Capacitor ESR and the dangers therein

[YNWaN]

Of course, the title is a bit tongue in cheek because there aren't any really serious dangers, but there are perhaps some issues.

A recent thread has thrown up an interesting issue regarding the ESR (resistance) of replacement capacitors.

I must admit that although I knew of ESR in caps and new that polyprop have significantly lower ESR than electrolytics I didn't really consider it when upgrading my own Yamaha crossovers!

Perhaps I should have though. In truth I didn't think it would impact to a significant level - but perhaps I was wrong. In my case the crossover is all second order (3 way); the tweeter and upper mid caps are apparently metalised paper (very cool for the time) but the lower mid and bass are electrolytic. I replaced all of these with same uF value polyprop and didn't notice any balance issues - nor measured FR issues after the event. Having said that, there are resistors in the form of L-Pads duplicating the role of adjustable pots for the mid and treble so perhaps differences were swamped.

Personally, I find it a very interesting subject. Have others any experience or views?

[The Barbarian]

Im keeping out of this one since practice & experience counts for nothing on AOS

[Gazjam]

Lower ESR decoupling cap in my Young PSU (One of Si's) made a great improvement, noticeably opened the dac up.

[istari_knight]

I've "re-capped" over twelve pairs of speakers now, the first ten I followed internet wisdom which was: always replace an electrolytic with a film. In some instances it really improved the sound, looking back this was always on simple designs with first order crossovers [Acoustic Research & Epos to name a couple] and yet on more complex designs such as IMF or Rogers replacement always resulted in an inferior sound. They would always end up bright & harsh sounding, out of kilter if you will... Like someone's turned the treble knob up too high ! I was at a loss to why this might be but disregarded it as flaws in the speakers design being highlighted by the higher quality capacitors.

Whilst researching how to re-cap a pair of Ditton 44's & replace the tweeters with modern variants I stumbled across a very knowledgeable chap on Diyaudio who kindly spent a good few hours calculating the necessary changes. When told all electrolytics were going to be replaced with films he was adamant this is very bad practice for the very reasons Paul outlined on the other thread [electrolytic=high esr / film=low esr.] He then calculated the necessary "esr mimicking resistors" if films were to be used to guarantee the crossover still performed as it was intended.

I have followed that advice ever since & have always been happy with the results which were much more predictable than before ! I dont want to speak for him but will say: Andr'e always replaces electrolytic with films but not any old film, only Ansar Supersounds. He has also experienced poor results when using other brands. He has a good ear so I intend to give his method a go next time, after all... Not to try is not to know, as they say.

[Marco]

Im keeping out of this one since practice & experience counts for nothing on AOS

Not true! ;)

Marco.

[walpurgis]

What exactly is wrong with using Electrolytics? If they were fitted to the original crossover, which would no doubt have been designed carefully and listened to during development, surely they would be the right thing? Seems to me, automatically looking for alternative types may not necessarily be the right move.

[YNWaN]

Im keeping out of this one since practice & experience counts for nothing on AOS

Oh, I don't know -I think part of the problem is that we are all very proud (justifiably) of our own experience and the knowledge it has given us - the problem is we all have different experience and have probably learnt different things from it :). When I first read Paul's (RFC's) thought on ESR I was immediately dismissive - but thinking (and reading) about it I regret my knee-jerk reaction. I need to read up on it a lot more, which is not easy as the Internet is not rich on the subject (thanks for the article link by the way, I have actually read it before but it bears further study). I think Paul and Co potentially have a very valid point with regard to potential issues with ESR. What I mean is I think they definitely have a very valid point it's just that I have to contextualise it.

[YNWaN]

What exactly is wrong with using Electrolytics? If they were fitted to the original crossover, which would no doubt have been designed carefully and listened to during development, surely they would be the right thing? Seems to me, automatically looking for alternative types may not necessarily be the right move.

Well, playing devils advocate, the problem with using electrolytics is that they perform and sound bad! They may well have been carefully auditioned (equally, they may not have been) but then polyprop caps didn't exist back then and if they did they were too expensive to consider. The problem with like for like replacing electrolytics with polyprop isn't that polyprop aren't technically and subjectively better, it's that they have a specific quality that is quite different and this may have significant knock-on implications (namely, their ESR).

