Oversizing capacitors (audio and power)

Hi ,
done some search inside groupdiy
unfortunately not found  info/helps about ,
than this thread as help - as well .

Frequently it happens to find not available  the right value of capacitors (audio and power type) required
from "preferred" shops, having not the necessary time to wait they become available ,
and do not want make/place another order only for a small quantity  on a different store because shipping cost ,
order manage-packaging , etc..

also in various audio circuits (cool) "Mods" (mixing desks ch strips , preamps , eqs, comp, etc.. modules..)
the capacitors size is increased for better sound performance ,
(eq and filters  frequencies apart)

but overall for the first reason and electrolytic type ,
the point of this thread is :

what can be the tolerance of the volts and uf ?

thanks in advance for any info about
r

It depends on circuit an application. Increasing the voltage rating shouldn't be a problem in most cases.  Increasing the size of coupling caps (DC blocking) is usually fine, but the size may have been picked for a reason. Be careful with caps in PSUs, they may be a certain size for a reason.

What I would be most careful of is replacing standard electrolytics with modern low-ESR types. This can seriously impact stability in many applications.

living sounds said:

It depends on circuit an application. Increasing the voltage rating shouldn't be a problem in most cases.  Increasing the size of coupling caps (DC blocking) is usually fine, but the size may have been picked for a reason. Be careful with caps in PSUs, they may be a certain size for a reason.

What I would be most careful of is replacing standard electrolytics with modern low-ESR types. This can seriously impact stability in many applications.

Click to expand...

thanks so much for post ,

it is also true that many devices sold as "pro" with no low prices have inside very low cost-short life  capacitors
(no 105° and 0nly 1000 hours life , instead of 4-5-6-7000 and more...)
that are responsable of many damages and issues .....

but come back to oversize factor ,

more info are so welcome
r

Classic electrolytic tolerance is -50%/+100%. A "40u" could be 20u or 80u.

Tolerances are much better today. But I would not think twice about putting a 60u (or 63u) in place of a "40u".

The old-old electrolytics "did" have problems when a 450V unit was used in a 4V job. (Multi-section all-450V cap, one section used for cathode bypass.) I think this is no longer a problem. But you also have a huge choice of low voltages now. Use at least double what the cap sees in normal operation.

PRR said:

Classic electrolytic tolerance is -50%/+100%. A "40u" could be 20u or 80u.

Tolerances are much better today. But I would not think twice about putting a 60u (or 63u) in place of a "40u".

The old-old electrolytics "did" have problems when a 450V unit was used in a 4V job. (Multi-section all-450V cap, one section used for cathode bypass.) I think this is no longer a problem. But you also have a huge choice of low voltages now. Use at least double what the cap sees in normal operation.

Click to expand...

Thanks "Prr" ,

about  capacitors placed on the audio  "paths" , like into the Neve eq sections (and  other sections too....)
where "poly" caps type " select/set "  the frequency ,
how much the capacitor voltage rate can affect on sound ?
(mean 250-400 volts type instead of the 50-63 volts)

r.

r2d2 said:

Thanks "Prr" ,

about  capacitors placed on the audio  "paths" , like into the Neve eq sections (and  other sections too....)
where "poly" caps type " select/set "  the frequency ,
how much the capacitor voltage rate can affect on sound ?
(mean 250-400 volts type instead of the 50-63 volts)

r.

Click to expand...

Typically the breakdown voltage does not change the sound unless the voltage is inadequate for the circuit conditions and the dielectric breaks down (conducts). .

There are a number of capacitor performance parameters that can make a difference, Voltage coefficient is one where it describes how much the capacitance value changes with applied voltage (this expresses as distortion). Other parameters are DF (dissipation factor), DA (dielectric absorption), etc... 

Film capacitors behave more like ideal capacitors than electrolytic which is why they are preferred for circuits like audio filters that impose changing terminal voltages.

JR

r2d2 said:

Thanks "Prr" ,

about  capacitors placed on the audio  "paths" , like into the Neve eq sections (and  other sections too....)
where "poly" caps type " select/set "  the frequency ,
how much the capacitor voltage rate can affect on sound ?
(mean 250-400 volts type instead of the 50-63 volts)

r.

Click to expand...

Capacitor voltage spec does seem to affect the high frequency impedance spec on the data sheet of the capacitor,  whether you can hear a difference or not is up to you...

user 37518 said:

Capacitor voltage spec does seem to affect the high frequency impedance spec on the data sheet of the capacitor,  whether you can hear a difference or not is up to you...

Click to expand...

I am not aware of a "high frequency impedance" specification regarding capacitors. Are you perhaps referring to ESL (equivalent series inductance)? The significance of all non-ideal capacitor specifications need to be weighed in the context of the application.

JR 

JohnRoberts said:

I am not aware of a "high frequency impedance" specification regarding capacitors. Are you perhaps referring to ESL (equivalent series inductance)? The significance of all non-ideal capacitor specifications need to be weighed in the context of the application.

JR

Click to expand...

It took me ages to respond to this post, but here it goes, if you look at the data sheet of Panasonic FM caps, you'll see theres a column labeled Impedance at 100Khz.

https://industrial.panasonic.com/cdbs/www-data/pdf/RDF0000/ABA0000C1018.pdf

That looks a lot like ESR. What the table appears to say is the larger volume the capacitor, the lower the ESR. Panasonic calling it impedance is a bit confusing, as everyone else calls it resistance, which is what you want to know.

