Need help with voltage & current. Working out mods..

[Alexontherocks]

Hi everybody. I would like to ask a question about current and voltage.

http://img509.imageshack.us/img509/1294/modun7.jpg (http://imageshack.us)
[URL=http://g.imageshack.us/g.php?h=509&i=modun7.jpg][IMG]http://img509.imageshack.us/img5

In this picture you can see schematically what my x1950 pro mod looks like now. I used a variable resistor to improve the precision of the mod which originally envolved shading the original resistance down. I soldered all the connections as shown but the mod is pretty unstable. The values I measured between pin and ground have not changed after the mod. I set 600kohm with my pencil markings across the resistor and it gave me a specific voltage. At those voltages the overclock was stable as a rock. I then cancelled the pencil mod, set the variable resisotr and re-set 600 kohm turning the trimmer. It should have achieved the very same result instead I get frequent screen blanks and any overclock beyond 621mhz is unstable (with the pencil mod I have reached 717mhz stable).

I do not know much about electricity (all I know I have learnt by myself) however I figured that either the trimmer was not working or something elese was going on. I replaced the trimmer and nothing changed so here are the questions

a) Is it possibile I am changing current values by setting the mod the way I have? Current sensing is somehow influenced by the added components?

b) even though electricity follows the easiest path is it possibile I have some sort of "spreading thin" pehnomenon by not having removed the original resistor? Would I gain anything by removing the original resistor?

c) would it help to exorcise my x1950 pro?

thanks in advance!

Alex

[xoqolatl]

FB pin on the voltage regulator chip is used to sense the gpu voltage. When voltage decreases, regulator chip increases output voltage to compensate. Current flowing into FB pin is near 0; resistor from FB to ground is used to set a default (target) GPU voltage. By applying a resistor in parallel you decrease resistance from FB to ground; given that current is the same, voltage on FB pin decreases also -> regulator chip thinks Vgpu has dropped -> it increases output voltage. So cables or trimmer should not influence your mod.
My guess is that you have a "cold solder" - a faulty connection, usually made if you heat the solder and not the PCB trace you are soldering to. Try soldering again, then turn the trimmer and measure Vgpu when the card is running - that way you can set correct voltage and be sure it's within reasonable limits.

[Alexontherocks]

thanks for the reply! I have already tried and tried turning the trimmer while the car is running. The mod works perfectly. I manage to get the desired voltage with a very good range of precision. However the problem is still there. Regarding the soldering I admit I have very little experience but I tried a couple of tricks to avoid making bad connections such as connecting the wire to the trimmer directly on the "metal foot" of the smd resistor just as in the picture. For ground I have soldered a drop of metal inside a grounded hole (small empty pcb hole) making sure the drop convered the metal ring around it.

I am at a loss. I will try to resolder everything.....

xoqolatl, can you confirm that as depicted the mod should be working if all connections were to be sound?

Thanks for the help

[xoqolatl]

Well I dont have an x1950 to confirm, but that's a general idea how Vgpu mods do look :)
You could cut one of wires soldered to your VR and put a switch there; that will allow you to go back to pencil mod without unsoldering the hardmod. Also check your temps, check various clock settings etc.

PS. your second pic doesnt work, fix the link plz

[STEvil]

did you use the correct size variable resistor? If you used a VR with high resistance and are only using the last 5-10% of the VR to control the voltage it will be very inaccurate.

[Alexontherocks]

I will check with a different trimmer but for now I am running a t 1.56v rock stable using a pencil mod. Core at 715 artifact free and memory at 844 mhz. The problem is that anything more is precluded by the OCP protection.

http://img515.imageshack.us/img515/8084/chippi4.jpg (http://imageshack.us)
http://img515.imageshack.us/img515/8084/chippi4.2e21a0fd3b.jpg (http://g.imageshack.us/g.php?h=515&i=chippi4.jpg)

Here is the chip's datasheet. How would you go about modifying the sensing capabilities of the chip? I would rather not cut the OCP pin since I've read that no matter how "off" the perception is once modded the OCP sensing mechanism actually supplies crucial control over the voltage regulators.

