Tweeter passive XO based on a transformer
Posted 1st May 2017 at 01:04 PM by abraxalito
As avid readers of my blog over the years may well know, I've modded quite a few active speakers. Mainly Swan/Hivi D1010s and D1080s of various generations along with some 3Nods. The entry level Swans are supercheap on Taobao nowadays with pairs of older generation D1010s going for about 375rmb.
As a result of all these mods and my learning by doing I haven't got a single pair of actives in working condition, primarily because the mods I currently want to do (including fitting transformers inside and swapping out the classAB chipamps on bass/mid to replace with classD) won't fit in the available space. All the amp and active XO parts have been hacked about almost beyond recognition and are of very dubious reliability. So I have come up with a plan  I've thrown all the electronics boards away and will convert the carcasses into passive speakers. Since I also prefer to feed speakers via stepdown transformers for better dynamics, I wanted to explore the possibilities of using transformers to replace inductors in crossovers. I figured if the transformer could do double duty I'd significantly reduce the parts count and BOM cost.
A typical 2nd order tweeter XO has a series cap and shunt inductor  at first sight it seemed at least worth considering whether the transformer's intrinsic shunt inductance could stand in for the shunt inductor in the XO. Normally a transformer is designed to maximize the shunt inductance so as not to load the source too much at LF, here I wanted to hit a target shunt inductance which is given to me by the online XO calculator.
Go here : Passive Crossover Design Equations Formulas Calculator  Two Way Second Order Network Chebychev Bessel Butterworth LinkwitzRiley and you'll find for a 3.5kHz crossover frequency the shunt inductance needs to be 820uH when the tweeter's 18ohms. This for a Chebyshevstyle highpass filter. I chose 18ohms as the target impedance as 18ohms roughly corresponds to the lowest impedance a TDA8932 can drive continuously at 35V without overheating. TDA8932 is my workhorse amp of the moment, delivering enough watts (1020W) for desktop use for a pittance.
With 820uH as a target shunt inductance for our tweeter stepdown transformer, how do we go about reaching this value then? First I settled on a core type for this application  PQ2020. Then I calculated the minimum number of turns needed for the transformer not to saturate given a 3dB level signal at the XO frequency (3500Hz). From the transformer design equations this requires 72 turns. Now we have what we need to calculate the necessary AL value for our core = (820/(72*72))  which very oddly, or perhaps by obscure physical means gives a very standard 160nH/rtturn value. The fact we've hit a standard AL value means we can read the necessary gap width right from the Ferroxcube databook (0.6mm). Reaching this spacing took 4 layers of some doublesided sticky tape  in fact I only need to have 0.3mm spacing because my crude method introduces a gap in both the centre and outer legs.
From plotting the frequency response manually with my siggen and AC voltmeter, it was clear from the peaking that I was in need of a damping network on the primary side. RC duly added. The first prototype trafo that I installed in the D1010 MkIVs made the tweeter too bright (significantly more efficient than the partnering 4inch bass/mid) and had a turns ratio about 2.1:1 so with the 2nd prototype I went for a 3:1 step down. Subjectively the tweeter's sounding excellent, much cleaner than it ever did with the inbuilt chipamp feeding it.
In a subsequent post I might talk about how I included a transformer in the lowpass XO for the bass/mid.
As a result of all these mods and my learning by doing I haven't got a single pair of actives in working condition, primarily because the mods I currently want to do (including fitting transformers inside and swapping out the classAB chipamps on bass/mid to replace with classD) won't fit in the available space. All the amp and active XO parts have been hacked about almost beyond recognition and are of very dubious reliability. So I have come up with a plan  I've thrown all the electronics boards away and will convert the carcasses into passive speakers. Since I also prefer to feed speakers via stepdown transformers for better dynamics, I wanted to explore the possibilities of using transformers to replace inductors in crossovers. I figured if the transformer could do double duty I'd significantly reduce the parts count and BOM cost.
A typical 2nd order tweeter XO has a series cap and shunt inductor  at first sight it seemed at least worth considering whether the transformer's intrinsic shunt inductance could stand in for the shunt inductor in the XO. Normally a transformer is designed to maximize the shunt inductance so as not to load the source too much at LF, here I wanted to hit a target shunt inductance which is given to me by the online XO calculator.
Go here : Passive Crossover Design Equations Formulas Calculator  Two Way Second Order Network Chebychev Bessel Butterworth LinkwitzRiley and you'll find for a 3.5kHz crossover frequency the shunt inductance needs to be 820uH when the tweeter's 18ohms. This for a Chebyshevstyle highpass filter. I chose 18ohms as the target impedance as 18ohms roughly corresponds to the lowest impedance a TDA8932 can drive continuously at 35V without overheating. TDA8932 is my workhorse amp of the moment, delivering enough watts (1020W) for desktop use for a pittance.
With 820uH as a target shunt inductance for our tweeter stepdown transformer, how do we go about reaching this value then? First I settled on a core type for this application  PQ2020. Then I calculated the minimum number of turns needed for the transformer not to saturate given a 3dB level signal at the XO frequency (3500Hz). From the transformer design equations this requires 72 turns. Now we have what we need to calculate the necessary AL value for our core = (820/(72*72))  which very oddly, or perhaps by obscure physical means gives a very standard 160nH/rtturn value. The fact we've hit a standard AL value means we can read the necessary gap width right from the Ferroxcube databook (0.6mm). Reaching this spacing took 4 layers of some doublesided sticky tape  in fact I only need to have 0.3mm spacing because my crude method introduces a gap in both the centre and outer legs.
From plotting the frequency response manually with my siggen and AC voltmeter, it was clear from the peaking that I was in need of a damping network on the primary side. RC duly added. The first prototype trafo that I installed in the D1010 MkIVs made the tweeter too bright (significantly more efficient than the partnering 4inch bass/mid) and had a turns ratio about 2.1:1 so with the 2nd prototype I went for a 3:1 step down. Subjectively the tweeter's sounding excellent, much cleaner than it ever did with the inbuilt chipamp feeding it.
In a subsequent post I might talk about how I included a transformer in the lowpass XO for the bass/mid.
Total Comments 3
Comments

Well that is a different approach!
I remember seeing auto transformers used in some old XO's. That is essentially what you are doing here but instead running the full power through the core.
You might want to look at the auto transformers approach  less flux and likely less distortion as a consequence.
Nice off the wall idea.Posted 4th May 2017 at 03:28 AM by googlyone 
Thanks for the feedback! From what I recall about autoformers, they save a lot of copper when the turns ratio is close to unity but have smaller and smaller advantages moving away from that. Here as I'm going for at least 2:1 and probably 3:1 the advantage is fairly small. But might be worth considering if they show improvements in leakage inductance.
Posted 4th May 2017 at 11:01 AM by abraxalito 
Great read, thanks!
Looking forward to the next episodePosted 21st May 2017 at 12:47 AM by Alexandre