|Elliott Sound Products||Cables, Interconnects & Other Stuff - Part 5|
Rod Elliott - Copyright (c) 1999/ 2000/ 2001/ 2002
Page Last Updated - 07 April 2002
Here we have another bunch of lies - or perhaps half truths is a better description. There are differences between capacitors, but they are not (generally) audible - despite the claims. I have seen reference to dielectric losses, the 'sound' of polyester is supposedly inferior to that of polystyrene, and on and on. The stupid part is that all these are true - at radio frequencies - at audible frequencies it is very hard or impossible to measure any difference (or hear a difference, using even a simple blind test).
At the frequencies you and I can hear, there is no audible or measurable difference between most capacitors, unless the equipment builder has done something monumentally idiotic, such as reverse bias an electrolytic. This is (fortunately) rare.
There are some capacitors that are inferior in some regards (but superior in others). For example, many ceramic capacitors have a temperature coefficient that causes the capacitance to vary with temperature (usually negative - N750 or N500 capacitors). NPO ceramic caps have a 'negative/positive/zero' temperature coefficient - i.e. close to zero). There are many claims that these should not be used in audio, but they are useful in audio and RF designs for decoupling (bypassing). The values are generally too low to be useful in most audio circuits (although ceramics are made in higher values, but not always easy to get), but otherwise they would almost certainly be fine - after all, the dielectric is a ceramic and not plastic, so they have low loss and very low self inductance. Having said that, ceramic caps usually have poor stability (they should never be used in filter circuits other than for RF suppression), and non-linearities are well documented at high AC voltages. I would not use a ceramic cap in the audio path for this reason.
Many other high stability or low loss/high power RF circuits (but not those using inductors - N500 or N750 ceramics counteract the temperature coefficient of the coils) will use silvered mica or the like - this is great at 400MHz, but quite unnecessary at 20kHz. Mind you, they are far and away the best low value caps you can buy, and if you can tolerate the expense, fine. Just don't expect to hear a difference.
Many modern opamps have such a wide bandwidth that ceramic caps (usually in conjunction with electrolytics) should be the only choice for bypassing, despite the negative comments of some audiophiles.
Then there are electrolytic capacitors. Claim upon claim has been made about their distortion and poor frequency response, particularly at high frequencies. I recently saw an article (which I would give reference to if I could remember where I saw it) where a standard electrolytic and an audiophile grade unit were tested in the same circuit. The standard electrolytic was actually better, having a distortion component at mid and high frequencies that was only marginally worse than the 'high end' unit, but was much better at low frequencies.
The audibility of an electrolytic cap is (to my mind) still highly contentious. At low frequencies, all electrolytics will start to introduce some distortion. The levels are quite low, but as the capacitor's reactance becomes significant, distortion rises.
The reactance of any capacitor is determined with the formula ...
Xc = 1 / (2 *π * f * C) where Xc is capacitance reactance, f is frequency and C is capacitance (in Farads)
If Xc is maintained at 1/10 (or 0.1) of the supplied load impedance, then this low frequency distortion will not be an issue, but in any case is far lower than that of a loudspeaker.
High frequency performance is affected by the capacitor's internal inductance and dielectric losses, which causes a rise in impedance as frequency increases. It is very common to see electrolytics bypassed with polyester or similar caps, and for RF this is essential. It is also needed when bypassing the power supply rails of an amp, since at the frequencies that amps like to oscillate at (typically above 1MHz), the electrolytic simply has too much impedance. For audio frequencies a bypass is not needed, but will do no harm.
The combined effects of internal resistance and inductance contribute to the electrolytics' equivalent series resistance, or ESR. This can be measured (I have an ESR tester), and is a good indication that a cap is failing. As electrolytics age, their ESR will rise until a point is reached where the component will be unserviceable.
As a test, I checked a few caps in my workshop. I could not measure any distortion created by an electrolytic passing signal current (as opposed to speaker current, which I did not test at this stage). I also checked the frequency response of a couple of electros, and found zero degradation at 100kHz - even a square wave was passed with no visible deterioration in rise time (which would indicate frequency limitation).
