Ping Calvin

[moray james]

Calvin: thanks for posting all the great information on the step down toroids over at diyaudio. That is great information and I will cross post it over at diyhifi.org as well as at gainclone.com.
I am wondering if you could help with some information on basic crossovers? I have very little knowledge on crossovers and was wondering how to go about building a network for a small ESL super tweeter? I had thought about building a tweeter with perhaps 1/32 of an inch spacers and some very thin PET on the order of 2 microns. I would like to have a two or three position crossover as well as some means of level adjustment. I had thought that first order would probably sound best but am not sure if 6 db per octave would be a fast enough slope. Seems that there is a big interest in ribbon super tweeters that extend out toward 100K and I wonder what kind of step up transformers would be required to get out that far? Perhaps a leaky transformer would boost the top end to good advantage. This would be a good project to post as I am sure there would be some interest. Thanks for any suggestions. Regards Moray james.PS this link shows the kind of thing I had in mind http://www.kingsaudio.com.hk/product.asp

[Calvin]

Hi,

well, the behavour of a crossover is not easily predicted, because of the great changes in input impedance of the speaker itself. Most of the time You´ll have to linearize the impedance first. This is done simply and sufficiently with R1/R2 (resp. R8/R7 in second circuit). The parallel-resistor linearizes the impedance and the smaller serial-resistor eases the pain for the amp at the upper end as well as corrects the frequency response (it lowers the highest highs. Remember that measured in the nearfield the response should be tilted down towards the upper end.) The effect is best seen in the impedance-, the netgain and freq-resp-plots.

Diagrams:
page 1: Panel only
page 2: Panel + Rs
..
...
....

[Calvin]

...
....

Adding a capacitor gives the 1st order HP-response electrically. You can see it in the impedance- and netgain-plot. Acoustically there is nearly no effect at all, because of the huge resonance at Fs.
You have to kill this with a Notch (R3L1C2).
This gives You a clear HP-response (acoustically 3rd-order HP!!). The effect can be seen in all diagrams. The Notch adittionally rises the freq-response above its center-frequency, thereby correcting the loss of SPL due to phase cancellation. By fudging with the Notch´s resistor You can tilt the slope of the curve a bit.


Diagrams:
page 1: Panel + Rs + C
page 2: Panel + Rs + C + Notch

..
...

[Calvin]

..
...

Adding a further Inductor + Resistor (R4/L2, the parallel part in a 2nd order Filter) rises the crossover frequency and lowers the slope to 2nd-order (as wished).

All effects can be seen in the snapshot diagram

Diagrams:
page 1: Panel + Rs + C + Notch + RL
page 2: Snapshots of various Filters

..
...

[Calvin]

..
...

when You take the resonance of the frequency response of the panel into Your calculations, You may design a 2nd-order-filter with lesser part count. You design the filter with quite some overshoot and with an fs thats positioned right at the dip in frequency-response between the panels fs and midrange. In the example the panels fs is at 180Hz, the filter´s fs at app. 350Hz. This way You achieve a good efficiency and a good frequency response, but keep in mind, that at resonance the panel exhibits quite a long decay. Too distortion will rise substantially. Even though it looks fine (fine enough at least for ML, who do it this way), it only works ok with panels having a high fs, i.e that´s diaphragm is stretched very very hard. You can correct for a linear response with the resistor (R5) in the parallel part of the crossover.
You can see the differences on the second page.

diagrams:
page 1: Panel + Rs + dampened 2nd-order-filter
page 2: Snapshots for different Filters

..
...

[Calvin]

..
...

at last the circuits.

A good simulator-program should be at hand to help estimate the values and topologies of possible filters.

jauu
Calvin

[dstockwell]

wow You got to be a rocket scientist to understand all this.

[moray james]

Jean-Luc Picard would just say something like "make it so" or "engage". I am sure it will take me longer than that to realize but I will post results so others can give it a go if they are interested. Thank you Calvin! Best regards Moray James.