ATX Power Supply Confused by Voltage Divider.

[Electronaut]

I'm using an ATX power supply (taken from an HP pavilion desktop) to power circuits that require 3.3V, 5V, or 12V, and have had no problems, with and without use of a dummy resistor, until recently. When I try to get +6V/-6V by placing the 12V line across a voltage divider (4.7k ohm resistors, for
example, with GND=center tap), the current drops to zero.

Interestingly, when I unplug the ATX power supply, current is delivered as expected for a few seconds before it powers down.

On the other hand, if I use separaate ATX outputs, +12V (yellow), -12V (blue), and GND (black), there are no problems.

Thus it appears that use of a voltage divider confuses the ATX power supply. Any ideas?

 

[dknguyen]

What are you connecting to the circuit?
Which current drops to zero? The current out of the supply? Or the current out of the divider?
What resistance in the divider are you using?

Using a high resistance divider makes it a high-impedance source (bad...you want sources to be low impedance), but using a low resistance divider dissipates a lot of heat and bogs down your power supply in the process.

Examine the necessary return current paths through out of the ATX supply, through the divider, through your circuit, BACK THROUGH THE DIVIDER, and back to the ATX supply. Did you catch how an increased current on one "rail" of the divider" unbalances the voltage on the other rail? Not good.

You'd get better results using a capacitive divider with massive capacitors, and since capacitors aren't very accurate, put a resistor in parallel with each capacitor to balance the voltage. Effectively, parallel resistive divider with a capacitive divider. It lets you get away with somewhat higher resistances to reduce power dissipation while lowering impedance somewhat. Still not great though.

 

[Electronaut]

The circuit is a negative resistance oscillator, and the +6V/-6V lines are connected to the Vcc and GND pins of an LM358 opamp. The circuit works (with divider) when powered by batteries or a wall wart (AC adapter), but does not work when the ATX power supply is used.

Indeed, I disconnected the circuit from the divider and simply measured voltages across two 4.7k Ohm resistors connected in series, with the ATX 12V line connected across them. I got the full 12V across only one of the resistors!

I also tried placing capacitors in parallel with each resistor in the divider (220 uF) but this doesn't help.

I guess it is best to use batteries (or a bench supply) for projects like this.

 

[dknguyen]

Indeed, I disconnected the circuit from the divider and simply measured voltages across two 4.7k Ohm resistors connected in series, with the ATX 12V line connected across them. I got the full 12V across only one of the resistors!
/QUOTE]

That's odd. I don't know why that would be. You sure the resistors are what you think they are?

 

[Nigel Goodwin]

Are you sure you're not connecting ground to multiple places?.

The ATX supply is already grounded, you don't want (and must not have) another ground connected to your target circuit.

 

[Electronaut]

I double checked everything including resistor values, and checked for proper voltage division when I use a wall wart (or batteries).

The ATX supply ground is connected to one side of the voltage divider, with the +12V (yellow) lead to the other. The center tap
is used as circuit ground, in order to get +6V/-6V needed to power the opamp.

What seems to be happening is that the "intelligence" built into the ATX supply (which may not be standard) is not so smart. The odd behavior may be due to the fact that current flows only through my oscilloscope/multimeter probes, and not through the divider resistors (until I unplug the supply).

 

[dknguyen]

I double checked everything including resistor values, and checked for proper voltage division when I use a wall wart (or batteries).

The ATX supply ground is connected to one side of the voltage divider, with the +12V (yellow) lead to the other. The center tap
is used as circuit ground, in order to get +6V/-6V needed to power the opamp.

What seems to be happening is that the "intelligence" built into the ATX supply (which may not be standard) is not so smart. The odd behavior may be due to the fact that current flows only through my oscilloscope/multimeter probes, and not through the divider resistors (until I unplug the supply).

Is your meter battery operated?

 

[Electronaut]

Multimeter battery is OK, and multimeter reading agrees with oscilloscope reading (DC coupling).

 

[JonSea]

You're grounding the centertap of the resistors. The +12v supply is already grounded. You're shorting out one of the resistors.

