in ATX power supplies the more capacitance the less voltage

[laser]

I have an Antec SP450 atx power supply in my PC and was very weak in the 3.3v line
the motherboard was reporting 3.1 v even sometimes the voltage dropped to 3.0v when I use my TV card or playing games
My PC is very power-hungry it have a core i5 95w,geforce GTX 560 (75w ), soundblaster X-FI ,TV card etc

OK trying to solve this I remember this power supply originally had on the 3.3 v rail comes with a fujityu 3300 uf buggy cap which I replaced with a rubycon 3300 mbz series

Then to solve the weakness in the 3.3v line the first step I did was removing the 3300 rubycon cap for a 4700uf chemicon KZE

now the voltage of the 3.3 v line was worst than before dropped to 3.0v even to 2.9v

Then automatically I estimated was so capacitance for such 3.3v line which my power supply couldn't handle
so I replaced the 4700uf cap by a 2200 uf chemicon KZE and voltage raised to 3.2v and I was surprised but yet nor very happy with the results I replaced the 2200uf cap by a 1500 uf rubycon MBZ

and yes...now the result is 3.27 v which is very good...yet it dropped to 3.2v on stressing the computer anyways this very good compared to using a 3300 uf cap

Resuming here is the capacitance/voltage comparison ratio in the 3.3v rail

4700 uf > 3.0 v
3300 uf > 3.1 v
2200 uf > 3.2 v
1500 uf > almost 3.3 v

conclusions:

1:the more capacitance the more ripple but the less voltage
2:sometimes you need to low the original capacitance in 3.3v rail to get descent voltage (maybe this apply to 5v rail as well)

[mariushm]

Re: in ATX power supplies the more capacitance the less voltage

I may be wrong but I believe this SP450 power supply is group regulated, meaning the 5v and 3.3v are "linked" together.
If the computer draws more power from 5v, the power supply may increase the voltage a bit and in turn the voltage on 3.3v will also increase slightly.

My PC is very power-hungry it have a core i5 95w,geforce GTX 560 (75w ), soundblaster X-FI ,TV card etc

Well, you see, that's not really that power hungry, and another problem is that allmost all of that power is actually taken from 12v which is basically independent of the 5v and 3.3v outputs.

The processor uses 12v, the video card uses 12v, the ram gets power from 12v, the hard drive motor uses 12v, the fans use 12v ...

Only the sound card and the tv tuner uses 5v, the electronics on the hdd use 5v, maybe the onboard network card uses 5v... but all this is very little... maybe less than 3A on the 5v. This is probably too little to get the power supply to actually boost its output on 5v which in turn would boost the 3.3v rail a bit.

The 3.3v rail is almost not used at all. Some motherboards still use it for memory but it's much easier in newer designs to just get the ram power from cpu voltage regulator.. some still use 3.3v for powering the chipset and so on .. but again, it's only 1-2A probably... it's too little for the power supply to actually detect some load and boost the voltage to proper levels.

The Rubycon MBZ is ultra low esr and is not recommended for power supplies... most designs don't like such ultra low esr.

The KZE is proper for power supplies and has a bit of esr (it's still very low esr)... that tiny amount of resistance could have been just enough to tip the power supply just a bit to increase the voltage to 3.2v

As for the 1500uF mbz.. again mbz is not good for power supplies, in this case the esr was probably a bit higher than the KZE esr simply due to the capacitor can size being smaller.

I would say you drew some incorrect conclusions.

I would recommend putting the 2200uF KZE capacitor back, it's enough for 3.3v usage on modern computers and it's the most suitable capacitor for power supplies.

As an experiment, I would suggest getting some 12v lightbulbs (motorbike/car bulbs are 12v 10-20w). Then, connect them to 5v on the power supply (black and red wires in a hdd molex connector)... each lightbulb would use about 1-2 A of power on 5v. You should see that as you add lightbulbs and increase the load on 5v, the 3.3v should go up a bit as well.