[walpurgis]

Are there no Electrolytics of predictable and extended life and good performance? I'm asking these questions as I don't know. I understand basic electronics, but I'm no 'wiz'.

[YNWaN]

Electrolytic capacitors inherently alter their properties with time and use - essentially they eat themselves very slowly (I simplify). In absolute terms the main advantage of electrolytic caps is their low cost and the fact that they are physically small for the value they are. It depends what specification aspects one considers important, but within the context of capacitors it is generally accepted that polyprop (also polystyrene and a couple of other) capacitor types are better than electrolytic equivalents and these technical benefits appear to be supported by subjective experience/assessment.

Edit: Somewhat ironically, the aspect of electrolytic capacitors that often changes with time is not so much their actual value (their uF rating), it is their series resistance (the ESR). In the worst cases a caps resistance can increase so much that it barely passes a signal!

Edit of edit: The life span of electrolytic caps has to be put into context though - large reservoir capacitors used in power amplifiers are likely to age a hell of a lot faster than the very much smaller items used in a crossover (particularly as the power amp caps are likely to be heated by the amp which the crossover ones are not).

[Gazjam]

Lower ESR decoupling cap in my Young Dac (One of Si's suggested mods) made a great improvement, noticeably opened the dac up.

...meant to say.

[YNWaN]

I think that low ESR is acknowledged to be a good quality for capacitors. The particular issue relates to speakers and specifically to the like for like value replacement of relatively high ESR capacitor (typically electrolytic) with very low ESR caps (typically polypropylene). Decoupling caps in electronic circuits is an altogether different matter.

[sq225917]

I'd always stick close to the recco value and original caps or ensure I measured them in room. My Yamahas are noticeably flatter from 8k upwards due to better tweeter and additional XO.

[fatmarley]

If I lower the ESR 1ohm from the capacitor in the zobel circuit of my tweeter filter it lowers the high frequencies around 2db (2db makes a big difference) and also knocks the phase tracking out. It's much worse if I do it to the cap in the bass filter.

Does anyone know what the typical ESR values of your average electrolytics would be? I know it depends on the type and value, It's just I don't bother to model crossovers with electrolytics, so wouldn't have a clue.

[Reffc]

If I lower the ESR 1ohm from the capacitor in the zobel circuit of my tweeter filter it lowers the high frequencies around 2db (2db makes a big difference) and also knocks the phase tracking out. It's much worse if I do it to the cap in the bass filter.

Does anyone know what the typical ESR values of your average electrolytics would be? I know it depends on the type and value, It's just I don't bother to model crossovers with electrolytics, so wouldn't have a clue.

Its not as simple as that unfortunately...ESR varies with frequency, temperature and even the age of the electrolytic but for modelling purposes, by far the most important variable is frequency.

For 200Hz and under, 1 to 1.3 Ohms is about right; 200 to 500; around an Ohm; 500 to 1KHz 0.7 to 1 Ohm; 1 to 2kHz, perhaps 0.5 to 0.75 Ohms and above 2KHz, at least 0.5 Ohms unless low loss electrolytics are used in which case, probably half the above values. This is very general but is a guide and only valid for normal ambient temperatures on new caps. However, it's sort of reverse engineering when it comes to filters as there's no substitute for measurement of actual response following rough value estimation through calculation. I don't know what you mean by "phase tracking" but phase isn't directly affected by reactive impedance.

[fatmarley]

Its not as simple as that unfortunately...ESR varies with frequency, temperature and even the age of the electrolytic but for modelling purposes, by far the most important variable is frequency.

For 200Hz and under, 1 to 1.3 Ohms is about right; 200 to 500; around an Ohm; 500 to 1KHz 0.7 to 1 Ohm; 1 to 2kHz, perhaps 0.5 to 0.75 Ohms and above 2KHz, at least 0.5 Ohms unless low loss electrolytics are used in which case, probably half the above values. This is very general but is a guide and only valid for normal ambient temperatures on new caps. However, it's sort of reverse engineering when it comes to filters as there's no substitute for measurement of actual response following rough value estimation through calculation. I don't know what you mean by "phase tracking" but phase isn't directly affected by reactive impedance.

Oh yes I remember now, of course it varies with frequency. Looking at LspCAD you can only enter a single value for ESR though, or am I missing something?

