What's all this Trapped Charge and Dielectric Compression stuff anyhow?

What's all this Trapped Charge and Dielectric Compression stuff anyhow ?

Yes I've 'hi–jacked' the "what's all this . . . stuff anyhow" title but hopefully Y'all already be aware of the articles ~ What's All This Soakage Stuff, Anyhow ? and Understanding Capacitor soakage to Optimise Analog Systems and What's All This Capacitor Leakage Stuff, Anyhow ? plus What's All This CR Stuff, Anyhow ? and forgive me for the expropriation of the titles from their author the late Bob Pease and for my different view on the subject

Bob Pease [RAP] focused on an imperfection of capacitors attributed to Dielectric Absorption which he shortens to DA and then also refers to as Capacitor Soakage suggesting that a capacitor dielectric soaks up or absorbs a charge in addition to that determined by its value and the voltage across it ~ He also describes the soakage effect as fairly linear without any qualification apart from presenting a linear multiple capacitor/resistor model for it

The test circuits RAP used to show capacitor soakage and his leakage tests only charge capacitors in one direction with d.c. ~ No a.c. tests for soakage appear to have been done although soakage affecting fast settling a.c. amplifiers and filters is mentioned and the ability of the so called and self proclaimed golden eared ones to distinguish audible differences between low and high soakage capacitors is questioned [and rightly so]

ESL ~ Dissipation Factor ~ ESR ~ Polarisation ~ Dielectric Constant Temperature and Voltage Coefficient and some chemical effects found in electrolytic capacitors are not being questioned here although I do have concerns about statements like ESR varies with frequency so cannot have a finite value ~ Perhaps we should use a term like Series Complex Impedance when that is what we intend

I know I'm not alone looking for a correlation between the inadequacies of different capacitor types and the perceived quality of sound reproduction when they are used in audio equipment ~ Could high soakage somehow change the tone of an amplifier as old paper in oil capacitors or electrolytic capacitors appear to do ?

The usual properties of a capacitor tend to exhibit distinct measureable effects and where possible these have been considered and discounted during my tests ~ Dielectric Absorption [DA] does not appear to defined by manufacturers and is not easily measurable in isolation ~ For certain types of capacitor I believe the DA effect may be due to physical construction and therefore not easily modelled or otherwise expressed in numbers

I was shocked by Capacitor Soakage ~ Or was I ?

During the 1980s I worked on many high power radio and TV transmitter sites where we stored large value high voltage paper in oil capacitors with wire shorting links across the terminals after testing the dielectric strength with a high voltage meter ~ This I was told ensured they were fully discharged and could not 'pick–up a charge' from the radio transmissions

Looking back I question the pick–up a charge effect because after years of storage with a shorting link some of those high kV capacitors once unlinked would start to develop a voltage across their terminals ~ If this was charge pick–up it would require a non linear or asymmetric effect to produce a d.c. voltage across the terminals due to RF pick-up ~ Or maybe some other principle

The voltage rising from zero was surely not due to traditional capacitor soakage because even the longest time constants of the dielectric absorption model would have fully discharged after such a long period and how does the voltage start to rise from zero ? Could a charge somehow be stored in a capacitor with zero potential across its plates for many years ?

From the results of some d.c. tests to check for 'Capacitor Soakage' I propose it could be the physical construction and dielectric compressibility that makes some capacitors perform differently in sensitive analogue circuits and in RAP's soakage tests ~ Whether audio amplifiers can 'sound different' may not be due to soakage or dielectric compression or any of the other known capacitor attributes alone

After many years of tinkering as RAP once said of me I still believe that ~ if everything else is kept equal and when used in a good design ~ poly[styrene] capacitors will sound ~ dare I say it ~ better than paper in oil type and that both sound better than standard electrolytic when used in the audio signal path ~ Many claim that paper in oil capacitors sound 'warmer' and 'musical' but why is this?

With paper in oil [PIO] and electrolytic capacitors ~ especially the vintage new old stock preferred by some ~ it is most likely high leakage and other measurable losses like ESR that affect the audio amplifiers they are fitted in ~ Almost all aged paper in oil capacitors like the 1959 Hunts L45 tested below show leakage and may now not be suitable for use in high impedance and high voltage circuits such as valve amplifiers

Voltage rises across a Capacitor after discharge to 0V ?

