• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Capacitor upgrade in crossover - Is it audible? - Part 2

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
Rated ripple current of bipolar 22uF/63V is only 95mA.

https://www.tme.eu/Document/68c09c71f2e1f7e96ea173ff2581ee56/e-uvp.pdf

In my test the capacitor current was about 1A for 9.5V test and proportionally less for 2.8V test, still about 3x the rated current. It is a wrong part at the wrong place, even if the bipolar electrolytic capacitor is used. Incompetent choice.
 

77SunsetStrip

Member
Joined
Jan 7, 2021
Messages
48
Likes
20
Incompetent choice is accurate in some cases. Most of the time electrolytic caps are used due to space constraints and cost. Engineers almost always loose when bean counters are involved. Andrew Jones and others have confirmed that situation. Never assume every loudspeaker at any price point from any manufacturer is the product of engineering at its finest. More accurately, good enough within an enforced budget.

Film caps are always larger than an equivalent value electrolytic, so limited space can force the use of electrolytic. Decent quality 100-200uF caps can be found that are not unreasonable in cost. However, each individual can make their own choice to spend a little or spend a lot. It is perfectly valid to replace electrolytic caps with film for no other reason than to assure long term stability.
 

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
I have made further measurements with improved setup to get rid of mains spuriae. Now the same MKT 22uF capacitor was tested against bipolar electrolytic capacitor 22uF. Input voltage was 10Vrms and load resistance was 6.8 ohm as before. Voltage across the resistor was measured again.

Bipolar capacitor performed better than the single electrolytic capacitor. However the stress and rise of all harmonics from 3rd and above was observed. No wonder because the bipolar capacitor is operated out of specs re maximum ac current which is only about 100mA.

On the other hand, AC permissible voltage of the 22uF MKT capacitor is about 20V at 1kHz.

1624308727179.png


Distortion with the MKT capacitor followed the test setup distortion. Below is the distortion with MKT which follows the setup distortion.
10Vrms_MKT22uF+6R8.png



In the next plot is the distortion with bipolar electrolytic capacitor
10Vrms_bipolar22uF+6R8.png

One can see rise of higher order harmonics starting with 3rd. If we forget the H2 which is a product of the setup (amp + card) all the distortion components starting from H3 are created by the bipolar electrolytic capacitor and indicate that the permissible ac current is exceeded and the component works out of specs.

Conclusion
Bipolar electrolytic capacitors do work in speaker crossovers but the ac current exceeds the bipolar capacitor rated current. This result in higher distortion and shorted lifetime of the capacitor. Bipolar capacitors make a low-cost solution at small dimensions but they are wrong engineering solution and should not be used in speaker crossovers.
 
OP
C

ctrl

Major Contributor
Forum Donor
Joined
Jan 24, 2020
Messages
1,616
Likes
6,087
Location
.de, DE, DEU
@pma Thanks for your more detailed measurements, that's a great addition!

Now the same MKT 22uF capacitor was tested against bipolar electrolytic capacitor 22uF.

The tested bipolar electrolytic capacitor has only a "dielectric strength" of 63V DC.

In crossovers, perhaps apart from mini loudspeakers, only electrolytic capacitors with 100V DC should be used.
This should lower the harmonic distortions with 10Vrms@1kHz again significantly.


all the distortion components starting from H3 are created by the bipolar electrolytic capacitor and indicate that the permissible ac current is exceeded and the component works out of specs.
Your observation is certainly correct. But here, too, context is important.
At 10Vrms, the HD3 of the bipolar electrolytic capacitor is now well below 0.01%, with only 63V DC "dielectric strength". With a 100V DC capacitor well below that.

If we take the manufacturer's data from what is arguably one of the best bass-midrange drivers currently available, the Purifi PTT6.5 (at least in terms of harmonic distortion), HD3 is in the 200-4000Hz frequency range, on average, around the 0.01-0.04% range at around [email protected].

Again, the 10V measurement only makes sense to find out if the capacitor still works reasonably at high sound pressure levels (we are in the range of 97-104dB@1m SPL for 5-6'' drivers at 10V), which it does (compared to the distortion a driver causes).

At 100dB@1m, even the best driver is many times above 0.01% HD3 - except perhaps for larger horn speakers.

In terms of the possible perceptibility of harmonic distortion, one needs to do the measurements at 1-2V to reach 77-84dB SPL@1m, where the human ear is much more sensitive to harmonic distortion (since the ear itself produces little distortion then).
 

capslock

Senior Member
Joined
Jun 19, 2020
Messages
316
Likes
145
I was wondering if bipolar electrolytics have a crossover region similar to a class B amplifier. I seem to remember there is a voltage in the single digit mV range associated with the electrochemical cell formed by the electrolytic and the electrodes, and you would get that in either direction with a bipolar. In that case, one should be looking at signals of about 100 mV.
 

bjmsam

Member
Joined
Feb 1, 2022
Messages
41
Likes
21
Location
Mount Airy, MD
Conclusion
Bipolar electrolytic capacitors do work in speaker crossovers but the ac current exceeds the bipolar capacitor rated current. This result in higher distortion and shorted lifetime of the capacitor. Bipolar capacitors make a low-cost solution at small dimensions but they are wrong engineering solution and should not be used in speaker crossovers.
Even as shunts (vs. in series with the driver)?

Are NPEs not suitable for these four caps in the example below?
  • 8uF 100V
  • 40uF 50V
  • 30uF 50V
  • 470uF 50V

AM-JKLWTDuDM7TTxudeIikM2FM9PBSN24QyOPRy3cACThqf_ltoPnkylF0eYHB9OwSfzkkTNj0g79QHYNvmcpuHTGaxzkfq-dGS7MstCvDonffTlEAqfm8UY_6Lrs5enjHWuse-W6P92u_u5DNX3rAUX14ipJg=w901-h1107-no
 
Last edited:

TurtlePaul

Addicted to Fun and Learning
Joined
May 19, 2020
Messages
631
Likes
1,027
Location
New York
You need to be able to read PMA's charts to understand his opinion.

Yes, electrolytic capacitors are a design compromise. But looking at the distortion (<0.1%) vs. the distortion of real-world speaker drivers at these levels (usually >0.5-1.0%), you realize that at certain price points, of course bipolar high-voltage electrolytic make all the sense in the world as the impact on total distortion is measurable but very small. Audibility of distortion below 1% is already very difficult for music material in listening tests.

The level of added distortion of electrolytic caps is very low relative to the hysteresis caused distortion in iron-core inductors but even those make a ton of sense at certain frequencies and certain price points.

But yes, if you are buying $20k speakers, a good marketing point for them would be to not use any electrolytic caps. At $200 speakers, of course it will be all electrolytic. For most speakers in between you see one electrolytic cap and one iron-core inductor for the two components where large values cannot be affordably achieved and all film caps and air-cores elsewhere.

Shunt vs. in-series doesn't have much impact as they are all effectively in the signal path. However, zobel networks can probably use lower quality components as they are truly outside the signal path.

edit: also, if you are recapping an old speaker, please be aware that the crossover designer may be expecting some series resistance to exist accross some of the caps and there could be a negative impact if the replacement caps have no ESR even if they are better caps.
 

bjmsam

Member
Joined
Feb 1, 2022
Messages
41
Likes
21
Location
Mount Airy, MD
I just finished recapping my Acoustic Research AR-3a speakers with M.D.L. NPEs purchased from PE for about $10 (project details are posted on CSP). Any difference in sound is too subtle for me to hear. Below is FR measured in my HT before / after recap using a calibrated UMIK-1 on axis aimed between tweeter and midrange at 0.5m and 1.0m from the same position (the right stand). One of the 54-year-old level potentiometers worked only near its maximum setting so I set the others the same for consistency, resulting in boosted mids and highs. The cloth woofer surrounds have yet to be resealed, resulting in attenuated lows. I did what I could in my non-laboratory setting to make capacitors the only significant variable. FWIW, I am happy to share the raw measurements with anyone wants to do more with the data (distortion, etc.) but a ZIP is too large to attach (~20MB).

AM-JKLU8JbDrGY3qy1Tr2Xn14WKLLE_mD-am9B0KeUKeJSuJUpwV8k0IALJcdk9E-sUiA6J4ClORlWg514V9ca_KFBCa56tIyrpa4Go7_bdqA_FLB74RuImveLjg-mYhTNPoidDFbZxLN1ZkQj0Kx-_JWg-gkg=w1413-h1060-no


AM-JKLUR2P8QfORUX25rsJ_-ix4CwLnBH5vabw7U7dykwIFvVLNp-O3BM2vwEmpRt6iDfbHekyi946fLQfATR9Y1eEW59KOtgV5H4dGN37U3F-QqW43zuSK1J2LFOMcCATBR8UMc251IgxPWQVSJjwogUM3KYw=w1024-h607-no
 
  • Like
Reactions: JRS

JRS

Major Contributor
Joined
Sep 22, 2021
Messages
1,158
Likes
1,003
Location
Albuquerque, NM USA
Normally, I am a numbers guy--the blown up graphs showing division tics of 0.001 corresponding to 0.1dB--what is the original scale in? If it is truly normalized, I don't understand the math. I get deviations more of 0.015dB, an order of magnitude difference. Could you show me how you arrived at these differences?
 
OP
C

ctrl

Major Contributor
Forum Donor
Joined
Jan 24, 2020
Messages
1,616
Likes
6,087
Location
.de, DE, DEU
Normally, I am a numbers guy--the blown up graphs showing division tics of 0.001 corresponding to 0.1dB--what is the original scale in?
Thanks for the hint! You are right and I have found the error why VCAS has problems with the scaling in the diagrams, in the measurement files the first line consists of a line with sound pressure level, frequency and phase with zero - VACS does not cope with this.

It should not be normalized to 1, but to 0.

I'll have to go through all the diagrams in the posts from part 1-3 of the series.

Will report again when the diagrams are updated.
 

JRS

Major Contributor
Joined
Sep 22, 2021
Messages
1,158
Likes
1,003
Location
Albuquerque, NM USA
Thanks for the hint! You are right and I have found the error why VCAS has problems with the scaling in the diagrams, in the measurement files the first line consists of a line with sound pressure level, frequency and phase with zero - VACS does not cope with this.

It should not be normalized to 1, but to 0.

I'll have to go through all the diagrams in the posts from part 1-3 of the series.

Will report again when the diagrams are updated.
It's a great series, and I learned a bunch. Thx.

PS: Number of dB is as a function of the ratio of Vo/Vi. So normalizing is fine. The dB of say 0.99 ratio is -0.087dB of 0.999 is 0.0087, etc.
 

JRS

Major Contributor
Joined
Sep 22, 2021
Messages
1,158
Likes
1,003
Location
Albuquerque, NM USA
I just finished recapping my Acoustic Research AR-3a speakers with M.D.L. NPEs purchased from PE for about $10 (project details are posted on CSP). Any difference in sound is too subtle for me to hear. Below is FR measured in my HT before / after recap using a calibrated UMIK-1 on axis aimed between tweeter and midrange at 0.5m and 1.0m from the same position (the right stand). One of the 54-year-old level potentiometers worked only near its maximum setting so I set the others the same for consistency, resulting in boosted mids and highs. The cloth woofer surrounds have yet to be resealed, resulting in attenuated lows. I did what I could in my non-laboratory setting to make capacitors the only significant variable. FWIW, I am happy to share the raw measurements with anyone wants to do more with the data (distortion, etc.) but a ZIP is too large to attach (~20MB)
Oh my lord, it's like opening a casket and exhuming the dead. Used to have a pair of AR5's so all looks familiar. I don't know that the FR is faithful to the original--been a while but I was able to find published data at the Boston Audio Society archives, and I don't recall such a huge hole in the midrange (nothing to do with the xover, maybe the midrange drivers need to be replaced?)
 
F

freemansteve

Guest
Help me out here.
Is anyone claiming that replacing a cap with one that has the same measurable RLC values makes a different to the crossovers?

I certainly believe that replacing a cap with one that has measurable values that are different to the one fitted, but better suited to the original intention of crossover design will change the sound. It may even be for the better.

I also know that caps age (as does silicon too!) and that the RLC values of the cap will change over time, and with temperature, and that some cap types and some manufacturing processes are more stable. Quality parts should last longer. You may need to upgrade your kit over time.

Is there something else I've missed?
 

bjmsam

Member
Joined
Feb 1, 2022
Messages
41
Likes
21
Location
Mount Airy, MD
Oh my lord, it's like opening a casket and exhuming the dead. Used to have a pair of AR5's so all looks familiar. I don't know that the FR is faithful to the original--been a while but I was able to find published data at the Boston Audio Society archives, and I don't recall such a huge hole in the midrange (nothing to do with the xover, maybe the midrange drivers need to be replaced?)
The thick traces represent my room measurements with level pots in less than ideal positions as explained. The thin traces represent AR's lab measurements. I do not suspect a driver problem; now that the pots are fixed, and once the woofer surrounds are sealed, I should be able to dial them in to better align the top and bottom ends with the AES trace (the middle already is close).
 
Last edited:

JRS

Major Contributor
Joined
Sep 22, 2021
Messages
1,158
Likes
1,003
Location
Albuquerque, NM USA
The thick traces represent my room measurements with level pots in less than ideal positions as explained. The thin traces represent AR's lab measurements. I do not suspect a driver problem; now that the pots are fixed, and once the woofer surrounds are sealed, I should be able to dial them in to align with the AES trace.
Oh OK, I just looked at the pics :oops:. If you want to keep it on the cheap, I'd consider replacing the woofers with something inexpensive. Here is a potential tip:

A vastly improved woofer that will perform as well or better in every respect assuming you are trying to clone an AR-3A would be the Peerless NE315W-08. (Paper cone version) In a stuffed 1.4 cu. ft. box you can get and FB and F3 around 40 Hz assuming .5 ohms or so of RG to cover typical passive cross over series resistances. QTC is in the .76- .8 area which is typical of what the AR3 woofer actually did. Clean out to 1.8K so easier crossover than the original. Built about 50 pairs of the original AR3A back in the early 70's when Heath Kit offered a kit version of the AR3A. Worked in a Heath-Kit store in Houston and built them for customers who did not, or could not build them. J




Boden

Member​

2010-03-02 9:29 am

 

AmadeusMozart

Member
Joined
Mar 30, 2020
Messages
14
Likes
9

Ageve

Active Member
Joined
Jun 13, 2021
Messages
192
Likes
640
Location
Sweden
Bipolar capacitor performed better than the single electrolytic capacitor. However the stress and rise of all harmonics from 3rd and above was observed. No wonder because the bipolar capacitor is operated out of specs re maximum ac current which is only about 100mA.

Conclusion
Bipolar electrolytic capacitors do work in speaker crossovers but the ac current exceeds the bipolar capacitor rated current. This result in higher distortion and shorted lifetime of the capacitor. Bipolar capacitors make a low-cost solution at small dimensions but they are wrong engineering solution and should not be used in speaker crossovers.

I don't know if I'm missing something, but the inexpensive bipolar Bennic 22uF 100WV cap (very common in crossovers), has a maximum ripple current of 454mA (axial) or 531mA (radial).

The larger 47uF version, has a maximum ripple current of 803mA (axial) or 903mA (radial).


You only managed to generate <0.01% distortion, with a high 10Vrms input, and a much "weaker" capacitor. I would say that's a pretty good result for the electrolytic cap.

Also, I'm not sure what you mean regarding "shorter lifetime of the capacitor". I've measured 25 year old Bennic caps, and they were still within spec. The Visaton caps measured by ctrl in this thead, were almost 40 years old.
 
Last edited:
Top Bottom