Bypassing of PSU lytics with foils

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Tubetec

Well-known member
Joined
Nov 18, 2015
Messages
5,951
Just wondering what peoples opinions are on this , I see many dismiss it completely .In one project with 5 filter sections ,modern caps etc I bypassed each lytic with a foil ,less than 1uf in value , It didnt make any difference to measured noise levels seen on my multimeter , what it certainly did do was change the spectum of the noise left behind ,it reduced some of the upper harmonics coming out of the rectifier appreciably. Now Im not saying i heard a difference in the sound of the amplifier as a result just that the backround noise was less perceptable, less fuzzy .. Im just putting together a big three transformer supply for my new project , I have tons of space in my new enclosure and a big bag of surplus foils . I could see why in usual supplies adding some tiny value of foil cap to bypass over one or two stages isnt going to do much . I have the space to make the foils around 1/10 of the value of the electros in my new build ,could I expect to get less midrange harmonics in the output of my supply by doing this ?
Could using metal canned foils be of any benefit or maybe keeping them screened from the transformers  ?
 
Power supply design is a complicated matter. Semiconductor rectifiers are so fast they can easily cause very sharp spikes or high frequency bursts due to resonance with mains transformer leakage inductance. I find that a 100nF film capacitor across the input to a bridge rectifier successfully stops these. From the noise point of view I have not found it necessary to add any further capacitors. However, in order to ensure the output impedance of the power supply remains low at very high frequencies it is often useful to add a film capacitor across subsequent dc power filtering.

Cheers

Ian
 
Why?

Slow rise time audio signals added loop area etc.

Anyway sometimes it can make thing worse but you need to measure
 
Parallel caps is rather common when you need broadband low impedance.  But you do need to measure and get the right combo for the intended application.
 
seen it a lot on stereos, like this Sansui G-7000,

maybe they had a stability problem or just wanted to squelch switching noise,  they use a round power transformer which usually has lower leakage than an EI stack,
 

Attachments

  • 7000.jpg
    7000.jpg
    83.1 KB · Views: 41
I don't know about opinions but good practice for a long time was to use both large electrolytic to supply current smoothing in parallel with smaller HF caps (like ceramic discs back in my day) for low impedance across the board.

The HF caps are generally a consideration for regulator stability and for local circuitry to provide stiff PS rails. Another subtle concern is noise from rectifier diodes.  These are separate design concerns but may be dealt with using similar parts.

Also in recent decades electrolytic capacitors have become lower impedance in response to switching power supply applications, and general technology improvement.

JR
 
Tubetec said:
Just wondering what peoples opinions are on this , I see many dismiss it completely .In one project with 5 filter sections ,modern caps etc I bypassed each lytic with a foil ,less than 1uf in value , It didnt make any difference to measured noise levels seen on my multimeter , what it certainly did do was change the spectum of the noise left behind ,it reduced some of the upper harmonics coming out of the rectifier appreciably. N

Did you measure the before and after with a scope or spectrum analyser?

The diode turn on current surge can be worst with film caps causing an issue.

FWIW I replaced all the power supply caps in one guitar amp with polypros the toroidal transformer was not happy
My best guess was the diodes were not matched well enough and the toroidal transformer had some DC
Added some resistance and things were OK
I did this to REDUCE the power supply caps values and to not have to change them again

I have not heard a change adding films across power supply electrolytics and IIRC I first tried this starting in the late 70s after reading an audio magazine article in a Hafler DH200 amp. I tried a few more time over the years and did not hear anything when used across a power supply cap.

 
A modern radial electrolytic typically has a flat and low impedance  out to pretty much 1 MHz.  A flat controlled impedance means no quirky behaviour from audio circuitry - all frequencies flow in the same manner and see the same impedance path.

Adding a circuit path with a very sharp notch impedance is imho a risk more than any identifiable benefit.

I reckon its more important to be concerned about starring each +/- local supply node, and hence where the bypass cap terminals connect.
 
trobbins said:
A modern radial electrolytic typically has a flat and low impedance  out to pretty much 1 MHz.  A flat controlled impedance means no quirky behaviour from audio circuitry - all frequencies flow in the same manner and see the same impedance path.

Adding a circuit path with a very sharp notch impedance is imho a risk more than any identifiable benefit.

Adding a film cap will not add a sharp notch. It will do exactly what any capacitor in an RC circuit will do (roll off at 20dB/decade) but just at a higher frequency because of its smaller value.

Cheers

Ian
 
This may be TMI but real capacitors have C, L, and R...  their impedance will fall with rising frequency (C dominant) until equilibrium with R (R dominant)  and then rising impedance due to L dominant.

Small HF caps begin with higher impedance but generally exhibit impedance reversal (sometimes called resonance) at much higher frequency. When placed in parallel with larger caps the impedances combine to extend the useful low impedance region up to higher frequency.

Back several decades ago it was fashionable in audiophool circles to parallel blocking capacitors in audio paths with smaller HF caps. My bench work measurements suggest this was only productive if HF cap was roughly 10% or more the capacitance of the LF cap. So most such efforts were not doing anything useful in practice because of too small HF caps.

In power supply decoupling there are probably other concerns like low impedance at very high frequencies, way above audio passband.

JR

 
Thank you all for for your insights.

Ok well , as you all know Im not a compulsive for logging and  noting my findings in a very scientific way , traditionally all I had was a multimeter , Its not practicable to noise hunt audio with just a simple meter or even a good one for that matter .
It'll lump all sources of noise together ,REW is a revelation to me in this respect

I do understand that even a simple piece of wire ,which we might like to think of as( for the sake of simplicity) 0 ohms 0 capacitance and 0 inductance could be far from that in reality .