Here's a thread called "The importance of phase tracking (http://www.diyaudio.com/forums/multi-way/211883-importance-phase-tracking.html)"

[Reffc]

Oh yes I remember now, of course it varies with frequency. Looking at LspCAD you can only enter a single value for ESR though, or am I missing something?

Here's a thread called "The importance of phase tracking (http://www.diyaudio.com/forums/multi-way/211883-importance-phase-tracking.html)"

RE LspCAD, I'm not familiar with it as I tend to use LTSpice4 for electrical modelling.

Ah, right, I see what they mean. Phase performance varies between speakers (even with the same order crossover) with variations in cabinet design and the drive units themselves. It is important, both on and off axis not to have too steep a phase angle between units or this will be heard quite audibly at crossover. The debate is usually "how much is too much". Phase, inevitably will vary irrespective of the units/crossover used when you start to look say two octaves above and below crossover, for a number of reasons, but if you aim for both units being crossed over to remain within say a 20 degree envelope for at least and octave below/above, then much beyond that, the summed response should be in the favour of the duty unit, especially with higher order. It's the off axis response that matters, and the critical point is often seen as the crossover point itself where both units should be closely phase matched AND have similar responses to at least 30 degrees off axis. Having one unit's phase cross steeply at 90 degrees but being zero at crossover is not much good if there is a large off axis step response within half an octave combined with a large phase difference. Just to confirm here we're talking about measured acoustic phase response, and measuring it is the only way of checking it, so a calibrated mic is a must. Electrical phase isn't always a reliable indicator of acoustic phase. In a nutshell, the closer the phase, the better, especially within an octave or two of crossover.

[Reffc]

I've tried not to get too deep into phase response as it is one of those dry subjects that cause eyes to glaze over as soon as it is mentioned!

However, it is a really important thing to get right in terms of definition since it is the most mis-understood and mis-quoted aspect of design.

When you here terms such as "phase tracking" they are on their own, meaningless and confusing terms, so better to start with what phase is, and what affects phase.

It is simple the time based relationship between the acoustic signal produced between drive units at all frequencies. In an ideal world, you'd have a pair of speakers where the input signal was reproduced in perfect time alignment and heard at the listening position without room based distortions such as reflections all in perfect harmony. Back in the real world, this is not possible.

I mentioned above that phase response varies with cabinet design as well as electrically; those two aspects are important. By cabinet design, I refer to the positioning of the speakers relative to one another, and the baffle design (flat, sloped or stepped). The electrical phase response is largely controlled by the crossover design itself; in general, the higher order the crossover, the more phase coherent the design at all frequencies but only if the initial time alignment is correct (ie the audible effects of phase diminish the higher the order of filter slope used providing that the acoustic time alignment is considered from the outset).

Consider a typical criossover frequency for a 2-way of 2.5KHz. lets have a look at some of the considerations that affect phase:

1. positioning. The wavelength of a 2.5KHz signal is approximately 140mm, so the distance between the units being crossed over should ideally be no greater than this otherwise a time delay, or phase shift will occur relative to the listening position. How relavent is this? Anyone with a TV which has a control for lip synch will know that even a 20msec time delay is audible (compared with the picture), so it is very important to get this as close as possible.

It can be done by sloping the baffle or offsetting the tweeter slightly to get the acoustic centres equal at crossover but then we have another problem: the time alignment will ONLY be accurate at crossover, and will vary with frequency at all other frequencies depending upon the slope used for the crossover filters. It is why in some cases it is better to have the tweeter in vertical alignment with the mid/woofer rather than offset (for the lobing pattern effect) and treat the diffraction issues separately (as with the Rhapsody speakers) as you end up with a more "time coherent" design, although at this juncture, it has to be stressed that is only one of many ways of achieving this.

2. Electrical: A First order slope is widely regarded as phase coherent at all frequencies and some snake oil vendors even try and flog their speaker boxes quoting this as gospel. It's actually not true. Whilst it is true to say that 1st order slopes on initial analysis shows that 1st order slopes are 90 degrees out of phase at all frequencies between low/high pass filters, the fact is that the high pass section leads the signal by 45 degrees so the low pass as steady state lags by 45 degrees. This can be corrected by shifting the phase of one relative to the other which cannot be done for all frequencies at the crossover so has to be done by acoustic time alignment.