[laser]

Re: in ATX power supplies the more capacitance the less voltage

@mariushm

I didn't drew some incorrect conclusions
please note that is very clear in this experiment that adding capacitance drops voltage in the 3.3 v rail
and it seems low esr or ultra low esr it does not change anything
in fact the 4700 uf KZE deliver less voltage than a 3300 uf MBZ and so on

despite you say that most of the things uses 12v, voltage of 12v is always here between 12.5v and 12.6 v ( never drops )

yet some things like TV cards uses 5v or 3.3v because the voltage drops when TV card is in use, also the soundblaster x-fi is very hungry in power, the chip of this soundcard works hot like the northbridge of my motherboard

you recommended me the 2200 KZE ....yes it works fine...(I tested it ) but the 1500 MBZ works better because deliver more voltage

now I'm convinced that if I insert a 1000 uf cap in the 3.3v rail the voltage yet will increase a bit...but I will not do it because I sense 1000 uf is not enough for a descent ripple

I think that every combination of power supply/ computer is unique and sometimes you must calibrate it to get descent capacitance with good voltage

it seems for the 3.3 v rail something like 1500 or 2200 uf is the ideal, you can not go further because the voltage will fall
3300 uf for the 3.3v rail is an exaggeration...Antec' engineers failed in this calculation, or they done this calculation with those lame fijituyu caps

[shovenose]

Re: in ATX power supplies the more capacitance the less voltage

well I am not the most knowledgable person on this forum but I think you are wrong.
Lowering the capacitance will simply cause you more issues - you should use the same uF and same or slightly higher voltage capacitors as replacement.

Rubycon MBZ is not good for PSU, as far as I can tell.

[ivtec]

Re: in ATX power supplies the more capacitance the less voltage

well I am not the most knowledgable person on this forum but I think you are wrong.
Lowering the capacitance will simply cause you more issues - you should use the same uF and same or slightly higher voltage capacitors as replacement.

Rubycon MBZ is not good for PSU, as far as I can tell.

I'm with you on this one bro.

[laser]

Re: in ATX power supplies the more capacitance the less voltage

well I am not the most knowledgable person on this forum but I think you are wrong.
Lowering the capacitance will simply cause you more issues - you should use the same uF and same or slightly higher voltage capacitors as replacement.

Rubycon MBZ is not good for PSU, as far as I can tell.

I used to believe idem theory as you until yesterday....it's a myth
lowering capacitance increase voltage which is good for stability

example: add more capacitance and my TV card don't works fine ( the TV image start to flash)
TV cards are the first ones to fail in a computer because a bad cap or low voltage

anyways
anyone can test what I'm saying...if anyone have a weak line in atx power supply can try adding or deducting capacitance and you will see my results are right

[mariushm]

Re: in ATX power supplies the more capacitance the less voltage

laser,

The power supply has basically two outputs :

1. 12 v

2. 5v + 3.3v

The power supply monitors the output voltage on 12v and as the current pulled varies, it automatically adjusts the frequency and other things to maintain the 12v to a consistent level.
There is a relation between 12v and the 5v+3.3v pair ... as the load increases on 5v and 3.3v, the voltage on 12v should slowly drift down towards the 12v.

You currently have a high voltage of about 12.5-12.6v because you don't have much of a load on 5v and 3.3v - if you add some load to one of those, you will probably see 12v dropping a bit.
The PSU is just an old design adapter for modern systems - it's still one optimized for 5v loads as you can see from the ratings on the case .. 30A on 5v and 32A on 3.3v. Compare that to modern power supplies that barely give you 20A on each rail.

I can give you an example from a review (which unfortunately doesn't focus on the voltages at various loads):

http://www.silentpcreview.com/article260-page4.html

1. VOLTAGE REGULATION was within the 5% specified by ATX12V throughout the test, and the +3.3V and +5V lines were typically within 2-3%. The +12V line was consistently high throughout the test, but the total variance from high to low was small. The biggest voltage drop on the +12V line was seen at 400W and above, but it still stayed above 12V.

+12V: 12.24 to 12.54
+5V: 4.86 to 5.04
+3.3V: 3.27 to 3.32

I have an old 100w power supply here that is basically the other way around, because the 5v is the primary output and 12v was less important back then : by default, the 5v is at around 5.3v and the 12v stays at about 11.7v BUT as I put a few amps of load on the 5v and at least 1A on 12v, the 5v will get closer to actual 5v (drops down from ~5.3v) and 12v will drift up from around 11.7 to about 11.9v.

So what can I say...

You saw some small variations in the 3.3v output voltage because with such minimal load on the 3.3v AND with minimal load on 3.3v and 5v pair, even the ESR of the capacitors starts to play a part.

