Im designing a regulator board and a suggestion from a designer was to have a capacitor at the start worth between 50k - 100k uf capacitance.
All of the capacitors i could find that met the capacitance and voltage were extremely expensive and enormous in size so i was wondering if its possible for me to put 3 capacitors in series to meet the requirements. The capacitance of each is high enough that when i put 3 in series its JUST under what im looking for and the voltage across all 3 is acceptable.
they are electrolytic capacitors however. i know you can put 2 in series in opposite polarities but will it work if you put 3 electrolytic capacitors in series? i cant really find any info about this.
All of the capacitors i could find that met the capacitance and voltage were extremely expensive and enormous in size so i was wondering if its possible for me to put 3 capacitors in series to meet the requirements. The capacitance of each is high enough that when i put 3 in series its JUST under what im looking for and the voltage across all 3 is acceptable.
they are electrolytic capacitors however. i know you can put 2 in series in opposite polarities but will it work if you put 3 electrolytic capacitors in series? i cant really find any info about this.
That sounds like an awful lot of capacitor, and unless you need many, many amps of load current is almost certainly gross overkill and will cause very high peak currents in the rectifiers.
What are the parameters of this regulator, (Input voltage min & max, output voltage, max output current, ripple and noise specs)?
You do know that three caps in series will give you 1/3rd of the capacitance of a single cap?
Series chains are possible, but you must use leakage equalisation resistors, and it is generally only worth doing in high voltage supplies (Where you would not likely be using anything like that much capacitance).
Regards, Dan.
What are the parameters of this regulator, (Input voltage min & max, output voltage, max output current, ripple and noise specs)?
You do know that three caps in series will give you 1/3rd of the capacitance of a single cap?
Series chains are possible, but you must use leakage equalisation resistors, and it is generally only worth doing in high voltage supplies (Where you would not likely be using anything like that much capacitance).
Regards, Dan.
caps in series do NOT add up in value! Google it.
You need to connect them in parallel to add the values.
In any case, 100,000 uF is an ENORMOUSLY large value in any circuit. What is the application? Your "designer" might have been having a bad day.
You need to connect them in parallel to add the values.
In any case, 100,000 uF is an ENORMOUSLY large value in any circuit. What is the application? Your "designer" might have been having a bad day.
Electrolytic capacitors, like batteries, are polarized. Ohms Law states that any number or resistors in series has a resistance of the sum of the resistors. Exactly opposite for capacitors!
Now Three capacitors of equal value, (say for example 10,000uF 50Volts) in series will give a value of 1/3rd capacitance at thee times the voltage. In other words 3,333.33uF at 150Volts. Three 10,000uF 50Volts in parallel will give 30,000uF at 50Volts. I won't explain the ESR value but you will have got the theory and practical from my example, I hope.
Now Three capacitors of equal value, (say for example 10,000uF 50Volts) in series will give a value of 1/3rd capacitance at thee times the voltage. In other words 3,333.33uF at 150Volts. Three 10,000uF 50Volts in parallel will give 30,000uF at 50Volts. I won't explain the ESR value but you will have got the theory and practical from my example, I hope.
A regulator board should not start with a massive capacitance. Assuming that is the main reservoir cap, it should be with the rectifiers not the regulator.
High power PSU design is for experts only. Experts don't ask this sort of question.
High power PSU design is for experts only. Experts don't ask this sort of question.
ok so let me explain...
the capacitors are 220k uf 6.3 volts each. with the 1/3 capacitance it comes out to 80k uf.
i didnt mean im adding 3 low level capacitors in series like everyone has assumed.
The regulators also have the rectifiers on board. im going from one transformer to several regulator boards and each of them have their own rectifier section at the start.
The designer who mentioned using the high capacitance figure at the start was john swenson from Uptone audio.
Here is the entire post:
4est has nailed it. Particularly for computer use you need a supply that can respond to a VERY wide range transient loads, from single digit Hz. up to many MHz. Audiophile capacitors simply don't do this. There is NO single capacitor that does this, it doesn't exist. It takes multiple capacitors of different characteristics to do this well.
Unfortunately the common response to this is to parallel several different capacitors right at the output, this doesn't work either. The inductances of the different caps interact with each other causing high impedance at multiple frequencies.
So what do you do? I call it the hierarchy of charge. Put small very low inductance caps right at the load, those get fed by a regulator that has a solid polymer cap in the mid hundreds uF as its output cap (no more than that, anything bigger is too slow). That regulator is fed by some large traditional good quality electrolytics. For a normal type cap input filter you want a pre-regulator to deal with what comes out of this filter.