We then need to look at the value of the amplitude (voltage) and power at various 0ctaves above and below the crossover point of the phase differences. Without needing to work out and quote the figures, which are actually easier to visualise graphically, it soon becomes obvious that low order slopes maintain higher amplitudes for the phase differences above the cutoff point, so the higher the order, the lower that audible phase shift becomes and the more phase coherent a system can sound but ONLY if the acoustic centres are time aligned to begin with (this does not refer to physical driver positioning but to relative phase between drivers). That is why, generally speaking, the more expensive designs use higher order or even active filters, because the difficulties in producing a well designed passive filter much beyond 3rd order get very complex not to mention cost. A well damped LF driver using a minimum of 2nd order is usually fine for most purposes. When people claim that they can "hear" the crossover, it is usually not a reflection on a higher order crossover being used, but on the designer not getting the acoustic phase alignment correct to start with AND due to steps in off axis response which brings us neatly to:

3. Off axis response: Now that we have looked VERY briefly at acoustic and electrical phase, the other important aspect of phase response is off axis response. It is essential for a good pair of speakers to work well off axis, or all that trouble getting time alignment and electrical phase spot on is wasted because at some frequencies, there will be large differences between listening on axis and off axis and this means that the reflection of the signal from on axis can sometimes bounce around the room and reach the ear at a higher amplitude than the off axis signal but crucially, also out of phase, so high phase distortion off axis is the result. This becomes more important at mid and high frequencies, so there are two points to remember: Firstly, the mid/HF units should be capable of very close 30 degree off axis performance across the all important mind band and for much of the lower HF region. Above that, our hearing becomes less sensitive. At lower frequencies, the effects of bass "timing" with the higher frequencies still dictates that the unit must be acoustically phase matched from a time alignment perspective but the cabinet design and driver design then can affect timing lag, which is more commonly referred to as "Group Delay" (not exclusive to bass but describes part of a signal or system out of step with the rest). Whilst sealed boxes give the lowest group delay (generally speaking), bass reflex can be carefully designed for very low group delay. Anything of 20msecs or greater can be audible so it's important that cabinet designers consider this aspect to get coherent bass timing.

For those feeling adventurous, there is plenty of in depth explanation on-line but a good place to start, and a resource that I'd recommend is on this link: http://sound.westhost.com/ptd.htm (http://sound.westhost.com/ptd.htm)

Now that all of the above has been considered, the big shocker (drum roll please) is that as phase accurate as you can get a design, the one way to undo all of that good work is to put the speakers into a room! A bit flippant, but the point being that whilst getting phase accuracy from a speaker design is important, the phase distortions caused by most rooms make the inaccuracies of many speaker designs pale into insignificance, so the message here is to really invest a lot of thought and time in getting the very best positioning in room. This generally means things like getting speakers well away from side walls, avoiding reflective surfaces in between the speakers and applying considered room treatments at first reflection points (soft furnishings are fine). It's often a compromise due to the room itself, but nothing introduces so much phase distortion as the room itself. The message is really not to get too hung up on the effects of filter design providing the basics are understood and attention has been paid to time alignment. This applies to capacitor types and values. Here, mimicking what the designer originally had in mind is rule of thumb but that is not to say it cannot be improved upon, as everything can. Hopefully this helps rather than adds to the confusion!

[fatmarley]

Thanks Paul. Loads of good info to digest there.

Should the drivers be time aligned for the listening distance? Mine use a waveguide and are 11mm out of alignment at 1m, so I assumed (probably wrongly) that they wouldn't be far out at the listening position of around 2.5m (don't think it matters with 4th order acoustic slopes though)
The crossover slopes are 4th order L/R acoustic according to simulations (2nd order electrical) , so I guess I should get my calibrated microphone out and double check that they really are 4th order L/R. Or would it be better to just reverse the polarity of the tweeter and check to see if the null matches my simulations at various listening heights? In my sims the biggest null is obviously at the listening height and it gets smaller above and below (as it should). If my measurement match this at the listening distance, then wouldn't that be a good way of checking? I know the room messes up measurement so I can measure outside if I can't get an accurate measurement indoors.

[Reffc]

Thanks Paul. Loads of good info to digest there.

