Introduction to designing crossovers without measurement

Eyes glazed over after re-reading this whole thread...I am refurbishing a friends old Scott 188T three ways. 10"/5" CTS, 1" Phillips. My question is about the tweeter leg connection point in the crossover. Some say it should come after the mid cap and others have all three legs in common. Amplifier will be ordinary receiver. Posts 178 and 342 show variations on the first approach. Discuss? Irrelevant?
 
diyAudio Moderator
Joined 2008
Paid Member
I assume this is one of those that cascades the tweeter after part of the midrange network, maybe after one capacitor? The main question probably is does the midrange capacitor have an effect on the tweeter in the band where it is audible.

Another question is does the presence of the tweeter and its circuit affect the midrange, and does the tap out of the middle of the midrange circuit involve it enough to have an effect on the tweeter? Since the reactance of the first midrange capacitor is likely low over the midrange and the tweeter bands, the answer might be not enough to cause a significant effect, unless it is proven otherwise with measurements, where it is still likely to be small.

Another issue is the effect of the extra capacitor on the tweeter at low frequencies, below its audible band. Maybe this is the appeal of this kind of circuit. The idea that unnecessary excursion can be limited, reducing heating of the voice coil, also reducing the potential for non-linear distortion as the coil ventures beyond design recommendations.

Harmonic distortion has been brought into question, especially recently. There is no doubt that we can hear it under certain circumstances. However, I think that it isn't responsible for most of the errors that we can hear in a speaker but can't seem to identify the reason for. Furthermore it has been shown that other specific things are responsible for a significant proportion of these faults.

With regards to voice coil thermal characteristics speakers have limitations, and 1" dome tweeters are no exception. There may be a couple of types of thermal concern. One that depends on the overall heat buildup, and one that may cause a minor effect regardless of how hot the voice coil environment is. How this can be addressed, and whether extra high passing has an effect or not can make interesting research. At one time it must have been thought worthwhile.

Back to the primary concern. Will changing the tweeter network input point from after the midrange capacitor, to before, bring a result that is significantly similar? Probably yes. Will calculating the value of the first tweeter cap and the first midrange cap in series, to get the effective capacitance, address much of the remaining discrepancy? Probably yes. Will it be perfect, maybe not yet .. but could you achieve the original intended response after bypassing the midrange cap, and after further work to determine the differences? Yes.
 
Post 342 has a simple common feed option, and another which has the tweeter paralleled with the midrange, sort of. So I was looking for a reason to cascade the tweeter after the midrange cap, an arrangement more often seen in older texts like the Phillips book on loudspeakers from the 70's, which is not exactly option two in the cited post. I'm inclined to feed the tweeter directly unless there's a reason not to. No electrolytics involved by the way. Thanks for the response and for this great tutorial.
 
Just got thru reading this great thread. Thanks Allen and all for the great information. Hope I could ask a question about an old Realistic Nova 8 crossover. My dad bought me a pair of these for me in the mid 70s, and a few years later I replaced all the drivers with top-end Philips drivers, with the intent of redoing the XO, which I never did (ah, the 70s...).

I recently got them out after 20+ years and recapped them (with original values), and was planning on just restoring the cabinets, but then thought about checking out the XO first.

Have you seen an XO design like this before? The schematic is right, but I cant verify the driver connections. One thing I would like to verify is the polarity/phasing of the drivers. It looks to me like the woofer should be reversed so the mids are reversed and woofer/tweeter in phase?

At some point I want to take all this good info and redo the XO, but should I scrap this and build a new one, or could I modify this old design? FYI - these guys actually sound quite good as is, but I know they can sound better.
 

Attachments

  • Nova 8 Crossover Schematic XSim.JPG
    Nova 8 Crossover Schematic XSim.JPG
    218.9 KB · Views: 302
diyAudio Moderator
Joined 2008
Paid Member
This crossover does look a little unusual. This is because in part it is a series crossover.

With a series crossover (eg the image below), a second order filter can be made with an inductor and a capacitor. Then one driver, say a woofer connects to the filter in the normal way, and the tweeter uses the filter upside down at the same time.

