Does anyone know the role of different voltages from the power supply? I think 5V is mainly used for logic, 12V for electromechanical parts and +5V SB for allowing power buttons to work, STR and such. I am not sure what 3.3V, -12V, -5V are used for. Anyone?
-12V, -5V
- Thread starter Jerboy
- Start date
<< i've heard some video cards use -12
3.3 might be regulated down to the cpu core voltage? >>
3.3 is for the CPU (it is cut approximatly in half by the 3-phase power supply on your mobo (3.3/2 =1.65)).
3.3V is for PCI and DIMM. -12V and 12V were originally used by serial ports, but now these are generated from the 5V. -12V and -5V were once used for the bias voltage of some chips, but almost always now the chips generate these voltages internally.
<<
<< i've heard some video cards use -12
3.3 might be regulated down to the cpu core voltage? >>
3.3 is for the CPU (it is cut approximatly in half by the 3-phase power supply on your mobo (3.3/2 =1.65)). >>
Huh? Do you KNOW what a 3-phase power is? Three phase power has three wires, each phase shifted by 120degrees.
Demon-Xanth
Lifer
- Feb 15, 2000
- 20,551
- 2
- 81
3.3V is typically used for I/O within the motherboard, some PCI cards, (there is a 3.3V and a 5V PCI standard), powering your RAM, the chipset, and stuff like that.
The -5V and -12V is used in only a few places.
The -5V and -12V is used in only a few places.
<<
<<
<< i've heard some video cards use -12
3.3 might be regulated down to the cpu core voltage? >>
3.3 is for the CPU (it is cut approximatly in half by the 3-phase power supply on your mobo (3.3/2 =1.65)). >>
Huh? Do you KNOW what a 3-phase power is? Three phase power has three wires, each phase shifted by 120degrees. >>
no, I don't know how the thing works exactly, but that is how I was told it worked roughly.
<< Huh? Do you KNOW what a 3-phase power is? Three phase power has three wires, each phase shifted by 120degrees. >>
You are thinking about 3 phase AC that is way different then the 3 phase voltage regulation used on mobos to efficiently reduce the 3.3 V supply down to that used by the CPU. If someone (pm?) on here has a good expanaintion of how the 3 phase regulation works I would enjoy reading it!
Generally DC voltage regulators are not very efficient, creating heat while droping the voltage to the required level, I believe that the 3 phase system gets around this proplem to a certian degree. The effeciency of this circiut could be a reason why some mobos (ECS K7S5a?) seem to require more power to operate the same CPU then others.
<< You are thinking about 3 phase AC that is way different then the 3 phase voltage regulation used on mobos to efficiently reduce the 3.3 V supply down to that used by the CPU. If someone (pm?) on here has a good expanaintion of how the 3 phase regulation works I would enjoy reading it!
>>
so I am right?
<<
<< Huh? Do you KNOW what a 3-phase power is? Three phase power has three wires, each phase shifted by 120degrees. >>
You are thinking about 3 phase AC that is way different then the 3 phase voltage regulation used on mobos to efficiently reduce the 3.3 V supply down to that used by the CPU. If someone (pm?) on here has a good expanaintion of how the 3 phase regulation works I would enjoy reading it!
Generally DC voltage regulators are not very efficient, creating heat while droping the voltage to the required level, I believe that the 3 phase system gets around this proplem to a certian degree. The effeciency of this circiut could be a reason why some mobos (ECS K7S5a?) seem to require more power to operate the same CPU then others. >>
What is the definition of three phase here? ECS K7S5A has a PWM chip and chokes to step down the voltage and this kind of regulator is called switch mode DC-DC converter. Others like LM317 is called linear regulator. Linear regulators are very inefficient.
It switches 5V right into capacitors through an inductor. Inductor doesn't allow sudden flow in current so, the PWM keeps the power on until it reaches 1.75V at the capacitor end and cuts it off, back on when capacitor discharges alittle, and cycle continues. . This is repeated thousands of times in a second.
