And outside the crossover point as well, witch is not negligible.
I prefer to use high slope crossovers, to minimize the range where the two sources interfere together (+- one octave with acoustic 24db/oct slope). Any phase/level error can be noticed.
Zobel, as well as motional impedance compensation makes the filter design much easier, as we said, and save a lot of time in curve response measurements and corrections. It gives a better chance to to DIYers without expensive acoustical measurement instruments.
Well, it brings a better coherency around the crossover frequency.
But it is beneficial for the upper transducer as well.
You will notice an apparent decrease in trebles. But they are those hifi trebles, not natural, that you never heard as such with acoutic real life musical instruments. A better coherency, making easier to separate dynamic details of some sounds, like hit of the stcks on a cymbal.
And this improvement is real, even if it change nothing but phases far outside of the tweeter bandwidth limit.
I use an horn, on my personal speaker system, and the cutoff of the driver is around 16Khz. I never succeed to add a super tweeter. It kills all the coherency.
Nobody, listening to my system never reported a lack of trebles, or details. On the contrary. On this driver, we worked to flat impedance up to > 40Khz. ANd every change around this frequency was noticeable.
A Zobel on its own has no effect on phase response. It tackles impedance variation by giving an alternate current path across a driver. Done correctly the total current is constant, it goes up through the Zobel as it goes down through the driver. In other words it flattens impedance. Still, current through the driver doesn't change and so phase response doesn't change.
Will it effect a crossover? Yes, if more network elements are put on the front side of it (Zobel and driver) then the altered impedance means the response of the added network will vary. That is, a Zobel (or any impedance correction) at the front of a multiway network and system can have no effect on phase or frequency response. If you impedance flatten drivers and then put additional network in front of them then it will have a slight effect as you have changed the load on the network.
Will an (internal) Zobel improve phase overlap? Sure, in some cases. In probably as many cases it will degrade it. There are too many variable (roll off points, roll off orders, driver depths, etc.) to ever generalize about this. By whatever amount it changes the individual section frequency response there will be a corresponding phase response. It is generally under first order so the phase correction has to be minimal.
David S.
David
Will it effect a crossover? Yes, if more network elements are put on the front side of it (Zobel and driver) then the altered impedance means the response of the added network will vary. That is, a Zobel (or any impedance correction) at the front of a multiway network and system can have no effect on phase or frequency response. If you impedance flatten drivers and then put additional network in front of them then it will have a slight effect as you have changed the load on the network.
Will an (internal) Zobel improve phase overlap? Sure, in some cases. In probably as many cases it will degrade it. There are too many variable (roll off points, roll off orders, driver depths, etc.) to ever generalize about this. By whatever amount it changes the individual section frequency response there will be a corresponding phase response. It is generally under first order so the phase correction has to be minimal.
David S.
David
Think we are mincing words here. It was implied that the Zobel affects everything down the line due to the impedance shift.
Have found that if the driver response is well matched to a natural rolloff on the top end with little peak (minor breakup) more times than not improves the situation, not degrade it. Now how this works with a driver with a peaky breakup is another story all together or if crossing over beyond this point in either direction ignoring off axis power response complicates the issue.
Have found that if the driver response is well matched to a natural rolloff on the top end with little peak (minor breakup) more times than not improves the situation, not degrade it. Now how this works with a driver with a peaky breakup is another story all together or if crossing over beyond this point in either direction ignoring off axis power response complicates the issue.
David, remember i'm talking in the situation where the driver is after a passive filter, and, of course, the zobel is paralleled to the driver.
Of course, Zobel will change the phase response when the impedance charging the filter (or the added zobel capacitance) change.
But there is an other point to consider. The amp side. The charge applied to an closed loop amplifier usually change its behavior during transients. (overshoot, phase margin etc.). And it can change-it in an non linear manner between small signals and large ones.
Of course, Zobel will change the phase response when the impedance charging the filter (or the added zobel capacitance) change.
But there is an other point to consider. The amp side. The charge applied to an closed loop amplifier usually change its behavior during transients. (overshoot, phase margin etc.). And it can change-it in an non linear manner between small signals and large ones.
