This post is to present a 3 design alternatives to the coupling cap approach that I would like to share.
The traditional coupling cap is a single cap based on frequency response and DC offset. For high voltages this means physically large caps. I call this One Big Cap in the attached PDF.
A couple years ago I move to 2 smaller caps in parallel which increased speed, reduced cost and somewhat reduced the space needed. This is the two in parallel approach.
Lately I have been using WIMA SMD films. Small in size but also small values. So while this ticks off the space and speed boxes. There still can be some leakage. Stacking the caps help reduced the DC offset but did not fully get rid of it. Since the output was going to another device this would not do.
Then it occurred to me to use a cap resistor cap resistor series approach. I call this the two in series approach. The first cap has to be 400 volts which the .1uf SMD WIMA Film fits the bill. The second cap only needs to meet the amount of initial DC surge and leakage. So a nice small .15 or .22uf WIMA through hole of 63 or 100 volts fits the bill. This allows me to add color as needed, immediately cut off the DC surge / leakage, have fast charge and discharges, while maintaining a smaller footprint (5 by 12mm ish) at a lower cost.
True this means a -6dB pole at 10 hertz. But from an overall system approach this is not an issue as most speaker are 30 hertz.
If you are wanting to reduce cost and the size of your design the series coupling cap approach is something to consider.
The traditional coupling cap is a single cap based on frequency response and DC offset. For high voltages this means physically large caps. I call this One Big Cap in the attached PDF.
A couple years ago I move to 2 smaller caps in parallel which increased speed, reduced cost and somewhat reduced the space needed. This is the two in parallel approach.
Lately I have been using WIMA SMD films. Small in size but also small values. So while this ticks off the space and speed boxes. There still can be some leakage. Stacking the caps help reduced the DC offset but did not fully get rid of it. Since the output was going to another device this would not do.
Then it occurred to me to use a cap resistor cap resistor series approach. I call this the two in series approach. The first cap has to be 400 volts which the .1uf SMD WIMA Film fits the bill. The second cap only needs to meet the amount of initial DC surge and leakage. So a nice small .15 or .22uf WIMA through hole of 63 or 100 volts fits the bill. This allows me to add color as needed, immediately cut off the DC surge / leakage, have fast charge and discharges, while maintaining a smaller footprint (5 by 12mm ish) at a lower cost.
True this means a -6dB pole at 10 hertz. But from an overall system approach this is not an issue as most speaker are 30 hertz.
If you are wanting to reduce cost and the size of your design the series coupling cap approach is something to consider.
I don't understand your ideas about speed and color.
Haven't single caps been used successfully at a wide range of frequencies (audio up beyond RF?) in the past?
Also, why didn't you change the cap values when using series connection to give the same overall capacitance?
(0.1 uF + 0.22 uF in series = 0.07 uF)
Haven't single caps been used successfully at a wide range of frequencies (audio up beyond RF?) in the past?
Also, why didn't you change the cap values when using series connection to give the same overall capacitance?
(0.1 uF + 0.22 uF in series = 0.07 uF)
Last edited:
For a given dielectric the volume of a capacitor is closely proportional to CV^2, paralleling caps won't change the overall size needed (or leakage) for a given value and voltage. The price may be more or less depending on the manufacturer's pricing structure (which depends on volume of sales as well as amount of stuff - common E3 values tend to sell in greater numbers).
Paralleling typically reduces ESL and ESR, sometimes useful.
Caps in series don't change the overall size or leakage either, its still the same amount of stuff. This will increase ESR and ESL typically, and you have to worry about voltage distribution across the series chain - for DC-blocking this is particularly difficult as you don't want a balancing series resistor string increasing the leakage.
Paralleling typically reduces ESL and ESR, sometimes useful.
Caps in series don't change the overall size or leakage either, its still the same amount of stuff. This will increase ESR and ESL typically, and you have to worry about voltage distribution across the series chain - for DC-blocking this is particularly difficult as you don't want a balancing series resistor string increasing the leakage.