Anyways I've never heard "3 phase" DC-DC converter.
<< Also, you will notice the now signature six MOSFETs present that reminds us of ABIT's 3-Phase power that is used on this board and many other Socket A boards they have engineered. >>
This quote is taken from This Hardocp review. Now by your definintion they must be using 240VAC on this board. Hummm.... I think not. Perhaps you can cast this review into the light of your knowledge and help all of us understand what is happening.
I have seen refrences in other places, including this fourm, to motherboard 3 phase power so perhaps you need to read up a bit more on specific motherboard circuitry.
People used to pay me good money to, among other thing, design power supplies. It's been awhile. To design an efficient 1.65 volt, 60 amp source would have been considered impossible not long ago. It is still not an easy thing.
To get a clue what was happening in this area, I hunted down some engineering pages that showed current approaches to powering CPUs. Wow. The problem has been right in front of me for quite a while, but I never paid much attention to it. It got the gears grinding.
This is my take: From a power supply guy's point of view, a CPU is the load from hell!
It is not just the very high current and fussy low voltage that is the challenge. It's what the CPU does. Here is the CPU loafing along at a few amps; suddenly it goes to work flat out and in a few nanoseconds is pulling an extra 50 amps.
No supply, even a few inches away, can cope with transient current like that in maintaining voltage. So, the first line of defense is lots and lots of high quality ceramic caps placed very near the CPU. That isn't even good enough. Some of the caps have to be right on the CPU itself. Well, that gets us through the first dozen nanoseconds allowing the demand for more current to reach the electrolytic caps through an unavoidably inductive path an inch or so from the CPU socket. This is the second line of defense. Their job is to back up the ceramic caps and provide current for many microseconds. Remember, we are talking about 50 amps here and very little voltage drop allowed.(very low ESR - effective series resistance)
Finally, at this point, the active parts of the motherboard power supply module begin to react to the demand. All this time, the CPU is supposed to be getting a nice even 1.65 volts. Well, maybe. The active circuitry will, at this point, probably be driven into a condition known as maximum slew. That is, all of the pass elements will be turned on for a short time, allowing the current to increase as fast as possible through the inductors - those little donuts with a couple of turns of wire on them. If the current in the inductors can build up fast enough, the voltage in the electrolytic caps will not drop excessively, and the CPU voltage will be adequately maintained.
The rest of the picture, the pulse width modulated mosfets, two phase, three phase is almost anti-climactic. Here is a description:
The basic circuit used is called a buck regulator. The high source voltage is switched on for a very short time through an inductor which feeds current to the output. When the input voltage is not connected to the inductor, the input end of the inductor is switched to ground to maintain the inductor current feeding the output. The output voltage is controlled by varying the percentage of time connected to the input voltage.
If you put two or three of these circuits side-by-side (parallel) and drive them sequentially - round robin fashion, you have the essentials of a typical motherboard CPU power supply.
Besides the obvious, more circuits - more current capacity, operating the circuit legs sequentially improves the response time of the supply.
Yes, Virginia, three phase DC to DC.
To get a clue what was happening in this area, I hunted down some engineering pages that showed current approaches to powering CPUs. Wow. The problem has been right in front of me for quite a while, but I never paid much attention to it. It got the gears grinding.
This is my take: From a power supply guy's point of view, a CPU is the load from hell!
It is not just the very high current and fussy low voltage that is the challenge. It's what the CPU does. Here is the CPU loafing along at a few amps; suddenly it goes to work flat out and in a few nanoseconds is pulling an extra 50 amps.