It looked like the bridge was being upset by the d.c. voltage ~ and to some extent it probably was ~ but with some high value electrolytic the capacitance value appeared to change with little or no measured d.c. but this was often with very old capacitors that had needed longer reforming so I just accepted it and measured them after a long re–form and very long discharge period to reduce the voltage close to zero

Then I had the crazy notion that in a practical capacitor ~ because the dielectric material is also used to keep the two plates a precise distance apart ~ the capacitance could 'actually' change as a capacitor is charged and discharged ~ If the plates could move when a voltage was applied across them they would attract due to the opposite electrostatic charges on each side ~ They also attract as the capacitor is discharged but in each case the dielectric should eventually relax back decreasing the capacitance to its nominal value

For any fixed charge [Q] on the plates of a capacitor the voltage will fall if the plates move together [V=Q/C] and will increase if the plates move apart ~ If a dielectric material can be readily compressed and relax back because it is made of fibrous paper soaked in oil and or the foil of the plates is loosely wound then the plates will move together as the capacitor is charged ~ The value of C will increase and give a greater charge than expected for the applied voltage

After a rapid charge and removal of the supply the voltage across a capacitor can rise beyond the supply voltage which is likely due to the compressed dielectric relaxing back to its natural position and its original value of C ~ After a rapid discharge ~ where the high current pulls the plates together ~ the remaining low voltage will rise as the residual charge on the plates is again subjected to decreasing C as the dielectric relaxes back ~ These effects are easy to measure at d.c. but not easy to isolate from other effects like chemical and molecular changes in the dielectric

Testing for Dielectric Compression in an a.c. circuit

The Full circuit used to test the 0.1 µF capacitors ~ L=6.2mH with 0.59Ω resistance so Q is very high ~ R=220kΩ to enable connection of a floating d.c. voltage at point A without affecting the resonant frequency or upsetting the analyser tracking generator ~ C are the two 0.1 µF capacitors of same type under test The idea was to measure any changes of capacitance due to an applied voltage but I did not have access to test equipment like a sexy C meter that could measure the very small changes of capacitance of a 0.1 µF capacitor due to about 30V d.c. applied across it

The change in capacitance could be due to a Permittivity change or a Piezoelectric effect both of which mainly affect high k ceramics and have not been ignored here ~ Such non linearity can often be measured as third and odd harmonic distortion on sine waves when passed through a simple CR stage but with these capacitors this was not seen in such tests ~ Experiments by others show changes of capacitance due to applied d.c. voltage but not for paper in oil and not when the voltage is reversed

The analyser was locked to a GPS reference because it helped speed up the measurements ~ The Marconi 2382 has very good ovened oscillator with good phase noise figure but I needed to make many measurements over a long time without having to account for the slightest drift and a maximum frequency error when locked to GPS of 10 parts in 10¹² I thought could be neglected and was

By connecting a floating power supply across point A of the test circuit a voltage is applied across the capacitors without any d.c. current [apart from leakage] flowing through the inductor or the test equipment and the applied voltage could be VARIED and REVERSED without any concern that it was affecting the test equipment or the frequency other than by changing the value of the two Cs

The idea was to test if the applied voltage changed the notch frequency noticeably thus indicating that the capacitor value had changed ~ The Q may also change but this was being simultaneously monitored ~ This simple test is very revealing and showed the Hunts L45 pair lowered the notch frequency by 3Hz when 30V d.c. was applied in either direction

The Q did not appear to change during any of the tests and the shift in frequency was always measured at the 3dB points on both slopes of the notch to better discriminate the small changes while also confirming the BW remained the same ~ The Russian K40Y9 0.1 µF 400V capacitors showed Q slightly better than the L45s and the frequency shift was less than –1Hz and so hard to determine even with the good markers and resolution of the 2382

The polystyrene capacitors showed no noticeable shift in frequency and the 3dB BW as seen in the trace above is close to the theoretical limit of 17.3Hz set by the Q of the inductor and the loading of the two resistors R which act as a 440kΩ resistor from the centre of the Cs to the input of the analyser thus do not form a true bridged T ~ The value of R chosen actually gives slightly better Qs for lossy high ESR Caps and worse Q for good Caps like the polystyrene RIFAs and so help keep the traces on the same scale