You basically had 3300uF MBZ , 2200uF KZE, 1500uF MBZ you don't say the voltages so I'm going to go with 6.3v

MBZ
KZE

4700uF 6.3v KZE : 3290mA ripple, 0.016 ohm esr

3300uF 6.3v MBZ : 2800mA ripple, 0.012 ohm esr

2200uF 6.3v KZE : 2150mA ripple, 0.022 ohm esr

1500uF 6.3v MBZ : 1540mA ripple , 0.026 ohm esr

Do you notice a pattern esr wise? It pretty much follows the voltage drop.

3300uF MBZ < 4700uF KZE < 2200uF KZE < 1500uF MBZ

Put a crappy 3300uF general purpose capacitor with higher ESR and you may find out the voltage stays at 3.4v... it's not about capacitance but other factors.
Put a resistor or something on the 3.3v output, put some load on it, and again it may stabilize closer to 3.3v.
Put some load on 5v and 3.3v rail may come up.

Capacitors don't deliver more voltage ... they just buffer energy and smooth out the DC coming out of them and lower the ripple.
But they act as a very small load due to their ESR/internal resistance.
Also, if you have something in your computer that pulls current in short bursts then when you have a large capacitor it will pull more current when it charges to the maximum, which could affect the behavior and the balance between 3.3v and 5v rails.

ps. and btw...the ATX standards REQUIRE all devices to accept 3.3v and 5v with +/-5% tolerance and 50mV ripple... so if your tuner doesn't work with 3.135v - 3.465v it's a shitty tuner. With 12v, you can have 11.4-12.6 and all's good.

ps2. .... and here's a fuhjkuytrtr capacitors series popular in power supplies :

3300uF 6.3v : 1870mA ripple, 0.035 ohm esr notice how it's larger than the other series above, larger even than the 1500uF mbz capacitor?

ps3 and be careful when advising people to drop the capacitance, because not all people have shitty group regulated power supplies.

I have a Seasonic X-650 power supply which only outputs 12v. The 3.3v and 5v are separately generated from 12v using dc-dc converters and have polymer capacitors on the outputs. It's a quality psu, no need to change capacitors, the voltages will barely vary even when pulling 650w from it.

[ben7]

Re: in ATX power supplies the more capacitance the less voltage

well I am not the most knowledgable person on this forum but I think you are wrong.
Lowering the capacitance will simply cause you more issues - you should use the same uF and same or slightly higher voltage capacitors as replacement.

Rubycon MBZ is not good for PSU, as far as I can tell.

Well the capacitance, if too little, will increase ripple a bit. This is because the capacitance is not enough to smooth out the pulsed DC.

If too large, the power supply feedback loop slows down, and the output voltage has more ripple or is too low/high!

Also, generally, capacitors with larger capacitance will have lower ESR. If the ESR is not where it was designed to be, then the feedback loop in the PSU becomes unstable. This is why bad capacitors commonly make your PC unstable - the voltage fluctuations cause errors in the digital signals and processing.

Although, yes, MBZ might be a little overkill for a PSU.

[laser]

Re: in ATX power supplies the more capacitance the less voltage

@mariushm

thanks the links and for the analysis
and yes all the caps I mentioned are all 6.3v

so your theory is the LOW ESR the less voltage.. ? is interesting
and you say that if I insert a 3300 cap high ESR will work great..OK
I will check this later
it happens in this PSU there is no place to fit caps larger than 1mm in size
and the ones at 1.25mm do not fit well

and in this moment I haven't a 3300uf cap high or normal esr to test this

the ideal could be something like a chemicon LXY 2200 uf >> 0.059 ohm esr
because is 10 mm size....larger caps are at least 12.5 mm

those lame fujitsu caps are very rare cause are rated at 3300uf but only 10 mm in size ???? seems there is not equivalent in 10 mm in the good brands

pending

[ben7]

Re: in ATX power supplies the more capacitance the less voltage

those lame fujitsu caps are very rare cause are rated at 3300uf but only 10 mm in size ???? seems there is not equivalent in 10 mm in the good brands

You mean fuhjyyu capacitors?