No supply, even a few inches away, can cope with transient current like that in maintaining voltage. So, the first line of defense is lots and lots of high quality ceramic caps placed very near the CPU. That isn't even good enough. Some of the caps have to be right on the CPU itself. Well, that gets us through the first dozen nanoseconds allowing the demand for more current to reach the electrolytic caps through an unavoidably inductive path an inch or so from the CPU socket. This is the second line of defense. Their job is to back up the ceramic caps and provide current for many microseconds. Remember, we are talking about 50 amps here and very little voltage drop allowed.(very low ESR - effective series resistance)
Finally, at this point, the active parts of the motherboard power supply module begin to react to the demand. All this time, the CPU is supposed to be getting a nice even 1.65 volts. Well, maybe. The active circuitry will, at this point, probably be driven into a condition known as maximum slew. That is, all of the pass elements will be turned on for a short time, allowing the current to increase as fast as possible through the inductors - those little donuts with a couple of turns of wire on them. If the current in the inductors can build up fast enough, the voltage in the electrolytic caps will not drop excessively, and the CPU voltage will be adequately maintained.
The rest of the picture, the pulse width modulated mosfets, two phase, three phase is almost anti-climactic. Here is a description:
The basic circuit used is called a buck regulator. The high source voltage is switched on for a very short time through an inductor which feeds current to the output. When the input voltage is not connected to the inductor, the input end of the inductor is switched to ground to maintain the inductor current feeding the output. The output voltage is controlled by varying the percentage of time connected to the input voltage.
If you put two or three of these circuits side-by-side (parallel) and drive them sequentially - round robin fashion, you have the essentials of a typical motherboard CPU power supply.
Besides the obvious, more circuits - more current capacity, operating the circuit legs sequentially improves the response time of the supply.
Yes, Virginia, three phase DC to DC.
<<
If you put two or three of these circuits side-by-side (parallel) and drive them sequentially - round robin fashion, you have the essentials of a typical motherboard CPU power supply.
Besides the obvious, more circuits - more current capacity, operating the circuit legs sequentially improves the response time of the supply.
Yes, Virginia, three phase DC to DC. >>
The circuit seems like an ordinary PWM controller to me. When I switch it on with the power supply AC cord disconneted, it attempts to pull current from whatever's charged in PSU's capacitor and mobo's regulator screeches, just like regulator PWM regulator does when its fed with grossly undervoltage.
Last time I checked, the PWM was operating around 160KHz and I believe there were six switching semiconductors. If I observe the ripples on three pairs of them, can I see that each phase has ripples at different time?
When you say 'ordinary PWM ' you are correct. The topology and control of the typical CPU PS is in principle, textbook ordinary. Pulse width modulation is the means of control. A constant frequency is used. You measured it at 160 kHz - typical for a small switch mode regulator. The PW as a percentage of cycle time is varied to give control. Wider on-time PW gives higher output voltage; narrower PW gives lower output voltage.
A single regulator could be used. That would equate to 2 mosfets and 1 toroid inductor. (In higher voltage applications, the second mosfet would just be a diode.)
The single regulator is not used in these apps because it had serious disadvantages, including very high ripple. This is like a single cylinder car engine - too much vibration. More cylinders, less vibration. More phases, less ripple.
Like engine designers, who never have all cylinders fire at the same time, the regulator designers make the multiple regulator legs take turns also. If you sync the scope on one point, you will see the sequential timing that is used.
A single regulator could be used. That would equate to 2 mosfets and 1 toroid inductor. (In higher voltage applications, the second mosfet would just be a diode.)
The single regulator is not used in these apps because it had serious disadvantages, including very high ripple. This is like a single cylinder car engine - too much vibration. More cylinders, less vibration. More phases, less ripple.
Like engine designers, who never have all cylinders fire at the same time, the regulator designers make the multiple regulator legs take turns also. If you sync the scope on one point, you will see the sequential timing that is used.
<< When you say 'ordinary PWM ' you are correct. The topology and control of the typical CPU PS is in principle, textbook ordinary. Pulse width modulation is the means of control. A constant frequency is used. You measured it at 160 kHz - typical for a small switch mode regulator. The PW as a percentage of cycle time is varied to give control. Wider on-time PW gives higher output voltage; narrower PW gives lower output voltage.