I then tested the ESR and capacitance of a 220µF and 10µF electrolytic, and a 1µF polyester capacitor.
|Type||Value||Meas Val.||ESR (Ohms)|
|Electro||220 µF||207 µF||0.17|
|Electro||10 µF||10.1 µF||1.2|
|Polyester||1.0 µF||1.02 µF||1.5|
I thought that this was quite interesting, personally. If we use 10µF electros where we might have otherwise used a 1µF polyester, the ESR is better. Will it make any difference whatsoever to the sound? Of course not. None of these devices introduced any measurable distortion or anything else that I could see. One thing I know for sure is that if I can't see any change on my distortion meter residual, then there is no change. A complex waveform does not affect the validity of this testing, since I can test distortion at any frequency I like, and the appearance of multiple frequencies at once does not affect any passive device.
Considering that I use the averaging facility on my oscilloscope to eliminate noise completely, I can see the most minute change in a signal waveform. If nothing can be seen here, then no-one, regardless of how good they think their ears are, will be able to hear the difference in a properly conducted test.
I was recently taken to task for not mentioning tantalum capacitors. I hate them! They are unreliable, and many tests have shown that their linearity is highly suspect. The only intermittent short circuit I have ever found in a cap was with a tantalum in a power supply circuit. It would fail for long enough to blow the fuse, and then work again. I strongly suggest that you don't use tantalum caps in anything more advanced than a dustbin.
Bottom Line on Capacitors
Various people have advocated passing pulse signals through two different sorts of capacitor, and subtracting the result, claiming that the non-zero residue proves that capacitors can introduce audible errors. In fact such tests expose only well-known capacitor shortcomings such as dielectric absorption and series resistance, and perhaps the vulnerability of the dielectric film in electrolytics to reverse biasing.
No-one has been able to show how these imperfections could cause capacitor audibility in an amplifier,and my own tests confirm this.
I must confess though, that perhaps we don't know how to perform the 'audibility' tests. I do not believe that there is a significant difference, but many do ... who knows?
Non-polarised electrolytics are a different matter, especially when used in crossover networks. These have a tendency to lose capacitance as they age, shifting the crossover frequency with disastrous results (sonically speaking). Because the loss is gradual, you may possibly not even hear it until the tweeter has almost stopped working, as you get used to the sound over a period of time. Unless all bi-polars age at the same rate (unlikely), you will start to notice a difference between the two speakers. This is your cue to head off to the electronics shop and buy some replacements (non electrolytic, preferably). There are (supposedly) some major audible differences between bipolar electrolytics and film dielectric (plastic) types. This is your chance to test the theory.
There is an audible and measurable difference between different dielectrics. It's less to do with dissipation factor, and more to do with dielectric absorption. There is no black magic about it - its very well documented throughout the entire range of electronics industries. Here's how to convince yourself that this is possibly the most insidious source of distortion in audio. Get a largish value electrolytic reservoir cap and charge it up to (say) 50 volts for a minute or so. Then, with a DVM attached to the terminals, discharge the cap so the voltage reads zero.
After removing the discharge resistor, watch the voltmeter reading climb back up as the cap miraculously charges itself back up from nowhere. In a signal coupling capacitor this would be bizarre enough behaviour to be a worry, but can you imagine what the effect must be if the cap is in a feedback loop? Give it some thought, and think about how much damage would have to be done to a signal to loose the ambience surrounding a quiet instrument buried in a large ensemble. Think of reverb at -60dB, or lower.
Dielectric absorption is significant, and ceramic caps suffer from it badly, as do electrolytics. Do some measuring, but more importantly, do some listening.
Fair comment, and it deserves an answer. I also know the 'stored charge' phenomenon of electrolytics (I actually demonstrated it to my son some time ago), and although this is absolutely real, it does not reflect the behaviour of the cap in a real world amplifier.
Most capacitors normally don't charge and discharge in this manner, but remain charged to some DC potential at all times. The charge recovery mechanism should never come into play in a properly designed circuit, regardless of programme material. There are exceptions! A capacitor used as the 'dominant pole' or Miller capacitance in the Class-A amplifier section of a power amp is charged and discharged fully with each cycle of the input waveform. Ceramic capacitors are commonly used in this role, but I suggest that polystyrene is probably much better. Will you be able to hear the difference? My own experience is that you should not hear the slightest change in the sound, but it is conceivable that with some amplifiers this may in fact be audible in extreme cases.
