Non-Polar Electrolytic vs. Polypropylene Film Capacitors

gtforme00

Member
2004-10-21 4:35 am
This isn't what you think! While there certainly are subjective preferences on these parts, I would like to present to this community the results of some tests I have performed today. The parts tested are all currently available, value oriented parts from Madisound, Parts-Express, or MCM.

-Measurements were made using the HP 4192A Impedance Analyzer (Calibrated this year).Three measurement frequencies were used: 100Hz, 1kHz, 10kHz.
-Signal voltage was 1VRMS.
-The series equivalent circuit model was used for calculations.
-Measurements presented are the capacitance, equivalent series resistance, and dissipation factor respectively.


  • 3.3uF Panasonic Polypropylene Film 200V (Madisound surplus)
    0.10 kHz 3.313 uF <0.1 Ohm <0.01% DF
    1.00 kHz 3.306 uF <0.01 Ohm <0.01% DF
    10.0 kHz 3.31 uF 0.005 Ohm 0.1% DF
  • 3.3uF Bennic XPP 250VDC (Which are on sale this month at MCM)
    0.10 kHz 3.313 uF <0.1 Ohm <0.01% DF
    1.00 kHz 3.306 uF <0.01 Ohm <0.01% DF
    10.0 kHz 3.28 uF 0.002 Ohm <0.1% DF
  • 4.7uF Non-Polar Electrolytic 100V (From Parts-Express)
    0.10 kHz 4.989 uF <5 Ohm 1.45% DF
    1.00 kHz 4.826 uF 0.93 Ohm 2.8% DF
    10.0 kHz 4.7 uF 0.264 Ohm 7.8% DF
  • 5.0uF GE Polypropylene 400VDC (Surplus from Madisound)
    0.10 kHz 4.93 uF 0.001 Ohm 0.1% DF
    1.00 kHz 4.92 uF <0.01 Ohm <0.01% DF
    10.0 kHz 4.93 uF <0.01 Ohm <0.01% DF
  • 10uF Madisound Surplus 100VDC
    0.10 kHz 10.1 uF <0.01 Ohm <0.01% DF
    1.00 kHz 10.07 uF <0.01 Ohm 0.04% DF
    10.0 kHz 10.1 uF 0.003 Ohm 0.2% DF
  • 10uF Dayton MPT 250VDC (From Parts-Express)
    0.10 kHz 10.005 uF <0.1 Ohm <0.01% DF
    1.00 kHz 9.98 uF <0.01 Ohm 0.03% DF
    10.0 kHz 9.99 uF 0.001 Ohm 0.1% DF
  • 12uF Bennic XPP 250VDC (Which are on sale this month at MCM)
    0.10 kHz 12.05 uF <0.1 Ohm <0.01% DF
    1.00 kHz 12.03 uF <0.01 Ohm <0.01% DF
    10.0 kHz 12.04 uF 0.001 Ohm <0.1% DF
  • 12uF Bennic Non-Polar Electrolytic (From Madisound)
    0.10 kHz 12.675 uF 3.45 Ohm 2.7% DF
    1.00 kHz 12.12 uF 0.40 Ohm 3.0% DF
    10.0 kHz 11.74 uF 0.073 Ohm 5.4% DF
  • 31uF Bennic Non-Polar Electrolytic (From Madisound)
    0.10 kHz 32.3 uF 1.60 Ohm 3.2% DF
    1.00 kHz 30.5 uF 0.241 Ohm 4.6% DF
    10.0 kHz 29 uF 0.057 Ohm 10% DF
  • 33uF Non-Polar Electrolytic (From Parts-Express)
    0.10 kHz 33.9 uF 1.78 Ohm 3.8% DF
    1.00 kHz 32.1 uF 0.170 Ohm 5.4% DF
    10.0 kHz 31 uF 0.047 Ohm 9% DF

Impressions.
Non-polar electrolytics are pretty lossy compared to polypropylene. I have always heard that, but never quantified it. Notice how the capacitance drops with frequency and the dissipation factor rises in the electrolytic capacitors.

Electrolytics are not very linear with frequency, we see a 10% change from 100Hz to 10000Hz! (this does NOT imply they cause non-linear distortion due to this fact, please don't misunderstand). The polypropylene capacitors by comparison are fairly linear in capacitance with frequency and most have dissipation factors too low to reliably measure.

Notice the high series resistance of the electrolytics compared to polypropylene. This does not seem to me to be a negligible amount of resistance in crossover design, after all, don't we model inductors with this much series resistance? I do not have experience with crossover design software beyond Speaker Workshop, and do not know if any of them can model the non-linear resistance of an electrolytic capacitor. It would seem to me that this would be particularly important for midbass drivers in a 3 or more way. The Qts of the driver-enclosure system will be affected by this series resistance.

Anyone else care to comment on this?

Regards,
David
 
I've never been a fan of non-polar electrolytics, in fact I think they're often inferior to standard electrolytics. As a general rule, polypropylene caps are similar to polystyrene, with very low DF. Teflon is better, but impractical in any value I'm interested in. Mylar (polyester) is worse, and many people can't stand them. It's easy to measure the differences in DF, but with a good bridge, not an impedance analyzer. Still, I find the wide range curves you can generate with the impedance analyzer more useful. Personally, I don't find the differences that great, though I prefer polypropylene. I avoid electrolytics of any type in the signal path, with the exception of power amp feedback circuits, just because most other types are impractically large. I use active crossovers, so there's no need for electrolytics there! I've yet to see convincing tests that show high end caps to be any better than similar type and value of common inexpensive caps.
 

gtforme00

Member
2004-10-21 4:35 am
Conrad,

Thanks for the response! All of this came about as a result of a current project to clone the NHT M3.3. Jack Hidley graciously provided me with the schematics and the drivers are available here. This is by far the most complicated crossover I've built, and I want to know exactly what I am getting with the components I am buying. I wonder if NHT made allowance for the series resistance of the electrolytics used in the midbass and midrange crossovers?