A single regulator could be used. That would equate to 2 mosfets and 1 toroid inductor. (In higher voltage applications, the second mosfet would just be a diode.)
The single regulator is not used in these apps because it had serious disadvantages, including very high ripple. This is like a single cylinder car engine - too much vibration. More cylinders, less vibration. More phases, less ripple.
Like engine designers, who never have all cylinders fire at the same time, the regulator designers make the multiple regulator legs take turns also. If you sync the scope on one point, you will see the sequential timing that is used. >>
I heard something about ECS K7S5A introducing pollution into PSU power. I have a screenshot from scope and it has spikes about every 8microseconds on every voltage. Isn't this caused by crappy board regulator?
Hiwire, jerboy Thank you both, excellent info, just what I was hoping to get.
Jerboy could you possibly post that screen shot?
Jerboy could you possibly post that screen shot?
Here is the O'scope displayof the power supply output that Jerboy sent me. Do you suppose that this is unusual or do all multiphase regulators show this type of spike?
Measurement technique can play a part here. Sometimes, to get rid of ground loop effects, it is necessary to use a differential mode - 2 scope probes - 1 on the voltage under test and 1 at a ground point close by. Put both probes on the voltage point first; if the probe are equally compensated, the sig will be quite clean. Then move one probe to a ground point close to the measurement and see what you've got. In any case, sync the scope externally from a point relevant to what you are looking for. If you are looking for trash produced by the CPU regulator, sync the scope to any one of the phases.
The computer power supply is a switcher, and always has some hash on it - usually 50 mv or so. The load from the board also drives some noise back to the input lines. You should see evidence of noise from both sources. If I clearly saw 3 subtle steps in the trace I could interpret it as being from the CPU reg.
In any case, noise in the range of 100 mv is not unusual. All switching regs produce some noise; a regulator with say 3 instead of 2 phases should produce a bit less noise in both the source feeding it and the load, other things being equal.
The computer power supply is a switcher, and always has some hash on it - usually 50 mv or so. The load from the board also drives some noise back to the input lines. You should see evidence of noise from both sources. If I clearly saw 3 subtle steps in the trace I could interpret it as being from the CPU reg.
In any case, noise in the range of 100 mv is not unusual. All switching regs produce some noise; a regulator with say 3 instead of 2 phases should produce a bit less noise in both the source feeding it and the load, other things being equal.
<< Measurement technique can play a part here. >>
Well I know what I'm doing and I did it right.
ECS K7S5A is a notorious power polluter relatively speaking.
K7S5A produced 75mV p-p spikes whereas an Abit P3 board only did 15mVp-p noise.
K7S5A is a terrible polluter as you can see on K7S5A mod plan
The power supply did not make any spike noise under resistive dummy load, so it is coming from motherboard regulator.
when loaded with K7S5A
[when loaded with Abit]hopefully coming soon[/lol]
Arent those spikes typical for switching regulators with RC-load. Whenever the voltage over the capacitor drops, it gives a spike in current (I = C*dV/dt), which results in a voltage spike in the resistors.
Correct me if 'm wrong,
Menel.
Correct me if 'm wrong,
Menel.
Yes, Jerboy, it's not about generalities here any more. Right? We are talkin' trouble-shootin' with a capital T.
I finally realize that you have been hacking away at this specific thing for a while. So, I took your scope shot and communed with it for a while. As a result, I think I know pretty much now what is happening and why. And, I take it, that what we are looking at in the shot is typical of most ECS K7S5A boards. OK so far?
There IS a problem. ECS may think it is minor. To pursue this further and clear up a couple of questions, it would be helpful if you took another good scope shot with a regulator phase event on the lower trace. The gate of a low side mosfet or the high side of any CPU reg inductor will due. Then a partial schematic of the reg is in order.
At this point though, maybe one should say adieu to the problem if it is ECS's problem. This is usually the point where the company is contacted and a research contract is negotiated - even if it is only for a few free boards. Or maybe it's for points in the next world - that sort of thing. In any event, schematics and other data are all belong to ECS. Engineering a fix for this board with only a scope and a screw driver could be much more work than it needs to be - or for the easily entertained, a hobby all by itself.