The loop between the transformer secondary windings ,rectifier, first filter cap and ground should be as small as possible , the bigger it is , the more voltage drop for the AC current and the more of it potentially ends up in your audio .
If we put a wacking great foil across our sub miniature 220uf in the vicinity of stray magnetic fields were certainly likely to do more harm than good ,by virtue of the fact that we have actually just created an induction loop for low freq junk from the TX .

Anyone who's had a chance to fool around with a dual rectifier (Boogie) ,where you can switch between solid state and tube rectifiers from the back panel; will know ,the character of the backround noise changes radically between tube and solid state .

Adding bypass 'small foils' to 5 big RC/LC stages  ,the high frequency mush is alot less noticable ,but low mid 'honk' takes centre stage at some point  long as you've attended to 50/100 hz adequately

If we make the lytics 10 times bigger than olden days ,and a tenth of the size ,then put in foils which in themsleves are good enough to do the job at 100hz by the time the lytics are 1/10 their original value it wont matter much anymore,we wont matter much either 

I see some nice oil filled 2-12uf,metal cased 'Kelvin' oil motor caps ,might just fit in with my balanced 6X4 heater/cathode scenario  ,if you cant beat em' cancel em' 'inductance and capacitance' ?
 
ruffrecords said:
Adding a film cap will not add a sharp notch. It will do exactly what any capacitor in an RC circuit will do (roll off at 20dB/decade) but just at a higher frequency because of its smaller value.
Ian,  certainly for electrolytics for valve amps (with higher voltage ratings and smaller cap values) the ESR is typically more than an ohm.  For ss amps, the electrolytic ESR may be down to 10-20mohm for large value caps.  Good electrolytics for both of those applications would have flat impedance from low audio to at least 100kHz.

A foil cap  will show a falling capacitance impedance at all frequencies pretty much up to its resonant frequency, where it will notch to a sharp minima.  For valve amp applications, the foil cap notch impedance can be more than  two orders of magnitude lower than the electrolytics impedance, and so the combination of caps exhibits pretty much a flat impedance out to the notch.
 
if you shop the digikey pages you will see that certain  pwr supply filter caps are rated for audio and others of the same value and voltage are not.

probably have to ask a Nichicon engineer what gives.

 
CJ said:
if you shop the digikey pages you will see that certain  pwr supply filter caps are rated for audio and others of the same value and voltage are not.

probably have to ask a Nichicon engineer what gives.
Probably marketing...  ::)

JR
 
trobbins said:
Ian,  certainly for electrolytics for valve amps (with higher voltage ratings and smaller cap values) the ESR is typically more than an ohm.  For ss amps, the electrolytic ESR may be down to 10-20mohm for large value caps.  Good electrolytics for both of those applications would have flat impedance from low audio to at least 100kHz.
I think you may be confusing impedance and capacitance. An electrolytic capacitor will have a flat capacitance up to about 100KHz. However, over the same frequency range its impedance drops with increasing frequency - it is not flat. After the internal resonant frequency, the impedance starts to rise due to the internal inductance.

Because a film cap has a much higher resonant frequency it's impedance continues to drop long after the electrolytic's impedance has passed resonance.

Cheers

Ian
 
Ian,  Assume a valve amp type electrolytic of up to 100-200uF - its impedance will will fall with frequency up to about 1-10kHz where ESR and temperature dictate a flat impedance level out to where capacitor package inductance starts to raise the impedance.

An ss amp could use a way higher capacitance electrolytic, and a typical cap impedance would again fall with frequency up to about 1-10kHz, as the lower ESR allows the 'impedance dropping' range to extend to a higher frequency. 

The simplest example plot I could find was from KMH series of caps - see the last plot of a 180uF 400V cap versus a 1,800uF 160V cap at 25C:
http://disti-assets.s3.amazonaws.com/tonar/files/datasheets/16654.pdf
 
tr>> for electrolytics for valve amps (with higher voltage ratings and smaller cap values) the ESR is typically more than an ohm.

Additional chart (prepared while you were posting).

https://media.digikey.com/pdf/Data%20Sheets/Epcos%20PDFs/B43501_Rev_Aug_2014.pdf
VR = 450 V DC
CR uF  --- ESRtyp mOhm
47 uF ---2820 mOhm  (3 Ohms)
68 uF---1950 mOhm    (2 Ohms)
100 uF --- 1330 mOhm  (1 Ohm)
 

Attachments

  • CapImpedance.gif
    CapImpedance.gif
    12.2 KB · Views: 13
Yes I thought it best to try and simplify the view for Ian.  The obtained ESR (and hence 'corner frequency') is very dependant on temperature, and datasheet curves are likely a typical bogey curve, and table values are usually a max value, and ESR may well change with ageing.

The main points I was trying to convey were that electrolytic ESR is typically flat in the mid-band to high frequency audio range, whereas a foil cap will have a notch style impedance minimum, and that minimum is likely to be an order of magnitude less than any electro used in the application.  Having both those impedance paths in parallel makes for non-flat curve, and a local high Q CLC  loop for a perturbing signal of the right frequency.
 
I am not quite sure what you guys are trying to convey to me. Both of the graphs presented show the actual impedance is inversely proportional to capacitance as you would expect. The limit of low impedance is set by the ESR but, from the graphs,  it is not clear if that figure is primarily dependent on capacitance  or voltage rating. There seems to be no evidence of ESL.

Cheers

Ian
 
ruffrecords said:
There seems to be no evidence of ESL.
Correct.  For a modern electrolytic in radial format, the ESL is so low that impedance rise due to ESL is typically pushed out beyond 100kHz, and often to past 1MHz.  As such, one electrolytic cap presents a stable low impedance "voltage supply" to the local amp circuitry.  Ie. there is no need for any supplementary parallel cap for a normal audio amplifier.
 
Back
Top