This is something I have been trying to research and still don't have answers so let see what everyone else does.
Lets say your building a power supply, maybe this
Capacitance Multiplier design. And as is often done, you use say 2, 3 or maybe 4 10,000 uF or some such arrangement for the main input filter caps. What is the best method for connecting them? Does it matter? If this were a regulated PS would the method matter more or less? I lean towards using copper plates, maybe 3-4mm thick with wires from one end into the circuit. What about snap-in caps with a PCB design? Any better?
Pictures of what your methods would be cool.
-John
Lets say your building a power supply, maybe this
Capacitance Multiplier design. And as is often done, you use say 2, 3 or maybe 4 10,000 uF or some such arrangement for the main input filter caps. What is the best method for connecting them? Does it matter? If this were a regulated PS would the method matter more or less? I lean towards using copper plates, maybe 3-4mm thick with wires from one end into the circuit. What about snap-in caps with a PCB design? Any better?
Pictures of what your methods would be cool.
-John
Most important for low ripple is to take the output closest to the caps plates. This is so you don't include I^2*R charging pulses. They actually make 4 terminal caps that do this (more for SMPS) These have seperate input/out with the common connections made internally.
So I would just replicate this with either wires or PCB traces.
For multiple caps just use a seperate common input and output. Best results would make use of identical caps with symetrical traces (perhaps using the same/similar pattern for top as well bottom sides with top being the input and bottom used for out)
So I would just replicate this with either wires or PCB traces.
For multiple caps just use a seperate common input and output. Best results would make use of identical caps with symetrical traces (perhaps using the same/similar pattern for top as well bottom sides with top being the input and bottom used for out)
assuming the the same total smoothing capacitance I would expect many parallel caps to perform as well as a single good cap in most respects.
In some respects the parallel set up will perform better.
But, to get the benefit of that improved performance the inductance of the leads connecting the caps must be reduced to an absolute minimum.
One can never get as low as the pin spacing of a single low ESL cap but that is what you are striving for.
Take three caps in parallel connect all the +ves together, connect all the -ves together. Connect the output of this triplet from the middle cap. The longest connection from the outlet to the farthest cap is one cap diameter. That defines the inductance of that leg. The other remote cap is also one diameter away. These two caps in parallel will have an effective inductance roughly equivalent to the pin pitch spacing of a single cap of the same size. Now add in the parallel inductance of the middle cap defined by it's pin pitch and the resulting triplet inductance will be less than any bigger (standard ESL) cap with that pin pitch or greater.
Similarly the ripple current capability of parallel sets is usually better than a big single. I suspect much of this id down to the extra surface area of the multi cap arrangement and it's ability to dissipate heat.
ESR is also reduced compared to a single.
Take all these obvious effects into account and you may find that parallel cheap (low spec) caps can perform as well as an expensive premium spec single cap.
comments please.
In some respects the parallel set up will perform better.
But, to get the benefit of that improved performance the inductance of the leads connecting the caps must be reduced to an absolute minimum.
One can never get as low as the pin spacing of a single low ESL cap but that is what you are striving for.
Take three caps in parallel connect all the +ves together, connect all the -ves together. Connect the output of this triplet from the middle cap. The longest connection from the outlet to the farthest cap is one cap diameter. That defines the inductance of that leg. The other remote cap is also one diameter away. These two caps in parallel will have an effective inductance roughly equivalent to the pin pitch spacing of a single cap of the same size. Now add in the parallel inductance of the middle cap defined by it's pin pitch and the resulting triplet inductance will be less than any bigger (standard ESL) cap with that pin pitch or greater.
Similarly the ripple current capability of parallel sets is usually better than a big single. I suspect much of this id down to the extra surface area of the multi cap arrangement and it's ability to dissipate heat.
ESR is also reduced compared to a single.
Take all these obvious effects into account and you may find that parallel cheap (low spec) caps can perform as well as an expensive premium spec single cap.
comments please.
After looking at the pictures, all I see are power supplies, using the transistors as some type of voltage regulator.
If you use a Darlington transistor as a regulator ( making the appropriate design allowances ) the ripple will be very very low.