A coupling cap is not supposed to "add colour". It is supposed to pass AC and block DC. A single cap is almost always the right solution.
Paralleling caps may reduce ESR and ESL, but for an audio coupling cap this is of no consequence so talk of "speed" just demonstrates lack of understanding.
Paralleling caps may reduce ESR and ESL, but for an audio coupling cap this is of no consequence so talk of "speed" just demonstrates lack of understanding.
My post was to present an alternative. We used to use mail for everything in the past and that worked well. We now use email and the internet. Times, ideas and approaches change. The circuitry following the coupling cap as well as the chassis size needs to be taken into account. For many tube circuits the old way of having some leakage and passing that on to a tube and an output transformer may be fine. And I am not saying that you cant continue with your 1 cap approach.
But take a headphone circuit. That 100mv of leakage may be passed into a solid state output stage and into the headphones. Now the 100mv of leakage and the initial startup surge is unacceptable. With the series approach the initial surge and DC offset is immediately shut down and there is no speaker pop.
Size is reduced by the usage of 2 smaller caps. A 400v .33uf WIMA is 26 by 11mm
MKP1J033305I00MSSD WIMA | Mouser
The series approach reduces the size to 1 10 by 7.6mm SMD
SMDIG03100VA00KR00 WIMA | Mouser
And 1 through hole that is 7 by 5.5mm
MKS2F032201H00KI00 WIMA | Mouser
Speed
Smaller caps charge and discharge more quickly than larger caps. Therefore, the transient response of a smaller cap is quicker. Which in turn affects your frequency response. This goes for both the parallel and series approach.
Color.
I like PCB caps. But they are no longer made and high voltage values are hard to find. With the series approach I get to sound coloration of a PCB cap without needing the high voltage rating as it is only dealing with the leakage from the first cap.
Overall capacitor value
What was described is not 2 electrolytic caps being combined to get a higher voltage rating. The series approach to blocking DC with 2 smaller caps results in 2 CR high pass filters and should be thought of as such. It is the effect on the low end response of the circuit that you should be concerned about and not overall capacitor value. Now if the next device in the signal chain is another tube then its characteristics need to be taken into account. At any rate the smaller the capacitance the better.
But take a headphone circuit. That 100mv of leakage may be passed into a solid state output stage and into the headphones. Now the 100mv of leakage and the initial startup surge is unacceptable. With the series approach the initial surge and DC offset is immediately shut down and there is no speaker pop.
Size is reduced by the usage of 2 smaller caps. A 400v .33uf WIMA is 26 by 11mm
MKP1J033305I00MSSD WIMA | Mouser
The series approach reduces the size to 1 10 by 7.6mm SMD
SMDIG03100VA00KR00 WIMA | Mouser
And 1 through hole that is 7 by 5.5mm
MKS2F032201H00KI00 WIMA | Mouser
Speed
Smaller caps charge and discharge more quickly than larger caps. Therefore, the transient response of a smaller cap is quicker. Which in turn affects your frequency response. This goes for both the parallel and series approach.
Color.
I like PCB caps. But they are no longer made and high voltage values are hard to find. With the series approach I get to sound coloration of a PCB cap without needing the high voltage rating as it is only dealing with the leakage from the first cap.
Overall capacitor value
What was described is not 2 electrolytic caps being combined to get a higher voltage rating. The series approach to blocking DC with 2 smaller caps results in 2 CR high pass filters and should be thought of as such. It is the effect on the low end response of the circuit that you should be concerned about and not overall capacitor value. Now if the next device in the signal chain is another tube then its characteristics need to be taken into account. At any rate the smaller the capacitance the better.
As to color. All parts add color. Carbon resistors sound different then metal film. Polyester caps different than Polypropylene. By understanding how parts color the audio you can change the sound of your design.