Anyway I thought of working on the resistors of the di divider shown in the picture. I have yet to start testing but I believe sapphire have not probably installed the chip with significant changes in the component layout so everything in the datashhet should be more or less there on the pcb..


Need help guys...I have a feeling this budget board can be pushed even more! (Would love to see 800mhz even for 1 3dmark06 run)

Alex

[celemine1Gig]

Please post a pic of your actual mod, i.e. the soldering job. I also tend to believe that you are suffering from bad soldering joints.

[Alexontherocks]

@celemine1gig I will try but I would need to resolder everything since I took the mod apart to use the pencil markings. However the trimmer mod was interesting in the hypothesis that I actually managed to disable the OCP/oVP protection. Right now I am currently running at 1.56v which is the maximum anyway so there is reall no point in installing the trimmers untill I figure out the OCP mod first. I know this sounds like I have been wasting time but really there is no point in having a trimmer if you guys tell me that there is nothing to be done regarding the OCP mod :shrug:

Alex

No ideas about the OCP? what about shading the resistors? If it is a voltage divider there should be a R1/R2 relationship. How should I go about working on that?

[celemine1Gig]

First of all: Are you even sure that it actually is OCP and not OVP?

[Alexontherocks]

well the datashett describes botha a reference voltage 1.6v (which makes sense since a get a blank screen @ above 1.57v ) and also a OCP circuit.

The data sheet also says: @ pin 8: A resistor divider connected from this pin to vref programs the current limit threshold. Th outputs of the internal current sense differential amplifiers are summed togther to represent the output current. This voltage is then compared to this threshold.

Any ideas?

[xoqolatl]

http://omgwtfbbq.ovh.org/files/1950ocpovp.jpg
BE ADVISED THIS IS ONLY A GUESS AND I SHALL NOT BE RESPONSIBLE FOR WHATEVER HAPPENS
To increase OCP threshold (allow higher current) you would have to decrease resistance between OCP pin and ground (decrease resistor circled in red)
To increase OVP threshold you would have to decrease resistance between PGSEN pin and ground (decrease resistor circled in green). OVP should kick in when voltage at PGSEN is about 0.96V.

[Alexontherocks]

thanks for the suggestion. I am fully aware I perform these mods at my own risk :)

I was just wondering wether sapphire has any sort of datasheet database for their own applications of this chip. Some resistances are different than typical use and it is a pain to check for continuity with my multimetre and map the whole thing step by step. I am not even sure I get the correct values.

Is there an easier way to map the smd components near the chip?

[xoqolatl]

There's no need to find circled resistors, just measure the resistance from OCP or PGSEN to ground and solder an appropriate VR in parallel. Though it might be useful to find alternate soldeing points, as it's easier to solder to resistor or capacitor than SMD chip leg.

[Alexontherocks]

tried to decrease resistance but no effect... what would happen if I blinded completely the ocp pin or PGSEN pin?

[xoqolatl]

Sorry, I was wrong about OCP mod. OCP threshold is set by voltage on OCP pin, the higher the voltage the higher the allowed current. This voltage is derived from Vref through a resistor divider. To increase max current, you would have to put a VR between Vref (pin 1) and OCP (pin 8).

First, try to check if it's OCP or OVP that hinders your overclock. Measure the voltage at PGSEN pin when the card is running: if it's around 0.96V or more, it should be OVP.

Also note that PGSEN has two functions, one is OVP which triggers if Vpgsen is more than 0.96V, and the other is Power Good detection - Power Good signal is triggered when Vpgsen is above 0.6-0.64V. Attempting to remove OVP by decreasing resistance from PGSEN to groud might result in card powering up too late or not at all.

I am quite sure I have not made such mistake this time, although all mods are your responsibility. You can and should check my reasoning.