As for connecting the caps....
in parallel....voltage rating stays the same, capacitance adds. HOWEVER......increasing the capacitance also increases the startup load on the transformer because of the 4Tau rule. ( the cap appears as almost a dead short longer, drawing more current from the transformer, making it work harder on startup ). For large caps you need a stout transformer. Using a big cap on a small transformer is cheating, and it will be only a matter of time till the windings pack it in.
In series...voltage increases, capacitance halves ( if you have say 2- 1000uF caps in series, then the net capacitance is 500 uF ).
Capacitance in series :
1/total capacitance = 1/ cap A + 1/ cap B + 1/cap C
where Cap A is capacitor A, cap B is capacitor B, etc, etc...
The other thing about large capacitors ( 10000 uF+ ), is that they discharge slowly.
That means they will deliver a lot of charge, but at a slower rate, because the internal resistance of the cap itself goes up.
The way to combat this is to use 10- 1000uF caps in parallel ( instead of 1-10000 uF cap ) if you need a lot of current in a hurry.
The bad side of this is that the caps appear as a hard short on startup and will need a VERY stout transformer, or some type of soft start circuit. .
The bottom line is you cant cheat and get something for nothing.
Large capacitances need large transformers to drive them.
Hope this helps...
Rob
If you use a Darlington transistor as a regulator ( making the appropriate design allowances ) the ripple will be very very low.
As for connecting the caps....
in parallel....voltage rating stays the same, capacitance adds. HOWEVER......increasing the capacitance also increases the startup load on the transformer because of the 4Tau rule. ( the cap appears as almost a dead short longer, drawing more current from the transformer, making it work harder on startup ). For large caps you need a stout transformer. Using a big cap on a small transformer is cheating, and it will be only a matter of time till the windings pack it in.
In series...voltage increases, capacitance halves ( if you have say 2- 1000uF caps in series, then the net capacitance is 500 uF ).
Capacitance in series :
1/total capacitance = 1/ cap A + 1/ cap B + 1/cap C
where Cap A is capacitor A, cap B is capacitor B, etc, etc...
The other thing about large capacitors ( 10000 uF+ ), is that they discharge slowly.
That means they will deliver a lot of charge, but at a slower rate, because the internal resistance of the cap itself goes up.
The way to combat this is to use 10- 1000uF caps in parallel ( instead of 1-10000 uF cap ) if you need a lot of current in a hurry.
The bad side of this is that the caps appear as a hard short on startup and will need a VERY stout transformer, or some type of soft start circuit. .
The bottom line is you cant cheat and get something for nothing.
Large capacitances need large transformers to drive them.
Hope this helps...
Rob
are you sure?musinteg said:
The short term overload capacity of a transformer is enormous. A few tens of milliseconds of overcurrent will not raise the winding temperature nor core temperature significantly.
are you sure?musinteg said:
Have you tried shorting across a charged capacitor with a screwdriver?
are you sure?musinteg said:
Many amps are built with transformers that appear to be too low in rating. Reasonable performance is ensured by carefull attention to the compromises that are set to achieve economy/weight/volume restrictions.
We don't have too many transformers that fail if the equipment is NOT abused.
Simply providing adequate smoothing capacitance is not abuse in my book.
Thanks for the answers eveyone. Great information..
So using 5-mm thick copper bars for connecting parallel capacitors may not be the best idea? Is this just because of I^2*R charging pulse issues and inductance and spacing or are there other factors? Its very easy to get caught up in the more is better routine and I'm trying to break that habit!
From what I understand then, preference would be to use a PCB to connect identical capacitors in parallel which effectively increases plate size and reduces ESR (and of course increased capacitance). Assuming the caps are too big to be on the same PCB as the PSU, you would then wire from the middle cap to the PSU circuit. Is it fair to assume that this wire should be kept as short as possible to keep inductance down?
If you use a PCB to connect caps in parallel, are there any guidelines as to how big the traces would need to be on a PCB? I know its a function of voltage and amps but is there more to consider such as inductance?
I see lots of pictures of people using bars between parallel caps so, without knowing the pros and cons, and thinking is actually looks kinda cool, I just did it the same way.
So using 5-mm thick copper bars for connecting parallel capacitors may not be the best idea? Is this just because of I^2*R charging pulse issues and inductance and spacing or are there other factors? Its very easy to get caught up in the more is better routine and I'm trying to break that habit!