The two regulators provide isolation between different capacitor types, so that you can properly to the "very big slow caps, big medium speed caps, small very fast, and very small VERY fast caps".
The very fast cap at the end can respond VERY quickly to load transients, but it can't do it for long. As it is running out of charge the upstream cap is starting to provide charge to keep things going, as THAT cap is running out, the next up the line can take over. This sequence can handle with ease anything the computer can throw at it.
For building a power supply feeding an existing digital system (computer etc) you don't have any choice about the lowest level, it is already done. For this system I would use a single solid polymer cap as the output of the supply. The regulator driving that should be a fast regulator, with 10,000uF or so cap at ITS input. The next level up should be a regulator fed by 50,000 to 100,000uF. The first reg does not have to be as fast as the second one.
The thermal characteristics of the system are important. If you are running 5A or something you need the raw voltage to be high enough so it doesn't get below the cutoff of the first reg. But the second reg is being fed by a reg so the voltage difference between them can be small. Thus the first reg tends to dissipate quite a bit more heat than the second one.
Something like this will be an extremely good source for a computer.
John S.
the capacitors are 220k uf 6.3 volts each. with the 1/3 capacitance it comes out to 80k uf.
i didnt mean im adding 3 low level capacitors in series like everyone has assumed.
The regulators also have the rectifiers on board. im going from one transformer to several regulator boards and each of them have their own rectifier section at the start.
The designer who mentioned using the high capacitance figure at the start was john swenson from Uptone audio.
Here is the entire post:
4est has nailed it. Particularly for computer use you need a supply that can respond to a VERY wide range transient loads, from single digit Hz. up to many MHz. Audiophile capacitors simply don't do this. There is NO single capacitor that does this, it doesn't exist. It takes multiple capacitors of different characteristics to do this well.
Unfortunately the common response to this is to parallel several different capacitors right at the output, this doesn't work either. The inductances of the different caps interact with each other causing high impedance at multiple frequencies.
So what do you do? I call it the hierarchy of charge. Put small very low inductance caps right at the load, those get fed by a regulator that has a solid polymer cap in the mid hundreds uF as its output cap (no more than that, anything bigger is too slow). That regulator is fed by some large traditional good quality electrolytics. For a normal type cap input filter you want a pre-regulator to deal with what comes out of this filter.
The two regulators provide isolation between different capacitor types, so that you can properly to the "very big slow caps, big medium speed caps, small very fast, and very small VERY fast caps".
The very fast cap at the end can respond VERY quickly to load transients, but it can't do it for long. As it is running out of charge the upstream cap is starting to provide charge to keep things going, as THAT cap is running out, the next up the line can take over. This sequence can handle with ease anything the computer can throw at it.
For building a power supply feeding an existing digital system (computer etc) you don't have any choice about the lowest level, it is already done. For this system I would use a single solid polymer cap as the output of the supply. The regulator driving that should be a fast regulator, with 10,000uF or so cap at ITS input. The next level up should be a regulator fed by 50,000 to 100,000uF. The first reg does not have to be as fast as the second one.
The thermal characteristics of the system are important. If you are running 5A or something you need the raw voltage to be high enough so it doesn't get below the cutoff of the first reg. But the second reg is being fed by a reg so the voltage difference between them can be small. Thus the first reg tends to dissipate quite a bit more heat than the second one.
Something like this will be an extremely good source for a computer.
John S.
Gross overkill is what that is!
By the time you have even a few inches of wire between the regulator and the processor board, you have so much lead inductance that nothing 'fast' is going to happen at the regulator, and the reg can do NOTHING about the many tens of mV of noise on the motherboard power planes simply because it is nothing like near enough (Fortunately PDN designers understand this and handle all the fast stuff locally on the board).
If you are having an off board power supply with many MHz of loop bandwidth you are almost certainly just going to be stuffed by the lead inductance, so there really is no point.
Large amounts of very low ESR cap can actually be a liability because even a few tens of nH of series L can form an unpleasantly high Q resonance at a surprisingly low frequency, you sometimes see a large cap with deliberate series resistance used to damp these resonances, or sometimes a cap with a carefully chosen ESR gets the job done.
You really want to be very careful about those sorts of cap values at a regulator output anyway, you might get away with it with a old school NPN pass transistor job, but a modern PMOS LDO is likely to be become unstable.
I will admit to having some trouble figuring out why you are not just using a switcher here, a computers PDN is so horrifically noisy simply due to the nature of the system, that the additional hash from a decently designed switchmode is lost in the noise.
Seriously PDN design for high speed logic is much more then simply designing some audio speed power supply with insane capacitor values, and most of the interesting stuff is on the motherboard (Because it has to be, a few mm of trace inductance stuffs you with modern logic).
