Overclocking

Now with the system running stable. I started to overclock the frontside bus.

After screwing up my Windows installation two times I made a bootable CD-ROM with Prime95 and other tools on it. When testing overclocks I always boot from this CD first, test if the system is stable, then boot into normal windows.

For some reason the highest FSB I could run this tB1 Tualatin at stable was 126 Mhz (=1512 Mhz). Since I was expecting a higher overclock I bought another processor- this time an tA1 Tualatin (Philippines, SL68P, FPO: 7226A585, Packdate: 09/11/02).

I modded it using the wire wrapping method, put it in the slotket and started the system. Set voltage to 1.75V. Booted at 100FSB. Now I got greedy and started pushing the FSB without even testing for stability. 100 - 133 - 135 - 137 - 140 - 142. At 142 i stopped increasing because this board has a max PCI divider of 1/4. Wanted to run the stability test and left the room while the system starts from CD. When I came back the screen was black. UH OH!

Motherboard dead?

The system didn't start up .. not even a power-on or error beep and the power switch didn't work for turning it off .. I pulled the board out and after some time I found out that Vcc was about 0.5 V. This couldn't be right. I remembered reading some posts about dead mosfets at overclockers.com Forums- that probably happened to me too. I also noticed one leaky and another bulgy capacitor (JACKCON brand) in the CPU power converter input area- replaced those. If you want to replace your failed caps make sure you use 120°C rating capacitors of about the same capacity and voltage. These can be easily found in old PSUs.

Introduction on MOSFETS

MOSFETs come in four different types. They may be enhancement or depletion mode, and they may be n-channel or p-channel. For this application we are only interested in n-channel enhancement mode MOSFETs, and these will be the only ones talked about from now on. There are also logic-level MOSFETs and normal MOSFETs. Either of these can be used.

Unlike bipolar transistors that are basically current-driven devices, MOSFETs are voltage-controlled power devices. If no positive voltage is applied between gate and source the MOSFET is always non-conducting. If we apply a positive voltage UGS to the gate we'll set up an electrostatic field between it and the rest of the transistor. The positive gate voltage will push away the ‘holes’ inside the p-type substrate and attracts the moveable electrons in the n-type regions under the source and drain electrodes. This produces a layer just under the gate's insulator through which electrons can get into and move along from source to drain. The positive gate voltage therefore ‘creates’ a channel in the top layer of material between oxide and p-Si. Increasing the value of the positive gate voltage pushes the p-type holes further away and enlarges the thickness of the created channel. As a result we find that the size of the channel we've made increases with the size of the gate voltage and enhances or increases the amount of current which can go from source to drain- this is why this kind of transistor is called an enhancement mode device.

More info here: http://homepages.which.net/~paul.hills/SpeedControl/Mosfets.html

MOSFET testing

I managed to unsolder one FET but the other one didn't want to come off. Now to test the unsoldered one i used the following procedure since i dont have the equipment to properly test MOSFETS:

Get a multimeter with a diode test range.
Connect the meter negative to the MOSFET's source.
Hold the MOSFET by the case or the tab if you wish, it doesn't matter if you touch the metal body but be careful not to touch the leads until you need to. Do NOT allow a MOSFET to come in contact with your clothes, plastic or plastic products, etc. because of the high static voltages it can generate.
First touch the meter positive on to the gate.
Now move the positive meter probe to the drain. You should get a low reading. The MOSFET's gate capacitance has been charged up by the meter and the device is turned on.
With the meter positive still connected to the drain, touch a finger between source and gate (and drain if you wish, it doesn't matter). The gate will be discharged through your finger and the meter reading should go high, indicating a non-conducting device.

Replacing the MOSFETs

Datasheets of interest:
Components used on the Abit BH-6:
CS5165 - 5-Bit Synchronous CPU Buck Controller
HUF 76129S 3S - 56A, 30V, 0.016 Ohm, N-Channel, Logic Level UltraFET Power MOSFETs

This MOSFET was no longer working. So i tried to find a replacement part for it. The datasheet for the CS5165 voltage regulator IC contains the following sample circuit:

I tried a quick and dirty fix: Since I had no comparable FET i took all FETs from my Fanbus project and soldered them in parallel onto a pcb and connected it to the motherboard. Now the system was working again - however i quickly turned it off again because current draw was probably too high for my mosfet construction.

The most important criteria on choosing a FET for this application are (in order of importance):

On resistance, Rds(on)
Switching characteristics, tON, td(ON), Tr, Td(OFF), Tf, TOFF
Input capacitance, Ciss

Maximum drain current is also important but it's basically derived from On resistance and packaging type. Logic level drive is not important because the CS5165 datasheet says that it can drive both logic and standard level FETs.

Replacement MOSFETs I came up with:
irf3205.pdf
irl3103.pdf
irl3713.pdf
irl3716.pdf
ISL9N302AS3ST.pdf
STD150NH02L.pdf
STV160NF02L.pdf

Since I didn't want to wait to wait till an order arrives I went downtown and started shopping for MOSFETs at Conrad. The only usable FET they had was the IRF3205 - At € 3 each a rip off.

Before leaving home i put the motherboard in the freezer hoping that i could break that other mosfet off when i come back. The mosfet did break off - actually it broke in the middle :) [INSERT PIC HERE]

I quickly soldered the two new mosfets on and the board worked just fine.
Now i carefully monitored mosfet temperature, this is where i noticed a weird thing: When using the tB1 stepping chip MOSFET temperature rises to about 75°C, but with the tA1 Celeron temperature skyrockets to 150°C- where i turned the system off because the max. junction temperature for the FETs is 175°C. Both CPUs were set to the same FSB and Vcore. During testing one of the FETs got so hot that the solder connection melted and the cable went off!!

The mosfets were screwed to the heatsink of an old PSU and duct taped together with a 40mm fan on the motherboard.

The board worked now but i wasn't happy with the MOSFET temperature when running the tA1 chip. So i started searching for better mosfets and found these ON Semiconductor NTP90N02 - 90A, 24V, 0.005 Ohm. Four of them were soldered (two in parallel) to the board and screwed to another PSU heatsink which was then attached to my case. The combination of low Rds and paralleling considerabely reduced heat output- passive cooling keeps the MOSFETS at about 50°C.