If you want/need ±6v, you'll need two voltage dividers. Two resistors between +12 and ground will give you +6 volts (at essentially zero current) and two resistors between -12 and ground will give you -6 volts (again, at essentially zero current).

 

[Electronaut]

Thank you. Sounds plausible, but I am missing something here. Everything works (with the divider) when I substitute a battery or a wall wart for the ATX yellow/black supply leads. The battery is "already grounded" as well, to its negative end, and this ground is connected to one end of the divider, with the positive end connected to the other side of the divider. The circuit common or "ground" is connected to the center tap, and this is not the same as the supply ground. Thus relative to the circuit common one side of the divider should be at +6V, while the other is at -6V, as desired. See Practical Electronics for Inventors, Fourth Edition, p.649, for example. (BTW, the circuit under study here appears on p.684 of this book, Figure 10.2)

 

[unclejed613]

if shutdown happens when you connect the oscope ground to your circuit ground, be aware that the oscope ground and the ATX supply ground are connected through the AC outlet ground wire, so if your circuit ground is connected to the scope ground, the scope ground is connected also to the circuit's -6V rail through the wall outlet.

 

[Electronaut]

The ATX supply does not shutdown. This has happened under different circumstances (and requires a cooling off period before it works again), but not here. Here the current simply drops to zero while running, and when the supply is unplugged the current mysteriously is non-zero for a few seconds. I have tried carefully to distinguish between earth ground (=scope ground) and circuit ground, and the circuit ground is isolated from earth ground as far as I can tell.

 

[unclejed613]

back when IBM PC, and PC-AT systems were common, the power supplies would shut down when there was not enough current drawn from them. ATX supplies, however have two pins that must be used properly to insure the power supply starts and remains running. these pins are PWR_ON and PWR_GOOD. PWR_ON must be driven low to turn on the ATX supply, and PWR_GOOD must be driven high after the power supply turns on for the supply to keep running.

btw, if your supply voltages can be allowed to be +/-12V, the blue wire is a -12V rail. this would allow you to use earth ground as circuit ground.

 

[Electronaut]

Thanks for the tips. Yes, using the +12V (yellow) and -12V (blue) leads instead of a divider works, as I noted previously in this thread.

I cannot get the divider configuration to work with this supply. I followed the instructions at **broken link removed**, which agrees with what you say here, except that it suggests PWR_GOOD should be connected to GND (not HIGH), through an LED with 220 ohm resistor. It also says that a low-resistance load is needed. With everything connected as documented, the supply turns on, and the LED stays ON, even while my voltage divider does not work. I tried your suggestion as well (PWR_GOOD high) but this didn't help. Thus it appears that the supply does not shutdown, yet the current drops to zero.

As I noted at the top, this supply worked well with several other projects (where no voltage divider appears).

 

[Electronaut]

JonSea is correct, the problem is that the ground lead of the oscope probe is internally connected to the ground leads of the ATX power supply, so by attempting to measure the waveform I was shorting out one of the resistors in the divider! I was confused by the fact that my continuity tester does not show a hard connection between the oscope ground and the ATX ground, probably because this connection is made electronically. To check JonSea's theory I had to be sure the oscope was not connected when I used by multimeter to measure the voltages. Thanks JonSea!

 

[Nigel Goodwin]

I refer you to post #5

 

[JonSea]

When I try to get +6V/-6V by placing the 12V line across a voltage divider (4.7k ohm resistors, for
example, with GND=center tap), the current drops to zero.

I think most of this thread could have been avoided if someone had asked for a sketch showing exactly what "GND" meant and how the voltage was being measured.

 

[Nigel Goodwin]

I think most of this thread could have been avoided if someone had asked for a sketch showing exactly what "GND" meant and how the voltage was being measured.

That applies to most threads here, but I suspect any such sketch wouldn't have shown what we needed to know.

Possibly a photo of EXACTLY what he was doing would have been more help, and easy in this age of mobile phones.

 

[throbscottle]

Electronaut, if you connect the 'scope's probe ground clip to your circuit ground via capacitor, you will eliminate this problem. Will prevent DC measurement though. Other thing you can do is disconnect the 'scope's earth wire at the mains plug. Best way is what JonSea said though, assuming your PSU does actually have -12v.