ATX Power Supply Extension Cable Limitations?

Do you have a question? Post it now! No Registration Necessary.  Now with pictures!

Threaded View

I have a case mod project in place - the design i have planned will involve
the PSU being placed at least 80cm + from the HD's CD-Rom's etc.

Is anyone aware if there are limitations on ATX Power Supply Extension
Cables? i.e. how far can i extended it?

Would this have any problems with supply to the HD's and data etc?

I will be going with a 600W+ PSU

Something like:



Re: ATX Power Supply Extension Cable Limitations? wrote:

Quoted text here. Click to load it

Pretty far. I rebuild a IBM Netfinity 5500 6U rack case to take a standard
ATX board and i think the wires bestwwen the PSU and the motherboard are
almost 1 meter long.

Quoted text here. Click to load it

My system is rock stable.

Lasse Jensen [fafler at g mail dot com]
Linux, the choice of a GNU generation.

Re: ATX Power Supply Extension Cable Limitations?

Quoted text here. Click to load it

Should be fine.

Quoted text here. Click to load it

There is no nice tidy limit. The problem is that you will get some
voltage drop down any cable. How much you get depends on the
current and the guage of the wire used. Should be fine for HDs
and CDs tho, just dont try to have just one wire for each rail if
you have a lot of drives.

Quoted text here. Click to load it

The problem wont be with the power supply cables,
the problem will be with the ribbon cable lengths.

Quoted text here. Click to load it

Re: ATX Power Supply Extension Cable Limitations?

Quoted text here. Click to load it
Quoted text here. Click to load it

First off, I'm not a power supply designer. But things I would be
interested in:

1) Static voltage drop. If you take the resistance of the wire
   in the extension cable, multiply by the current being carried,
   you get the voltage drop. You can look for AWG (American Wire
   Gauge) resistance tables, to find a value. Say the value
   was 0.004 ohms per foot for a certain gauge of wire, you
   have three feet of wire, and the current is 6A (the max for
   an ATX 20 pin connector, per pin). The drop would be
   0.004 * 3 * 6 = 0.072V. Compare this to, say, the 3.3V
   spec of +/- 5% at the output of the power supply. That is
   +/- 0.165V . That gives you some idea of the magnitude of the
   drop that could occur. The thicker the wire gauge in the
   extension cable, the lower the drop. Presumably, some ATX
   specification allocates margin for static voltage drop,
   because not all rails on an ATX supply have a remote
   sensing feature.

2) Transient response. The processor current changes rapidly.
   The power supply attempts to respond to the load change, limited
   by the switching frequency of the ATX power supply. The long
   cable becomes part of an R-L-C network at the load. The
   motherboard has electrolytic capacitors which bypass the
   ATX power supply rails, and that feature helps prevent the
   power supply wires from being an issue. While I don't think
   this would be an issue, it is another aspect of the analysis.

Some supplies have the addition of remote sensing. In remote
sensing, a sensor wire is attached to the final connector at
the motherboard. For example, this is how many power supplies
do their 3.3V. If you look carefully at a typical ATX supply,
pin 13 has two wires. The thinner wire senses the 3.3V and
improves the voltage regulation at the load. The remote_sense
"nulls out" the voltage drop of the cabling.

    ---------------+   remote_sense (thinner, brown wire)
                   |                                      |
    Power_Supply   |   3.3V wire (orange, pin 13)         v
                   |--------------------------------------+--- Mobo

This Jeantech supply has remote sense for all three rails
(3.3V, 5V, 12V). The modular cable assembly separates the
sensor wires. Antec Truepower also had remote sense on
all three supplies, as another example of remote sense
on all heavily loaded rails.

If you have a supply with remote sense on all three rails,
and you can route the sensor wire so it connects to the
_final_ ATX connector, the remote sense allows the regulator
in the power supply, to null out the drop in the cable.
You would custom wire to look like this. This requires
cutting the remote_sense wire, and rewiring it so it
connects as near as possible to the motherboard final

    ---------------+   remote_sense
                   |                                      |
    Power_Supply   |   power_wire                         v
                   |<------------------><---------------->+--- Mobo
                   |    First cable        Extension

Here is how it works. The remote_sense picks up the voltage
at the load. If it is too low, the regulator turns up the
voltage. In the example below, there is 0.1V drop in the
first cable, and 0.1V drop in the extension cable. The power
supply ends up turning up the voltage to 3.5V at the regulator,
in order to meet the 3.3V at the load.

    ---------------+   remote_sense
                   |                                      |
    Power_Supply   |3.5V               3.4V               v  3.3V
                   |<------------------><---------------->+--- Mobo
                   |    First cable        Extension

Without rewiring the cable assembly to have the above shown
wiring, the remote_sense would only sense the voltage at the
end of the first cable, meaning the drop in the Extension is
not compensated for. The voltage where the first cable meets
the second is 3.3V, and at the load 3.2V. This is what
blindly slapping an extension onto an ordinary power supply
(with remote sense on 3.3V) looks like.

    ---------------+   remote_sense
                   |                   |
    Power_Supply   |3.4V               v  3.3V             3.2V
                   |<------------------><---------------->---- Mobo
                   |    First cable        Extension

For cheap supplies without remote sensing on +5V and +12V,
an extended cable would look like this:

    Power_Supply   |5.0V                  4.9V             4.8V
                   |<------------------><---------------->---- Mobo
                   |    First cable        Extension

Another thing to note, in "real" power supplies, there is a
specification for that "3.5V" voltage value. It is called the
"compliance voltage limit". As the extension cable becomes
longer and longer, the 3.3V regulator has to lift its output
higher and higher. There is some limit to how much a supply
with remote sensing will lift its output voltage. The 3.5V
shown above would be a reasonable design, for example.
ATX power supplies _never_ list the compliance voltage,
while many lab power supplies (say from Agilent) would state
how much the remote_sense can compensate for the cabling.

You can slap an extension on the output of any supply,
and it could well work. I would test with Prime95 (one
copy per core on a dual core processor), and see if
the system is stable. For SLI video cards, run 3DMark06 or
something, so you get a heavy loading on both the processor
and the SLI video power cables. You can use that as a test, in
lieu of having a proper spec from a standards body,
for the allowed voltage drop in the cable.

An ATX power supply has 5% load regulation. The typical supply
behaves better than that. If the supply only varied by
3% (from no load to full load), you would have a
2% budget (3.3V * 0.02 = 66mV) for an uncompensated cable
drop. This is why blindly slapping an extension on many
supplies can work. As long as the supply is not so crappy,
as to let its output drop by the whole 5%, the remaining
margin can be used for an extension cable. Again, use
Prime95 or 3DMark06, use the motherboard voltage monitor,
and monitor the diff in voltage between 100% CPU load and
idle. This will give you some idea how good the supply
regulates, and whether it can take the additional drop
of a cabling scheme that doesn't use remote regulation.
Do the test both without the extension, then with the
extension, to see the impact of what you've done.


Site Timeline