+1
And now consider you are using a 3rd order network. The inductance of the driver simply* adds to the inductance of the second coil. So why bother with impedance flattening?
* I know that the driver is not a simple RL-Load, but that doesn't matter
I imagine you are talking of a low pass filter ?
Well, if you calculate them right from formulas, the good response curve is for a constant inpedance of the charge. In a 12db/oct filter, depending of the impedance curve of the driver and the cutoff frequency, you can simply increase the paralleled capacitance, but why bother for two capacitances and two resistances ?
The first thing i do when i got a new speaker is to determine their compensation networks. And consider them as a part of the speaker.
Then, i work on my passive filters. For industrial production, of course, if i can simplify and save some components, i do-it. But, i home, i don't bother and leave the impedance compensations components on board.
I will leave them on board even with my active filters. Will see if removing-them change something only by curiosity.
yes, a lot of people do that...and it is totally silly.
This notion that adding a Zobel is inherently good procedure is totally unfounded. In the end if you impedance flatten individual units it just becomes a minor response shaping excercise. It may help or it may hurt your final response.
If you ever create crossover network with software using an optimizer, it is instructional to properly Zobel the drivers then allow the optimizer to vary all components when chasing a target shape.
The stupid optimizer will always drag the Zobel components away from the impedance-flat values. That is, best response will seldom occur with "ideal" Zobel values. How do we interpret this?
Now I will frquently use what looks a bit like a Zobel circuit on a woofer. That is l in series, r and c to ground. Frequently I'll have a resistance value of 1 or 2 ohms in the shunt r. This is not a Zobel in that no attempt is made to make the impedance flat at a particular point. It helps to get a reasonable corner shape with high series inductance. The high inductance flattens the midrange. The point is that passive networks need to be optimized to achieve individual driver targets, plus the phase blend between drivers must be considered. Impedance flatness at some point in the circuit has absolutely no benefit and may be at odds with achieving good response.
For the case of designing a total system and then adding an additional network at the front to flatten total impedance, unless you have a bad amplifier, I don't see any benefit. You did promise us measurements on that???
David
No. I never added any compensation network *In front* of an enclosure. I use to add both motional and Zobel in front of my speakers *AFTER* the filters. Means paralleled to the speaker's voice coil.
Yes, of course a low pass filter.
I usually use a 3rd order electrical filters for their benefits: effectively reducing the bandwidth of the speaker, therefore increasing power handling, while having enough degrees of freedom to shape the response properly. So, obviously, I don't need a Zobel in those filters, because I use the Le of the speaker together with the second coil.
For a 2nd order electrical filter a Zobel can be useful, but as I said before, it was until now seldom necessary. I could nearly always get a linear response with near-to-perfect crossing.
I knew an engineer at McIntosh that liked to add sections to a crossover network like he was designing with a graphic equalizer. A little tank circuit notch, then another, each to whittle little dips into speaker frequency response. When I pointed out that he could replace these various sections with a simpler but well optimized topology he had no interest. He had found a way that worked for him and didn't care whether it was efficient or not.
I see that frequently in this forum. People swear by the procedure of "Zobel first, then crossover", or "add a bass shelving circuit, then crossover". They may have good success with the approach but they are not universal solutions!
You mention a scientific approach. There is only one scientific approach and it is as follows: divide the target acoustical response by the unfiltered response. This creates the electrical filter target. Determine the degrees of freedom and topology required to hit the filter target shape. Optimize to find the best circuit values to hit the target shape.
That is it.
The degrees of freedom concept is not well understood by most speaker designers. If you do a lot of automatic optimization with the more sophisticated programs it will eventually make sense. I have experience with the original KEF software, XOpt and the Speaker Shop suite. They all work similarly. It is instructional to set up a topology and see if the automatic optimizers converge on a solution (find network values that work). If your crossover order is too low then you will not converge on a solution, or will only converge over a limited frequency range. If your order is too high then you may find that some values "blow up", become quite large or quite smaller. The optimizer is telling you it doesn't need them.
In general the degrees of freedom (network order) can be equated to the number of shape inflection points needed: first order can only roll off, second order can bump then roll off, third order can sag then bump then roll off. Ultimate slope needed and inflection points needed will tell you what crossover order is required.
Where do Zobels work into that thinking? Generally they don't. They are an innefficient approach to crossover design and usually the common shapes can be achieved without resort to them. Many shapes can most efficiently be achieved with a standard ladder network topology.
A final thought on the scientific approach. If we brought in an outsider, say a mathematician quite experienced in theoretical filter design, I can't imagine them pointing us towards Zobel networks. They would fully understand the "required order" "degrees of freedom" approach and see Zobels or bass shelving as an unnecessary intermediate step.
My opinion at least.
This is one of the rare cases where I have used impedance flattening. In the Snell XA Reference the low crossover point interacted with the vented bass cabinet upper impedance hump. Working out a network of the desired order I was getting a response error from the crossover point being too close. I did put an impedance flattener over the upper bump and was able to achieve the right response because of it.
So I am not opposed to using the circuits when they seem to be required. I just wouldn't go out and preach their universal benefits afterwards. This was a specific case.
You started this thread with a claim that the majority of amps performed much better when faced with a flattened load. I would still be interested in some real data on that. Otherwise it is just "opnion masquerading as fact".
David
Last edited:
Who pretend it is an universal solution ?
I just said it helps, and, can bring some benefits i listed.
There is so many solutions, different approaches on how to design an enclosure...
Some prefer full range speakers, despite membrane's fractionning, some prefer 6db/oct filters, despite the diplet on a large band of frequencies... Some prefer active filters, despite some dynamic losses of coherencies...some passive filters, despite efficiency looses and less good damping etc.
Some prefer horns, others condensers speakers, soft domes etc... Each ones with their evils.
Some are interested in group delay, other in flat response curves, others in efficieny ...
When you published this damping material experience, did-i answered it was fake ? (Or even OOT ?). No i was just curious and interested to know better about YOUR experience.
About measurements, just refer, for the moment, to the Les menus du chapitre : Filtres site, until i have time to put in line my article. I have the same kind of results (of course).
And about the benefits a motional impedance correction can bring to your listening experience, how can-we talk of that if you don't listen to it ?
Last edited:
I opened this thread because we had, in an other, an out of topic discussion about i wanted to bring elsewhere.
I first explained the theoretical reasons why such networks can bring a simpler way for DIYers to build they filters, because filters formulas works with constant impedance..
And how motional impedance compensation can bring in the trasiant response of the basses loudspeakers and flatten filters for medium and treble units.
.
Then reported my listening experience with them to try to precise what you can expect of them.
And tried to explain what i suspect as the reasons for.
Gave a link where you yet have some measurements available, a tool to calculate Zobels and motional impedance networks in line and near identical listening reports. I precise i have no connection with the owner of this site.
As all my previous works on enclosures was not intended for publication, i have to recover my Bruel & Kjaer paper measurements, if they still exists, in the middle of tons of papers, then write a full article...Everybody can understand it will be kind of a work that i cannot do in less than one month, as it is not my only occupation and i'm not home for the moment...holidays till the end of August.
So, please, be patient.
In the middle, you can experiment by yourself, or, if you are fully satisfied with your enclosures, why even bother to read this topic ?
I first explained the theoretical reasons why such networks can bring a simpler way for DIYers to build they filters, because filters formulas works with constant impedance..
And how motional impedance compensation can bring in the trasiant response of the basses loudspeakers and flatten filters for medium and treble units.
.
Then reported my listening experience with them to try to precise what you can expect of them.
And tried to explain what i suspect as the reasons for.
Gave a link where you yet have some measurements available, a tool to calculate Zobels and motional impedance networks in line and near identical listening reports. I precise i have no connection with the owner of this site.
As all my previous works on enclosures was not intended for publication, i have to recover my Bruel & Kjaer paper measurements, if they still exists, in the middle of tons of papers, then write a full article...Everybody can understand it will be kind of a work that i cannot do in less than one month, as it is not my only occupation and i'm not home for the moment...holidays till the end of August.
So, please, be patient.
