Crossover Component Upgrade Pitfalls?

[gootee]

I was ready to jump on the bandwagon and upgrade the coils and capacitors in my Magnepan MG-12/QR crossovers (schematic: http://www.integracoustics.com/MUG/M...ver/mg12xo.jpg ) to better-quality components of the same values, replacing the steel-core coils with air core, and replacing the electrolytic capacitors with polypropylene film.

But then I did a few simulations to compare the old and the new bass circuits, because I was wondering how a lower coil DCR might affect the response of the system. And of course it can significantly raise the power delivered to the speaker, in the low-pass bass filter's pass band. If I lowered the 2.8 mH coil's DCR from 0.36 (measured) to 0.19, it would result in a power increase to the speaker in the bass pass band of 0.7 dB, which would be a 17.5% increase.

I could insert a high-quality resistor in series upstream from the coil to restore the original characteristics. But I'm wondering if it would then be worth doing at all, just to get rid of the possible hysteresis distortion caused by the steel-core coil (satuation of the coil shouldn't be a problem).

Replacing the electrolytic capacitors with polypropylenes turned out to open a much bigger can of worms:

By lowering the ESR (equivalent series resistance) when changing from an electrolytic to a polypropylene, the crossover point and the shape of the roll-off of the filter are significantly changed!

A 50 uF electrolytic might have an ESR of about 2.3 Ohms (just a typical value), at 120 Hz. But I measured the ESR of the 50 uF polypropylene that I bought (Solen) and it is about 0.06 Ohms, at 120 Hz.

The change in capacitor ESR would raise the filter's -3dB cutoff frequency by 7%(!), and would also significantly change the SHAPES of the amplitude and phase of the roll-off of the filter, as seen in the plot that I attached.

I could insert a high-quality resistor in series, ahead of the capacitor, to restore the original crossover plot shapes. But would it then be worth doing it at all, just to have less dielectric absorption? (Or what other improvements might there be?)

It seems that most of the "dramatic improvements" that people have been hearing after they replaced crossover components with higher-quality ones that have the same values MIGHT actually MOSTLY be due to changes they have unwittingly made in the actual crossover characteristics, by using parts with lower DCR and ESR.

My goal is the most-ACCURATE reproduction of the source material.

I definitely do _NOT_ want to base any of my tweaking decisions on my "personal preference", or on subjective listening impressions.

Does anyone think that the accuracy of the reproduction of the sound could be significantly improved, EVEN IF appropriate resistors were inserted in series with the higher-quality coil and capacitor in order to keep the original crossover's amplitude and phase characteristics?

Cheers,

Tom

[Conrad Hoffman]

Tom, I don't have a clue but I'm so happy to see somebody actually looking at these issues before diving in that I'm just giddy! I've believed that most of what people hear (if they hear something) when they do mods are the result of not considering ESR and what happens to the resonances it might affect. Here's one more piece of info- you can't duplicate the losses of a high loss capacitor by adding a resistor to a low loss capacitor. The math doesn't work. Well, the math always works, but it tells you that the response differs. Most capacitors maintain more-or-less constant dissipation factor over a wide frequency range. If you convert that (there's a useful utility on my website) to Rs you'll find you need a varying value. Not sure about Rp. That's also a factor in simulations. There are ways to simulate the real performance of a cap, but its a pita and hardly anybody does it.

[tinitus]

Originally Posted by gootee

I could insert a high-quality resistor in series, ahead of the capacitor, to restore the original crossover plot shapes. But would it then be worth doing it at all, just to have less dielectric absorption? (Or what other improvements might there be?)


Tom

resistor in series with paralel caps, definately yes, a must to me
tho I really don't know the theory behind it

resistors in series with paralel inductors, same thing

tho, series resistor on series inductors, I think is a nono

series resistor on series cap is ofcourse ok, and attenuates signal

question is whether the original crossover is optimal

[tuxedocivic]

You are replacing a 2.8mH with .36 DCR for one with .19? Why not get an air core with .36 instead of .19. It'll be cheaper and do what you want. Besides this, the 0.7db increase may actually increase the accuracy, depending on the original frequency response. If it was 0.7db too low in this region, then the lower DCR will help.

As for caps, I dunno.

[DrDyna]

You could always try the copper foil air core inductors like Goertz or Erse, I think those have higher DCR than the solid ones like Solen.

[Arty]

actualy you should simuklate the original xover, and design a new one that does have close to same parameters.

You did find out quite cleverly that replacing a component of higher quality in a wellmade xover where the components REAL WORLD properitys where factored will yield in negative results.

[gootee]

Originally Posted by Conrad Hoffman

Tom, I don't have a clue but I'm so happy to see somebody actually looking at these issues before diving in that I'm just giddy! I've believed that most of what people hear (if they hear something) when they do mods are the result of not considering ESR and what happens to the resonances it might affect. Here's one more piece of info- you can't duplicate the losses of a high loss capacitor by adding a resistor to a low loss capacitor. The math doesn't work. Well, the math always works, but it tells you that the response differs. Most capacitors maintain more-or-less constant dissipation factor over a wide frequency range. If you convert that (there's a useful utility on my website) to Rs you'll find you need a varying value. Not sure about Rp. That's also a factor in simulations. There are ways to simulate the real performance of a cap, but its a pita and hardly anybody does it.

Thanks, Conrad. I have only considered the bass crossover filter, so far, and it seems to work (mathematically) to use a series (ESR-substituting) resistor to restore the original bass crossover characteristics, _IF_ I can assume that the ESR would stay constant over the narrow range of low frequencies, maybe up to about 600 Hz. Beyond that, I would guess the response is so attenuated, either way, that it wouldn't matter.

I actually did a lot of work, a few years ago, to find methods for having good frequency-dependent (and temperature-dependent) spice models of electrolytic capacitors. So I know what you mean about the pita (unless you use capacitors from a company that already supplies the needed spice model parameters).

I still need to find out if better components will be helpful-enough even if their series resistances are maintained or added back in. It seems like I should make the new crossovers as outboard units so that I can leave the original ones in place, and should at least temporarily add switches to enable switching between them. Furthermore, I should add switches to enable switching between having and not having any esr-substituting resistors, wherever they might be needed. I have a very flat measurement microphone that I've never used so it looks like this would be a good time to finally put together a decent computerized measurement system, too.

Cheers,

Tom

[DrDyna]

Originally Posted by gootee

It seems like I should make the new crossovers as outboard units...
Tom

I always build any crossover on a piece of wood that sits behind my speakers. It's usually bristling with gator clips and BS so I can mess with it till it sounds right.

It never fails, however that as soon as it's in the box it won't sound right. Luckily, Maggies don't have that problem!

[gootee]

Originally Posted by tinitus

resistor in series with paralel caps, definately yes, a must to me
tho I really don't know the theory behind it

resistors in series with paralel inductors, same thing

tho, series resistor on series inductors, I think is a nono

series resistor on series cap is ofcourse ok, and attenuates signal

question is whether the original crossover is optimal

Tinitus,

Your last point is especially good. I had considered that, too, but didn't mention it above. After I realized that whether or not the original crossover was optimal was what should determine whether or not I should worry so much about changing its characteristics, I decided that unless I could perform good measurements (and knew what I was doing) or had information to the contrary, then, for now, I had to assume that Magnepan knew what they were doing and that their crossover characteristics should be maintained.

Tom

[DrDyna]

Originally Posted by gootee

Magnepan knew what they were doing and that their crossover characteristics should be maintained.

Tom

And, they should be. Chances are it's a great crossover...replacing parts will most certainly bring rewards as long as you are careful not to differentiate from parameters too far while experimenting.

DCR is probably something you'll have to pay close attention to if you're changing inductors. A few tenths of a percent are within tolerances of good parts, which you should be able to stay within so long as you're careful to check parts before ordering them.