As to speed the capacitance size greatly affects the the ability to respond. Take for example a switching power supply. The 100kHz frequency means that a 100uf cap is useless as a filter. But a 1000pf to .01uf cap is effective because is can respond quicker.
The same goes for a coupling cap. As the audio signal moves up and down the cap must also respond with charges and discharges. The smaller the cap the more quickly it will charge and discharge. But it will be less effective at blocking DC. This is why you don't see 10uf coupling caps which are great at blocking DC but are sluggish to respond.
Again I am just proposing an approach. You can try it or not.
As to speed the capacitance size greatly affects the the ability to respond. Take for example a switching power supply. The 100kHz frequency means that a 100uf cap is useless as a filter. But a 1000pf to .01uf cap is effective because is can respond quicker.
The same goes for a coupling cap. As the audio signal moves up and down the cap must also respond with charges and discharges. The smaller the cap the more quickly it will charge and discharge. But it will be less effective at blocking DC. This is why you don't see 10uf coupling caps which are great at blocking DC but are sluggish to respond.
Again I am just proposing an approach. You can try it or not.
Just wondering where these hair brained ideas came from. Very close to snake oil.
The opposition of a capacitor to AC depends on how early on in the AC cycle the back emf across it's plates builds up to oppose the applied voltage.
At a given frequency, the smaller the capacitance the sooner the back emf across the plates will build up to oppose the applied voltage. For the remainder of the charging part of the cycle no current can flow and this is what accounts for capacitive reactance.
Note that the back emf may reach opposition quicker, but the capacitor does not 'speed' things up as we must wait for the applied AC to complete its cycle.
A coupling capacitor is chosen so that its capacitance provides the desired reactance at a particular frequency - it can have no effect on the 'speed' of the signal.
Well, physically larger caps do have more inductance that small caps. This is why large capacitors aren't suited for high frequency stuff. Perhaps that is what OP means by "faster to respond".
Speed is quite a strange term to use in the frequency domain.
I disagree with the statement about small caps being better than large caps in DC blocking. How well a capacitor blocks DC is based on its leakage and isolation voltage. Not on it's capacitance.
Besides that, coupling capacitors form a high pass filter. You'd want the high pass filter to be outside of the audioband which means you'd need a larger capacitor!
No. At least, not at audio frequencies. You may have been reading too many audio websites and not enough electronics textbooks.Ajcrock said:
A 100uF cap will work fine as a filter in an SMPS. The only snag is that it will overheat due to ESR and other losses. You are confusing two quite different things.
No. A correct size coupling cap will not charge and discharge very much. The smaller the cap the more it needs to charge and discharge. Any value will block DC. The reason you don't see 10uF coupling caps in valve circuitry is that they would give too low a LF rolloff and they would either by huge film caps or leaky electrolytics.
You really need to get out less and do more reading. Learn from others before presuming to teach them.
No that is not how a capacitor works. First, there is no back emf; back emf is a term used for motors and inductors. Any charge causes a voltage change; there is no delay before things start happening. The current does not fall to zero for part of the cycle - maybe you are confusing coupling caps with reservoir caps?Galu said:
I suppose I was thinking more in terms of DC. I was simply trying to get a handle on this 'speed' versus 'capacitance size' relationship mentioned by Ajcrock.
Anyway, I'm going to brush up my knowledge of capacitors and AC by studying the tutorial below. Hope it doesn't 'phase' me further!
https://www.electronics-tutorials.ws/accircuits/ac-capacitance.html
Nonsense.
Signal within the passband will behave exactly the same; lower frequencies will be attenuated 6dB/oct below a crossover defined by the load resistance and capacitor value.
As long as capacitance value stays the same, no matter how you combine caps to reach said value, crossover frequency will stay the same, as well as behaviour below it.
In an nutshell, signal behaviour will be the same, either above or below RC defined points.
No "speed" variation of any kind.
