What capacitors in cross-over filters?

[jacques54]

Hello,

I need to replace the capacitors on the cross-over filter in one of a pair of old Bowers & Wilkins DM6. There are 40 and 25 uF bipolar capacitors.

I have replaced them by new ones, but it seems I suffer from a reduced bass range on this speaker, confirmed by switching them -- so, not only a matter of room acoustics etc. Can it just be an impression?

Would polyester capacitors (4 x 10uF and 2x10 uF+1x5uF, I presume 63V AC) perform better than replacement bipolar capacitors? Or am I just wasting my money?

Thank you for your thoughts,

Jacques

[Ap]

You say you have replaced capacitors already - are you sure you havent wired them incorrectly ie. series instead of parallel? What is the measured value?
As for poly sounding better they do, but probably only really worth doing if they are in series with the woofers, if they are in parallel then adding a small high quality poly (5-10%) is enough.

[sreten]

Hi,

its possible that B&W selected particular values (e.g. 38uF and 27uF)
from batches of 40uF and 25uF capacitors. Bipolar values are not
accurate.

Cheap ones are +/-20% i.e. 40uF=32uF to 48uF, 25 = 20 to 30.

Probably you need to measure the old capacitors and match them.

/sreten.

[RJ]

I hope you paralled all them caps. If you seriesed them than they would be about a 1/4th of the value you are trying to replace.
See pic...

[jacques54]

Hello everyone,

Yes, I believe I still remember that you have to put capacitor values in parallel in order to sum their values :-)

Please find attached a schematic and a PCB picture. The LF contour control (with its inductor) is not present in the speaker, just replace it by a straight connection. No, it wasn't there when I bought them new in 1978.

I have replaced the leaking capacitors by new bipolar ones. There are 3 electrolytics: 2 x 40 uF and 1 x 35 uF (sorry, from bad memory I wrote 25 uF in my original post).

I have had a look at the PCB in the other speaker. My multimeter measures capacities of 39.2 uF, 39.1 uF and 34.0 uF (from the still-functional original board) while the new replacements (on the repaired board) measure 33.6 uF, 48.2 uF (summing up to about 80 uF -- yes, they are definitely in parallel) and 34.0 uF.

The values are about right, I presume. Maybe these new capacitors (NITAI) themselves aren't of good quality -- bought them in good faith from the shop around the corner.

[sreten]

Hi,

Nitia appear to be cheap and nasty.

You need something like ALCAP low loss bipolar types.
(From falcon acoustics in the uk http://www.falcon-acoustics.co.uk/pl28p7.htm)

/sreten.

[kevinkr]

More likely that B&W designed the x-over network to compensate for a known range of ESR in the capacitors used in the x-over.

This was definitely the case with some AR58S I used to have, sounded much leaner and somewhat lacking in the midbass as well after I changed the stock bipolar caps for Solens. Overall resolution was much better, but the tonal balance was significantly worse.

I would try adding a small amount of resistance in series with the caps and see if this restores the proper tonal balance. (Anywhere from a couple of tenths of an ohm to a couple of ohms at the most.)

I recently designed a set of x-over for my Onken project and deliberately designed around low ESR caps and low resistance inductors. Adding R to either of these components in the simulation had a big effect on Q and inflection point of the x-over.

Two large loudspeaker manufacturers in the New England area I have worked for in the past definitely took ESR in cheap bipolar caps and dcr in inductors into account in their x-over designs.

[hooha]

Kevin, you're absolutely right - and the two speaker companies you worked for aren't the only ones who look at design the same way...

I personally would lean more towards the cost factor of bipolar vs. type-x cap in this particular design given the sizes used. It was probably the bipolars they worked with in their R&D from beginning to end, hence the sound change when they were replaced.

That's the problem with "tweaking" x-overs from commercial systems. Most of us think that by replacing a lower cost, lower quality component like a bipolar cap would automatically result in superior sound when in many cases it has an opposite effect.

Mark

[sreten]

Originally posted by kevinkr


I would try adding a small amount of resistance in series with the caps and see if this restores the proper tonal balance. (Anywhere from a couple of tenths of an ohm to a couple of ohms at the most.)

Hi,

Pointless for the 80uF. The 80uF is already in series with a resistor.

The 35uF is connected to the midrange and it is possible midrange
levels have been lifted slightly, easily fixed by the contour control.

/sreten.

[kevinkr]

I don't entirely agree, if the original 80uF had an effective ESR of say 0.5 - 1 ohms and was replaced by one having an ESR of 0.05 ohms instead there should be a discernable effect on the tuning. Have you simulated this? I haven't for this case but in my own x-over designs this is an issue.

I would recommend simulating the effects of different amounts of resistance in series with that cap from say a low of 2 ohms to a high of 4 ohms and see how much the x-over tuning varies.

Your comment about the midrange x-over is overly simplistic, the ESR of the capacitors used in this network WILL change the filter Q, and that level control will not change the shape of the resulting response. The effect may be subtle or not. If you don't believe me try the simulation with "perfect" caps and inductors and then add some R to one or the other or both - the change in response will be quite illuminating.

Of course as has been alleged poor quality capacitors may also be contributing to this issue as well. (High DF/DA)

I had similarly lousy results replacing presumably poor quality no name brand bipolar electrolytics with very good quality Solens in the afore mentioned AR58S.