What value cap to bypass power supply cap?

Is there a way to calculate the value for a bypass cap to parallel a power supply cap? To bypass a coupling cap or crossover cap, I've heard you should use 10 to 20% of the value cap to be by passed. For example, you should use 0.5 microF cap to bypass a 5 microF cap. However, I've heard you should use a standard 0.1 microF cap to bypass power cap, no matter the value (I have power supply caps up to 760 microF).
Here’s some answers from some of our members

Short answer ...

05-03-05: Herman
Of course different designers have different opinions on this but a good place to start is to bypass with a value 1/100th of the big cap. You can bypass again with another cap 1/100th of the smaller and see if that does anything.

example, original electrolytic 4700 uF bypassed by a 47 uF film cap and a .47 uF film cap.

Long Answer ...

05-04-05: Sadownic
The values of bypassing caps are determined by the amount of ripple rejection desired and the frequency range. Any capacitor will have a self-resonant frequency calculated by
Above the self-resonant frequency, the capacitor will start to look like an inductor and its impedance will increase. For optimum ripple rejection, you want the shunt impedance to be as low as possible. Impedance is calculated by


Where: f=frequency (Hz) C=capacitance (farads) L=inductance (Henrys).

As you can see from above, the self-resonant frequency will vary depending on the value of capacitor used and its inherent inductance. The physical construction, size and value of the capacitor will determine the amount of inductance. So in order to maintain maximum ripple rejection across a large frequency range you’ll need to add additional (smaller value) capacitors in parallel with large value caps. Find the spec sheets for the capacitors you plan on using to determine their inductance and self resonant frequency and then calculate the values you’ll need for the additional bypass values. It’s better to understand why things are done instead of using “rules of thumb”. I hope this helps.
Sadownic (Answers | This Thread)

Complete Thread

Thanks! That was very helpful.
Dracule1, Take into consideration that bypassing large inductive electrolytic caps with small non-inductive cap crates parallel resonance circuit that will ring in response to rapid current changes. Designer did not put them for a reason.

I would speculate that you might not like the result but it isn't very expensive to try.
Kijanki, thanks for that info. However, all my power supply caps are film caps which according to the manufacturer have ESR and inductance which are a fraction of electrolytic. I know the ESR is 1 mOhm at 10 kHz, but I don't know the inductance. Are film caps considered to be inductive enough to have parallel resonance?
Dracule1, I don't think so. Looks like you got great stuff there. I second 1% minimum recommendation.
Davehrab and Kijanki, I asked an experience electronics engineer who has written papers on bypass capacitors. His response to your statement:

"As you can see from above, the self-resonant frequency will vary depending on the value of capacitor used and its inherent inductance. The physical construction, size and value of the capacitor will determine the amount of inductance. So in order to maintain maximum ripple rejection across a large frequency range you’ll need to add additional (smaller value) capacitors in parallel with large value caps."

was this:

"Yes, theoretically the self resonant frequency of the capacitor can affect the ripple rejection of the power supply filter. HOWEVER, you have to remember that the amplitude of the harmonics of the rectified AC power decrease with frequency so at a few kilohertz and above, the power supply noise is pretty small."

He recommended bypassing the amplifier end (using 0.1 microF bypassed on 10-22 microF bulk bypass cap) and directly across each power supply caps (using 0.1 microF bypass caps).

Not trying to start any flames here. Just trying to learn the differing points of view on this subject.
Dracule1, I speculate that main reason to add bypass caps is to improve frequency response and to lower the noise. To combat ripples you need to increase main capacitor and/or reduce its ESR.

As for self resonance - you want it as high as possible because it is the point where capacitors starts behaving like inductor (inductive reactance larger than capacitive reactance) meaning start of impedance increase with frequency (lowest impedance point). You don't have much control over it unless you use different capacitor type.

Increasing main capacitance will reduce impedance but also reduce self resonance point. It might be necessary step to combat ripple but does nothing to improve frequency response needed to lower the noise or improve load response.

If your main capacitors are film type then you most likely won't have any problems with load response within audio frequencies hence bypass cap will be oriented toward defeating high frequency noise coming from power supply. For that you need much better capacitor with much higher self resonance point. We are not talking about harmonics of rectified AC but about high frequency noise on power line.

I'm not sure what he means by amplifier's "end"
Amplifier's end means where the output transformer goes to ground, I think.

I'm a little confused.

"Increasing main capacitance will reduce impedance but also reduce self resonance point. It might be necessary step to combat ripple but does nothing to improve frequency response needed to lower the noise or improve load response."

"Increasing main capacitance will reduce impedance but also reduce self resonance point. It might be necessary step to combat ripple but does nothing to improve frequency response needed to lower the noise or improve load response."

