Which Impedance for Crossover Calculation? The Nominal, the One at Crossover or?

[T101]

Which Impedance for Crossover Calculation? The Nominal, the One at Crossover or?

The topic is weakly addressed in this thread: http://www.diyaudio.com/forums/analog-line-level/233169-calculate-crossover-re-impedance.html
Generally the Internet is lacking serious read on the matter or at least Google fails to present in the first pages with results. The thread from the link above is one of the few sensible results at all.

There are some self propelling myths on the matter though.
One myth is that the impedance of the driver at the crossover frequency should be used for the calculations.
Another myth is that Zobel networks can be avoided.

According to my humble results, in order not to use a Zobel on a conventional driver and not have anomalies in the summing at crossover frequency, one should use some very heavy R&D department in order to have the exact high Q crossover alignmet requied to fill the holes.

Actually it is either flat impedance or nothing. It occurs that it is that simple and this is yet another constraint to add to the already commonly known ones. In order to avoid impedance equalization crossovers should be placed at fequencies where drivers show relatively flat impedance.

Now my simple experiment.
I ran simplified simulations of a symetric Butterworth crossovers at 2000 Hz as follows:

- 1st order with values for 8 ohm loads loaded with 14 ohm load in the LP section;
- 2nd order with values for 8 ohm loaded with 14 ohm in the LP;
- the same as above, but instead the resistor, the LP is loaded with a 0.45 mH inductance with dcr for 6.2 ohm;
- 3rd order with values for 8 ohm, LP section loaded with 0.45 mH 6.2 ohm dcr;
- and in the end, the same 3rd order as above but with values of the LP section for 14 ohm load.

On all simulations are shown the curves of the response of ideal 8 ohm loads and the close to real load for comparison.

(a typical "8 bass driver with not very long 1.5 in (37-38mm) two layer voice coil would have about 0.45 mH Le and 6.2 ohm Re)

As it becomes obvious, you can't avoid Zobel networks and if you take the impedance at the crossover frequency it becomes worse.

Originally the thread is posted in Bulgarian language here: Bulgarian language thread on impedance at crossover frequency

Best regards!

 

[cradeldorf]

the one at crossover.

 

[Inductor]

As the passive components are reactive (caps and coils) with the inductance of the woofers drivers having some undesired effects (peaks) when making crossovers, Wayne Parham gives a good rule of thumb...

...So a good “rule of thumb” is that woofer compensation RC damper circuits may be omitted if voice coil inductance is less than crossover inductance.
Pi Speakers - unmatched quality and state-of-the-art performance

 

[Pano]

If you use some software tools like the Passive Crossover Designer (or other) where you can virtually try a crossover topology on a real impedance curve, you'll understand quickly what's going on. It makes a difference!

I often don't Zobel my woofers, just design with the rising impedance in mind. It can often help achieve the roll-off I want.

 

[sreten]

Another myth is that Zobel networks can be avoided.

Hi,

Your wrong, the myth is that they are needed, which they are not.
When something similar is needed, values are not Zobel values.

Analysis of electrical x/o functions is a waste of time as they are
all wrong. Right is acoustic functions taking into account driver
and box effects diffraction effects, everything else is wrong.

e.g. 4th order L/R is acoustic, not electrical. The electrical
functions are the difference between 4th order L/R and
the acoustic responses of the drivers in the box design.

rgds, sreten.

 

[Pano]

Right is acoustic functions taking into account driver
and box effects diffraction effects, everything else is wrong.

Yes. This is the very important fact that most people miss. It is the acoustical result that matters. You use whatever electrical filter function you need to achieve the acoustical results. That's why passive crossover values that have been optimized often look odd.

 

[T101]

Please bear in mind that the "optimization" and "experimentation" capabilities of the ordinary hobbyist are very limited especially when inductors are concerned.

I thought of two examples when impedance equalization is not required, this is when the case is LP crossover of bass driver between 100 and 600 Hz and the second is a HP crossover for ribbon driver.