Should the drivers be time aligned for the listening distance? Mine use a waveguide and are 11mm out of alignment at 1m, so I assumed (probably wrongly) that they wouldn't be far out at the listening position of around 2.5m (don't think it matters with 4th order acoustic slopes though)
The crossover slopes are 4th order L/R acoustic according to simulations (2nd order electrical) , so I guess I should get my calibrated microphone out and double check that they really are 4th order L/R. Or would it be better to just reverse the polarity of the tweeter and check to see if the null matches my simulations at various listening heights? In my sims the biggest null is obviously at the listening height and it gets smaller above and below (as it should). If my measurement match this at the listening distance, then wouldn't that be a good way of checking? I know the room messes up measurement so I can measure outside if I can't get an accurate measurement indoors.

The design of the baffle and positioning of the speakers should account for time alignment, but as previously mentioned, dont think of time alignment as purely the physical distance of the driver voice coils or diaphragms relative to the listener, as that is not what is meant. It is a combination of physical and electrical properties ensuring that 1. drive units have the same phase at crossover and 2) they exhibit close phase throughout the frequency range that they handle. It's not an easy concept and further reading is almost mandatory to grasp the fundamentals. When you say "11mm out of alignment" do you mean in the horizontal "Z"plane (ie voicecoil centres not aligned) or horizontal "Y" plane (ie relative to the centreline of the speaker unit). If the latter, then you will have a time alignment issue and you should be able to work out the phase difference from the distance and speed of sound. It can be corrected with offset tweeters, partially within the crossover and also by using a sloping front baffle.

Be careful with measurement. When you speak of "nulls" at the listening height you need to be sure that you're referring to acoustic phase matching and not summed cancellations due to reflections or out of phase response.
Never used a sine sweep test tone for measurement as you will never be able to rid the measurement of room interaction and whilst measuring the latter can be helpful, it doesn't tell you if you've got the first bit right (ie the speakers). You can take the speakers outside and measure them with the mic at between 1.5 and 2m from the tweeter (I never use 1m as I consider it too close) or even at the listening distance which to me makes most sense. In fact if using a test sweep tone, it's the only way you can measure them but you do have to be quite a way from any reflective boundaries, or up a tree! Better is to use pink noise and impulse measurement. Room reflections then become far less of an issue since the measurement is made up via a series of micro-measurements of the pink noise, so having the speakers in the middle of the room is probably fine for this type of measurement.

I would be surprised to see an acoustic 4th order from a 2nd order filter slope except for bass roll-off in a reflex cabinet. When measuring for this, you need to measure one driver at a time with its respective filter in circuit. If a true acoustic 4th order, then it will roll off at 24dB/octave. If it doesn't manage this, then it's not 4th order. I don't know of any accurate electrical simulations which can determine acoustic slopes. The ONLY way to determine these is by measurement so my scepticism remains about a 2nd order filter resulting in a 4th order acoustic slope.

[fatmarley]

Sorry I meant the acoustic centres are 11mm out at 1 metre (I'm not very good with the terminology), but according to my sims everything seems ok - Phase tracking good, big null etc.

When I get a chance I'll measure the slopes and to be honest I'll be surprised If they're not 4th order acoustic, but being an honest person I'll happily admit if I'm wrong. Either way I'll upload some screen grabs because I'm curious myself.

[dimkasta]

For anyone wondering why electrolytics are worse in audio signal, another significant aspect is the dissipation factor, which is a measure of how much energy is converted to heat or otherwise absorbed by the capacitor.

Electrolytics tend to have big a dissipation factor, meaning some of the signal is lost in there (more pronounced in line level, and perhaps not so much in speaker level). On the other hand, caps like teflon or mica tend to have almost 0, making them very nice for signal levels ( although not very practical for crossovers due to size limitations)

About the significance of ESR in a crossover, even on first order design, adding an electrolytic in series with an 8ohm FR is an optimistic ~5% increase in impedance and double that for a 4ohm driver, so altering that can easily mean throwing SPL matching with woofers or tweeters out of the window. And this can easily add up in designs with L pads or notches and stuff...

I do not know if we are talking just about crossovers, but ESR can be very significant in PSU design as well. One of the biggest pitfalls being the use of low ESR caps at the output of series regulators, which leads 99% to oscillations. I love them for local decoupling though :)

[Firebottle]

Great posts Paul, it's good to learn a bit every day.
:thumbsup:

[fatmarley]

On a similar note, there's a guy on another forum who replaced all of the inductors in his crossovers (there's quite a few) with air-cores but he didn't match the dcr of the original inductors :doh: