Do It Yourself PLL

How to build your own TurboPLL device. — FrankisGER

This article is based on what I learned from visiting Japanese overclocker
sites and trying it out myself. This is not for beginners, and is not risk-free, so
I’m disclaiming any responsibility for any damages you might incur doing this. Always
be careful and try to act logically. All names and trademarks used are the property of
their owners.

I got interested in creating my own clock generator about six months ago. I was using
a KT133 board at the time, which was pretty FSB-limited. Like most people, I couldn’t
get the board to run higher than about 110Mhz, and figured the board had been somehow
crippled.

I started researching how Turbo PLL devices worked at Japanese hardware sites. I ended up
looking at many homemade versions, and began to understand the basics behind them.

While these devices were often used to tweak BX boards, almost every motherboard can be tuned this
way.

This would have been written earlier, but the German army found better things to do with my time. 🙂

How A PLL Works

It gets an input frequency of 14.318 MHz and generates a number of frequencies out of it:

Clip

That 14.318Mhz is generated by a crystal tuned to that frequency. You might think all you need is
a crystal tuned to a higher frequency, but that makes ALL the speeds higher.

Clip

You might think that good until you find out your USB devices don’t work anymore and your clock
records an hour passing in just 53 minutes.

What A PLL Needs To Do

What you need is a PLL that increases the frequency to the areas that can use it, but not to the
areas that don’t.

Clip

You use a faster crystal to speed up the items you want to speed up, but you disconnect the items
you don’t want speeded up by removing their impedence resistors and use another crystal to generate frequencies for
those.

How To Do This

I won’t go into too much detail, since I don’t want people who aren’t familiar with this trying it.

The core components of my system at that time were a KT133 mainboard (ENMIC 8TAX+), a Duron and 128 MB PC-133. I couldn´t get above 110 MHz FSB with this setup.

This self made version of Turbo PLL allowed me to run my system at speeds as high as 131 MHz FSB. While I didn´t use it long at these speeds, that was only because my harddisk didn´t like 44 MHz PCI speed.

These are the parts I used to run FSB, AGP and PCI clocks faster:

  • 2 pieces 74F04 hex-inverter (serve to amplify the crystal oscillators outputs)
  • 1 piece crystal oscillator 14,318 mhz
  • 1 piece crystal oscillator 24 mhz
  • 1 piece crystal oscillator 48 mhz
  • 20 piece smd-resistor 22 ohm
  • Some caps to stabilize the power inputs of the crystal oscillators and hex-inverters

    These are the parts I used to make the rest of the system happy:

  • 1 piece crystal socket, 3 pin
  • 1 piece crystal 14,318 mhz (100 mhz -> 100 mhz) just if something goes wrong
  • 1 piece crystal 16 mhz (100 mhz -> 112 mhz)
  • 1 piece crystal 18 mhz (100 mhz -> 125 mhz)
  • 1 piece crystal 18,43 mhz (100 mhz -> 128 mhz)

    The first step is to find out which PLL was on the motherboard. In my case, it was a Cypress W230

    I went to the Cypress homepage, and looked up the datasheets for it.

    From this, I found out, that Pin 48, Pin 26 and Pin 25 were the pins necessary for my “operation”. I also found out after measuring on the motherboard that KT133s don´t use 24 mhz output.

    The datasheet showed me the right PLL pins, but the impedancy resistors also had to be removed. I found these impedancy resistors with the marking “220” on them (=22 ohm resistance) by following the traces from the clock output pins of the PLL.

    I removed these impedancy resistors as well as the 14,318 mhz crystal:

    I connected the external frequency generators to the mainboard.

     

    What Happened?

    After doing this, the maximum FSB speed I could reach went from about 110Mhz to 131 mhz.

    Here’s a WCPUID screenshot of my mainboard at 131MHz (with an FSB setting of 102MHz in BIOS) using a 18.43 mhz crystal and I/O voltage of 3,8 volt. My harddisk was holding me back a bit; it only booted into windows without errors when set into pio-3 mode.

    If you want to build “something like Turbo PLL” yourself, this little utility, OC Function (written by my friend Jan-Hendrik Diedrich, Thanks!!!), may help you in calculating the speeds:

    If you would like to research this subject further, here’s a text file with website URLs I found useful.

    Good luck!

    FrankisGER AKA Christoph Jadanowski

    Questions and Answers

    Q: Can I run my PCI/AGP at 33/66 Mhz and my FSB at whatever speed I want with this modification?

    A: No. The “self made Turbo PLL” only lets you reach higher FSBs by modifying some input and output signals of the mainboard´s PLL. The PCI and AGP remain at the same ratio compared to the FSB as before (example: PCI = 1/4 * FSB).

    Running the PCI/AGP at asynchronous speeds from the FSB is only possible with a few boards. Here is a Japanese site that talks about the necessary modifications to a BX board. Can’t say if it would or even could work with other boards.

    Here is my old idea to change the PCI divisor, but I’ve never tried it. I would if somebody gives me a mainboard to play with 🙂

    Q: Can you provide step-by-step instructions on how to build one?

    A: I could, but I don´t want totally unexperienced people to try doing this. It´s simply too dangerous. That´s why the article didn´t include a exact step by step how-to.

    Q: Can you build one for me?

    A: No. Learn the necessary electronic knowledge. It´s easy, really.

    Q: Can you provide more pictures?

    A: Yep. First, some more information:

  • The circuit shown uses 74HC04 inverters instead of 74F04 users as my DIY guide recommends. 74HC04 is slower, use 74F04!
  • I do not use A 24 MHz crytsal oscillator because this mainboard´s PLL has no 24 mhz output.
  • The PVC hoses serve to prevent short-circuits when the PCB is attached to the case.
  • I took the +3.3 Volt from the I/O Voltage/
  • The caps I used between +3,3V and GND are 100nF SMD cap, 10 µF electrolyt cap.
  • The crystal oscillators and the hex-inverters are designed to run at 5 volts. I run them at 3.3 volt, otherwise, they would kill the mainboard.
  • My circuit design isn´t very good because the traces are too long and too thick. This decreases the output signal quality. If you do this, make it smaller. Don’t use very long cables to connect the outputs to the mainboard.

    That said, here is the front of my PLL device:

    Here’s the back:

    This shows the crystal oscillator’s four pins

    This illustrates the hex inverter’s fourteen pins:

    A is input

    B is output

    For example: when I put a frequency to Pin 1 (1A), the opposite (inverted) signal will come out of Pin 2 (1Y).

    Pin 7 is ground (GND)

    Pin 14 is +3.3 Volt (Vcc)

    In my circuit, all Inputs were connected together (all pins with a A in their name; 1, 3, 5, 9, 11, 13).

    Finally, a picture explaining the various parts:

    If you have additional questions, visit the DIY: Build yourself something like Turbo PLL” Thread in [H]ardOCP Forums.

    FrankisGER AKA Christoph Jadanowski

  • Be the first to comment

    Leave a Reply