In your case the woofer and midrange(s) are in this configuration, meaning L1 and C1 are also responsible for cutting the lows from reaching the midrange drivers. The tweeter takes its signal from the same place as the mids, but its response is likely more the result of R1 and C3 than on C1 and L1 because it covers frequencies away from their effective band.

The series crossover has certain benefits. Naturally it saves parts but, it also has the quality of reducing the crossover sensitivity to impedance variations in some speakers.

In your crossover, C2 and L2 make a notch filter (a bandstop filter) for the midrange. This is designed to suppress a region around 8.5kHz. With new drivers it becomes uncertain whether this is still needed. You could connect a clip lead across them to see if you hear a difference. You could disconnect the tweeter first to make the difference more obvious. Possibly you'll want to consider a larger inductor in place of the two components, low-passing the upper midrange.

The midrange level is easily adjustable. The tweeter level can be changed through the resistor. When you increase this value, you also lower the tweeter crossover point. This can be countered by changing the capacitor value, but for small changes it's up to you.

Another option would be to use an L-pad on the tweeter. Normally these are set to a certain point and left alone, they are not meant as treble tone controls. However as you have new drivers you will need to find the correct level for the tweeter and mids.
 

Attachments

  • series.jpg
    series.jpg
    26.8 KB · Views: 298
Last edited:
and the tweeter uses the filter upside down at the same time.

The tweeter takes its signal from the same place as the mids, but its response is likely more the result of R1 and C3 than on C1 and L1 because it covers frequencies away from their effective band.

In your crossover, C2 and L2 make a notch filter (a bandstop filter) for the midrange. This is designed to suppress a region around 8.5kHz. With new drivers it becomes uncertain whether this is still needed. You could connect a clip lead across them to see if you hear a difference. You could disconnect the tweeter first to make the difference more obvious. Possibly you'll want to consider a larger inductor in place of the two components, low-passing the upper midrange.

The tweeter level can be changed through the resistor. When you increase this value, you also lower the tweeter crossover point. This can be countered by changing the capacitor value, but for small changes it's up to you.

Another option would be to use an L-pad on the tweeter. Normally these are set to a certain point and left alone, they are not meant as treble tone controls. However as you have new drivers you will need to find the correct level for the tweeter and mids.

Many thanks for the reply.

So if the tweeter uses the filter 'upside down', does that mean the current polarity/phasing of the connection is correct? Also, R1 in my diagram is actually a 10 ohm pot (same as R2/R3) used as a variable resistor (side/middle pole). So varying this changes the XO point and not just the level to the tweeter?

So is the phasing for woofer-mids-tweeter ( - - +) correct?

I'll test out removing the notch today and see how it sounds removed.

BTW - the original XO points for this were 800 and 5K.

The other mystery here that I thought I had bought 8 ohm woofers when I replaced them, and they are stamped on the magnets "/W8" Philips 8 ohm designation, but both these measure 3.6 ohms. (Total box DCR is 4.6 ohms). The woofer has a fairly broad range (30-1500hz) - with this series filter, do I now need to change C1-L1 to compensate?

Thanks again, I lost all notes to what I originally intended on doing with these - that was a long time ago!
 
diyAudio Moderator
Joined 2008
Paid Member
So if the tweeter uses the filter 'upside down', does that mean the current polarity/phasing of the connection is correct?
Changing the polarity of one driver will flip the phase by 180 degrees in any case. It is fair to assume it was set right originally.

The series connection should give a result approaching second order, which more often likes a reversed polarity. I can see the woofer has its positive connection attached to the positive output of the amplifier, and vice-versa for the midranges.

Beyond this, the phase may wander. If you are dissatisfied trying to get the drivers to blend and have tried both polarities, your next step is to try to blend the response slopes. Can I assume you have them from the manufacturer?
So is the phasing for woofer-mids-tweeter ( - - +) correct?
Presently, it is (+ - +).
Also, R1 in my diagram is actually a 10 ohm pot (same as R2/R3) used as a variable resistor (side/middle pole). So varying this changes the XO point and not just the level to the tweeter?
Yes. The reason is more to do with the first order crossover. It may not sound obvious and may not be a problem.

The following is only a suggestion. It may be of interest to use an L-pad instead. This way you could vary the level without the frequency change, and there would be a reduced response variation in general, as you do. This would also reduce any response bump introduced by the tweeter resonance impedance peak.

So, how would you decide which setting of the potentiometer works best (so that you can decide this value works well with the fixed capacitance, when you remove and replace the pot..), because when changing the pot it also upsets the level and makes it hard to decide? Maybe it would be easier to vary the capacitance after removing the pot from its original location and configuring an L-pad.
woofers ... measure 3.6 ohms ... do I now need to change C1-L1 to compensate?
In your case, the output of the woofers is likely to be reduced at its top end around the crossover. Reducing the value of L1, or increasing C1 are also going to affect the mids. Maybe the sympathetic nature of a series crossover will be partially self correcting. Maybe your ability to change the mid level is fortunate?
 
Thanks. What I couldnt understand was the midrange bandstop (I had to look that one up).
I didn't get a chance to test a bypass of that ckt, but I think I may take your advice and remove that cap and figure a higher value for that low pass L2.

Really appreciate your help!
 
Btw - I do have the specs and freq graphs for all Philips drivers. I have found some great old docs on these, and also found a gem of a online book by M.D. Hull of Philips ‘Building Hi-Fi Speaker Systems’, from 1980, which includes specs on my drivers, acoustic theory, loudspeaker design, and example systems.

In doing some tests, I’ve found that one mid (Mid#2) is damaged, as its output is about nul and distorted. So I am replacing them with the same type as the Mid #1 (AD5060) which I feel better not having two different mids to worry about with different responses. I’ll remove the bandstop add a simple bandpass and should have a much better speaker. Thanks again.
 
We'll need to use one resistor and one capacitor (per woofer).
I have not read the entire thread, and, may-be some one has even said this, but this is far to be optimal.
To linearise a speaker, you need to compensate both its inductance, by a RC network in parallel, and its peak of impedance at its resonance frequency, by a RLC network, in parallel too.
See:
La correction d impédance RC série
La correction d impédance RLC série
La correction d'impédance RC et RLC série

As the good speakers manufacturers usually provide the impedance curve of their speakers, no need to any measurement on your side.
After the correction done, you will have a flat impedance curve, usually at the DC resistance of the coil (6 Ohms for a 8 ohms speaker, 3 for a 4 Ohms). And you can easily calculate your filter on the theoretical values, it will work as expected.
Just my two cents..
 
Last edited:
diyAudio Moderator
Joined 2008
Paid Member
This is a good tool. I have used this technique of RLC compensation on a woofer's impedance when I introduce a finite source impedance. For example a crossover or an amp. It can help to get the wanted result.

I considered this when starting this thread. My thoughts were that typical woofer resonance is around or below 100Hz. Typical crossovers of this kind, are around or above 1kHz. This is an order of magnitude difference. In other words, this is far enough apart in frequency that the interaction can be considered insignificant for most purposes.

I also find RLC compensation is something that can be applied to a tweeter at its resonance.
 
This is a good tool. I have used this technique of RLC compensation on a woofer's impedance when I introduce a finite source impedance. For example a crossover or an amp. It can help to get the wanted result.

I considered this when starting this thread. My thoughts were that typical woofer resonance is around or below 100Hz. Typical crossovers of this kind, are around or above 1kHz. This is an order of magnitude difference. In other words, this is far enough apart in frequency that the interaction can be considered insignificant for most purposes.

I also find RLC compensation is something that can be applied to a tweeter at its resonance.
This practice (RLC for all speakers) has 4 major benefits.
1- It helps to have theoretically calculated passive filters to work as expected, not having to tune them on the bench with a mic.
2- It helps the speakers , as they have now a flat impedance curve, to keep their response curve unchanged, whatever the serial impedances of the cables and passive filters.
3- It helps to improve the damping of the speakers: you can easily hear the difference (power amps off), by hitting the membrane of the boomer with your finger.
4- It helps the amplifier(s), especially their power supplies, to get more coherency all over their frequency range, having no more to deal with a schizophrenic Voltage:current behavior. May-be too, we will enjoy a benefit on the micro dynamic looking at the levels in their feedback paths to compensate the difference between voltages/currents VS frequencies.

About the "insignificant", OMHO, the best speaker assembly is the one with the less ways as possible, to reduce the acoustic problems of crossovers (phase turns etc.).
So, in my case a two ways (with horn), the crossover frequency is not so far from the resonance frequency of my horn driver, and close too to the first break of the membrane of the boomer. I can hear the difference, despite my filters are 48dB/oct, comparing with/without RLC comp. on each of the two speakers.
 
Last edited:
diyAudio Moderator
Joined 2008
Paid Member
I agree that the impedance of a compression driver, and a horn with a cutoff frequency, can be a significant problem. Also these often have to be cut close to their limits to suit other needs.

I have some concerns with other points you have made, but at this level I have no problem with them.
 
Thank you Allen B. I have studied your thread and made sure I understood everything before I took the plunge into XO's. I love it. I have been very successful on my first attempt to make really good sound. I am still tweaking my design but I have found that is the best part. I tweak something and then listen, maybe minute, maybe hours before I think it may need more tweaking. I really feel like I have accomplished a lot! Such a great thread!!!
 
Hello Allen B, it has been 8 years since you published this tutorial, thank you and thank you, I wish I have a better words to express my appreciation.
XO design has been a show stopper for a newbie DIY like me to embark to the next journey, and you have removed this barrier.
Great job and you shall be proud with your work.

Chandra.
 
Hello Everybody,

I used this tutorial to finish my small speaker project, many, many thanks to the original author for simple and understandable language in explaining problem during designing passive crossover!

I did use on occasion Boxsim program from Visaton website, it was a nice software tool to play around with changing parameters to death but I felt you are never sure how is going to turn out on the end.
To help myself and shorten the developing time I decided to make kind of universal crossover and play with parameters live while listening.
Inside of blue box was hidden 12db or 18 db adjustable high-pass crossover for tweeter with attenuator, polarity switch and choice of 3 different coils… each capacitor bank could go from 0,5uF up to 22uF… in short I could emulate in real-time and listen to the effect.
I have to say it was revelation as you could accurately fine-tune without guess work, but I did not measure final result. This will have to wait until I buy proper microphone.

Mini speakers that I worked this time were 2-way closed box design inspired by BBC Research done in the 70’s designing small monitor speakers for broadcast. I did not copy any of drawings from LS3/5a, just followed designed guidelines and choose myself off the shelf drivers I could afford.
Mini monitors are based on Visaton W130S woofer and Dayton Audio RST28A tweeter as the best candidates without breaking the bank.

I finally came to conclusion that 12db crossover sounded best all-round but 18db was a tad more compressed in comparison. Point here is that you would not have idea how what sound unless is accessible on flick off a switch.

Still, the work done on software program was crucial in finding sweet spot to start with, especially low-pass part of crossover, so once I settled on low-pass, I further spent time with variable high-pass to great delight as you could hear every adjustment and spend some quality time in fine tuning and experimenting.
I could post drawing of crossover if there is interest, for time being just a couple of photos of the speaker and adjustable high-pass crossover.
 

Attachments

  • Mini 1.jpg
    Mini 1.jpg
    980 KB · Views: 453
  • Mini 2.jpg
    Mini 2.jpg
    1 MB · Views: 2,233
  • Front & back panels.jpg
    Front & back panels.jpg
    878.4 KB · Views: 433
  • Final crossover.jpg
    Final crossover.jpg
    1,008.5 KB · Views: 2,050
  • Variable 12db:18db crossover copy.jpg
    Variable 12db:18db crossover copy.jpg
    549.8 KB · Views: 1,543
Last edited: