5V vs Vcore: what about Remote Sensing?

[GigaFrog]

I know this problem has being addressed before.
Vcore fluctuations limit stability and therefore overclocking.
Some power supplies don't have a very good regulation on their 5V line.
How 5V fluctuations will affect Vcore fluctuations, do 5V even affect Vcore?
When do we have to worry about the 5V line? at 4.8V, or at 3V?

I did some Excel processing on the MBM log file, to compare 5V and Vcore results.

This first graph show Vcore and 5V with no processing. http://www.geocities.com/alochin/temp/Vcore1.gif

On the secong graph, I added 2 on each Vcore sample, so that it would show up at 4V, and allow a more detailed volt scale. This is Excel processing, there is no change in the Vcore value. http://www.geocities.com/alochin/temp/Vcore2.gif

On the third graph, I multiplied Vcore to 2.4, in order to amplifie its fluctuations and also bring it close to 5V, again to allow a more detailed voltage scale. http://www.geocities.com/alochin/temp/Vcore3.gif

There is a long iddle time in the middle of the graph, and both 5V and Vcore jumped synchronously by a significant amount.

All fluctuations of Vore are not due to 5V fluctuations, as we can see on the graphs, but the most extremes are. Having a good regulation of the 5V line do impact Vcore stability. Even low fluctuations, like drops to 4.7V, do affect the Vcore.

Considering the current draw, even the best power supply will not prevent fluctuations that will be due to line loss in the wires (these line loss will changes as the current draw changes, resulting in voltage fluctuations). The only way I see to reduce these fluctuations is Remote Sensing, and I think I will be investigating that soon.

Remote Sensing is forcing the power supply to regulate its 5V not at its outputs, like most power supplies do, but at the load connections, in our case at the input of the Vcore switching regulator. To do so, 2 wires (+ and -) must be added between the load and the sense inputs of the power supply. The sense inputs do not draw any current, therefore the remote sensing connections will not be subject to any voltage fluctuations. This is sometimes called a 4-wire connection (in opposition to a conventional 2-wire).

Now I have to open my PSU and find out these sense inputs.

Sorry for geocities inconvenience.
Hope this was informative.
Giga.

[grep]

Wow, someone is thinking, and thinking good. Your 10 steps ahead of me. But I am getting there slowly. Keep up the work on this idea and please keep posting results of your efforts.

Grep

[GigaFrog]

After doing some research, I found out that most power supplies use the TL494 PWM controller that suits well this application. My PowerTronic uses it too.

Its datasheet can be found here: http://www-s.ti.com/sc/ds/tl494.pdf

This links shows some schematics on how the TL494 is used. http://www.nutsvolts.com/PDF_Files/PSRepair.pdf

Not much.
I think the error amps are where the feedback is sent.
I'll have to re-open the PSU, and this time, measure signals at the pins of the TL494.

I processed another MBM log file that was logged while various applications were runing (Sandra, mpeg4 and toast). More fluctuation is shown, with still a clear pattern between 5V and Vcore. http://www.geocities.com/alochin/temp/log1b.gif

I hope I'll figure out hot to remote sense the 5V, I am sure it would stabilize the Vcore.