In this blog post we look at implementing a passive high pass filter for the stock Fostex T900A. Since featuring the tweeter in a recent blog, I was asked to develop a proper high pass filter for the super tweeter.
The goal is a 12dB/octave high pass slope at 7kHz to mate up with the midrange horn.
First we will look at what Fostex recommends for this tweeter and see how that does in terms of frequency response. Fostex call for a 1.00uF capacitor and 0.15uF inductor for a 2nd order filter.
Below is my measured frequency response for the recommended high pass filter with blue showing the tweeter's raw response and green showing with the Fostex high pass filter.
We can see that this did nothing to deal with the peak centered at 5.7kHz.
If we look at a CSD plot of the resulting frequency response we can see that the peak is a result of stored energy.
What we are after is a textbook 12dB/octave 7kHz high pass filter slope which I've hand drawn in red below.
To do this we will need to implement a conjugate filter to flatten the peak in the impedance curve. This will create predictability in our high pass filter design.
Below is the stock impedance sweep for the tweeter. Notice the severe peak centered at 5.8kHz. If this is not dealt with it will be very difficult to create our target slope.
Throwing the test data into Vituix I was able to simulate a conjugate consisting of the following schematic.
Here is an overlay on the predicted impedance and phase response. The dashed lines are the non-conjugate responses for phase and impedance.
Building and testing the conjugate with the DATS speaker tester is shown below.
I'm probably a little aggressive with my resistor value as we can see that it dips to 4 ohm, but let's continue with the high pass filter elements and see what happens.
Below is the Fostex Suggested High Pass Crossover (green) with the conjugate shown in red.
We can see the conjugate helps in the 5kHz region but it is still not enough. We will have to look at other options.
At this point I decided to try a variety of schematic configurations within the Vituix simulation software. I adjusted the values of each component until I came up with something resembling my target response. I ended up on a fourth order configuration along with a resistor to ground to shape the steepness.
The simulation shows a gradual slope (brown) against the raw response (dashed blue).
I then rigged up the crossover and measured the response.
This is much more agreeable! Subjectively the tweeter took on a much sweeter sound.
Comparing against the Fostex suggested schematic is shown below on the left.
The new impedance sweep is shown below.
Conclusion
The Fostex T900A presents some challenges for a proper high pass filter implementation. Even after attempting a conjugate filter to flatten the impedance curve we were not able to achieve our target response. In the end, I found a solution that involves a rather unorthodox 4th order configuration. The specific component values are not even close to anything textbook calculated, but seems to work in shaping the response. For example a textbook calculated 4th order would have C1 at half the value of C2, however in my design C1 is actually larger!
I was told by a few members that they ultimately sold their T900A after unsatisfactory results. I can see why if one relies on the suggested schematic provided. The 2252 schematic certainly improved the the sound quality to the point where I was very happy with what I was hearing. I expect to get more listening time on the modified version using the 2258 horn lens.