Regards, Dan.
By the time you have even a few inches of wire between the regulator and the processor board, you have so much lead inductance that nothing 'fast' is going to happen at the regulator, and the reg can do NOTHING about the many tens of mV of noise on the motherboard power planes simply because it is nothing like near enough (Fortunately PDN designers understand this and handle all the fast stuff locally on the board).
If you are having an off board power supply with many MHz of loop bandwidth you are almost certainly just going to be stuffed by the lead inductance, so there really is no point.
Large amounts of very low ESR cap can actually be a liability because even a few tens of nH of series L can form an unpleasantly high Q resonance at a surprisingly low frequency, you sometimes see a large cap with deliberate series resistance used to damp these resonances, or sometimes a cap with a carefully chosen ESR gets the job done.
You really want to be very careful about those sorts of cap values at a regulator output anyway, you might get away with it with a old school NPN pass transistor job, but a modern PMOS LDO is likely to be become unstable.
I will admit to having some trouble figuring out why you are not just using a switcher here, a computers PDN is so horrifically noisy simply due to the nature of the system, that the additional hash from a decently designed switchmode is lost in the noise.
Seriously PDN design for high speed logic is much more then simply designing some audio speed power supply with insane capacitor values, and most of the interesting stuff is on the motherboard (Because it has to be, a few mm of trace inductance stuffs you with modern logic).
Regards, Dan.
Most Power Supplies for Computers use a SMPS and detects very small and fast changes in the load, topping up almost instantaneously the voltages. These voltages must be very stable and indeed are, why do you think there are so many wires. The maximum valued capacitor used in this 920 Watt version I am now looking at is 2u2F. Why go backwards when others have a better solution? Use error detection at the power drain by running a connection from the sensitive points ensuring the fast acting regulator maintains the power where is should be, not at the power source.
I don't see the point of having huge caps and a regulator. They are more likely to fight each other than aid each other, as regulator output impedances are often inductive. The piece quoted from John Swenson reads like a set of audiophile myths arranged end to end.
Regulators do DC to lowish frequencies. Add smallish value caps to do high frequencies. Job done.
The lowest frequency present in a computer power draw is not "single digit Hz" but the inverse of the time since it was last booted - typically microHz. The highest frequency could be a few GHz. Fortunately, you don't need to worry about this as the computer designer knows how to ensure data integrity and that is all an audio computer needs to do. If you rely on a computer for timing information then you have already decided to use the wrong system so no point then in trying to make a silk purse out of a sow's ear.
Regulators do DC to lowish frequencies. Add smallish value caps to do high frequencies. Job done.
The lowest frequency present in a computer power draw is not "single digit Hz" but the inverse of the time since it was last booted - typically microHz. The highest frequency could be a few GHz. Fortunately, you don't need to worry about this as the computer designer knows how to ensure data integrity and that is all an audio computer needs to do. If you rely on a computer for timing information then you have already decided to use the wrong system so no point then in trying to make a silk purse out of a sow's ear.
Caps can be used as singles, or in series, or in parallel, and finally in a series/parallel combination.
For a 20V 100mF rating you could use a 25V 100mF, or two 25V 50mF, or two 15V 200mF, or four 15V 100mF.
The combinations will perform differently.
I would use the paralleled combination where massive capacitance is required.
I use back to back in series where an AC rating is required from a polarised electrolytic.
I have used back to back in parallel with front to front for AC coupling and cannot tell them apart from metallised film passing audio signals.
I would be very resistant to using a series connected for increased DC rating because it relies on calculated voltage balancing resistors to help ensure no capacitor is exposed to excessive voltage.
22mF and 33mF capacitors are available for voltage from 50V and lower. These can make good financial solutions and perform better than one enormous capacitor.
For a 20V 100mF rating you could use a 25V 100mF, or two 25V 50mF, or two 15V 200mF, or four 15V 100mF.
The combinations will perform differently.
I would use the paralleled combination where massive capacitance is required.
I use back to back in series where an AC rating is required from a polarised electrolytic.
I have used back to back in parallel with front to front for AC coupling and cannot tell them apart from metallised film passing audio signals.
I would be very resistant to using a series connected for increased DC rating because it relies on calculated voltage balancing resistors to help ensure no capacitor is exposed to excessive voltage.
22mF and 33mF capacitors are available for voltage from 50V and lower. These can make good financial solutions and perform better than one enormous capacitor.
Last edited:
Series electrolytics are commonly used for high voltage supplies, where a single cap with enough voltage rating is either unavailable or too expensive. No need for low voltage supplies.
- Status
- Not open for further replies.