So, because I have film power supply caps which are nearly double the capacitance than specified in the original schematic and have very low ESR being film caps, you're saying I really don't need to worry about reducing ripple as much? Then how do I improve the frequency response needed to lower the noise or improve load response? (How does improving frequency response lower noise or improve load response?)

Do film caps in general have higher self resonance than electrolytics?

I'm going to use 0.1 microF Mundorf M-Cap Supreme as bypass caps, which are noninductive. But I have no idea what the self resonance point of these caps. Do these seem like a good choice for intended purpose?

Sorry, lots of questions, I know.
No, you most likely don't have to worry about ripple.

Film caps of equivalent capacitance have much higher self resonance point and much lower ESR than electrolytics.

Capacitor that has better frequency response offers low impedance at higher frequencies (higher self resonance point) becoming more effective filter for high frequency interference/noise. Also it can respond faster to rapid current demand having lower inductance (inductance resists current change).

Mundorf M-Cap Supreme is high quality metalized polypropylene cap. It might work fine but you can also try metal film polypropylene or Teflon. Metal foil offers lower connection resistance. Try to keep short connections since even straight wire has inductance.

You mentioned 760uF therefore it has to be low power electronics (preamp or DAC?). We might be dealing with regulated power supply greatly reducing ripple. If you have voltage regulators and plan to improve any caps behind them check first what type of regulator it is because some of them don't like low ESR load (oscillate). Give me more info.
They're not for preamp or DAC. They're for monoblock amps. However, there are a total of 12 of these. Still not a lot by electrolytic standards, but they are huge cans, each about 7 x 5 inches and weighing around 4 to 5 lbs. According to the design of the amp, huge values found in electrolytics are not needed.

The Mundorfs Supreme, although a metalized film cap, is noninductive by design (noninductive winding geometry). Shouldn't that result in very good frequency response?

Teflons are too expensive. I would need 12 of them for the power supply caps. The amp side would need only one 0.1 microF Teflon cap (I have a couple of V-Caps from another project left over). But the metal foil polypropylene caps may be viable solution for the power supply caps.
I'm not sure why you need to bypass each capacitor. Why not use just one 1uF cap at the PC board entry. M-Cap like that should be most likely around $20.
Dracule1 .. my post was a attempt at painting a picture with broad stokes ... and suggesting the number you were seeking was 1/100th or 1% as Kijanki stated

My post got you in the ball park but seated you in the third deck .. Kijanki's guidance has give you field level box seats 3 rows behind first base

6 years ago I upgraded the PS caps in my D500 Phase Linear .. I bumped the main caps from the factory stock 35k to 140K of Nichicon’s and bypassed with 1% Rels .. I also had a dual diode bridge installed and eliminated the stock factory attenuators by wiring the inputs directly to the outputs eliminate the old carbon pots and about 4 feet of wiring

The difference was far from night and day ... much more like dusk to dawn ... but the improvement was positive with no down sides .. tightening up the base and opening up the treble very nicely

I think the Rel’s helped extend and open up the treble and the Nichicon’s gave some boost to the bass

I have found greater gains in working with room acoustics .. power deliver and vibration control .. but have no complaints about the amp upgrade to this day .. that amp isn’t going anywhere nor are the double run of Ridge Street PIII speaker cables driving my NHT3.3’s

Dracule1 .. I’m sure you’ll agree with me that they should put Kijanki and Almarg’s picture on money for all their contributions and time donated to this forum

ATB Dave
Thank you Dave for good word but I'm only trying a little to repay all great thing I learn on this forum. Almarg is is the one to be praised for sharing his knowledge and constantly helping people. He is pretty much a guru for all technical things.
Hi Kijanki, I think what you're talking about is putting a single cap at the input as an AC line filter. That is a great place for Mundorf Supreme. People have already tried this out and got good result:


However, as you can see in the link, there a caveat with Mundorfs as AC filter...It is note rated for use in AC application.

I have been told by several amp designers that bypassing each power supply cap with a 0.1 microF (or other small value cap) is good design practice to filter out noise in the hundreds of kHz to megaHz range, which I was told can cause unwanted oscillations in the amp.

Thanks for hanging in there to guide me.
Hi Davehrab, thank you for your input. When I started this thread, I was confusing several things about bypassing. Here is what I've learned so far from helpful people like you and Kijanki. Please correct me if I'm wrong.

1) The 1% bypass cap recommendation is to reduce ripple in the power supply. This helps with the frequency response of the amp (not exactly sure why) and decreases noise generated by the power supply.

2) Film caps have higher self resonant frequency than electrolytic caps, which is a good thing because it is farther out from audible frequency range. Film caps have much lower ESR (and ESL?) than electrolytics. Do these properties of film caps mitigate the need for the 1% bypass cap recommendation that is employed on electrolytics?

