - except to illustrate the need for a detailed schematic for the circuit or better yet, entire device.

Also, bypassing a fault protection circuit would tell anyone with a decent knowledge of electronics that that would just be dumb.

Possible this will work?

Sure. But it is also possible it will not work - catastrophically. Without a complete schematic, it is impossible to be sure ESPECIALLY since we don't know what else might be wrong with that PSU.

Capacitors don't have to bulge or leak to be bad. And electronics failing when there is no obvious damage is common - typical actually, in low-voltage devices.

 

[ArdWar]

Hardly the point. If anything, it makes it even more apparent shorting pin 3 to ground would create a short to ground. Ohm's law is clear. When the voltage remains constant and resistance drops, current goes up. And when current goes up, so does heat. After all, that heat is what blows (melts) the fuse, right?

Wait, what? I was commenting about identifying what's in pin 2 itself, not about shorting anything into anything else...
There are several common practices when designing a PCB, and copper pour on GND is one of the most widespread one.

I do agree modifying things without understanding how it works is dumb, or in this case can be outright dangerous.


By the way..

The bottom red arrowed pin had 4.9 v

I'm not sure what you pick as reference here, but if your GND is not 0V, your PSU got big problem..

 

[Bill_Bright]

Wait, what? I was commenting about identifying what's in pin 2 itself, not about shorting anything into anything else...
There are several common practices when designing a PCB, and copper pour on GND is one of the most widespread one.

I was not accusing you of anything - sorry if I came across differently. I was just trying to point out the knowledge and skillsets of anyone viewing that device are immaterial when it is not known how it fits into the circuits.

Reminds me of: Check engine

 

[Beemer Biker]

View attachment 284432
View attachment 284434

Found this on some Vietnamese forum, people are right, you are connecting ground to what seemingly is a pin for fault protection

Yes, they are correct.
I spent some time looking for st9x313 pinout and the best I could find was a link that was mislabeled and went to an st9s429

I retook the voltage measurements as I had misnumbered the pins

The following is with the supply plugged in and turned on but no connection to the motherboard
1 PGI 0.0
2 GND 0.0
3 FPL 4.4
4 PDON 4.4
5 V33 0.0
6 V5 0.0
7 VCC 4.5
8 PGO 0.0


earlier I had verified the mechanical relay was was working.
All the caps look just fine.

I then connected the test bridge to the 24 pin connector normally plugged into the motherboard. This turned on a 12v fan which spun down and stopped after about 15 seconds. ie: the power supply shut down.

The following readings were taken

1 PGI 0.0
2 GND 0.0
3 FPL 4.4
4 PDON 0.0
5 V33 0.0
6 V5 0.0
7 VCC 4.5
8 PGO 0.0

[edit] Also want to point out that the problem is not related to cheap "chinese" parts.
Latest Nvidia boards are tripping the crowbar on other manufacturers power supplies

www.reddit.com/r/nvidia/comments/k6axtc/fyi_seasonic_and_rtx_3000_users_w_shutdowns/

 

[Panther_Seraphin]

Pin 5 and Pin 6 are seeing 0 volts where as they should be seeing the reference voltage of those rails when the PSU is set to power on.

Pin 7 Should see 5VSB when in standby and should see 12V when in normal operation.


So something is seriously funky is going on!!!

 

[Beemer Biker]

Pin 5 and Pin 6 are seeing 0 volts where as they should be seeing the reference voltage of those rails when the PSU is set to power on.

Pin 7 Should see 5VSB when in standby and should see 12V when in normal operation.


So something is seriously funky is going on!!!

5 and 6 are transients and go back to 0 with the shutdown. My Fluke meter has a min-max function but I did not see more than 1v on pin 5 while watching it. I need a clip to attach to the IC to hold the leads as I do not have enough hands. There should be a way to have the fluke determine how high the 3.3, 5 and 12 got before it was shut down.

 

[ArdWar]

My Fluke meter has a min-max function but I did not see more than 1v on pin 5 while watching it. .... There should be a way to have the fluke determine how high the 3.3, 5 and 12 got before it was shut down.

Once you set it in the minmax mode, pushing the "MINMAX" button repeatedly should make it cycle between showing max, min, and avg reading.
Although I doubt it can properly capture the transient. The glitch can be as narrow as 100us.

 

[Beemer Biker]

I ordered a test clip with cable and I have several small regulated power supply PCBs. Going to supply the correct voltages to those pins on the chip and see if I can eliminate either the chip or any of the voltages as being bad.

When I bridged pins 2 & 3 the supply stayed on and the 12 and 5 were 12.2 and 5.2 at no load. I think the ATX spec is +/- %5 so those are OK. The VCC is labeled "12" and had only 4.4 volts which is the standby voltage and seems rather low for the 5vSB. I do not use the 3.3v or 5v. This supply will never be connected to a motherboard. I need only the 12v to work with the graphic cards. If I supply the correct voltage to (for example) the 3.3 pin and the system works then at least the protection is enabled for the 12 and 5. With the info supplied by R00kie and advice here, I can debug the problem. Thanks for everyones help.

 

[Panther_Seraphin]

5 and 6 are transients and go back to 0 with the shutdown. My Fluke meter has a min-max function but I did not see more than 1v on pin 5 while watching it. I need a clip to attach to the IC to hold the leads as I do not have enough hands. There should be a way to have the fluke determine how high the 3.3, 5 and 12 got before it was shut down.

Use a clip to say a chassis ground will make it easier

The specs I can see mentioned is
3,3V
2.2 - 3.9 V is acceptable range

5.5V
3.5 - 6.1V is acceptable range

IF you have a problem on the 3.3 or 5V rails they are still a hazard as most modern designs run tap off the 12V rail to supply them vs rectifitying them directly off the mains voltage as in older designs.

 

[ArdWar]

Going to supply the correct voltages to those pins on the chip and see if I can eliminate either the chip or any of the voltages as being bad.

Make sure you disable the PSU's 5V and 3V regulator, or simply cut the sensing trace first (you can bridge back with solder later). Otherwise injecting external supply will make them try to pull each other into destruction, or at the least making your voltage injection useless. Most post regulator controllers have EN pin that you can use to disable them.

The VCC is labeled "12" and had only 4.4 volts which is the standby voltage and seems rather low for the 5vSB.

It's about right for 5VSB with 0.6V diode Vf drop.

 

[qubit]

That's a 1600W PSU right? Bodging it is like connecting something to the mains without a fuse. I wouldn't do it. Either fix it properly, or replace that PSU, it's a fire hazard. Especially so as modern high end graphics cards pull a lot of current with big power spikes.

I've got a newish 1600W PSU in my system that works perfectly and I'm thinking of derating it with its control software for safety as my system doesn't need anywhere near that much power. Otherwise it can just pump endless power into a partial short that simply looks like a big load to it and potentially start a fire.

 

[Beemer Biker]

Found problem. The 3.3 volt is not being generated by the v3.3 daughter board. Only 0.3 volts are created. The 5v daughterboard works ok in addition to the 12v main.
Going to try substituting a 3.3v linear for the daughterboard's 3.3. I tried connecting a 3.3v supply to the v33 pin but it did not work. I am guessing the chip did was looking for a jump from 0 to 3.3 which did not happen. Probably need to take the board out. Note that I will not be using the 3.3 nor the 5v.

 

[Panther_Seraphin]

Make sure there isnt a possible short dragging the 3.3V down. The fact its at 0.3 makes me think its possibly generating the 3.3V but its being dragged down.