I have Visaton's free software Boxsim. I took the first 8 inch bass driver which happened to be W200S, put it in 30L sealed enclosure and applied a textbook second order Butterworth crossover for 8 ohm load.
See what happened... I apply SPL curve and another simulation showing the exact influence of the crossover. The scale of the simulation is the same as the ones from LT Spice.

Optimisation is nonsens yet for another reason, it means that the crossover quality factor would be altered. Then the syne wave response might be satisfactory and the curves and phases might seem alright as well, BUT the impulse response and the crossover induced distortion would be unacceptable.

Now let us see what will happen if we design a Zobel and apply it without any other optimization or whatsoever. I designed it with this calculator: Impedance Equalization (L-Pad) Circuit Designer / Calculator using the published data: W 200 S - 8 Ohm

Wow! The ugly peak disappeared!

Values: LP Butterworth crossover 7.03uf and 0.9 mH, Zobel: 7.5 Ohm and 39.1 uf.

Guys, you will have to elaborate harder on this.

Why would I put the labor to "optimize" when I can design a Zobel with inconsiderable cost within seconds while in the same time I will have the benefit of keeping the crossover Q factor with the desired value. I might be a fan of LR crossovers for their great impulse and transient response, "optimization" would deprive me from the opportunity of using them.

I am not going to bite the LR-4 bait. But... with such steep slopes, summing is not a problem even if the acoustical slopes are deviated for some reason... Phases can be aligned in a matter of putting spacers beneath one driver flange... and the words "transient perfect" are pure majic
Sreten, please open another thread if you are so very interested in the topic.

Best Regards!

 

[Mario Pankov]

The one at and above crossover frequency. Zobel can be avoided if inductance of the voice coil is very low, which is not the case with the bulk of cheap drivers in the €10-100 range. If not using Zobel your filter function will change with frequency and reduce its steepness above the selected crossover point, which may not be desireable if breakup nodes are present. Textbook values are useless 99% of the time, unless a Zobel is used and DC resistance equals the reactive one.
I would also go for a LR2 function or even a sub Bessel instead of the Butterworth.

I see you`re from Bulgaria, I can lend you ( free ) my measurement rig if you promise to be good with it - calibrated mic, Behringer mixer for phantom power/mic amplification and a Behringer external sound card. Can also give you an impedance measurement device ( pretty simple to construct on your own ) and an RLC meter. You`ll need to download HOLM for acoustic and Limp for impedance. I`m pretty busy in the last weeks but if I have more time I could help you with the crossover, should be a weekend, otherwise impossible for me. Beer comes from you

 

[SY]

Please bear in mind that the "optimization" and "experimentation" capabilities of the ordinary hobbyist are very limited especially when inductors are concerned.

What you're not listening to (but should) is that what matters is target acoustical transfer functions. The crossovers might look a bit odd and will generally NOT resemble 1950s textbook topologies. That's one of the reasons that modern hobbyist speakers are so much better. If you want anything even approaching acceptable results, you need to:

1. Measure the SPL response of the drivers in the baffle/box.
2. Measure the complex impedance of the drivers as a function of frequency.
3. Determine target acoustic transfer functions.
4. Using a decent CAD program (e.g., Soundeasy), load in the SPL and impedance data, then simulate different crossover topologies and run optimizations on them to try to get close to the targets.
5. Then build and measure the complete system to see how close you are to the target (generally, you'll be very close).
6. Tweak component values to hit the target more exactly or to adjust the sound to something closer to your ideal.

Now, all of this ignores other critical issues like polar pattern (which is a function of driver choices, baffle shape, and driver placement), but good CAD software will even let you simulate that as well- and these days, the sims are pretty accurate.

Yes, you will need tools. If you want to design and build automobile motors, you need milling machines and lathes; you can't get good results with chisels and sandpaper! The tools you need to do good speaker design aren't cheap, but they're no more expensive in total than a pair of good woofers. You'll also want to read and understand a few basic texts like "Loudspeaker Design Cookbook" (Dickason) and "Testing Loudspeakers" (d'Appolito) so that you get meaningful results from your measurements and simulations. A good digital crossover (e.g., the emulations in Soundeasy or a Behringer DCX2496) can be extremely helpful as well.

