Capacitor upgrade in crossover - Is it audible?

[ctrl]

If you wonder why the phase of the measured capacitors (which should be at -90°) suddenly becomes "inductive" (phase at +90°) from a certain frequency on, this is due to my laziness.

The correct calibration of Arta-LIMP is time consuming and was not necessary for these measurements (up to 1kHz the capacities are correct) - only in the second part it was necessary to do the calibration exactly - there the measurements look like expected.

 

[lmaobrah]

So from my understanding, crossover affect the whole sound of the speaker a lot.
But that's purely from the crossover design, not the brand/material of the capacitor(s) used.

 

[solderdude]

Yes, and the tolerances of the used components as well.
4.7uF cap A may not have the same value as 4.7uF cap B, regardless of brand/type.

 

[restorer-john]

Replacing 20 year old caps is fairly silly. There is an enormous difference in the construction, chemistry and lifespan of a cap made in the 90s vice something of 60s/70s vintage.

OK. Right. Tell me more...

And then we'll do some tests on 10, 20, 30, 40 and 50 year old electrolytics. I have plenty here.

 

[pma]

Replacing electrolytic capacitors in the crossover, especially the old ones, by MKT caps like these attached makes sense, brings better and time stable parameters. And they are not too expensive.

 

[thewas]

It doesn't harm but benefit can't be generalised, I just replaced all caps on my 40 years old high quality Technics SB-10 and measured the old caps and all of them (mainly foils and few electrolytics) were in their nominal specs.
As an additional proof I measured also the FR response before and after and it did't really change:

Hope the new ones will also be as constant for the next 40 years, but somehow I have a doubt there...

 

[Killingbeans]

Could be because (potential) buyers kind of expect/demand this.
Could be the manufacturer believes in the powers of these parts.
Could be they do not want to be caught with their pants down if they were using cheap parts.
Could be sighted listening tests were performed as well.
If something sells better why not use it (certainly if you can mark the price up)
Looks well in advertisements ?

The one thing about Purifi Audio, that has had me puzzled for a long time, is how they insist on the use of woodoo crossover parts in their speaker demonstration designs. It directly clashes with the stated values on the front page of their web page

I can only imagine that they figured advocating both Class-D AND non-sexy crossover parts would scare away a dangerously large part of the potential clientel. I hope that's the reason

 

[pma]

It doesn't harm but benefit can't be generalised, I just replaced all caps on my 40 years old high quality Technics SB-10 and measured the old caps and all of them (mainly foils and few electrolytics) were in their nominal specs.

I often find not much change in capacitance of the old electrolytic caps, however ESR usually rises significantly. This is neither good for power amplifier filter capacitors nor for crossover capacitors. May have not much effect if higher series resistor is present in the circuit.

 

[FrantzM]

THANK YOU @ctrl !
I had to shout. .
Fate , had it that this threaded was revived and appeared on "New Posts".
I recently acquired although not yet in my house, a pair of Yamaha NS-1000M and was already scourging the Internet for crossover refresh and upgrades... some costing more than the speakers!!
That saved money will go toward a bit of room treatment and speaker cabinet refresh!

 

[D+S Tech Labs]

Thank you so much for doing the research and writing this article. I've been building speakers since the 70's and this has been an age old controversy among my "audiophile" customers.
If you're interested, my process for building crossovers is here.
--Bill Daniels

 

[Rick Sykora]

So the OPerative question here is capacitor audibility. Amir will soon test the premium crossover in the X-LS Encore and will have some more data, but the crossover was changed in other ways and so am considering a more controlled experiment...

Am building the Bagby Mandolin speaker and am just starting the crossover. The tweeter filter has only one capacitor and would be fairly easy to build with a switch to allow selection between the base kit capacitor and a “premium” one. As additional controls go, do not plan to tell Amir which position is base vs. premium.

So, the question is which premium cap to use? Unless there is a real desire to fund a something really pricey, am budgeting about $30 for a 4.7 uF cap. I have found a Clarity CSA, Mundorf Supreme, and Jantzen Superior that meet this criteria. The first 2 are upgrades via Meniscus Audio, but did not feel a need to limit to only theirs.

Any suggestions on the capacitor brand or other questions/considerations?

 

[R Swerdlow]

Any suggestions on the capacitor brand or other questions/considerations?

Make sure that both base and premium capacitors have the same capacitance, or are as close as possible to what the crossover schematic specifies.

Make sure that the test is done under blind conditions. Each listener must not know the identity of the crossover caps. It doesn't have to be double blind, but listeners must not communicate with or see the testers.

Perform negative and positive controls as described:

https://www.audiosciencereview.com/...dle-the-truth-part-3.12555/page-4#post-506233

 

[Rick Sykora]

Make sure that both base and premium capacitors have the same capacitance, or are as close as possible to what the crossover schematic specifies.