In the middle, you can experiment by yourself, or, if you are fully satisfied with your enclosures, why even bother to read this topic ?
Last edited:
This thread has taken a turn toward personal remarks. That is against the rules.I have removed the off-topic personal posts. Please discuss the topic, not each other - or this thread will be closed.
Don't Talk About It, Just Do It....
What I have not seen mentioned is that IME the improvement is not nearly so effective when installed away from the driver and at the crossover assembly.
IME experience, every amplifier markedly improves when driving a clean resistive load....just adding a quick and dirty (.47uF-8ohm) zobel directly across the speaker cabinet terminals yields a noticeable improvement.
Dan.
Hello Christphe, in my experience going to the time and expense of optimising zobel and 'motional' networks directly at the driver terminals is very well worthwhile the effort.
What I have not seen mentioned is that IME the improvement is not nearly so effective when installed away from the driver and at the crossover assembly.
There are many here who have every theory justification why a technique cannot work or is not necessary, but have never actually tried a mod such as this.
IME experience, every amplifier markedly improves when driving a clean resistive load....just adding a quick and dirty (.47uF-8ohm) zobel directly across the speaker cabinet terminals yields a noticeable improvement.
Dan.
This is a very usual problem in HiFi: adding something, hearing a change (frequently called "improvement") and saying that one single component now works better.
Do you have considered that such a Zobel can change the frequency response, depending on the output impedance of the amplifier and the inductance and resistance of the cable? That's simple physics, measure it, and report.
And do you have considered that you might just fooling yourself? Do you have made a real blind test?
I never tried this, Max.
My first feeling is we will lose one of the advantages that such network can bring, the flatten of the response curve due to the serial cable impedance making a bridge with the variable impedance of the enclosure. It would be interesting to compare square waves, in order to figure out better what's happens, amplifier side.
Because this thread was turning too controversial, i want to precise my points.
Did such networks makes a difference you can easily hear ? Yes.
Are they measurable yes.
Are the changes drastic ? No. It is only interesting with high end equipment.
Dit-it goes against laws of physic (V=RI) ? No.
Did they always improve existing enclosures ? No, it depend of them, It is obvious that, if you miss basses, as an example, the motional impedance compensation will reduce bass-medium levels a little more. And the added damping advantage can be negligible in front of this loses.
Dit-it improve the quality when the filters are designed with the compensations in place ? Yes, always in my listening experience. As i cant make-you listen and constat by yourself in a forum, you have to try to make your own opinion.
Did-it help to calculate the filters using traditional formulas, and get the expected levels at the out of them ? Yes.
Nothing more, nothing that justify too much controversy.
OK, now let's be even more precise on these two points. What exactly did you do?
- have you designed a the same speaker with and without the compensation network?
- have you matched both very carefully to have the same frequency response on- and out-of-axis?
- have you compared them in a real double-blind test?
See, if I design a speaker, and decide to use a Zobel or something, than it will surely sound a little different from the same speaker without the Zobel, because I will never get the same (literally!) frequency response. I will always have slight differences, and depending on the crossover point and the hearing thresholds at that frequency range, this differences might be audible.
But is it an improvement? Isn't it just a small change?
a subject I am trying to learn about, seems there are only a few drive units designed not to need correction circuits-
for example, I'm building a 2 way using aluminium units -
The 2" TB W2-800s I'm using above 750hz, they are rated 4ohms but have an impedance spike of 90 ohms! So I need a notch or 2nd order at least.
Compare with the 2.5" Bandor 50's (a full range designed specifically for a 2 way 1st order), the 8 ohm version has an impedance 'bump' of 16 ohms.. the 4 ohm hits just 10. No circuit necessary just a single cap, saving time and money
for example, I'm building a 2 way using aluminium units -
The 2" TB W2-800s I'm using above 750hz, they are rated 4ohms but have an impedance spike of 90 ohms! So I need a notch or 2nd order at least.
Compare with the 2.5" Bandor 50's (a full range designed specifically for a 2 way 1st order), the 8 ohm version has an impedance 'bump' of 16 ohms.. the 4 ohm hits just 10. No circuit necessary just a single cap, saving time and money
Last edited:
1- Of course yes, several.
2- I think on vs out axis responses are speaker (+box+room) dependent, not source or filters. When i design enclosures, total energy is one of my main preoccupation. The way i decides of the cutting frequencies and diameter of my two way speakers was in order to get the same radiation surface at the cutoff frequency (or the same space response,if you prefer).
3- Who have time to make 'blind tests' at each step when you design ? What can i say ? I'm very suspicious about my hearing experiences. Sometimes, you can fool yourself or turn crazy when you try to figure out subtle changes.
I use often a 'non concentrated' way to listen. (sometimes from an other room 🙂. And i don't care about suble changes you are not able to figure out immediately and repetitively.
4- I tried to answer this question. Difficult, as it is psycho acoustic. I'm very sensible to get as much separated instruments as i can, and high dynamic response. I hate false treeble, you know, this "tssstsss" produced by much of soft domes in multy-ways enclosures. As a sound engineer, i try to get instruments as realist as possible. I want cymbals as they are in real life: the hit of the stick, the weight of the metal. Did they produce trebles ? You never feel this way with real cymbals. See what i mean ?
This will be purely subjective, and you'll have to make your own opinion listening by yourself
About Zobel, let take an practical example i know very well: my own enclosures.
I use a 1 inch JBL Driver charged by a wooden spherical wave horn (close of the JMLC ones, a previous François Delamare design).
It gives an ascending response curve i had to correct. and i needed a 24db/oct filter, Butterworth. Done in a first time without Zobel and motional impedance compensation. The best i could.
Then, because i was not fully satisfied, i had redone the work with the two networks.
The difference is this one. no more harshness or "horn" aggressive response, better coherency and natural. Real improvement ? For me, yes. For you ? How can i know.
What i'm sure of is the difference is noticeable by everybody. Less teebles, more natural sound. I precise the acoustic roll of of the driver is around 16KHz.
On my bass reflex side, motional impedance compensation network brought suppression of lot of Boom-boom effects (resonances). Obviously, you can separate better kick drums and basses. Kick drums are more natural, "faster", lighter. You can feel the hit of the kick on the skin. They all sound different in various records, while they were similar before.
Real Improvement ? Again, it is purely subjective. My response is i will never build an enclosure without any more. This time, as it do not interfer with filters, it is easy to compare with and without. It was Obvious to everybody i made the demonstration, and for hundred of clients of Francois Delamare at the "Maison du haut parleur". And Francois is not a snake oil vendor, but a very clever designer, objectivist side.
It is worth to try-it, on my opinion, reason why i opened this thread. I don't understand why it is so controversial. Nothing to compare with some Bybee alien filter 🙂
Last edited:
Just Do It.....
I got past the 'expectation' factor decades ago...when I make a change, I measure it for interest and confirmation reasons, but the real proof is in the listening, short term and long term, in the sweet spot, from another room, from outside where there are natural sounds to reference to.....etc.
If it ain't musically 'right', that is obvious quickly, through experience.
Extended time and wide variety of music and levels confirms if a construction or modification is totally 'right' or not.
The reason for adding the likes of zobel networks is indeed to change the frequency response, duh !, both static but more particularly dynamic behaviours.
I suggest you try impedance correcting a loudspeaker and report back....the experience won't hurt you 😉.
Dan.
Mr BBB, I have been dealing with audio repairs, recording, live mixing, experimenting and musical enjoyment for over 35 years.....my life's work.
I got past the 'expectation' factor decades ago...when I make a change, I measure it for interest and confirmation reasons, but the real proof is in the listening, short term and long term, in the sweet spot, from another room, from outside where there are natural sounds to reference to.....etc.
If it ain't musically 'right', that is obvious quickly, through experience.
Extended time and wide variety of music and levels confirms if a construction or modification is totally 'right' or not.
The reason for adding the likes of zobel networks is indeed to change the frequency response, duh !, both static but more particularly dynamic behaviours.
I suggest you try impedance correcting a loudspeaker and report back....the experience won't hurt you 😉.
Dan.
- Status
- Not open for further replies.