Switching frequency, phase multipliers, motherboards. Too many questions

  • Thread starter Thread starter ssb2245
  • Start date Start date
  • Tags Tags
    Frequency Phase
AI Thread Summary
The discussion centers on the intricacies of switching frequency and phase multipliers in motherboard voltage regulator modules (VRMs). It highlights the relationship between switching frequency, current delivery, and transient response, noting that smaller VRMs with fewer phases often have higher switching frequencies but more bulk capacitance. A specific example is given of a Gigabyte motherboard with a 24-phase VRM that utilizes a 6-phase PWM and phase doublers to achieve its configuration, raising questions about the impact on transient performance and switching frequency. The conversation also touches on the challenges of identifying electronic components and the need for clarity in technical discussions. Overall, the complexities of VRM design and performance are emphasized, reflecting the technical depth of the topic.
ssb2245
Messages
12
Reaction score
0
Switching frequency, phase multipliers, motherboards. Too many questions!

So i want to introduce myself, i went to Georgia Tech, and I got my BS in Biology, but ECE has always been kind of alike a hobby of mine. So i really like computers, but after interning in a engineering firm i decided to just leave it as my hobby. I have taken a handful of ECE courses so I understand the basics. I have a lot of questions that i need answered.

First off, switching frequency of buck converters. Is there anyway switching frequency could increase the amount of current reaching the processor? As i know is increased switching frequency decrease ripple, it also decreases overshoot, and it decreases efficiency and increase temperatures as a result. As i also know it in modern day motherboard VRMs switching frequency is directly related to transient response, and that capacitor series resistance is a huge problem to transient response. So this means that smaller VRMs(less phases) usually have higher switching frequencies and higher bulk capacitance? it doesn't make sense, to me that i see smaller phase vrms and they have much more bulk output capacitance than does a board with let's say 24 phases which might have a 75% less bulk output capacitance. Could someone clarify this more for me?

Now i saw this board, its made by gigabyte, and it has a 24 phase vrm, it has physically 24 driver MOSFETs, 24 inductors(chokes), and only a 6 phase PWM. Now the PWm is made by Intersil and they use a phase doubler to achieve 24 phases. I want to know how this works? There are 18 of these phase doublers, so the initial 6 each can do two and then those can do 2 more each. I want to know what type of tech they use? do they hurt transient performance so much by dividing teh switching frequency each part of the way? from 1mhz to 250khz? OR i read in one of their patent applications (intersil) like 50 different ways to make a phase doubler. One of the most appealing concoctions was for a IC that puts the full switching frequency to one phase until it reaches output current then automatically switches to the other phase, so it basically switches really quick and provides double the phases, so each phase still works out of phase.

here is the patent application: http://ip.com/patapp/US20100079175



another question, is there anyway to preserve switching frequency when only doubling the number of phases a PWM can handle? Like let's say we want a 12 phase VRM on a 6 phase PWM but we want to keep the switching frequency at it's highest.


one more thing, i suck at identifying electronic compoenets. Who makes this inductor? supposedly can deliver 50a of current:
attachment.php?attachmentid=34042&stc=1&d=1301981109.png


Thanks for all your help! nice forums BTW!
 

Attachments

  • r80.png
    r80.png
    39.6 KB · Views: 811
Engineering news on Phys.org


no responses? not even one of my questions?
 


ssb2245 said:
no responses? not even one of my questions?

Yeah, the volume of questions in your post is kind of daunting. It might have been better to start off slow with just a specific question or two. And you are using some terms that I'm not familiar with -- like what is a "24 phase vrm"?
 


Sorry :( so what shoudl i do? I don't want to make new threads just asking a specific question becuase i am scared i will get banned or deleted.

A 24 phase VRm, is a 24 phase voltage regulator module. On a motherboard, it consists of the following:
A pulse width modulator, basically outputs a PWM frequency to drivers, the drivers control high side and low side MOSFETs(one pair) and then output into one inductor and then capacitors.

Each phase of the 24 phase power supply has all of the above, but its a bit different.
First we have DriverMOSFETs(DrMOS) which consists of a pair of MOSFETs and Driver all in one IC, so we have the PWM sending a PWm signal to the DrMOS, which then send a driver signal to the integrated high side and low side FETs, and then output to an inductor and then a group of capacitors. There are 24 inductors, 24 Driver MOSFETs, 1 PWM and a bunch of input and output capacitors which all make up the power supply(VRM).

The problem is that the PWM is only a 6 phase PWM, meaning it only has 6 channels and is meant to be used with only 6 sets of DrMOS and inductors, so the company that makes the PWm patented a phase doubler scheme, so there are 18 other phase doubling ICs, i wanted to know how they work. The patent i linked has been put into use already and the manufacturer has a white sheet that resembles schematics in a patent application.

this is what the 24 phase vrm looks like:
[PLAIN]http://img573.imageshack.us/img573/2489/sidephasessmall.jpg

and PWm is not pictured its an intersil isl6366
 
Last edited by a moderator:


Why do they have separate output voltage sources? What do they power?
 


They power 130watt CPUs that operate at 1.3-1.9v, and that 130watt number is a thermal number, i have a wattage meter hookup to that 8pin 12v input from the computer's PSU, and I have pulled upto 200watts, that is with watercooling, when you use dry ice or liquid nitrogen the CPU will pull upto 300-400 watts at something like 1.8-1.9v.

All 24 phases work and output the same voltage for the processor cores. at stock voltage this is 1.2-1.35v depending on what intel specifically bins the processor at. Same model the stock voltage can vary as much as 200mv.
The power supply pictured , the 24 phase vrm, gives a variable output voltage, but every phase outputs the same voltage and current. Another output voltage requires another buck converter, and the motherboards have many, one for the uncore parts of the processor, one for the RAM, one for the IOH, and other things like sata controllers use LDOs.

basically that power supply takes 12v from the computers main power supply and changes it into min 0.9-max 2.1v(no one goes that far).

each DrMOS can only output 35 amps at 0.9-2.1v.

each DrMOS outputs to its inductor, and all the inductors output to the same PCB power plane.
 
Hi all I have some confusion about piezoelectrical sensors combination. If i have three acoustic piezoelectrical sensors (with same receive sensitivity in dB ref V/1uPa) placed at specific distance, these sensors receive acoustic signal from a sound source placed at far field distance (Plane Wave) and from broadside. I receive output of these sensors through individual preamplifiers, add them through hardware like summer circuit adder or in software after digitization and in this way got an...
While I was rolling out a shielded cable, a though came to my mind - what happens to the current flow in the cable if there came a short between the wire and the shield in both ends of the cable? For simplicity, lets assume a 1-wire copper wire wrapped in an aluminum shield. The wire and the shield has the same cross section area. There are insulating material between them, and in both ends there is a short between them. My first thought, the total resistance of the cable would be reduced...
I am not an electrical engineering student, but a lowly apprentice electrician. I learn both on the job and also take classes for my apprenticeship. I recently wired my first transformer and I understand that the neutral and ground are bonded together in the transformer or in the service. What I don't understand is, if the neutral is a current carrying conductor, which is then bonded to the ground conductor, why does current only flow back to its source and not on the ground path...
Back
Top