I finally realize that you have been hacking away at this specific thing for a while. So, I took your scope shot and communed with it for a while. As a result, I think I know pretty much now what is happening and why. And, I take it, that what we are looking at in the shot is typical of most ECS K7S5A boards. OK so far?
There IS a problem. ECS may think it is minor. To pursue this further and clear up a couple of questions, it would be helpful if you took another good scope shot with a regulator phase event on the lower trace. The gate of a low side mosfet or the high side of any CPU reg inductor will due. Then a partial schematic of the reg is in order.
At this point though, maybe one should say adieu to the problem if it is ECS's problem. This is usually the point where the company is contacted and a research contract is negotiated - even if it is only for a few free boards. Or maybe it's for points in the next world - that sort of thing. In any event, schematics and other data are all belong to ECS. Engineering a fix for this board with only a scope and a screw driver could be much more work than it needs to be - or for the easily entertained, a hobby all by itself.
Yay! I just got my own hosting space. (WARNING - Its reallllllly slow)I can't make any specific right or wrong statement here so I'll let you guys be the judge.
This is a relative comparision
loaded with K7S5A
5V line of same power supply loaded with Abit and PIII
<< Yes, Jerboy, it's not about generalities here any more. Right? We are talkin' trouble-shootin' with a capital T.
I finally realize that you have been hacking away at this specific thing for a while. So, I took your scope shot and communed with it for a while. As a result, I think I know pretty much now what is happening and why. And, I take it, that what we are looking at in the shot is typical of most ECS K7S5A boards. OK so far?
There IS a problem. ECS may think it is minor. To pursue this further and clear up a couple of questions, it would be helpful if you took another good scope shot with a regulator phase event on the lower trace. The gate of a low side mosfet or the high side of any CPU reg inductor will due. Then a partial schematic of the reg is in order.
At this point though, maybe one should say adieu to the problem if it is ECS's problem. This is usually the point where the company is contacted and a research contract is negotiated - even if it is only for a few free boards. Or maybe it's for points in the next world - that sort of thing. In any event, schematics and other data are all belong to ECS. Engineering a fix for this board with only a scope and a screw driver could be much more work than it needs to be - or for the easily entertained, a hobby all by itself. >>
This is a relative comparision
loaded with K7S5A
5V line of same power supply loaded with Abit and PIII
<< Yes, Jerboy, it's not about generalities here any more. Right? We are talkin' trouble-shootin' with a capital T.
I finally realize that you have been hacking away at this specific thing for a while. So, I took your scope shot and communed with it for a while. As a result, I think I know pretty much now what is happening and why. And, I take it, that what we are looking at in the shot is typical of most ECS K7S5A boards. OK so far?
There IS a problem. ECS may think it is minor. To pursue this further and clear up a couple of questions, it would be helpful if you took another good scope shot with a regulator phase event on the lower trace. The gate of a low side mosfet or the high side of any CPU reg inductor will due. Then a partial schematic of the reg is in order.
At this point though, maybe one should say adieu to the problem if it is ECS's problem. This is usually the point where the company is contacted and a research contract is negotiated - even if it is only for a few free boards. Or maybe it's for points in the next world - that sort of thing. In any event, schematics and other data are all belong to ECS. Engineering a fix for this board with only a scope and a screw driver could be much more work than it needs to be - or for the easily entertained, a hobby all by itself. >>
Oh new links
http://www.geocities.com/jerboiat/AT/
Pic 1 is ECS K7S5A
Pic 2 is Abit running P3
The power supply in both tests are Enlight 340W running on 120V 60Hz United States domestic power.
http://www.geocities.com/jerboiat/AT/
Pic 1 is ECS K7S5A
Pic 2 is Abit running P3
The power supply in both tests are Enlight 340W running on 120V 60Hz United States domestic power.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter