Capacitor upgrade in crossover - Is it audible?

[audio2design]

Most of the capacitors in amps and receivers are not in the signal path and are there to pick off or shunt noise to ground. So when people replace all of them with "audio grade" they are mostly replacing non-signal path caps. It looks nice having snazzy colored caps though.

My theory is that most of these are just rebadged power supply capacitors. For large audio companies, they get the order cause it looks nice in marketing photos and for the smaller companies, they charge a premium for the different shrink wrap.

 

[audio2design]

I keep meaning to recreate the mechanical resonance test from the AES paper from a number of years back. It showed an electromechanical resonance in theory large enough to have the potential to be audible. Sloppy methodology on listening tests that indicate the potential for portions not blind, but also testing fatigue. Depending on the instantaneous impedances, changing capacitor types could have a subtle change on cross-over characteristics, but expect almost all reported changes from electrolytic to film are likely due to value tolerances. There was another AES paper not that long ago that did these same distortion tests and they were almost down to the instrument floor, however, if the electromechanical resonance is real, then the tests being applied are insufficient to claim lack of audibility. They may only show up with a tone specific IM test with one of the tones at the resonance. THD would not show it. Without investigating the resonance and investigating how to stimulate it and measure it for potentially audible effects, I don't think we can claim 100% certain inaudibility.

 

[pogo]

I don't think we can claim 100% certain inaudibility.

Here is a possible measurement method:

3D Step Response Measurement - The sound quality of the reproduction made visible

The 3D Step Response measurement shows the behaviours of the sound reproduction in the non swung-in state. This measurement is very important to judge the reproduction of impulses. As testsignale preferably half sinus oscillations of different frequencies at the same time are used, since acoustic events are essentially based on these modes of motion.

View attachment 83902



The frequency response is totally inadequate for the judgement of sound quality. The frequency response is measured with non-changing (Static) signals such as sinus waves or calculated noise signals. Against that, a music signal consists of changing (Dynamic) signals. The study of music signals shows that the sound quality of a musical instrument is determined by impulses. Especially the first impulse, the stroke of a guitar string, the impact of a piano cord, the hit on a drum and the blow of an organ pipe or brass instrument is important for the sound. In the Step-Response, all acoustic parameter information of the sound reproduction is contained. This consists of the frequency response, phase response and swing-in behaviour. There for the display is so complex, that the sound quality judgement based on the measurement is not possible. That is why the 3D Step-Response measurement was developed. By the measurement, the Step-Response is analysed and displayed in a 3D graph. The 3D measurement shows the impulse reproduction for each single frequency with use of an additional frequency axis. The measurement makes it for the first time possible to analyse the sound quality through measurement technology.

(Source: Kirchner elektronik)



Testing is not incorrect in what it measures, but in what it does not measure. You have to swim against the stream to reach its source

 

[audio2design]

But there is still the issue of audibility. A perfect step response sounds great acoustically, but does it matter psycho-acoustically when we process bands of frequencies with different processing algorithms (in the brain). Step response may matter from 200-1500Hz, but not be critical outside of this band, hence the 3D step may not illustrate actual perception.

 

[pogo]

That will be the case, but at least differences should be discernible.

 

[pogo]

These algorithms could also bring you closer to the truth:
Sound quality algorithms

 

[TrevC]

I have just replaced the crossover electrolytics in an old pair of B&W 601 with Panasonic polypropylene types and the speakers are transformed according to my ancient ears. Perhaps I should have increased the series resistor to compensate for the lower ESR, but what the heck!

 

[Grooved]

Six years ago I renovated my Yamaha M65 power amplifier and replaced almost all capacitors. I used audio-grade capacitors with low ESR and high temperature resistance (105°C).

After a few years the input and driver circuits were dead. A friend who did the repair said that the low ESR of the new high-grade capacitors was probably partly responsible for this - this is such a moment when you know that irony has shown you the finger.

Did the capacitors had the same values than the old ones, except the ESR ?

Just an idea that comes to mind regarding this test or one you would conduct to test ESR specifically : wouldn't the best way to test ESR be to replace each capacitor by two capacitors with half capacity in parallel ? This way, it would the same component, same value, except ESR being lower compared to only one capacitor.
I know that it can bring something in other contexts than audio, but not sure if it would change something here

 

[Wolf]

If you have 2 caps in parallel they add to double over the single unit's value.
ESR also does not work the way you speculate, as it is not like that of resistors at determining value.

 

[Grooved]

- If you have 2 caps in parallel they add to double over the single unit's value.
- ESR also does not work the way you speculate, as it is not like that of resistors at determining value.