3) The 0.1 microF bypass on each power supply cap and also on the amplifier end helps filter out the really high frequency noise (hundreds of kHz to MHz) range that can cause unwanted very high frequency oscillations in the amp. Some have argued just filtering out the high freq hash may even affect sound quality (not to sure about this argument).

I hope someone will correct me if any of my statements are incorrect. I am learning so much on this forum. Thank you.

BTW, I have replaced the carbon pot on my amp with stepped resistor attenuator. A significant improvement. But the most significant improvement came from treating the room acoustics...took me over two years with various combinations of absorption and diffusion. Phase Linear?...Bob's solid state amp. I'm a fan.
Dracule1, no I would not recommend Mundorf for AC line filter. What I meant is that all your caps are most likely connected in parallel and then to the amp's circuitry. At the input to this circuitry I would put 1uF cap.

When you connect caps in parallel wire going to circuit should be from the very last cap and not the one close to rectifier.
Dracule1, 1% cap won't do much to reduce ripple but will deliver current when demanded during high frequency transitions where main caps is too slow (inductive).

It will also work to suppress high frequency interference. Imagine that high frequency interference contains 90kHz and 100kHz frequencies. When they appear on non-linear element like output transistor (noting is perfectly linear) sum and difference of them is created (modulation). Sum is inaudible 190kHz but the difference is 10kHz. It gets worse when interference contains a lot of frequencies making a lot of new audible junk. Almarg posted few times on this subject stressing importance of protecting audio system from high frequency interference.
Oh I see. That makes sense, but I thought you want the bypass caps as close to the power supply caps as possible? The further away the bypass cap is, the less effective the filtering, no?

I thought the 0.1 microF cap will handle those high frequency noise you just talked about.
I would connect 12 capacitors parallel cap #1 to #12 to connect back to #1 (form circle). Now connect rectifier's output to cap #1 and load to cap #7 (opposite side of the circle). High frequency bypass cap should be, in my opinion, at the load to eliminate any wire inductance that can slow down response. If not then at least 1uF cap at main cap #7. Don't spend too much money since you already have very good (film) main caps and might not gain much from Mundorf caps. What kind of wire you use for connection?

Forgot to ask you question about 938s - I found that marked absolute phase doesn't correspond to membrane going forward when connecting battery plus to red and battery minus to black terminal. I tried opposite phase and got worse leaner sound. Is it possible that this phase inversion was done by Hyperion on purpose?
Hi Kijanki, I will need to think about your wiring scheme to make sense of it. I'm not an experienced DIYer.

I think I will use high purity copper wiring with Teflon jacket.

As for the Hyperions, I'm not sure if I can answer that question. However, there was some controversy with phase of the midrange drivers on the 968s:

Dracule1, the only purpose of this scheme is to increase wire current delivery since capacitors are charged in narrow current spikes of very high amplitude. You can connect them in series but be sure to use as thick wire as possible and to connect load to the last capacitor in the chain.
Let's turn this up a notch ... as long as you have the hood up and soldering iron out .. you may want to construct a Snubber or Tank filter


As your diodes convert AC to DC they switch on and off ... this cycling creates a ringing which can be tamed/toned down by placing a Snubber filter between the diodes and the PS caps ... read the PDF for field level box seats behind first base ... as again I’ve merely gotten you in the stadium

Here is an interesting read/point of view from one of my Sensei’s who likes to frequently poke me in the eye with a sharp stick to keep me in align

“Lets talk about FACTS. The AC powered audio devices in fact DON'T USE AC VOLTAGE AT ALL !!!
They are all DC devices and hence they need clean DC not AC. So creating "pure sine" is a most stupid, utterly heretic waste of time.
The AC, after entering our gear, becomes straight away brutally rectified into DC and that is the end of the road for the sine. Then the big electrolytic capacitors accumulate the electrons in a "bottle". These electrons become a source of energy for the active devices in our gear. They are being released slowly as a DC current.
THESE ELECTRONS HAVE NO MEMORY of their "childhood". They do neither remember nor care if they arrived in a form of perfect sine, cosine, square wave, or chaotic ripples. As long as they arrived, their sins are forgiven in the capacitor and they are purified. So in other words - the hi fi gear is indifferent to AC shape.

An absolutely another story is RFI. This frequency is millions times higher than the AC frequency. It behaves like radio-wave and it can penetrate the gear.

The problem is that the RFI can enter gear DESPITE the galvanic separation by transformer and despite filtering effect of capacitors in power supply. The RFI becomes the Trojan Horse of AC grid power system in our gear.

Getting rid of these is a bit tricky. But my advice is - that our intervention is focused on hundreds of kilohertz and not close to 50 Hz AC. So the intervention is very gentle, very low order, and nothing to do with our regular current feeding the gear.

