Author: Paul Hunting - TNT UK
Published: June, 2023
The job of a series crossover capacitor on the tweeter is to allow only the high frequencies to pass. The frequency at which this occurs is a user determined capacitor value that is also set by the tweeter resistance and the crossover order type. The capacitor values can be made up either by a single capacitor or paralleling a number of smaller capacitors with values that when added together make up roughly the same value as the single capacitor. Parallelling is said to improve capacitor performance.
Here we compare sound between just using a single capacitor, a two capacitor bundle of the same brand and a three capacitor mix where one capacitor was of a different brand.
TNT-Audio has previously looked at single capacitor solutions and recapping a single capacitor.
In a non-polar capacitor current flows from a nominated input conductor through an insulating dielectric to an output conductor. The signal enters one side of a capacitor and comes out with losses on the other side. The unintentional loss and added noise may result from, among other things, capacitor dielectric charge retention, dielectric movement, resistance (ESR) and inductance (ESL).
Paralleling capacitors will lower resistance (ESR), lower inductance (ESL), lower ripple and will have less dielectric movement. As smaller value capacitors are used, charge and discharge rates will be faster and capacitors will have better response time to high frequency signals. A bundle of smaller parallel value capacitors have a higher resonant frequency than one larger value capacitor. And small value capacitors are more likely to discharge fully resulting in less possible hysteresis errors and there should be less dielectric absorption.
Paralleling capacitors can give rise to antiresonance that may result in impedance peaks and lead to non-linearities in losses and errors over the capacitors frequency range. Antiresonance may be avoided or limited by parallelling capacitors of the same capacitance value.
In addition, there may also be issues resulting from different throughput speed between dissimilar bundled capacitors even that have the same capacitance. Dielectric material holds a charge, a thicker dielectric holds more charge and has a different charge and discharge rate compared to a thinner dielectric. Hence the throughput speeds may be different for some or all of the capacitors bundled and the output signal might include phase shifts and won't necessarily be a good copy of the input signal. You wont notice this with small ensemble reproduction but you might notice this, as harmonic distortion, with massed top octave vocals or with very top octave instrument note ringing, for example in piano reproduction, both can be brought on by smearing phase shifts. The sound of dielectric retention issues might be more apparent with a detailed playback system as there will be less noise masking.
Factors that impact on dielectric thickness and capacitor charge and discharge rates are the capacitor voltage rating, dielectric material, if the capacitor is foil or spray metalized, capacitor build and its capacitance. To limit these small errors it is best practice to parallel capacitors of the same voltage rating, the same dielectric material, the same build and the same capacitance. Though you can relax the criteria, and anyway rules are there to be broken, so mix and match but be aware of dielectric thickness variations and watch out for smearing.
The DAC (TDA1549T) I'm using has second order quantized noise filtering and lets through top octave noise, while the opamp in place, the LME49723, has a relatively low Common Mode Rejection Ratio (CMRR) at the top octave. CMRR is a measure of the effectiveness of the distortion nulling circuit of an op amp. A low rate at the top octave implies that the op amp does little to help in lowering top octave noise. Another way of viewing capacitor selection and paralleling is as a way of managing capacitor signal losses and which parts of the signal to degrade to get the sound you want. And this in part is what I'm looking for to balance my DAC/op amp signal. The capacitor mixes that sounded good to me did not let through too much treble extension and that seemed not to add to top octave noise.
For listening I used an aluminium Peerless tweeter (A25BG08-06) with added diffraction ring, and felt added to the hollow pole piece. Note the sound will differ depending on the crossover order used, the addition of loss components like series resistors and with different tweeters.
These capacitors I assume are sprayed splattered metalized Polypropylene film.
With a single Audiophiler (220uF) in place gave some ringing with top octave piano notes. Mid top octave violin played with tone. Slow material sounded a tad too slow though there was nothing wrong with pacey music. Eco-Pavan by Holt features piano and percussion, and this Audiophiler capacitor played this track just passably, but with little body and some hardness and resonance to top octave piano notes. There was not enough separation between instruments for an interesting performance. Boots of Lead (Holt) for soprano and ensemble was unlistenable and uncomfortable in the high treble region. Again there is not enough body to instruments played at the very top octave and treble rich material suFfered. Top octave brass instruments, Dance (Turnage), came across with little body and some annoying squawking. Listening to Takemitsu piano (Serkin playing), there was a mismatch in balance between the very top octave and lower octaves, it sounded as if the piano had been tuned wrong. Though with familiar ensemble material this mismatch comes across as a welcomed different view of the recording.
Five capacitor mix: paralleling five Audiophiler capacitors (5 x 0.47uF, 630V, total 2.35uF). The top end did sound more coherent. I found this mixture a bit too sophisticated for me but it was generally fast, detailed and yet had a bit of grit. Treble was still too extended in places. But top octave instruments did have more body. Piano in Eco-Pavan (Holt) was better separated and had more body. This piece was playable, though in places there was still treble exposure. “Boots of Lead” (Holt) for soprano and ensemble was played without issue with good vocals and with instrument separation, the performance here was better than expected. The high octave brass in Dance (“Turnage”) came across with tone and body and I liked the performance of this CD with this mix in place. Takemitsu piano (Serkin playing) was just passable but sounded too plodding. This five capacitor mix gave a safe performance, too safe for my tastes, but saying that it was coherent and played most material fine, there was nothing special about it.
Three capacitor mix: Audiophiler/Hovland tone mix. Two Audiophiler MKP (2x1uF, 630V), and one Hovland PB Blue series (0.22uF, 400V). Adding the Hovland improved the tone. High octave piano (“Pompili playing Castiglioni”) was more of a felt hammer striking strings, and there was more space or air between violins (“Departure”, George Crumb). The reproduction was generally not so extended. This Mix held together playing the difficult “Elliot Carter Sum Fluxae Pretium Spei” (Knussen conducts) and played massed high frequency vocals (“Early Voices” Philip Glass) with detail though this Mix did warm up some mid top octave voices to distortion, the rendition of this CD was captivating though and I carried on listening to the end of the CD. I could live with this Audiophiler and Hovland mix.
The Visaton is a polyester film between solid foil electrodes.
On listening to a single Visaton MKT-A, 220uF, there is not enough tone and body to top octave instruments. violins come across with too much sheen and gloss, and piano sounded washed out. Massed vocals are distorted. Takemitsu at the top octave piano was weak in tone and had, on occasions, shrill treble overtones to notes, the sound was not exposed. But the top octave piano rendition, for this CD, was not so good. On “Eco-Pavan” (Holt) there was a good balance and separation between high frequency percussions and piano, the best balance so far, and though high octave piano had shrill overtones these did not dominate and the reproduction of this piece was listenable and good. The soprano in “Boots of Lead” (Holt) played with no issue, OK some of the very high octave soprano notes were shrill, I liked the presentation of top octave cymbals with this single capacitor. Is the sound stage flatter? It might be, whatever it is, I'm liking the forward presentation of background instruments. Dance (“Turnage”) played without issue.
Two capacitor mix: inserting instead two Visaton MKT-A (2x1uF 250v). Takemitsue piano (Serkin playing) was very listenable, maybe a tinkle to very top octave notes. But the sound had good integration and balance and a good tone. This improvement was also apparent with Eco-Pavan (Holt) and high octave piano played well. And here the build up of musical tensions was expressed better than with other mixes. Instrument separation was good as was top octave instrument tone. This piece played well. “Boots of Lead” (Holt) played with no issue even with the very top octave soprano. There seemed to be a deeper sound staging now and background instruments were pushed back or there is a slight loss in detail. It's not slow. I could live with this two capacitor Visaton mix.
Three capacitor mix: Visaton/Mundorf tone mix. Two Visaton MKT-A (2x1uF 250v) and one Mundorf MCap ZN (0.22uF 250v). No top octave glare but the violins at the beginning of “Departure” (Crumb) sounded subdued but I felt these were played with tone. The mix played fine for other violin content CDs. “Scelsi Piano Works” were played with detail and the changing rhythms and structure of the pieces were well presented. Piano tone, for this CD, was luxuriant with this mix. Overall these Scelsi piano pieces sounded good, high octave notes were presented with body and with an insight into Bessettes piano playing. With most material the Visaton/Mundorf mix is listenable. There is nothing much to object to here. I could live with this mix and it needs an extended listen.
One capacitor only: Ampohm/LCR (2.20uF, 630v) FP-CA-AU good instrument detail, this was apparent listening to Drowned Out by Turnage. Mid top octave tone was fine however this capacitor lost very high frequency warmth and very top octave solo instruments came across as toneless with too much treble extension. At times there is, what sound like, an echo type distorted element added to top octave notes and this spoils some performances. Slow pieces could come across as too slow. Adding a WIMA MKP 10 (0.033uF 400v), I still found the Ampohm/Wima top octave bright but less so and this combination was listenable and high octave instruments were not so annoying and did not have an echo type distortion and did have more tone. The mix did give a reasonable sense of hands on keys and that solo piano was being played by a person. High frequency solo violin was not washed out by a toneless sheen. This capacitor combination handled the vocal harmonics of Proverb by Steve Reich, but only just. Though I felt that there was still a slight loss to dynamics playing slow music.
Two capacitor mix: Listening to two Ampohm/LCR (1uF, 630v) FP-CA-AU in parallel. Lots of tone with this Mix. Takimitsu piano pieces are very listenable, full of high frequency gravitas and tone. There is a slight tinkle overhang and hardness to very top octave notes but mid and low top octave excel. Delivery, pace and integration were good. Eco-Pavan (Holt) was also played with gravitas. Top octave hardness and tinkle overtones also crept in during loud and congested passages here. There was instrument detail and separation and again delivery was good. Seems to give an insight into background instruments and a marimba was heard in places. Surprisingly given the detail, the capacitors maintained a warm presentation. This parallel capacitor mix gives more pace, more dynamics, better tone and detail than any other mix tried. Perhaps I'll add the Wima MKP 10 to see if this takes down the very top octave congestion hardness a notch. I could live with this mix and it deserves an extended listen.
I tried a number of 0.22uF capacitors (Audyn, Hovland and Solen Fast Cap) in a three parallel mix with the two Ampohm/LCR (2x1uF, 630V) FP-CA-AU and none worked or could not produce a better sound over the two 1uF capacitor mix.
Monacor MKPA (250V), I had no single capacitor to compare the mix with. The Monacor capacitor mixes did not work for me.
Jantzen CrossCap (400V). Again I had no single CrossCap to compare mixes with. I tried various mix configurations and none of them worked for me. The capacitor was just too warm, rounded and rolled off and in the end dull in this set up.
Audyn Q4 MKP (400V). I tried a mix of two 1uF Q4s plus a 0.20uF Q4 and this mix worked well. I noted that some slow music did sound a little too slow and could come across as boring. Mid high octave notes (Marimba) were also warmed up. However the top treble was not extended. Very top octave piano notes had a nice dulled felt hammer sound. The mid top octave violins in “Departure” by George Crumb (Stanislas Quartet) were played with tone and this was an exciting performance, the best I've heard so far for this piece. I could live with this mix and it justifies an extended listen.
JB JFGC (630V). I could not find a combination that worked for me. Also I had no single capacitor to make a comparison with.
Though absolute proof is wanting, there is support that paralleling a number of similar value capacitors to make up the desired crossover capacitance does improve performance. Just paralleling budget capacitors may improve the sound but will not necessarily output the tone and detail that your system is capable of or that you want. The benefits of paralleling may increase with the capacitor quality used.
What is difficult to say is how much better paralleling say two capacitors is over the sound of just using one. A two capacitor solution will be pricier than a single larger value capacitor. Is it worth it? What paralleling implies is that you might be able to get good results from paralleling affordable but quality capacitors like the Visaton MKT-A, Audyn Q4 MKP, the Solen PB-MKP-FC FastCap or the ClarityCap PX. Also, this limited study did not identify the optimum number of capacitors to parallel. In designing parallel mixes you need to be aware of the thickness of the dielectric of the paralleled capacitors.
Capacitors are not perfect and will result in signal losses. Paralleling capacitors may reduce these losses. System performance at the speaker crossover stage is largely dependent on preserving the audio signal. And capacitors here won't improve on the audio signal but will cause losses. As signal losses decline as a result of better capacitor solutions these solutions will, system dependent, sound alike. Note: if the signal is already compromised by other components then changing capacitors may have little effect.
Another way of viewing capacitor selection and paralleling is as a way of managing capacitor signal losses and which parts of the signal to degrade or highlight to get the sound you want. The Mixes that sounded good to me did not let through too much treble extension and were capacitor combinations that did not add to top octave distortion. You need to think about your signal in particular by comparing the high frequency sound you are getting, as you change capacitors or tweeters, to your ideal sound.
© 2023 Paul Hunting - email@example.com - www.tnt-audio.com