This is not an easy thing, but some diligence will pay great rewards- doing things right with cheap drivers will get you hugely better results than doing things poorly with expensive drivers.

 

[sreten]

Guys, you will have to elaborate harder on this.

Hi,

No. Having completely the wrong end of the stick is your
problem. Analyse the designs here : Zaph|Audio

rgds, sreten.

 

[tinitus]

What you're not listening to (but should) is that what matters is target acoustical transfer functions.

easy to say, but not knowing how it works maybe not so easy to understand

but you are right
there really are no shortcuts
it will be easier and faster to learn and use measurements, than trying to learn doing it without

 

[SY]

This is why I recommended the two books. If you want to design and build auto engines, you need to understand what torque, power, and friction mean and how they're measured. Shortcuts are a guarantee of lousy results.

 

[tvrgeek]

Pano, SY. sreten are telling you the truth. MEASURE. Use a simulation like PSD or Sound Easy. It will get you close. Then build a pair and listen, listen, listen. Check what you think you hear in the simulator. Looking at John's designs as sreten says is a very good start. Might also look at Siegfried's designs. (Linkwitz.com) I may differ in my opinion on what is perfect to my ears, but these folks know what the heck they are doing.

My electrical crossover is usually an order lower than my target. They are frequently asymmetrical, and frequently the centers are not the same and the Q varies HP and LP. Not by much though. I can usually do the BSC without additional parts, and almost never need a Zobel. If I think I need a notch for cone breakup, I move the crossover or get a better suited driver.

One word of caution with simulators. You can get some pretty wild Q's if you are just twiddling the values. I did a quick spreadsheet to be sure my predicted results don't go crazy.

 

[T101]

You are all offtopic.

The agenda in this thread is how to derive the load value (in Ohms) to be used for calculation of passive crossover components values.

I am not asking how to design speakers, nor how to design crossovers.

There is this self propelling myth about taking the impedance at the crossover frequency which I busted at least twice with evidence without anyone acknowledging it.

And what is this mantra with the "optimization" and the fine tuning? Normal even advanced hobbysts are not the R&D department of B&W.
Do you optimize active crossovers as well?
Yes, people can take measurements for reverse dip, polar response, on axis extension and etc.

But this all is offtopic.

When we consider crossovers, there are two main characteristics, frequency response and Q factor. Most people in this lot Mr. Linkwitz, choose crossover type for it's steepness and Q factor.
While the Q factor governs frequency response before the crossover frequency which is the so exploited by the "optimization" activists, the Q factor governs transient and impulse response as well. After FR optimization the Q is random which is unacceptable.

Further, the "measure, measure, measure!" is anti scientific and non DIY friendly approach. Not that I do not measure. I even do not measure myself, but rely on engineers with the appropriate equipment and training for measuring.

It is so obvious that it does not need such and comprehensive explanations.

Best regards!

P.S. I apply simulations fully supporting my statements, please be so kind, when you are opposing my conclusions to use some kind of evidence not only your non supported wisdom.

P.S.2. Question to the management: Does the forum have some kind of commercial relationship with Mr. Zaph, Sreten or another person or entity interested in the commercial success of Zaph Audio?

P.S.3. I didn't expect that in this place I would be referred to with such a slang as the one with the wrong side of the stick!

 

[SY]

You are all offtopic.

No. The OP wants advice on how to head down a path that is unlikely to produce good results. Trying to help him down that path is not doing him a favor.

The equipment and knowledge to do it right are inexpensive and attainable; this is not 1975. If one doesn't want to get into the nuts and bolts of the design and measurement process, published designs and kits are a better choice.

 

[Pano]

You are all offtopic.

Or you are past your depth? Those who have posted here (to help) have long and painful experience with crossovers. Nothing said so far is really advanced, just the basics of modern filter design. As SY says, this is not 1975 anymore. The tools are available. Many are free. Come into the 21st century.

 

[AllenB]

There is this self propelling myth about taking the impedance at the crossover frequency which I busted at least twice with evidence without anyone acknowledging it.

I'd say this sounds reasonable, but I couldn't answer the original question.

Do you optimize active crossovers as well?

Absolutely, and this is the difference. The driver impedance is an inconvenience.

 

[T101]

I'd say this sounds reasonable, but I couldn't answer the original question.


Absolutely, and this is the difference. The driver impedance is an inconvenience.

The original question does not have an answer. Unless the case is one of the exceptions, whatever impedance you take, the result will always be as one of my simulations.
It might be a trick question, but it is very often asked one and very often the answer is the one about the impedance at crossover frequency.

Obviously, there are two solutions to the problem, not the question.

The problem is that drivers with voice coils misbehave when combined with passive crossovers due to their regular impedance irregularities.

The first solution is to apply Zobel network and cure the impedance curve to a level that it can be considered to be a flat resistive load. Then we can choose the crossover Q factor for it's properties as well as the slope steepnes. Very easy, very fast and very effective.
This approach saves us from the dangers hidden in the "Me being worth at least as the whole B&W R&D!" and we can use our simulation software, Mathcad, online resources, tone generators, LC meters, B&K microphones, second hand Clio systems and etc. for learning, reverse null, polar response, front to back difference in an OB and etc.

The second solution is to defy laws of nature with optimization, to waste time in wiinding and unwinding inductances, to pretend we are acoustic engineers who get paid to build the perfect sound source, to alter Q factors of crossovers pretending we never saw the impulse response of a high Q crossover.

As for the diffraction taken into account, diffraction is a simple phenomenon, it has known constant sub 1st order slope, a peak and a 6 db plato towards treble. It can be addressed without BSC in a three way system or in a FAST. I am pretty sure I have several posts on the matter in this forum.
Diffraction is baffle and position dependent, not driver or crossover dependent.

But this is all offtopic because the thread is not for speaker designing, but for crossover values calculation.

You are representing the whole hobby as an elite activity reserved for people with extensive knowledge, great patience and lots of free time.
Nope, there should be clear and simple guidelines for the designing of every part of a speaker.
Someone mentioned engine design. Does the dyno stand help in the design process? - you are wrond, it does not! it is a supplementary proof tool. Dyno helps us to tune the already designed engine. Tuning is not equal to your so called optimization. In an engine, your optimisation would be equal to milling the camshaft profile after it is installed in the engine.

And you introduced all this offtopic just to refuse the discussion on the Q's and impedance.

Maybe I should stop visiting here and read Zaph's site daily without rising voice?!

 

[SY]

But this is all offtopic because the thread is not for speaker designing, but for crossover values calculation.

Crossover design is an integral element of speaker designing.

Yes, it takes some time and effort to learn how to design good speakers. No, it doesn't take a PhD and $10,000 worth of equipment. If one just wishes to build some speakers, there's plenty of well-engineered designs out there to copy. If one wishes to gain some knowledge (and incidentally learn why most of these objections to correct design are nonsense), speaker design, construction, and measurement can be a rewarding use of time and very mind-expanding.

 

[sreten]

Hi,

Seems pointless posting in a thread when the OP is in
love with his own answer to his own original question.

The original question does have an answer. Neither.
And there are no set classic electrical filter functions
that require Q calculations and known impedances.

There is nothing "elitist" about presenting the subject
as a case of simply "knowing what you doing" than
waffling on about half-baked approaches to the problem.

FRD Consortium tools guide

Designing Crossovers with Software Only

Doing it properly doesn't require extensive knowledge,
(though it helps with some of the details, e.g. see
Zaph|Audio and Zaph|Audio - ZDT3.5
for less than usual crossover design details.)

And far from taking extensive time, proper simulation
and then tweaking is by far the quickest method, if
you cannot measure. You need decent files though.

Having a good idea what is going on with a simulator
makes tweaking much easier, and prevents some of
the howlers people can make, like an impedance
curve that is far from what is a sensible curve.

If simulation is too complicated for you, don't think
some other "quicker" method will work, generally
it won't. Your better off building a good design.

What does take extensive experience is trying to
wing it with simple approaches where you must
have the right drivers for the job and a good
idea of what is going on without simulation.

rgds, sreten.