Best practice for creating capacitor values in a cross over.

[DarrenHW]

Further to this thread: http://theartofsound.net/forum/showthread.php?40237-modifying-and-recapping-ditton-44-s&p=699115#post699115 where the intention is to create a value of 72uF from a range of capacitors (Mundorf ECap http://www.hificollective.co.uk/sites/default/files/mundorf_ecap_ac_pages_0.pdf) not containing 72uF.

Taken from the DIYAudio Celestion 66 thread:


Hi sba,

well you've jumped the gun and ordered caps before I got to that, but since I'm going to Post my recommendations later for the other guys you will find out eventually, this I may as well say now :-

68uF + 4uF is not the best audible way to achieve 72uF.

I don't know how much it will matter for those 72uF caps in parallel with the woofers, but with two widely spaced cap values a double resonance is set up. This is not as smearing of details and dynamics as the Dielectric Absorption of electrolytic caps cause, but better can be achieved by parallelling equal or near-to values.
For 72uF I would parallel 39uF + 33uF.

For Series connected caps, such as the 24uF to the MD500, and the 4uF and 6uF to the HF2000, it is audibly more critical to pair carefully if one has to use parallels, because the differences in the Pulse Rise Times between the caps cause signal to in effect pass through one cap of the pair faster than through the other, and this audibly changes the timbre {tone character of an instrument or voice}.
Larger value caps have slower pulse rise times than smaller value caps, as do caps of different Dielectrics, and even if use the same type of dielectric but different internal construction, or even different Voltage rating {in some cases} pulse rise times can be different.

Yes, I remember that Post where you wrote some of what you had read about combining various types and values of caps, and I only commented on that briefly then.
Basically that can work in Power Supplies when one needs to filter out wide-band noise, but to apply optimally one has to measure the results to ensure there has not been created one or more Resonant modes which couple into the circuit.
It is not a simple subject and I didn't want to type hundreds of words to explain it all, nor will I now, but ... for people who like to experiment, anything can and will be tried, and some experimenters chance upon sounds they like, and other experimenters don't, but simply keep on buying components to try every idea some-one has posted about a path to Audio Nirvana, but without understanding what they are doing.

I'm sorry that I don't have sufficient time available to address quickly or always comprehensively every point someone raises in this Thread, but as it seemed that every-one was not in a hurry I would have got to cap selection after getting through the relevant interim matters.

I don't know what you've ordered for 24uF, but I hope it is simply a 24uF single cap, or two 12uF caps.
Similarly, for the 4uF I would use a single 4uF or a single 3.9uF {which is only -2% difference}.
Some Brands have a single 6uF cap, or a near to such as 6.2uF, or a close values way of pairing.
I will be listing some of these when I list the reasonable value lower priced brands.

Also, for audilble consistency, presuming one has available a well made cap, one should use the same Brand and type and model {including voltage if possible} in all positions in at least each Filter arm of the cross-over, and maybe in the entire cross-over.
Thus I would never advise paralleling different brands or dielectric types, but any-one is free to try that and listen - who knows, one may chance on something one likes for the music types one listens to mostly.

Similarly the physical sizes, 100 Volt size is adequate if well made, and one does not need the larger 250 volt nor 400 volt sizes in cross-overs for domestic use loudspeakers that will not accept such large voltages, such as these old Celestions.
High voltage caps are only of audible benefit when there are large voltage stresses.
More important is how well the cap is manufactured, and you have already read something about this - I will explain more when time is available when I am recommending the alternate brands.
Some caps are manufactured to be Signal caps in amplifiers where there is small signal AC and large DC voltages present simultaneously, and this does not occur in loudspeaker cross-overs.

Well, if you have ordered I suppose you can't cancel now.
What are the cap values in your Dahlquist DQM-9 ?
You may be able to use some of the Solens more effectively there.

Any thoughts on this, should pairing high and low value capacitors be avoided, could a lower quality (or at least cheaper i.e. Alcap) capacitor of the right value produce better results than a higher quality (or at least more expensive i.e. Mundorf ECap) combination of values and what about using higher values capacitors in series?

[DarrenHW]

Whatever you do, try and get the +/- tolerances as low as you can, using a pair can sometimes be beneficial in that respect.

Yes, I did see this as potentially advantageous too, with a selection of low value and subsequently low cost caps very close values could be achieved.

[Reffc]

The advice to avoid mixing low/high values goes against many tried and tested commericial designs which if using caps in bypass suggest to use higher quality bypass valves to no more than 5 to 10% the value of the larger cap where one wishes to economise on the larger value (ie it's the smaller bypass that often seems to affect resultant SQ more as odd as it seems but this is from the standpoint of reasonable quality caps in the first place). The reason for bypassing is often not for reasons of SQ alone using multiple caps but is three-fold; Firstly (and obviously) for making up larger values, secondly, reducing effective ESR and finally, achieving a slightly better average tolerance although the latter works best when they are of equal value! It is not something that can be generalised upon from a SQ POV save achieving the right value at the required tolerance. Just use reasonable quality caps and concentrate on ensuring that total value and tolerance are as good as you can make them plus they're matched left and right channels.

This all pre-supposes that you KNOW that your drivers are equally matched of course (most newish drivers are generally within 1% from the factory these days which used to be what was described as "reference" standard, hence why it's more important these days to ensure for newer speakers that crossover circuits are equally close tolerance). If the drive units aren't closely matched, then it's all a bit academic anyway and you should't lose sleep over it nor waste money trying to achieve the unachievable unless you can tweak crossovers for individual circuits to accurately match the acoustic response of each driver between channels.

[DarrenHW]

Thank you very much for the reply Paul, not only very informative but exactly what I wanted to hear :).


The advice to avoid mixing low/high values goes against many tried and tested commericial designs which if using caps in bypass suggest to use higher quality bypass valves to no more than 5 to 10% the value of the larger cap where one wishes to economise on the larger value (ie it's the smaller bypass that often seems to affect resultant SQ more as odd as it seems but this is from the standpoint of reasonable quality caps in the first place).

With this in mind and with the intention of creating 72uF using 68uF + 3.9uF, as I read it there could be an advantage in using an lytic 68uF and a polyprop 3.9uF?


The reason for bypassing is often not for reasons of SQ alone using multiple caps but is three-fold; Firstly (and obviously) for making up larger values, secondly, reducing effective ESR and finally, achieving a slightly better average tolerance although the latter works best when they are of equal value! It is not something that can be generalised upon from a SQ POV save achieving the right value at the required tolerance. Just use reasonable quality caps and concentrate on ensuring that total value and tolerance are as good as you can make them plus they're matched left and right channels.

Achieving a better tolerance as identified in the posts above and the 44 thread and as you also point out matching left and right channels is one of the big advantages I can see to this approach.


This all pre-supposes that you KNOW that your drivers are equally matched of course (most newish drivers are generally within 1% from the factory these days which used to be what was described as "reference" standard, hence why it's more important these days to ensure for newer speakers that crossover circuits are equally close tolerance). If the drive units aren't closely matched, then it's all a bit academic anyway and you should't lose sleep over it nor waste money trying to achieve the unachievable unless you can tweak crossovers for individual circuits to accurately match the acoustic response of each driver between channels.

Yes the drivers are all closely matched, my speakers are made up from drivers taken from various pairs of speakers to achieve the closest values, my hallway's a Celestion boneyard :mental:.

Thanks again for the post :).