The dielectric absorption process is present to some degree in all capacitors, and although some are definitely worse than others in this respect, I have conducted some tests with my Sound Impairment Monitor (SIM), which has never been able to detect any degradation of the type you might think should happen.
The claim that there will be an effect similar to reverberation at -60dB is complete nonsense. No such effect is measurable or audible. I do think that after all these years, someone would have worked out a way to prove this effect if it existed at all. No such proof has been offered, but I have seen 'proof' that a ceramic capacitor (pushed way beyond its voltage rating if I recall correctly) can introduce some measurable distortion. The solution is easy - don't run any capacitor at above its voltage rating, and all will be well.
Feel free to test the theory. Make sure there is no DC in the signal line, and connect a bipolar cap (say 10µF) in series with the interconnect between preamp and power amp. Wire a switch across the cap, and have someone operate the switch while you are listening. You have to be able to hear a difference at least 75% of the time (and accurately identify whether the cap is in or out of circuit). If you can do this, then the probability is that the capacitor is audible (unless you do something nasty with the switch wiring that gives audible clues - this is cheating ).
There are some who insist that the instantaneous current output needs to infinite (or at worst, half this value), and that amplifiers with limited current sound terrible. This is another piece of nonsense.
Let's assume that a nominal 8 Ohm loudspeaker load has an impedance minimum of 1 Ohm at some frequency. This is a bad design, but a valid assumption. This means that the amplifier must be able to supply a maximum of 8 times the normal current. A 100W amplifier would then supply a normal peak current of a little over 3.5 Amps. At the frequency where impedance falls to 1 Ohm, this becomes just over 28A.
So let's have a look at the very worst case possible, where the load is fully reactive and returns all supplied energy 180 degrees out of phase (at this point, the load is performing no work, so if a loudspeaker, is making no sound). The amplifier now has to deal with two lots of current - that supplied to the load, and that returned from the load. Even if it were possible, the worst case above would require a current capacity of 56A, however a loudspeaker that presented such a load to any amplifier will not last long in the market, since it will blow up nearly every amplifier that is attached to it.
There is no audible benefit whatsoever in creating an amp that can supply 100 or 200A, since the load will never need this current and is incapable under any circumstance of drawing more than the applied voltage and minimum impedance will allow (allowing for the reactive component of the load).
Class-A amplifiers are generally capable of a very modest current, usually barely above that theoretically needed to drive the speaker. I have not heard anyone claim they are rubbish, because of the low current capability.
The one exception is with extreme crossover networks or other speaker configurations that create a difficult impedance load. It will often be found that some amplifiers cannot drive these speakers well, and others have no problem. An amplifier capable of high current may sound better with these loads, but I suggest that the speaker design is flawed if the designer is incapable of creating a crossover that cannot maintain a respectable impedance.
Someone has managed to convince a sizable segment of the audiophile fraternity that to achieve acceptable channel separation, completely separate amplifiers must be used. Considering that it has been shown  that 20dB channel separation is quite sufficient for a full stereo image to be appreciated, it is nonsense to claim that infinite separation is needed or desirable.
It is not at all difficult to design an amplifier with better than 50dB separation, even using valves, and any more than this is of no audible benefit. The 'cross-modulation' effect that a shared power supply supposedly introduces is drivel. If an amp is so heroically ill-conceived as to suffer from cross-modulation, then simply sticking it into its own case with a separate power supply certainly won't fix it. I might suggest that it is most ideally suited as a boat anchor, since the design is so seriously flawed that it is beyond salvation.
A common power supply is a sensible (and far cheaper) alternative, and will cause no crosstalk in itself. Most amps have a very high ripple rejection, and if they reject ripple, they will also reject any signal frequencies that happen to get onto the supply line.
In fact, the conventional power supply capacitors will filter out all but the lowest frequencies anyway, and since bass is almost invariably recorded onto disc as mono, a minor amount of crosstalk at low frequencies is of no consequence - even if it were possible by this means, which it generally is not.
Alternatively, they are useful if you want to have the shortest possible speaker leads. The amp can be installed next to the speaker, and a very short lead used to connect the two. Then we create a problem with the low level interconnects, which will be of significant length. There is far more chance of interference and high frequency loss in long interconnects than in speaker cables, so ideally the interconnects should be low impedance balanced circuits. Sadly, most monoblocks do not offer this essential option.
I have seen reviews, and claims by amplifier designers, that for this amp to sound good, it must have a fast power supply. A power supply does not have speed - high, low or otherwise ... unless it is regulated.
Anecdote: A regulated supply will have a finite ability to maintain its output voltage as the load varies - this is well known in engineering circles, and can cause problems if it is not fast enough.
I have worked on power supplies (many years ago) that were used to power the head positioning amplifiers in the old washing machine sized computer disk drives. One of the tests that we needed to run was a switched load - varying the load from about 0.1 of the rated current to full current repeatedly.
Detectors were used to measure how far the output voltage dropped when the full load was applied, and I designed the circuit to do this. It was fairly fast, and would latch a 'Fail' LED if the output fell below a predetermined limit for more than about 1us. BTW, the low voltage limit was set at only about 1V below the rated output voltage, which was 24V if I remember correctly.
Bear in mind that there is no audio signal that can cause an amplifier to go from quiescent current to full output in anything near 1us. A 20kHz signal has a period of 50us, and full power at 20kHz will fry your tweeters in seconds.
So, speed is valid for a regulated supply for a critical application, but is completely meaningless for a power amplifier with an unregulated supply - which is 99.9% of them. There is no audio signal that is so fast that it will demand power from perfectly ordinary electrolytic capacitors faster than they can supply it. It is not necessary (other than for radio frequency bypassing) to use polyester caps in parallel with the main electrolytic filter caps, and nor is there any valid reason to specify that 100,000µF (or any other outlandishly expensive number) is needed to power an amp so the supply will be 'fast'.
A standard electrolytic (say, 10,000µF) will have an equivalent series resistance (ESR) of perhaps 0.01 Ohm, which means that it has an internal time constant of about 100us. More significant is the fact that discharge current is limited by ESR, so if charged to 50V, the maximum current available is 5,000A peak - this is a lot of current! In fact it is so high that it can destroy the cap itself - this is a very good reason not to use a screwdriver to discharge a power supply, well apart from the fact that a decent amount of capacitance will take the end off a small screwdriver.
If this 50V supply is connected to a 8 Ohm speaker via an amplifier, the maximum current the speaker will draw is 6.25A (although some speakers will demand more at certain frequencies). In reality it will be far less than this most of the time. I can make a power supply 'slow', simply by placing some resistance in series - the caps will no longer be able to discharge at their maximum rate. Will this affect an amplifier? Only in that the maximum power will no longer be achieved, but this will also happen if the AC mains supply is 10% low. Does this somehow degrade the sound of an amplifier? I think not.
Massive capacitance and 'audiophile' grade caps are not going to improve the sound of the DC from your supply, regardless of cost or claim. The power supply is a passive part of the amplifier, and has little or no influence on the sound, unless grossly and ingeniously poorly designed. I say 'ingeniously', because it would take spectacular incompetence to so badly design a power supply that it audibly affected the sound at any signal level below clipping. From some of the posings I have seen on various bulletin boards, such incompetence may well be rife, since just changing a power lead makes them audibly better .
The editorial page has a pair of prime examples of 'Special Designs', including a more detailed examination of the sample below. See I am as Mad as Hell for more info.
I recently saw information on the web about an amp whose claims to fame (infamy, more like it) were along the following lines (this is taken from the site)
Oh, wow, and ... I mean ... like ... who gives a toss! This amplifier sold for some astronomical sum, and as near as I could tell from the advertising blurb, seems to use a couple of IC power amps as the entire circuit. The power supply was separate (and had an even sillier name than the power amp).
So the amp has few components and a short signal path. What about the several hundred metres of standard professional class cable and very long signal paths that are common in the mixers that were used in the recording studio? Is this 'magical' short signal path going to somehow make that all go away - somehow I doubt it. Since this amp does use a power opamp, the manufacturer obviously does not count the 30 or so transistors inside the device - why not? They are real, whether you acknowledge them or not.
As for the 'world's shortest negative feedback path'. So what? The claim was made that by doing this, power supply bypass capacitors that by some mystical process ruin the sound were not needed. What rubbish. My 60W power amp has a negative feedback path that is about 50mm long - in other words, typical. Because of its design, it will operate perfectly happily with no power supply bypass capacitors too - the result is greatly reduced power because of the resistance of the power leads. Do you want that? Does this sound like a good idea? No, I didn't think so either.