You sure about that?
I bet a .33 x 400 SMD film cap is about exact same body size as a .33 x 400 through hole film cap, all you save is those bulky pesky leads
No leakage is acceptable, period,so kludges allowing it but reducing is value are Bad Engineering.
You have now turned a, say, 1M grid reference load into a harder to drive 50k one but worst is that now you NEED to use a huge coupling cap .
You need .33 x 400 as initial cap driving a 50k load while physically 20 times smaller .015 x 400 driving a standard 1M grid resistor would achieve same response.
See that applying simple standard Engineering beats hare brained ideas by 20:1
No it does NOT
.1uF is 3 times less than .33uF
You might have started with it and reduced size to 1/3rd ... but then why even *talk* about a .33uF cap to begin with?
No it does NOT
If first cap is .33uF you have two option s:
* second cap is at least 10X to 20X larger than .33uF so its value is not much affected .... or:
* second cap is, say, 1uF and first one is increased to, say, .47 x 400 so total value is near .33uF
Now have a nice cup of tea while you think how much space you "saved".
I suggest a spray can, you have a HUGE choice of colours, even metallized ones, sparkle, stone or hammered finish, you name it.
the DC surge will last as long as the RC constant states; your approach won´t change it if same RC values are used.
You´ll have *one* carge when you turn amp ON; *one* discharge when you turn amp OFF and *none* while amp is in operation.
No smaller footprint at all and higher cost since you need to increase your .33uF cap to a higher value.
a small .015uF x 400 cap in series with a standard 1M grid resistor is also a 10Hz highpass and 20X smaller than your kludge
In fact, since you are saving so much space (and money) , you can splurge and upgrade it to 630V, go figure.
Try to imagine what that does to leakage.
Maybe better forget it.
Once again this is only a suggestion. Why don't you just build one of each of the models as shown in the PDF with no following circuitry. Hook it up to a scope and apply 300 volts. See which one settles the fastest.
As to the comment about snake oil in relationship to switching power supplies, perhaps you should brush up and read some white papers from TI and Analog devices on building a switching power supply. And then purchase a module and try to filter it. What you will find is that as you use larger caps and inductors the switching noise decreases in frequency and get larger. You will also find that there is a direct relationship to the ability of a cap to respond and its capacitance. You will also find the C0G 06 size capacitors are preferred. Larger caps make too much noise.
Sometimes it pays to step outside the norm and tey something different.
As for me, I made my post and I will go back to work.
As to the comment about snake oil in relationship to switching power supplies, perhaps you should brush up and read some white papers from TI and Analog devices on building a switching power supply. And then purchase a module and try to filter it. What you will find is that as you use larger caps and inductors the switching noise decreases in frequency and get larger. You will also find that there is a direct relationship to the ability of a cap to respond and its capacitance. You will also find the C0G 06 size capacitors are preferred. Larger caps make too much noise.
Sometimes it pays to step outside the norm and tey something different.
As for me, I made my post and I will go back to work.
Capacitors don't have back EMF, that's inductors.
For capacitors the defining equation is I dt = C dV
(or more simply charge = capacitance x voltage, so dV/dt = I/C.
This makes voltage the integral of current (scaled by 1/C) - this is the time-domain description - for instance this is often used in analog integrator circuits where a signal is turned into a current fed into a capacitor, then the capacitor voltage is the output signal.
Higher frequencies have higher rates of change, dV/dt is higher for the same amplitude, so I is higher - capacitor current increases with frequency (so impedance falls with frequency)
This leads to Z = 1/i2πfC, the frequency domain description of impedance.
This is just another way to say dV/dt = I/C, in terms of phasors for sinusoidal signals.
No need to do that, on the grounds that there is no need to try to understand something which is not true.Galu said:
On second thoughts, as you seem to misunderstand what you read maybe you should read less.Ajcrock said:
This thread is pointless and would be best deleted.