The filtering is achieved by two opposite effects: parallel capacitive shorting and series inductive blocking.
The kindergarten explanation for non-engineers is that for high frequency - a capacitor presents a SHORT, and coil - presents a high resistance (impedance).
So a RFI noise rides by the AC cabling to our house very happy, like a Porsche on the autobahn. But when it arrives to the filter - the road becomes blocked by the coil. Oh shit thinks the noise, getting through is gonna be real tough. But then it notices an alternative route - an easy one - via a cap. It chooses the easy route and ends up in the ground. Dead. That's our noise trap. That applies of course to very dumb Porsches only. This method does not affect the AC performance because the cap is VERY small and the coil has a DC resistance of milli-ohms. Literally - of a foot of thick wire. So the impulse response of the AC source to the amplifier demand is not limited. We say that the supply is NOT CHOKED.

The audible effect depends on many many factors - the circuits in gear, the internal filters commonly built in, the ground quality and arrangement in the house, the type of devices playing together etc. It is IMPOSSIBLE to predict the degree of improvement, but strangely - the results are very consistent


Lijanki .. OK I shall petition the federal government and the Treasury Department to put Almarg picture on the hundred dollar bill and your picture on the fifties ... the denomination value has nothing to do with your actual worth ... but only that the $100 are already occupied by Al’s picture

You are both the best at removing any self interest when giving a subjective opinion and Audiogon is lucky to have these two crown jewels in their in their collection

Keep up the good work

Dave, thanks again. I found nice drawing that shows how rectifier diode behavior introduces EMI:


Also notice at the "VDC Bus" arrow, that ripple is very big. When ripple gets very small charging current gets very narrow. That's the problem of linear power supplies that in reality are switchers (SMPS) operating at 120Hz but generating a lot of high frequency noise.

Also, SNAP shown at the bottom drawing is not perfectly vertical. Let's imagine that it comes back slowly. If we make vertical line from this small negative peak we can divide it into two times "ta" to the left and "tb" to the right of this vertical line. tb/ta ratio is called "Softness" of the diode. It means that diode pictured here has very little softness. Good diode is fast to switch off (short ta) but slowly snapping back (long tb).
Guys, I feel like a monkey being taught how to speak sign language. Dave, thanks for the watered down analogy of the tank filter. I'll see if this can be incorporated in my amp.
Dave and Kijanki, your very nice words are much appreciated. Thanks!

Dracule, I don't have anything much to add to the excellent inputs Dave and Kijanki have been providing. I would just emphasize Kijanki's point that there is most likely no need to spend a lot of money on the 0.1 and/or 1.0 uf caps, given that your present high value caps are good quality film types, and that the role of the lower value caps would presumably just be to provide filtering up to perhaps the low MHz region, and to provide supplementary short-term energy storage.

Do be sure to check that the voltage rating ("WVDC," for working volts DC) and the operating temperature range of the caps you choose provide large margins relative to the requirements. In typical amplifier applications, I would think that in general the commonly seen 85 degrees C temperature rating should be avoided, in favor of something in the 100 to 125 deg C area. And I would choose a WVDC rating that is at least twice and preferably three or more times as much as the voltage that will be handled.

Best regards,
-- Al
Thanks Al, for your wisdom. I will make sure I follow the guide lines provided by you, Kijanki, and Dave.
From many I have talked to, most agree to use bypass caps as close to the amplifier end as possible to filter out those ultra high freq noise in the hundreds of kHz into MHz. So can everyone agree on one sequence of capacitor values that makes sense? I'm not talking about bypass caps for ripple reduction across the power supply caps.

How about 0.01, 0.1, 1, and 10 microF?

Or even closer spacing?

Any chance of unwanted resonance creeping in using these series of values?
1/50 spacing 0.01, 0.47, and 22 microF also seems reasonable.

1/100 spacing 0.01, 1, 100 doesn't seem to reasonable because 100 microF film caps can get expensive.
I'm not sure why you need sequence of capacitors. It would make sense if you got fast and slower caps but your proposed caps vary only by value. I'm not even sure you need any caps. My vote, as I proposed before is just one cap of 1uF at the amplifier. Don't complicate it too much - just wire main caps with good heavy gauge wire and make sure amp is connected to the last cap since it is common mistake people make.
Sequence of bypass caps to cover broader range of filtering from several hundred kHz to 10s of MHz. They will be at the amplifier end as you recommend.
I agree with Kijanki that using three or four different value caps seems like overkill. Admittedly we are going on instinct, because there are many variables and unknowns involved, relating to the design of the amp, the characteristics of the capacitors, and the characteristics of the incoming AC. And because issues involving noise tend not to have a great deal of predictability.

My vote would be for at most two values, either 0.1 uf or 1.0 uf or both.

Best regards,
-- Al
Thanks. Will try with fewer caps to keep things simple.