The one I liked best (or least, depending on how you think I read this nonsense) was the 'world's smallest filter capacitors'. What possible benefit - other than profit maximisation - does this infer? I honestly have no idea. I could run my amp with 1000µF caps too. Anyone can. The immediate result is a dramatic reduction in power, as ripple voltage is very high at any reasonable power level, and you start to get clipping as the ripple voltage encroaches on the audio signal. Ah, but we also have 'powerful voltage regulation' and a high capacity transformer. Big deal. I can run my amp off a 10kVA transformer if I want to, and it won't change the sound one iota. Anyone can make a voltage regulator (assuming that one is actually used, which I doubt), but why? Extra heatsinks, more stuff to fail, and zero sonic benefit.
I won't even bother discussing the 'dual mono construction', but I am intrigued by the 'rigid and compact aluminium chassis to release vibrations smoothly'. Quite apart from the fact that being rigid and compact in no way ensures that vibrations will be released smoothly or otherwise, I am at a complete loss as to why anyone might think for an instant that this was important. This is an amplifier, not a speaker cabinet. Left to their own devices, amplifiers don't normally vibrate - this is not one of their characteristics. Are we supposed to believe that a power amp is in some way microphonic?
Try this (if you dare). Place your ear as close as possible to the speaker, and have someone drop the power amp a short distance while powered on and connected. What do you think is the chance that you will hear anything from the speaker (other than if the amp destroys itself when it is dropped)? I will tell you, to save the embarrassment of having to explain to the service guy what happened to the amp. Nothing, that's what. If these clowns have managed to make an amp that is microphonic, then I definitely don't want one.
I thought about this one for a while, and it finally made it into my 'Hall of Infamy' - the editorial. You can read more about it there (see above).
Many is the claim that opamps have a distinctive sound, and can readily be heard in audio equipment. Discrete designs supposedly sound superior, regardless of the fact that in many cases they will measure worse than even a cheap opamp.
I have never been able to measure an opamp's distortion, because it is so far below my equipment's limits that it cannot be detected. Devices are available with distortion as low as 0.00008% - this is close as you can get to the ideal 'straight wire with gain'. The bandwidth of the better devices is so wide that significant gain is available at 100kHz, so phase irregularities and response problems are non-existent in sensible designs.
Considering the fact (and in the vast majority of cases, it is fact) that the final mixed down signal you get from a CD has passed through up to 100 opamps at various stages of production before you even get to listen to it, it is ludicrous to assume that not using opamps in the last 1% of the audio chain will have any audible effect.
Valve amplifiers are back, with units in all sorts of configurations selling for astounding sums.
The valve sound is one phenomenon that is real. It has been known for a long time that listeners sometimes prefer to have a certain amount of second-harmonic distortion added in, and most valve amplifiers provide just that, due to huge difficulties in providing good linearity with modest feedback factors.
While this may well sound nice, hi-fi is supposedly about accuracy, and if the sound is to be modified in this manner, it should be set from the preamp front panel by a control (Douglas Self suggests a 'niceness' knob).
Valves offer some advantages - their overload characteristics are smoother than solid state designs, so even when clipping the sound is less harsh. While this is most desirable for a guitar amplifier that will be operating into clipping for much of the time, it is unhelpful for hi-fi, where clipping should be avoided altogether.
Valve amps also have much higher output impedance than transistor amps, and this makes some speakers sound better. It also makes other speaker sound worse, so the results are unpredictable.
There are few modern transistor amps that will measure worse than any valve amp, regardless of cost. Indeed, the vast majority are so superior in all respects that it is difficult to justify using valves in anything other than guitar amps, where, despite much advertising hype, no transistor amp has ever been able to sound exactly the same as a valve unit. Close - but not the same.
The rash of single-ended directly heated triode monoblock amplifiers of late is something that astonishes me. These will typically have a distortion of 1 to 3%, are of low power - typically less than 10W, and have no redeeming features (IMHO).
Such an amplifier generates large amounts of second-harmonic distortion, due to the asymmetry of single-ended operation, and needs a very large output transformer because the primary carries the full DC anode current, and core saturation must be avoided. The inherent distortion of an iron cored inductor or transformer is ever-present, and only global feedback can remove it.