[mockingbird]

Re: in ATX power supplies the more capacitance the less voltage

laser --

There's an excellent thread on re-capping the SP450, I suggest you take a look at it, and do a thorough job with the correct parts. If you use MBZ-class capacitors on the SP450, you will cause a massive increase in ripple that will be passed on to the filtering caps on your motherboard and other components, and you will kill them eventually.

[Pentium4]

Re: in ATX power supplies the more capacitance the less voltage

I lower the capacitance when recapping sometimes if the ripple rating is the same or higher

[mariushm]

Re: in ATX power supplies the more capacitance the less voltage

@mariushm

so your theory is the LOW ESR the less voltage.. ? is interesting

No, what I'm saying is THERE IS NO RELATION between voltage and something specific in a capacitor.

Less capacitance means more ripple, more capacitance means less ripple. BUT more capacitance also means slowing down the feedback which could make the feedback system unstable and in turn cause MORE ripple thereby you killed all the benefit you gained by raising the capacitance.

Just the same, the ESR of a capacitor plays a part into the feedback system. Too little ESR can make the feedback system unstable and cause more ripple, too much esr can also cause problems.

On some power supplies, even the LEADS of the capacitors change things. Long leads can act as inductors which can make the feedback oscillate.
Let's say you try to shove a 12.5mm (diameter) capacitor there by leaving the leads half an inch or so above the pcb for clearance and you might draw the conclusion that the voltage changes with the diameter of the capacitors.

The multimeter or tool you use can also read the voltage incorrectly, depending on how many times it measures and how it averages the readings... there's still ripple on the output due to the high frequency used by the power supply. An oscilloscope is often useful to measure the ripple and output voltage at various loads.

So the basic idea is that you can't put an equal sign between the voltage and a particular property of the capacitor. Several things contribute to what you observed.

and you say that if I insert a 3300 cap high ESR will work great..OK
I will check this later

What I was saying is that the 3.3v rail is used so little in your computer that just the small increase in the ESR of a crappy 3300uF could alter the feedback system behavior enough to slightly change the voltage you see on the screen. It was just an idea to prove to you that there's no direct relation between capacitance and voltage.

Your system probably uses so little power that even 1000uF of capacitance would be enough to keep the output voltage between that +/-5% and in general, more than 3300uF of capacitance just isn't needed.

That's why I initially recommended that you put there a 2200uF KZE capacitor and leave it like that - 2200uF KZE is somewhere in the middle, both capacitance and ESR wise

Put some load on the 3.3v (AND 5v), a few Amps on each, to give the feedback system for those rails some meaningful values to work with. Then, play with various capacitors.

it happens in this PSU there is no place to fit caps larger than 1mm in size
and the ones at 1.25mm do not fit well

and in this moment I haven't a 3300uf cap high or normal esr to test this

the ideal could be something like a chemicon LXY 2200 uf >> 0.059 ohm esr
because is 10 mm size....larger caps are at least 12.5 mm

I know, it sucks that you can only fit 10mm diameter caps. But you don't need more than 3300uF and you don't need very low ESR capacitors.
With such minimal load on the 3.3v rail these days you would be fine with 1800-2200uF, which you will find in 10 mm diameter.

In extreme cases - and I'm not advocating this - I've seen electrolytic capacitors replaced with let's say 2700-3300uF 4v polymers which also had a very small resistance in series with a lead to fake a higher esr.

those lame fujitsu caps are very rare cause are rated at 3300uf but only 10 mm in size ???? seems there is not equivalent in 10 mm in the good brands

pending

Don't confuse that fuhjyu$%$6 with fujitsu, which made some good polymer capacitors.

[Wester547]

Re: in ATX power supplies the more capacitance the less voltage

Well the capacitance, if too little, will increase ripple a bit. This is because the capacitance is not enough to smooth out the pulsed DC.

I thought it's more ESR and ripple rating that determines the end ripple voltage (how much ripple can be passed so it doesn't reach the load) and not so much capacitance (which I thought was important for keeping the rails in spec and how long the capacitors can remain in regulation in the event of a power spike or such, but I know voltage regulation depends heavily on group regulated design too or a lack of). But I understand your point.

If RAM now adays uses DC-DC converters from the +12V rail and not the +5V/+3.3V rail then it doesn't seem like those rails are vital for much of anything. I wouldn't be surprised if +12V only/+12VSB power supplies were to usurp power supplies with +5V/+3.3V/-12V/+5VSB outputs sooner than late.