From what I understand then, preference would be to use a PCB to connect identical capacitors in parallel which effectively increases plate size and reduces ESR (and of course increased capacitance). Assuming the caps are too big to be on the same PCB as the PSU, you would then wire from the middle cap to the PSU circuit. Is it fair to assume that this wire should be kept as short as possible to keep inductance down?
If you use a PCB to connect caps in parallel, are there any guidelines as to how big the traces would need to be on a PCB? I know its a function of voltage and amps but is there more to consider such as inductance?
I see lots of pictures of people using bars between parallel caps so, without knowing the pros and cons, and thinking is actually looks kinda cool, I just did it the same way.
jscherer said:
Works for me and alot of other DIYs (hey when you can't pass up good deals on surplus snap-in caps.)
jscherer said:
Big computer grade caps with bus bars look very cool and industrial looking. Looks are just as important for DIY as working well, so then I say you got the best of both worlds then.
I don't normally discharge 10000uF capacitors with a screwdriver. And yes, if you work on something like a nice Adcom 555, with 30000 uF of capacitance on each rail, charged up to 90 volts DC, you would get a face full of molten copper and steel if you did discharge them with a screwdriver.
NOW....lets not be silly. No-one in their right mind does such a dangerous thing Period. Thats what alligator clips and resistors are for.
By discharging slowly, I am talking in the amount of current that can be delivered within the first RC time constant. Smaller capacitors, have inherantly lower internal resistances.
Paralleling resistances gets you a smaller one in total for the capacitor network. Therefore you are able to deliver the same amount of current as a bigger cap in a much shorter period of time. Amplifiers that can do that have ( usually ) very good bass control.
As far as capacitances and transformer windings go, when you have a nice piece of old gear, such as a Carver or Mac, or an old HK, taking a chance and cooking the windings on a transformer that you can no longer get is foolish at best.
The idea ( as far as I can see ) is to improve the sound without overstressing the components. I have several unhappy carver owners, with cooked transformers that are no longer readily replaceable for a reasonable price. And that is due to heat and old age. All power supplies, are designed from the factory to deliver the rated current , at a specified price point.
Can they be improved.....yes. Is there a penalty involved.....yes. The penalty is longevity.
Its like my 2l honda engine. It is 150 horsepower. If I just do proper routine maintenance, it will last for 300,000 kM or more. If I hot-rod the engine up to 300 horsepower, with a nitrous shot to 550, the short term performance will be spectacular, but the engine will be burnt out much sooner.
I can see no need to risk burning out a transformer by connecting it to huge capacitances that it was never designed to drive. Personally, I would rather have good sound quality on a daily basis, than a spectacular but expensive memory.
RG
NOW....lets not be silly. No-one in their right mind does such a dangerous thing Period. Thats what alligator clips and resistors are for.
By discharging slowly, I am talking in the amount of current that can be delivered within the first RC time constant. Smaller capacitors, have inherantly lower internal resistances.
Paralleling resistances gets you a smaller one in total for the capacitor network. Therefore you are able to deliver the same amount of current as a bigger cap in a much shorter period of time. Amplifiers that can do that have ( usually ) very good bass control.
As far as capacitances and transformer windings go, when you have a nice piece of old gear, such as a Carver or Mac, or an old HK, taking a chance and cooking the windings on a transformer that you can no longer get is foolish at best.
The idea ( as far as I can see ) is to improve the sound without overstressing the components. I have several unhappy carver owners, with cooked transformers that are no longer readily replaceable for a reasonable price. And that is due to heat and old age. All power supplies, are designed from the factory to deliver the rated current , at a specified price point.
Can they be improved.....yes. Is there a penalty involved.....yes. The penalty is longevity.
Its like my 2l honda engine. It is 150 horsepower. If I just do proper routine maintenance, it will last for 300,000 kM or more. If I hot-rod the engine up to 300 horsepower, with a nitrous shot to 550, the short term performance will be spectacular, but the engine will be burnt out much sooner.
I can see no need to risk burning out a transformer by connecting it to huge capacitances that it was never designed to drive. Personally, I would rather have good sound quality on a daily basis, than a spectacular but expensive memory.
RG
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.