Make sure that the test is done under blind conditions. Each listener must not know the identity of the crossover caps. It doesn't have to be double blind, but listeners must not communicate with or see the testers.

Perform negative and positive controls as described:

https://www.audiosciencereview.com/...dle-the-truth-part-3.12555/page-4#post-506233

Yes, of course. Should have stated, I do not plan to test these myself (except for some design/build validation).

Currently, the target test listener is solely Amir.

 

[R Swerdlow]

Currently, the target test listener is solely Amir.

If only one listener is tested, it is critically important that at least a negative control be included. Because this listener also claims to be more sensitive to speaker distortion than most, positive controls would be highly informative.

The topic of designing blind listening tests has also recently come up on AH. Here are 3 links which flesh out my ideas of negative and positive controls.

https://forums.audioholics.com/foru...eaker-wire-does-it-exist.119346/#post-1418492

https://forums.audioholics.com/foru...wire-does-it-exist.119346/page-2#post-1418598

https://forums.audioholics.com/foru...wire-does-it-exist.119346/page-4#post-1418986

 

[DualTriode]

OP,

You persist in your assumption of minimal phase system. The assumption is incorrect.

If you swap out a capacitor with a significantly different ESR there will be a phase shift between the tweeter and woofer in a two way system.

With your assumption you do not measure the phase of the tweeter relative to the woofer at the crossover frequency. That is not even on your radar.

DT

 

[Rick Sykora]

OP,

You persist in your assumption of minimal phase system. The assumption is incorrect.

If you swap out a capacitor with a significantly different ESR there will be a phase shift between the tweeter and woofer in a two way system.

With your assumption you do not measure the phase of the tweeter relative to the woofer at the crossover frequency. That is not even on your radar.

DT

I had considered this as I seek out a premium cap to test, but finding published specs/data sheets on these caps is not like other components.

In a real world application, how high does ESR get? For polarized caps, ESR is measured in milliohms...

 

[DualTriode]

In the real world of crossovers an old non-polar electrolytic capacitor could easily have an ESR of 0.500R, working with a 4Ohm tweeter that is significant.

Perhaps the OP should measure some capacitors on his network analyzer and show us the phase and impedance differences.

DT

 

[ctrl]

If you swap out a capacitor with a significantly different ESR there will be a phase shift between the tweeter and woofer in a two way system.

In a real world application, how high does ESR get? For polarized caps, ESR is measured in milliohms...

Perhaps the OP should measure some capacitors on his network analyzer and show us the phase and impedance differences.

In the second part of my mini series you can find comparisons of electrolytic capacitors with film capacitors.

There an electrolytic capacitor was compared with a film capacitor with the capacitance of 8.2µF - see section a). The resistive part of the electrolytic capacitor @1kHz was four times as high as that of the film capacitor.
This is several times higher than replacing a "normal" film capacitor with a high-end capacitor.

How much did this affect the phase frequency response of an tweeter? (The capacitor in series with the tweeter)

At the position of the largest deviation, it was an "outrageous" 0.1 degree - which is actually nothing.

To emphasize it again, the biggest problem in these comparisons is the capacitance deviations of the capacitors among themselves.
It is not enough to buy two capacitors for a comparison, you have to make sure that the capacitance values really match - not on paper, but with the most accurate measurements possible.

If there is then a significant deviation in the measurements, it is really due to differences in the quality of the capacitors.

 

[Wolf]

First post here- I want to say I really appreciate the no-nonsense of this forum, and I feel I have information to offer on this subject. I've done a lot of experimenting and swapping of parts, even dialing in that of component values to feel I have a decent grasp of the subject at hand. I've even conducted some testing of my own on groups for nothing more than something someone could possibly take personal notes on.

While I do not see anything that would favor a result that is distinguishable via measurement conditions posted above, I have experienced easily discernible results that close peers have also stated without knowing the conditions of the same trial. The test chosen here, I would think could possibly be audible, and yet not measurable- even though the measurements don't show it. I've also done measurements that did not show a variance in frequency response, and yet perception was an audible result by several.

Let's examine the caps here- all are of at least a film variety, and likely low-ish ESR. So at least they are reputable components. However- all of these are of the metallized variety and not a differing construction. Sometimes this can matter, and sometimes it doesn't.

Instances where I found things most notable:
- In my Nephila speaker project, I had a 4.7uF cap in the AMT tweeter circuit. This Airborne RT5002 is a low HD AMT, and resolves this difference easily. It had a 4.7 no-name poly cap of unknown construction though likely polypropylene, and a North Creek Zen-Cap (standard metallized poly) as the opposition. In the application of an SPDT switch, the Zen had massive soundstage, while the no-name was a bit more focused. This was the most easily discernible cap-swap I've entertained.
More than one instance of this swap was done, and many told me the same thing I experienced.

- In my Cecropia Supreme project, I had to use a funky L-pad and contour network I call a cross-pad or Xpad. It involved 1st order xover sectors with separate attenuation on either leg to fix an up-shelved response resulting from a greater than 180* baffle angle, or somewhat of an anti-waveguide. Initially I had a 15uF Eton Cap in the low-leg, and I swapped in an AuriCap for a much better result than I anticipated. Neither of these are cheap caps, but that is immaterial. Both are that of metallized poly construction and likely low ESR. The lower-treble came through a lot more open than it had been prior.

- In my Attitudes project, which most of my initial playing around with capacitors of different flavors was done, I had a nonpolar electrolytic (NPE) adjacent to a Solen FC and a Jantzen CrossCap across my woofer circuit. The NPE was clearly inferior and choked off the sound. The other 2 were closer, but the Solen sounded like more detail came through the woofer. I will not use a Solen in series with a mid or tweeter as they can sound gritty, but across the lower range they seem to be well placed. I wonder if this is not a function of the 'dirty' sound that they portray in series, and actually shunt the dirty crud down the chain when going to ground. This was a standard tanked and damped second order circuit.

In the same project, I had a couple 20uF caps; an ASC X386 poly-in-oil motor-run style capacitor, adjacent to an (older now) Clarity SA capacitor. These are out in front of a 3rd and 3rd order midrange bandpass, electrically speaking. This was more similar to the Nephila results in terms of what I heard, but not as night and day. The Clarity was the less spacious soundstage unit.
Statistical analysis said nothing was more notable than a chance situ in the results of my group testing, but I did make converts of doubters as well as solidifying skeptics' viewpoints.

- In my Glucinium Amor, I had a tweeter circuit of standard second order electrical. The padding setup was split with one across the tweeter, and the other out front. I had a 0.1uF Mundorf Supreme cap across the front resistor to lift the top-end relative to the rest of the response. As I wanted to experiment with the top-end a smidge, I paralleled a Cornell-Dubilier 0.22uF cap directly across the Mundorf to see if lifting further was better for my tastes. The result I could not measure, no matter how hard I tried, but it smeared the upper treble to where I was hearing echos of the words sung in the music where there should not have been any.

- I took the cap-test performed by Bob Cordell back in 2004 (I was 26) utilizing a cascaded 1st order 3-way bookshelf called the M3, consisting of a SoniCap polypropylene and industrial grade GE mylar or similar opponent. This test took a half hour, and was an ABX type. Upon conclusion, I only guessed 2 correctly, but that means I had 8 wrong, or 80% just guessing a cue incorrectly. There was a preference slot also, and I reportedly preferred the SoniCap when I thought I was picking the mylar. The person next to me had exactly opposite answers (no cheating was involved),and picked the mylar, much to his chagrin. I still believe I was able to discern a lot in that test, learned a lot from it, but perspective and my guesses/answers were at odds with each other.

If I can be of any help and bring some closure to this long debated subject, I am a willing participant.
Later,
Wolf

 

[R Swerdlow]

- I took the cap-test performed by Bob Cordell back in 2004 (I was 26) utilizing a cascaded 1st order 3-way bookshelf called the M3, consisting of a SoniCap polypropylene and industrial grade GE mylar or similar opponent. This test took a half hour, and was an ABX type. Upon conclusion, I only guessed 2 correctly, but that means I had 8 wrong, or 80% just guessing a cue incorrectly. There was a preference slot also, and I reportedly preferred the SoniCap when I thought I was picking the mylar. The person next to me had exactly opposite answers (no cheating was involved),and picked the mylar, much to his chagrin. I still believe I was able to discern a lot in that test, learned a lot from it, but perspective and my guesses/answers were at odds with each other.

I was at that cap test too, in Fairfax, VA in the fall of 2004. Bob Cordell, Dennis Murphy, and a third guy whose name I don't remember had organized it.

It was primarily a test to see if listeners could hear differences in speakers with different capacitors in the crossovers. There were about 40 listeners there, most of them DIY speaker builders with widely varying amounts of experience. Most listeners did about 10-12 repetitions of the test.

In all there were over 400 repetitions of the test – and the overall result was that listeners were able to correctly hear differences between capacitors 50% of the time. Of course, they were also wrong 50% of the time. This was no different than if they had randomly guessed. That result was unmistakably clear. But that didn't prevent a big debate from breaking out. It continued for many days online.

I later spoke with the organizers about running a Negative Control along with the listening test, where listeners listened to the same speaker with the same capacitors in them. Think of a Negative Control as a measure of how many 'false positive' answers there were. You cannot assume people can always tell when capacitors were identical – you have to measure how many false positive answers there were for each listener.

At the time, they all replied that running negative controls would have cut down on the overall numbers of listening tests they could run. Later, after those excessive arguments, they admitted they wished they had done some negative controls, as it would have silenced a lot of the debate. The take-home lesson is that listening tests are no better than the experimental controls done at the same time.