Crossover nightmare!!!!!!!

[tinitus]

Originally Posted by alan-1-b

Hi tinitus,

Is it the 2.2mH inductor in electrical Parallel with the Series network of 72uF + 72uF = 2.2mH//36uF resonant circuit -{a Tank circuit, the old Radio fellows used to refer to such as}- that is worrying you ?

Lets not go there again
He is happy with things as they are
Why not leave it at that

But the 2x 72uf are certainly NOT in paralel as they are seperated by series inductor
I have seen it done that way a long time ago
I tried it too
Didnt work too good
Two equally sized caps on either side of an inductor is just some fancy idea
Is ESR really 1ohm on bipolars?

I use relatively big "shunt-series" resistors on ALL paralel component, caps AND inductors, so it really doesnt bother me
Cant bare to listen to it any other way
Tricky to make work properly, but WHEN it works its good

But please, lets not spoil the otherwise happy ending
Maybe just wait fore the next, "inductor-nightmare"
cheers

[alan-1-b]

Correction :-

I should not have used the term Lvc for the effective driver inductance in my last Post, because Lvc is the term used for the Inductance of the voice-coil.
That inductance is only relevant when the driver's cone is not moving.
As soon as cone motion commences all the mechanical and electrical Impedances in the complete Magnetic + Electrical + Mechanical system that comprises the driver add, and in accord with their respective Phase relationships to each other.
Similarly, the effective Resistance is no longer the DCR of the driver only, but is a sum of all the relevant resistances.
Both the effective Resistance and the effective Inductance vary across the frequency bandwidth of the driver.

I will refer to the driver's effective inductance at a particular frequency as Lde.

I will refer to the driver's effective resistance at a particular frequency as Rde.

Thus the second Tank resonance is caused by Lde//72uF and is reduced in onset and magnitude by Rde + the ESR of the 72uF cap.

Originally Posted by tinitus

Pardon me if wrong, but are there not TWO equally sized value 72uf in paralel on woofer ?
Its pretty daft
If it at least dont ruin response too much, it certainly will screw up phase
I dont understand how they left it like that
I bet almost anything else works better
But at the time I believe it was one of those slick ideas some believed in

Man, there have even been very skilled guys who made 2ways 12db with equally sized mirror imaged reversed values on woofer and tweeter, or whatever we should call it

All multi-component cross-overs cause a Resonance, at at least one frequency, as result of the different Time Constants of each of their electrical paths combining in their different respective Phase Angles to each other.


"screw up phase":-

If the driver was a fixed constant Resistive load and had a flat frequency response then that bass filter would "screw up phase".
The driver actually presents a non-constant Impedance load {to the filter} of Rde + Lde + Cde
-{Cde is a mechanically caused Capacitive effect from various resonances in the cone, surround, frame, etc ... and is thus part of the Impedance of the driver system, and is detectable in perturbations in the measured driver's impedance plot}.
The Phase Response of the system is the sum of the phase relationships of each of the components in the bass filter plus the phase response of the Impedance load of the driver and of its frequency response.
My guess is that Celestion's engineers chose the closest to the optimum capacitace from the available cost-effective capacitors, then calculated the inductance values to part-compensate for what-ever residual small capacitive errors, and for all to compensate for the varying driver Impedance load and the non-flat frequency response of the driver through the cross over region of about 250Hz <---> 1kHz -{this allows for at least one Octave on each side of the cross-over design frequency of 500Hz}.
Why am I guessing that ?
Because a lot of listeners still like the sound from these speakers - in preference to a lot of other designs from that era - and after many years of further developments and new designs - thus it seems that design, and a few others, was got sufficiently close to optimum to reproduce sufficient of the recordings of the types that those listeners like.


"slick ideas":-

Indeed there were some "slick ideas" at that time, and there are some now, and likely there will always be some ... at least till we as a species succeed in annihilating ourselves ... as result of application of some of our slick ideas.
"Its pretty daft" !

[alan-1-b]

Originally Posted by tinitus

Is ESR really 1ohm on bipolars?


cheers

The ESR can be calculated from the Dissipation Factor of the capacitor.

For all capacitors, Dissipation Factor varies at least a very little with the applied frequency,
and with Electrolytic caps it varies substantially !

With typical Bi-polar caps of that period the Dissipation Factor was approximately 0.25 across the upper bass <---> lower midrange, and rose to approximately 0.5 through the upper midrange and into the lower treble region of the audio spectrum.
At higher frequencies it could be anything !

Different types of electrolytic caps were then {and still are and to better degrees now} designed so far as possible for their Dissipation Factors to be lowest in the areas of the electrical bandwidth they were primarily intended to be applied to.

The 1ohm I stated refers only to 72uF and at one particular frequency within that bass to low mids crossover region, but the deviation from 1ohm around the general 500Hz region will be small.

************************************************** **********

There will be replies to others' Posts next time - I have to go now.

[alan-1-b]

Originally Posted by wintermute

Hi Alan, I've only played around with impedance compensation on my current project, and didn't care for it that much. Sorry if my post came across as supporting the idea of putting in impedance compensation, that was not the intent.. The intent was simply to state that adding it would more than likely change the sound, and possibly also necessitate making other crossover changes. Something I didn't think Lucas really wanted to do.

Tony.

Hi Tony,

The problem with a number of the Academics' thinking amongst loudspeaker designers is they are considering the driver as a passive device which receives, transduces, and then emits the audio signal -{even thoughmost of them know it isn't purely "passive"}.

The driver is Active -
(1) - parts of it resonate, and those resonances are coupled back via the magnetic system to the voice-coil and into the cross-over.
(2) - the cone is a diaphragm connected to a coil that moves in the gap of a magnetic, and that is basically a Microphone circuit.
Even though a speaker driver is a fairly inefficient microphone, it is still sufficiently a microphone to pick up the sound in the room and feed it back into the cross-over -as in (1) above.
{Next all from (2) and (1) arrive back at the amplifier and get into it's Feedback Loop, but that is another topic, and not entirely for here now.}

If one designs a passive x-over to work into a fixed resistive load, and designs a passive impedance compensation network to connect between the variable impedance load of the driver and the x-over filter, one has to accept that the Active driver will drive that impedance compensation network, and via it will drive the x-over filter.
When all that is occurring, there are a lot of resonances arising, caused by the interactions of the passive components with each other, in a similar manner to those I described in #112 and an earlier Post above,
AND all those interfere with the signal from the amplifier, AND the resultant mess of all that is fed to both the speaker driver, AND to the amplifier via its feedback loop.
Thus I suspect this is what you may be hearing with some of your circuit experiments.

How does one minimise all this from interfereing with the the signal ?
Use a simpler circuit, because it will resonate less, and it will have less Time Constants in it if less signal paths through it.
{Time Constants cause the different action times plus the different frequencies which resonate or cancel each other in part, depending on the circuit. That is a Simplification - study more about Time Constants elsewhere.}

What I prefer to do is combine the Filters with the Impedance Compensation into a single circuit rather than use two separate circuits, and thereby usually I can use less components and have fewer signal paths,
thus fewer problems ... but there are compromises which have to be made ... as always, so that aspect is not a problem, and I state it only to say that what I and some other designers do does not solve all problems associated with trying to get drivers to work with passive crossovers.

[spd]

Hi, From the photo it looks like the black wires from the mid and bass drivers are not
connected to -ve input from the amp. Or is this just the way it looks on the photo. If so this would certainly cause a problem

[LucasAdamson]

Originally Posted by spd

Hi, From the photo it looks like the black wires from the mid and bass drivers are not
connected to -ve input from the amp. Or is this just the way it looks on the photo. If so this would certainly cause a problem

Thanks. No, they're fine. The "nightmare" was resolved a long way back. It was a total short circuit caused by the aluminium plate into which the speaker terminals fitted. Doh!

[grantnsw]

Hi all, (sorry I couldn't resist!)

I hope you don't mind me 'chiming-in'. Please correct me if I'm wrong on any of this. I'm still learning, of course, and don't want to step-on-anyone's-toes!

But to me, that 40? year old crossover no doubt worked beautifully, FOR ITS DAY!

It has no tweeter or mid attenuation, so the drivers must be well SPL matched to each other, and/or the xo points have been deftly aligned to equalize response.

It's 3rd order electrical on the tweeter, 2nd order bandpass on mid and effectively a 4th order electrical cascaded Butterworth on the woofer.

The matching 12dB per octave cascaded sections on the woofer yield 24dB/Oct
electrical roll-off, which again, ...correct me if I'm wrong... was advanced in those days, being prior to Linkwitz-Riley and other modern topologies. But hey, it still worked well back then.

However, that old xo, no doubt CAN be greatly improved upon by modern techniques, and by examining each drivers polar responses, by measurement if you have the 'kit'.