For my self-designed projects, anything over 2-way has received active crossovers. I whole heartedly agree that they are superior to passive crossovers! With the current prices of passive components, they are a whole lot less expensive as well!
 
I'm fortunate to have periodic access to an Agilent E4980A. That has capabilities out to 2MHz, and I did all the spreadsheet measurements using that. At home I use a combination of the GR1615 standards bridge, which excels at accuracy and low DF, the GR716C, which can get to near 1MHz, but none of my antiques can measure large values at high frequencies like the $16K Agilent.
 

gtforme00

Member
2004-10-21 4:35 am
You are quite welcome. This information was gathered with a personal interest, but I considered it valuable to share as I have not found it available in condensed form, especially for "everyman" style capacitors.

I have noticed an error in my original post, and have not the ability to change it. The Madisound 10uF surplus capacitor should read 10.1 uF at 10kHz, not 100.7uF.

Regards,
David
 

gtforme00

Member
2004-10-21 4:35 am
Would a moderator please make the above correction?

An update to my inquiry on capacitor ESR and crossover design.

First, I was a bit redundant in specifying both ESR and DF, as they are inextricably linked for a given frequency. DF = 2*PI*F*ESR*C*(100 for percent). However, it is nice to see the ESR to get a grasp of the resistance each capacitor represents.

Second, in private conversations with a loudspeaker engineer (The above mentioned Jack Hidley from NHT) he confirmed that an experienced designer will take into account the ESR of both capacitors and inductors. Furthermore he stated that the ESR was taken into account in the design I am cloning (NHT M3.3) and that I should try to match the Dissipation Factors given in the specifications. This was good for me to discover, as I had purchased at least one film capacitor to replace an electrolytic, a substitution which I will no longer be using.

Now, it is very common to hear discussion of Inductor ESR, but this is the first confirmation I have seen that capacitor ESR should be considered when designing loudspeakers. This does NOT automatically invalidate all designs which have not taken this into account. The differences among the ESRs of various types of film capacitors are fairly small, however there is a significant difference between the ESR of film type capacitors and that of electrolytics. The practice of swapping electrolytics for film capacitors to "improve" a design is immediately called into question.

There is no paradigm breaking news here, and to some this may even be old news. This is just one more small block of knowledge to add to the vast amount of information available on this forum.

So what practical lessons should we take away from this experiment? If one is designing a crossover with electrolytics, one should be aware that the ESR of the capacitor will have a small effect on the drivers actual response, and that some software is not capable of modeling this (in my experience Speaker Workshop). Similarly, if an existing design has an electrolytic, be aware that there will be a deviation from the design if a film type capacitor is substituted. Whether this deviation will be perceived subjectively as an improvement is outside the scope of this discussion.

I hope that this discussion has been as informative for the readers as it has been for the author.

Regards,
David

P.S. Next up, can I measure "voltage coefficient of capacitance", and how do different types of capacitors compare?
 
David, yes you can measure the voltage coefficient, and the various types are definitely different, with some ceramics being pretty easy to see. This is one reason I like my antique bridges, as they can go up to several hundred volts, unlike modern solid state bridges that are doing well to get to a couple tens of volts.

You should also measure dielectric absorption. It's sort of a PITA, with accurate timing required. I bought a computer interface with this exact thing in mind, but haven't had time to set up the experiment.

Regards,
Conrad
 
originally posted by Conrad Hoffman
You should also measure dielectric absorption. It's sort of a PITA, with accurate timing required. I bought a computer interface with this exact thing in mind, but haven't had time to set up the experiment.


In this article by Bob Pease are described two test fixtures to measure DA:

Understand Capacitor Soakage

A simple one with a switch which has to be engaged manually and more a much more sophisticated with timers etc.
 
poobah, the carpet is long gone, the walls are torn out, everything is a mess, but I cleaned the seaweed and dead fish off the test bench, replaced the power strips that were under water, and at least I'm up and running! Fortunately I didn't lose much, though I did spend some time washing the components that were in the lower drawers of my parts cabinets. I also discovered you can recover some amazing things, if you act quickly. I opened up a nice dial indicator that was completely submerged, cleaned and relubed it, and it works better than new. I also have some mylar caps that were under water for a week, so I ought to check the DA on 'em! No, wait, this is 2008- just call them minty and sell them on eBay!
 
Thank-you and kudos on presenting some real data. Your efforts are to be applauded. As we all know, from data comes knowledge, from knowledge, wisdom.

The esr of capacitors isn't so critical because their reactance is so high at low frequency. For instance 3.3uF at 100Hz has a reactance of 482 ohms so an ohm of ESR doesn't really make much difference. Even at 10KHz, the reactance is 4.8ohms. Not to say it's not important, it needs to be taken in context and of course, it's the opposite for inductors, ESR is the essence of inductance.

As far as I know, bipolar electrolytics are formed by making a pair of double value electrolytics in series and so you'd expect their ESR and DF to be 2X worse than regular electrolytics. Let's face it electrolytics were created to do one thing well, dump lots of low frequency energy. They are not intended for audio duty.

When you give a value of <0.1ohm for instance, what do you mean? Is 0.1ohm the limit of measurement for that test condition?

I vote to make this a sticky that we should all put our data. I, for one, would really like to know if a particular manufacturer knows what they're doing!