I totally wrote by mixing everything, edited it but thanks for flagging this
- the half part was on the capacity side, I was thinking of two capacitors with half the capacity of the original to get the same value in the end
- for the ESR, it's not half value but it's lower with two capacitors in parallel than with only one, isn't it ? (I'm talking about all capacitors having the same ESR)
For example : 2 x (1800uF/18mOhms ESR) in parallel give a lower ESR value than one 3600uF/18mOhms ESR, right ?

 

[Wolf]

The measurement of ESR involves that of a known resistor, and can be quite complicated. I was told by someone with more knowledge than I have that ESR is not as simple as halving those in parallel ESR values. This is due to how capacitors work opposite of coils and resistors do for values.

I am not certain that 2 in parallel yield a lesser ESR over one unit.

 

[Watsonian]

A lot of Neve style microphone preamps are marketed because they feature "original" capacitors specified in the original Neve designs. Weirdly the argument for using capacitors (of a specific type) in Neve style preamps is that they provide harmonic distortion in the low end.

A couple of times I've nearly made a few purchases simply microphone preamp manufacturers advertised the fact that they use capacitor type x, y or z and they're expensive. Okay, that's two reasons.

Fortunately I settled for the middle ground and just bought an expensive microphone preamp because it was...expensive. And I liked the name.

Yes, I'm shallow and unscientific and probably vain. Still got the preamp, it's a keeper.

I did find the OP's argument a good read though - very interesting.

 

[Grooved]

The measurement of ESR involves that of a known resistor, and can be quite complicated. I was told by someone with more knowledge than I have that ESR is not as simple as halving those in parallel ESR values. This is due to how capacitors work opposite of coils and resistors do for values.

I am not certain that 2 in parallel yield a lesser ESR over one unit.

Forget about the half, I didn't wrote at first what I was thinking, and put the "half" part in the ESR sentence instead of the capacity. It's corrected now.
So sure, I didn't wanted to say that the ESR will be half, but that it would be lower.
Tested it in the past (long time ago so I forgot the details) and pretty sure it was lower, but it was done to optimize a power supply, especially from a battery, and not used on the signal path. Now, I may have a memory problem and it's possible that it improved something but that the ESR was not lower

 

[58bry]

I have changed Capacitors in a Couple of Sets of Kef speakers. Concertos born in June 1970 and Cadenza born August 1974. I unsoldered one leg of each elcap black and red capacitor from the PCB and measured it with an ESR meter. In both cases the smaller value caps 5yf and 7yf had wondered up to 10 to 15 yf. Whilst the larger 60 and 80 yf units were within 10% of their nominal value but the equivalent series resistance was way over the top of the maximum amount specified by their manufacturers. I simply purchased a complete like for like replacement set from Wilmslow audio in the UK. I also checked out the 30 + capacitors in Kef C80s born in 1986. Every original blue elcap capacitor was within specification of value and esr so I resoldered all and put them back....

 

[deano454]

Replacing 20 year old electrolytic should be mandatory, especially midrange and tweeters

 

[thewas]

This can't be generalised as there have been measured many that are over 40 years old and still in their tolerance spec, but if someone cannot measure it is safer to change them all.

 

[SmokinJoe]

If anyone wants to dispute this information or other components impact on the audio stream, they should do a double-blind test. If you can consistently discern the difference between the components, then you have something to consider. Otherwise, you're just fooling yourself.

 

[Jukka]

Nice post, I was afraid no IMD measurements, but you delivered Did you try non-linearity by signal level? Also I'd like to see measured how far electrolytes can drift after decades of use (not far fetched), how they compare against those film caps, new and aged. After all, elko is one of the fastest aging component in (audio) electronics.

 

[fpitas]

If anyone wants to dispute this information or other components impact on the audio stream, they should do a double-blind test. If you can consistently discern the difference between the components, then you have something to consider. Otherwise, you're just fooling yourself.

But they're having fun. Honestly if capacitors have a "sound", the logical thing to do these days is to throw them all away and go active.

 

[ctrl]

Nice post, I was afraid no IMD measurements, but you delivered Did you try non-linearity by signal level? Also I'd like to see measured how far electrolytes can drift after decades of use (not far fetched), how they compare against those film caps, new and aged. After all, elko is one of the fastest aging component in (audio) electronics.

Unfortunately, I won't be able to do that.
Actually, I've been wanting to do another capacitor comparison with audio samples for ages. If I find the time, this will probably be the conclusion of the series.