about capacitor values

bzfcocon said:
Is there any reason why paralleling is bad ? I've heard people RECOMMENDING either bypassing with a smaller capacitor(in the hundreds nF range) or paralleling larger values with very high quality smalle ones (1uF)

Paralleing is fine to achieve the required value.

By-passing large caps with small ones maybe useful in chip supply decoupling.

In crossovers and any kind of filter network in the audio band it is not relevant and its advocates do not understand what they are doing!
I used the crossover calculator avalible on lalena dot com. It dsn't ask for specific driver values other than the drivers impedance (one scan speak revelator-mid @4ohm and scan speak 7100 tweeter @4ohm) and it also asked for the x-over freq which was 2500hz. it gave me a cap at 15.9 and an inductor at .25 i'm basically making a "clone" of the sonus faber cremona auditor using the 1st order crossover and scan speak components. but using the dayton audio curved box design that is maybe 1 inch off size from the original cab design.
Sounds like you are going to pair a couple of the worlds finest drivers with the worlds poorest crossover.

At least if you must use that useless online calculator, you must enter the impedance at the frequency you are crossing! Impedance plots are included in the datasheet for the drivers.

I would highly recommend using a proven design designed by one of the several diy experts around the net.
Those impedance values are nominal. It is necessary to find the actual impedance at the intended XO point.

I may be wrong but I have always believed that if you are paralleling caps try to use two the same. ie: it is better to use two 4.4's to get 8.8 than it is to use a 6.8 and 2.0 Not sure if that's correct though.

EDIT: Posted at the same time as DcibeL
Cal you may be thinking in the right direction... 2uf cap will have higher ESR (generaly) so even though you reach the same capacitance, it is at a higher impendance. Sometimes however we jsut need to get to a value quickly and anything will do...

Hope you have some way of measuring the cap....

The 15uf cap could have a physical value from 13.5 to 16.5uf if it is common 20% tolerance models...
Capacitors in these value ranges are usually 10% or 20% accurate, for real money you can get 5%. If you parallel a 15 and a 1 or 2 x 8 you get 16 uF an error 0f 0.6%.

I tend to favor two capacitors of 1/2 value each, based on personal experience one of the two should be connected anti-parallel i.e. backwards from the other (assumes a non-polar design).
I will look at the data sheet for the impedance values... I don't think that I will be finding any schematic for the Sonus Faber Xovers, I'm going off of what they listed (which may be vague?)

"2.5khz, non-resonant acoustic first-order, optimized acoustic phase response, highest-grade components"

maybe sombody can tell me what this means and give me a little guidance on my X-over design, since my values are wrong then maybe I won't have to parallel any Caps!


2006-12-30 11:55 pm
StefanPoPolsku said:
"2.5khz, non-resonant acoustic first-order, optimized acoustic phase response, highest-grade components"

maybe sombody can tell me what this means

The crossover is done by requiring the drivers response to be first order, and using whatever electrical components are required to do the job. This doesn't necessarily mean you have a first order crossover in the text book sense.

Unless you're prepared to graph the electrical and acoustic properties of the drivers, you're only guessing.

If you want to produce a makeshift crossover, you could develop a zobel impedance compensation network for your woofer first, then add a textbook first order crossover. Tweak until you find something you like.

bzfcocon said:
Is there any reason why paralleling is bad ?

There are some good threads here about this, some get quite technical. The issue in a nutshell is resonances between the caps involving stray inductances. IMO this is less of an issue with speakers, but still a matter of trial/taste.

I am thinking that the accuracy of the extra 0.9uF is a little pointless in this case due to the expected inaccuracy of the crossover design at this point.


2006-12-30 11:55 pm
You can take a published frequency response plot and add the effects of the enclosure on the bass, and the effects of the baffle (baffle step). These both can be simulated with reasonable results.

Rarely do you find published phase data, and with baffle effects it would probably be inaccurate for individual cases, anyway.

Published impedance data correlated with simulated in box impedance data at the low end, should get you close. You probably wont find impedance phase data, though, and this will make your crossover design inaccurate, however you can measure this data with a normal sound card and no mic, so it isn't as painful.

Iain McNeill

diyAudio Moderator Emeritus
The natural driver roll off is 2nd order (-12dB/oct) so a 1st order XO will morph to a 3rd order at some point. Not quite sure how you "optimize the phase response" using a 1st order XO but hey, I'm not a marketing type.

The driver impedance is very important as this will change the performance of a 1st order electrical XO greatly. As will the actual value of the capacitor/inductor.

Don't believe what is written on the component. Measure it. You'll be surprised how "off" the value is. Personally, I buy several parts of values around the target as well as a few "top-up" values. If you need 14.7uF and your 15uF cap is actually 13.9uF you can add a 1uF cap that is actually 0.8uF to get it dead nuts. Make sure you balance left & right.

As for paralleling/seriesing caps, if they're the same brand/product line then you can pretty much mix any values you like. That's all the mfg is doing when they make bigger caps, really. The issues come in when you mix brands and one cap has low esr over one freq range and the other has low esr over a different range. But this is not usually a problem in the audio band - with good quality parts of course. If you're cannibalizing that 1950's radio for parts well then, your performance may vary.


2006-12-30 11:55 pm
Iain McNeill said:
Not quite sure how you "optimize the phase response" using a 1st order XO

Amongst the options, you could create a second order electrical filter with an especially low Q factor tuned above the crossover point, which is chosen to be below the woofers natural rolloff. If done right this can behave like a first order acoustic for an octave or so.