Both these sources say ESR is commonly measured/quoted at 100kHz:

https://www.avnet.com/wps/portal/abacus/resources/engineers-insight/article/understanding-esr-in-electrolytic-capacitors

https://en.wikipedia.org/wiki/Equivalent_series_resistance

scott2000 said:

What about changing these 33u caps to 47u?

Why would this have a negative side effect?

Think the 12ax7 plates are 90v

Click to expand...

These are used for couple some external input to the grid. Any "negative" side effect might be a larger/longer lasting 'pop'/offset if the thing connected has a DC offset. with those resistors (3.3k in series to the grid, 10k from the grid to ground) it already has a flat response below 20Hz, so 33 vs. 47 doesn't much matter.

What's a problem is not the exact value, but that an electrolytic is used to couple two things that are basically the same voltage (0V on both sides of the capacitor). From what I've read and seen, an electrolytic in such an application will lose capacitance faster than one with significant DC voltage across it (as in maybe 10 years vs. 30 years).

Interesting.

It seems no one has addressed why you would want to increase the capacitance of a decoupling cap.    Basically the cap determines the L.F response.  If you double the cap value i.e  20u to 40u you halve the L.F response  for example 20Hz would become 10Hz.  Ths is because the decoupling cap forms an H.P filter.

user 37518 said:

It took me ages to respond to this post, but here it goes, if you look at the data sheet of Panasonic FM caps, you'll see theres a column labeled Impedance at 100Khz.

https://industrial.panasonic.com/cdbs/www-data/pdf/RDF0000/ABA0000C1018.pdf

Click to expand...

Ohms at 100k is spec data sheet speak for ESL (series inductance) in series with ESR (series resistance). Showing it as ohms at a (relatively high) spot frequency makes comparison easier (but only if everybody specs their caps that exact same way...).

ESL and ESR are the actual specifications. The ESL will continue rising above 100 kHz. A typical combined capacitor impedance vs frequency plot will look like the familiar bathtub curve, so it will first fall with increasing frequency as capacitance dominates, then flatten out (while ESR dominates) , then start rising as ESL dominates.





JR

Rob Flinn said:

It seems no one has addressed why you would want to increase the capacitance of a decoupling cap. 

Click to expand...

Agreed.

  Basically the cap determines the L.F response.  If you double the cap value i.e  20u to 40u you halve the L.F response  for example 20Hz would become 10Hz.  Ths is because the decoupling cap forms an H.P filter.

Click to expand...

It changes low-frequency distortion too. Since a 40uF coupling cap sees half the AC voltage of a 20uF, its distortion contribution (all other factors being the same) is also halved. Generally this results in a cleaner low-end, unless the extended LF response generates LF oscillation/instability (motor-boating). This kind of mod should be accompanied with a similar increase in the value of rail decoupling caps. Re-capping is a good opportunity to take advantage of the progress in capacitor manufacturing and increase in volumic capacitance.

JohnRoberts said:

Ohms at 100k is spec data sheet speak for ESL (series inductance) in series with ESR (series resistance). Showing it as ohms at a (relatively high) spot frequency makes comparison easier (but only if everybody specs their caps that exact same way...).

ESL and ESR are the actual specifications. The ESL will continue rising above 100 kHz. A typical combined capacitor impedance vs frequency plot will look like the familiar bathtub curve, so it will first fall with increasing frequency as capacitance dominates, then flatten out (while ESR dominates) , then start rising as ESL dominates.

JR

Click to expand...

+1

I think the graph of impedance vs frequency is probably the best way to explain / visualize how capacitors actually work.  The same principles apply to other components as well.

I would add that ESR can also be a bit misleading being called resistance,  since strictly speaking it also has some frequency dependence.

john12ax7 said:

I would add that ESR can also be a bit misleading being called resistance,  since strictly speaking it also has some frequency dependence.

Click to expand...

Not some. It is frequency dependent.

ESR= DF/2*pi*f*C

DF: Dissipation Factor.

sahib said:

Not some. It is frequency dependent.

ESR= DF/2*pi*f*C

DF: Dissipation Factor.

Click to expand...

Yes,  but even that can be misleading (DF is not a constant).  In practice ESR or DF are a way to describe loss, which implies the real resistive part. But also a real resistive part that varies with frequency,  as opposed to a strict reactive frequency dependence.

john12ax7 said:

I would add that ESR can also be a bit misleading being called resistance,  since strictly speaking it also has some frequency dependence.

Click to expand...

Would you say that the resistance of a wire is misleading because it has some frequency dependence? Yet it is the case, due to skin effect. The essential character that defines resistance is that current and voltage are in phase.
However I agree that calling ESR a resistance is misleading as it's a mathematical abstraction that encompasses several elements that are frequency-dependant. I would think the problem is not with words it's about knowing what these words hide vs. what they show.

sahib said:

ESR= DF/2*pi*f*C

DF: Dissipation Factor.

Click to expand...

That is a very simplified formula. The more complete is:
ESR=DFR/2.pi.f.C+DFD/2.pi.f.C+DFL/2.pi.f.C = Rc+DFD/2.pi.f.C+1/RL(2.pi.f.C)²
Where DFR, is the dissipation factor associated with contact resistance, DFL with leakage losses, and DFD the dielectric losses.
It is important to remember that ESR is a mathematical abstraction, just like DF. Contact resistance losses are proportional to current, when leakage and dielectric losses are voltage dependant. The use of DF simplifies calculations related to power losses, but when one needs to simulate or analyse the actual behaviour of the component in-circuit, a more detailed model is needed.