High values of feedback around a transformer are extremely difficult, because the phase irregularities generally cause the amplifier to oscillate. This may have been the state of the art 50 years ago, but there is no sensible reason to go back. Next we will hear someone extolling the virtues of the wax cylinder as having superior sonics to vinyl or CD (needless to say these superior sonics will be "very subtle" and "only audible with the finest (i.e. most expensive) single ended triode monoblock amplifier").
In one review, a single ended triode amplifier yielded 3% THD at 9 Watts, at a cost of $3400 . This is an appalling result for a very expensive single channel amp. The amplifiers in powered computer speakers are better than that!
Despite all of the above, I have no doubt that many of these amps sound delightful. Not exactly my cup of tea, but having used valve amps of many types over the years (including those I designed and built myself), I still like the sound of them. They also don't blow up with difficult loads - they may stress out a little and give less power than normal, but they survive. The majority of valve amps are far less forgiving of open circuits (no speakers connected), and some will fail if pushed hard into an open circuit. The typical failure mode is a high voltage flashover, which either carbonises the valve socket or base (or both), or causes the insulation in the output transformer to fail.
The trend towards having these hot 'bottles' out in full view, and able to be touched (and / or broken) by age challenged persons (the rug-rats) is a definite safety hazard. I would not like anyone's kids to be able to burn and then electrocute themselves in one small mishap.
.... However - I do (or did until recently) use a valve preamp in my own system, and I have no idea what that says about me. It does sound nice, but I am probably deluding myself in thinking that it is better than my solid-state preamp. That's fine for me, because I designed and built it, so it didn't cost me a king's ransom.
There are many very fine loudspeakers available, and interestingly, although these have a far greater effect on the sound that you hear than the amplifier, there is nowhere near the controversy with loudspeakers as seems to be evident with the rest of the audio chain.
Certainly there are proponents of various crossover alignments, the benefits or otherwise of vented boxes versus sealed, but otherwise this seems to be a reasonably sensible (even if intimidatingly expensive) field of endeavour.
Most audiophiles have their favourite speaker system(s), and these will all have some undesirable characteristics, for such is the state of the art. The perfect loudspeaker does not exist, because of the physics of making electro-mechanical objects with finite mass react in a completely predictable manner at all frequencies. This (of course) is something that speakers cannot do.
A flat frequency response is desirable, and rapid decay of internal resonances means that the loudspeaker contributes a minimum of its own sound to that from the source. Good quality drivers and well braced, non resonant cabinets, combined with high quality components in crossover networks and a sensible approach to ensure that phase irregularities at the crossover frequencies do not cause response or impedance peaks and dips are common in most quality systems.
The listening room and the recorded material has a very great influence on the final sound you hear, vastly more than a few interconnects or a mains lead. No-one is going to make the listening room anechoic, and nor would you want to. The positioning of speakers is one thing that can have a profound effect on the sound, but this is so often completely ignored.
One problem is that the optimum placement of speakers for sound quality will often be completely inappropriate to the layout of the room, meaning that a livable area is no longer available, and causing much friction between the listener and s/he who must be obeyed.
Bear in mind that building a speaker system without measurements is futile. The ear (and attached brain) is easily fooled, and has a very short memory for what you hear. Speakers can have huge anomalies in response, and within a few minutes the brain has made the necessary adjustments, and everything will seem to sound as it should.
Try this experiment. If you have a graphic or parametric equaliser, reduce a band somewhere in the midrange area (say, between 500Hz and 1kHz). Listen to the system for about 15 minutes, then restore the missing frequency range. Suddenly, the system will sound as if it has a huge peak in the midrange, and for a time will sound awful. Within another 15 minutes or so, everything will have settled back to normal.
|Copyright Notice. This article, including but not limited to all text and diagrams, is the intellectual property of Rod Elliott, and is Copyright (c) 1999/2000,2001/2002. Reproduction or re-publication by any means whatsoever, whether electronic, mechanical or electro-mechanical, is strictly prohibited under International Copyright laws. The author (Rod Elliott) grants the reader the right to use this information for personal use only, and further allows that one (1) copy may be made for reference. Commercial use is prohibited without express written authorisation from Rod Elliott.|