The old cascaded C and L values seem to have been computer optimised since, see the 4th order here:

2-Way Crossover Design / Calculator Help

so the original (and identical) 72uF caps and coils on each branch are no longer optimal, just my 2cents. I won't get into the cap quality debate, but imho, its all irrelevant if the xo is sub-optimal.

If I'm way off track...please let me know! cheers , Grant

[alan-1-b]

Hi Grant ,

"chiming-in" - as long as that is not a resonance occurring in your tweeters you may be lucky !

Given the frequency responses of the specific drivers used in the 44,
and together with the cross-over frequencies chosen,
the final roll-off slopes for the drivers are almost certainly as follows:-
18dB/octave hi-pass for the tweeter, all from the electrical filter.
18dB/octave lo-pass upper midrange, from summation of the 12dB/oct. electrical filter + the 6dB/oct. natural roll-off of the upper mids of the driver.
24dB/octave hi-pass lower midrange, from the summation of the 12dB/oct. electrical filter plus the in-back-enclosure shaped 12dB/oct. low frequency roll-off.
24dB/octave lo-pass for the woofer, from that specifically designed, cascaded electrical filter combining with the electrical impedance of, and the non-flat frequency response of, that woofer.

See my Post #91 on Page 10 which includes more about that.


Linkwitz-Riley, and other people, published details for types of filters which a few of the previous eras' designers may have discovered after experimentation, and some of which would not have been able to be published because it would have been considered as the intellectual property of the Companies that employed those designers.
That does NOT in any way diminish the achievements of Linkwitz & Riley, etc ... but it does include that there was a lot known to only a few people,
and such still occurs in the current era ... some designers do not give away their secrets !

See also my Post #114 on Page 12 for more about the likely reason for the two 72uF caps.

Keep in mind that this is NOT a 4th Order filter designed for a fixed resistive load - it is a complex filter designed to include frequency response and impedance variations across the relevant bandwidth of the load.
It is likely very close to "optimum" !
The second 72uF cap, including its ESR, is a combination of a smaller cap in electrical parallel with another smaller cap + small resistor impedance compensation network.
Don't only look at identical digits when evaluating circuits - instead, consider the entire circuit, and what is interacting with what-else in it - as I have explained in a previous Post referred to above.

************************************************** ***

Please, new readers, start from the beginning of this Thread so that you can read why certain things have been stated in the later Posts,
because otherwise it simply causes confusion to less knowledgeable readers who find new later Posts stating things that have already been explained.

************************************************** ***

That Calculator in the Link for cross-over design will NOT calculate optimum values for anything except fixed constant resistive loads !

[AndrewT]

Originally Posted by alan-1-b

That Calculator in the Link for cross-over design will NOT calculate optimum values for anything except fixed constant resistive loads !

and that is not what a driver does in loading a crossover.
The crossover components must be adjusted and added to, to correct for the changing impedance of the driver.

[alan-1-b]

Originally Posted by LucasAdamson

The only way to suspend the resistors in air on all surfaces is to place them close to the drivers. Is that ok?

If the leads are the thick, which most are for physically large wire-wound resistors, they will be strong enough to hold the resistor's weight with no risk of electrical fault occuring when the resistors are hanging in air between the drivers and the crossover board - which it seems is what you are describing here ... ?

But I am wondering how you have the circuit correct for all of the resistors positions ?

The resistor for the 33uF mid's cap has to electrically part of that cap only, thus it can be connected in Series between the input to the crossover and the 33uF cap, or after the 33uF cap in Series to the junction between the .34mH coil and 6uF mid's cap on the board.
There is nowhere for this resistor to be directly connected to the mid-driver.

Similarly, the resistor to the 6uF mid's cap has to be in direct Series with it,
either at the end where it meets the .34mH coil or at the other end where it connects to Ground and that can be to the negative terminal of the mid-driver, but still there must be the direct to Ground wire from that driver terminal back to the board or to the -ive connection at the back of the cabinet.

The 0.5ohm resistor for the treble can be connected between the tweeter and the 6uF treble cap output from the crossover board.

Some experienced people will say its better to not hang resistors suspended by by wires in mid space in the enclosure, but if nothing nearby can get caught on the wires or the resistor I doubt there will be any problem, and these resistors are not so heavy as to break the wires when the cabinet is moved around, unless very thin strands in the connecting cable in the cabinet.