It was also noticed that the frequency shift of the L45s and K40Y9s [and some other PIO caps] took much longer to recover following longer charge periods which is similar to the effect noted by others and attributed to Dielectric Absorption or Capacitor Soakage as measured by a rise in open terminal voltage after discharge but in this test the voltage should not affect the frequency or its measurement

Rather than calculate figures for these test capacitors and become tied down with small % changes please accept that for some reason the value of a capacitor appears to increases when a voltage is applied across it sausing current to flow through it ~ This Dielectric Compression effect appears proportional to the voltage MAGNITUDE and should not be confused with 'momentary dielectric relaxation' which is another dynamic effect akin to hysteresis

These tests do not prove that the increase in capacitance is due to the plates moving together under electrostatic attraction but other effects and some knowledge of capacitor construction suggest that it is ~ My other idea of a Captive Charge is just another untested idea but it appears to explain additional effects attributed to DA like the very long term storage of charge in a shorted capacitor which should be dissipated if the soakage models are correct

If it can be demonstrated that the distortion produced by those capacitors that have a warm sound or show high soakage is mainly 2nd and even harmonics ~ due to the plates dynamically compressing the dielectric at twice the rate of the applied signal voltage then it looks like some capacitor soakage effect may actually be due to the dielectric compressing and perhaps be visualised [Einstein] without the need to measure results scientifically

Can charge be trapped indefinitely in a shorted capacitor ?

Dielectric compression ~ where the plates of a capacitor are free to move may explain some of the stored charge often attributed to dielectric absorption when measured by the rise in open terminal voltage after discharge test but often the voltage will rise higher than the small changes of C ~ that I have measured ~ would indicate

Measuring a rise in open terminal voltage after discharge is not measuring stored charge no matter what mathematics are applied ~ The voltage may be partly due to residual charge absorbed by the dielectric ~ but how how long can this DA or soakage charge be held in a short circuited capacitor ?

I think it is very likely that when the plates of a capacitor are shorted and kept shorted that some residual charge ~ that has not converted to heat or radiated as sound or an EM pulse ~ could be forced to the centre of the dielectric where it will remain held captive ~ Repelled equally by the plates either side which are now at the same potential

The Captive Charge or the charge trapped in the dielectric will not be subjected to normal discharge by leakage as it cannot flow to either plate so it will just sit there between the plates waiting for someone like Pieter van Musschenbroek to open the jar and release it or Georg Christoph Lichtenberg to 'figure' how we could visualise it ~ It's soakage Jim but not as we know it

If this Captive Charge exists it should also be seen even in Vacuum capacitors but it should be more easily measured and appear as a problem in very high permittivity dielectrics which are after all intended to 'hold' more charge but these also suffer other problems like changing permittivity with voltage ~ Low permittivity PTFE or Teflon is noted for its particularly low DA ~ Maybe even a charge won't stick to Teflon !

The dielectric of High–K ceramic capacitors appears to cause problems in say a sample and hold circuit with fast discharge ~ Once the capacitor is shorted and the plates are at the same potential there are only a few places the Captive Charge can go ~ Maybe a compensating bootstrap circuit ensures the plates are never at the same potential until the desired level of discharge is reached

The findings of RAP that The accuracy of a sample and hold capacitor tends to get better as you go faster could well be explained by the physical compliance of the dielectric material ~ as the frequency increases the rate at which a dielectric can physically compress and relax will reduce and ~ depending on the construction ~ the plates may even experience resonance effects

Referring back to another article by RAP where a Mylar capacitor was heated to 150˚C ~ Did the experimenters really measure the charge that flowed from the capacitor ~ or just the voltage attributed to the charge on what they thought was a stable capacitance value ~ Was their additional charge actually stored on the molecules of the dielectric or simply between the capacitors plates which were being expanded apart and contracted causing a change of C ?

Capacitor Construction and Dielectric Compression

RAP showed that silver mica capacitors have high dielectric absorption errors which goes against most electronic engineers' intuition ~ including my own at one time ~ This could be explained by the fact that mica can be readily compressed ~ Mica compression variable capacitors were once very common for trimming radio circuits

Two types of construction used for fixed value mica capacitors are silver 'metallised' mica plates and separate layers of mica and metal foil ~ Both types are normally crimped and held in place only at the edge terminals ~ Nowadays most are the metallised type which are very stable in value but only for single plate construction

Silver Mica in the RAP soakage tests would suggest the 'best' type was used but high value silver mica capacitors are made with many layers and subject to dielectric compression ~ Also the coating can affect the performance with hard resin impregnated blocks being more stable than ceramic or resin dipped ~ Until about 1970 mica and foil types were often just dipped in wax to keep moisture out yet they were very stable in radio frequency [RF] circuits

Mica is known for its good insulation and low loss dielectric and stable capacitance at RF so why does it perform badly in low frequency [LF] tests and sample and hold circuits ? ~ Is it because the dielectric is compressible but at RF and HF the compression effects do not apply because the plates cannot physically move so fast ?

The porous paper spacer holding apart the plates of standard electrolytic capacitors is soft and easily compressed as it needs to hold the electrolyte ~ Similar is the waxed or oiled paper used in PIO capacitors

Pictured are cores of a standard electrolytic and a solid Sanyo OScon capacitor ~ The OScon could not be unwound and is fully embedded in black resin and is generally considered to sound much better than the standard one on the left ~ Other electrolytics such as Cerafine and Black gate also use incompressible material in the dielectric and again are considered good for audio amplifiers and measure well

As electrolytic capacitors 'dry out' the above effects appear to become more prevalent ~ Even if the capacitance value measures correct its ESR and Dielectric absorption are greater than when new ~ I have repaired Hi–Fi power amplifiers and guitar amps by only changing electrolytics and although the electrolytics replaced were technically much better than the old originals the sound of the amplifiers according to the owners in most cases was changed

Tantalum bead and Solid Aluminium capacitors are well respected in audio equipment by those who really know ~ and the smaller lower voltage sizes appear to 'sound' better and are cheaper than large overrated or wet construction types ~ It should be no surprise that their construction is extremely solid and incompressible in both cases ~ But what about polycaps which according to RAP were not all the same

Polystyrene and PTFE capacitors are almost always made to very close tolerances ~ When making polystyrene capacitors a separate foil and polystyrene film are tightly wound to a slightly higher value than required and with an overwinding of several layers of polystyrene ~ The already tightly wound part is then heated for a time at about 106˚C to compress it to the required value ~ further compression or relaxing is very limited once made

Most precision and timing capacitors are made by very tight winding and are often potted to ensure they stay compressed to the correct value ~ Extended foil capacitors made this way are more stable still because the ends of the foils are soldered over which makes them very rigid as well as lowering the self inductance giving a very high Q

Pictured is a 100nF 2% 100V RIFA216 potted extended foil polystyrene capacitor ~ Unfortunately no longer made as they are an excellent capacitor for audio ~ Note the creases at each end where the polystyrene has shrunk down and compressed the extended ends of the tin foils

It is clearly established that both PTFE and Polystyrene are among the lowest recorded DA error capacitors so empirically I may be correct to assume some of the dielectric absorption or DA attributed effects may actually be due to Dielectric Compression or the plates able to move due to internal electrostatic forces while effects in discharging circuits may be due to Captive Charge maybe in conjunction with dielectric compression

PTFE and polystyrene capacitors have long been regarded as excellent in audio circuits ~ Some may argue that their low tolerance is seen by 'Pseudo quality seekers' as better and this translates to 'sounds better' ~ However there is no doubt that both PTFE and polystyrene caps appear in many of the best audio amplifiers and there is a strong correlation between low DA and capacitors which are good for perceived accurate audio reproduction

Capacitors in Audio circuits

Most tests for dielectric absorption or capacitor soakage 'biased' toward d.c. conditions ~ Further tests that I and others have made to measure distortion produced by capacitors in audio circuits have not proved conclusive ~ especially when looking for the generation of clearly audible [>0.1%?] even harmonics to explain why some caps like paper in oil are preferred by the golden eared due to the concordant nature of the distortion

An advert in June 1992 Electronics World claiming 'proof for the golden ears hypothesis?' showed that even harmonic distortion can be introduced into a power amplifier already producing mainly odd harmonics by using back–to–back electrolytics to de–couple the negative feedback in place of the single capacitor originally fitted ~ I believe the featured amplifier later sold on eBay for a mere £350

The same technique using 2 discrete back–to–back electrolytics as opposed to a single bipolar or tantalum is often used to couple the inputs and outputs of audio components from makers like TEAC and Sansui where there is 0V across the caps ~ Many of these 'high end' audio components also have 'discrete' op–amp elements and multiple parallel DACs which may be 'seen' to sound better so sound better they do

Some ceramic capacitors show significant often odd distortion ~ Mainly due to their dielectric constant not actually being 'constant' with voltage ~ Many paper in oil or electrolytic capacitors also cause distortion in audio amplifiers but due to leakage changing the bias of the following stages or simply due to being very lossy ~ Placing 'poly' caps in series with leaky old PIO caps can often reveal such problems but may not prove 'revealing' to the golden eared

It is likely that the leaky or very lossy capacitors found in vintage audio equipment change the frequency response inside what little negative feedback loop they have ~ Many current designs getting 'a good press' feature multiple RC coupling circuits with some 'odd' time constants coupled with low open loop feedback or no negative feedback

With push pull valve amplifiers where the coupling capacitors are leaky [as original vintage parts will always be to some extent] the distortion can actually sound pleasant due to the normally cancelled even harmonics becoming prominent and more concordant ~ especially over time as the amp warms up and the leakage increases

Unfortunately the output transformer may also start to saturate making the amp sound worse ~ I wonder how many audiophooles actually bother to measure changes in the operating conditions of their equipment before and after swapping capacitors and valves etc.?

With single ended [SE] audio amplifiers which naturally operate in Class A capacitor leakage can affect the asymmetric top or bottom clipping which again increases the even harmonics ~ The best SE amplifiers often produce enough distortion to be considered 'distorting' by Hi–Fi standards but generally SE amplifiers have to be very bad due to cheap transformer construction or valve intolerant design to actually sound bad

Audio folklore shows a strong correlation between close tolerance and well made capacitors and how they sound in audio equipment ~ I believe as I state above that this may be due to the fact that these capacitors [irrespective of the dielectric type] are tightly wound to achieve their specification so have minimum dielectric compression and they often have extended foils to reduce the series inductance and this also strengthens them against compression

Well made NOS Russian Paper in oil capacitors are a good replacement for old leaky ones in British vintage amplifiers but many of these capacitors are reported as sounding 'clinical' and more like poly types ~ Russian K40 are made with individual foil and paper layers ~ The K42 type are made with metallised paper and are smaller for the same rating ~ Both K40 and K42 have extended 'foils' but do they sound different ?

Now in the 21 century it is beneficial for capacitor manufacturers to make technically good capacitors ~ Many are very tightly wound with thin plates and thin dielectric to reduce material cost and many use extended foil because it is easier to machine test and reject bad parts before fitting expensive copper or plated leads

There are still many manufacturers making cheap capacitors like the crushed Mylar above and these can actually work well in many audio circuits providing the a.c. and d.c. are kept within the component rating and they are not required to pass high currents ~ Do they sound musical because they produce the right type of distortion ?

This is probably a bit of a let down because I have not demonstrated any reasons why certain capacitors are preferred for audio other than by spectral evidence which suggests a correlation between capacitor construction and audio reproduction ~ Old leaky PIO can sound nice and musical and 'in the other corner' tight wound polystyrene and PTFE caps sound precise and natural ~ So either even harmonic generation is required or no harmonics at all

One thing capacitors do ~ or should do when used for coupling audio amplifier stages ~ is ensure that no d.c. component is present in the audio analogue ~ d.c. is initially removed at the microphone along with 'low frequencies' that are 'higher' than many audiophooles would like because they believe capturing the absolute air pressure at time of recording is required for complete reproduction

Transients and other features like high frequencies superimposed on low that make up music ~ whether 'miked' or from electronic instruments ~ can introduce very low frequency components and or momentary d.c. offsets which may require long time constants to reproduce accurately and which may be sufficient to momentarily change the characteristic of your 50 year old paper in oil capacitors which are probably already d.c. coupling part of the signal to the following stages or the output transformer

See Here for further tests to show that dielectric compression may exist even if it cannot be shown to produce nice even harmonics or other noticeable audible effects

" Who inspires your fabled fool ~ That's my claim to fame "