Fostex T900A Bullet Style Super Tweeter

In this post I look at the T900A bullet style super tweeter from Fostex. This "Laboratory Series" is made in Japan unlike lesser models. The tweeter uses an alnico magnet along with a magnesium ring radiator diaphragm. Other than that, there's very little information online for this particular model.






IMG_7217_480x480.jpg
IMG_7216_480x480.jpg


The front of the tweeter is made from solid brass which has been lathe turned and polished.

Disassembling the tweeter reveals clock work precision internals.









IMG_7290_480x480.jpg








Measurements​

I began by measuring the impedance sweep. The Fs is actually quite low at 1885Hz with a second peak at 5685Hz.



Frequency Response

The response shows a large peak centered around the driver's Fs however this is below the usable frequency range of the driver. The response is flat from 7kHz-10kHz with a peculiar +4dB HF shelf. Response is then flat again until 21kHz where we see a mild breakup mode in the diaphragm. (click to enlarge below)



If we compare this against published we do not see the HF shelf at 10kHz. We also do not see extension to 30kHz with my test data.

Screenshot_2023-09-01_222333_480x480.png


The off-axis at 0,15,30, and 45 degrees off-axis shows a considerably wide coverage centered around 7kHz but quickly narrows above 11kHz. By 15kHz the listening window has narrowed to only 60 degrees.



Burst decay shows a mild resonance at 21kHz, otherwise we are trouble free.





The CSD plot shows the typical stored energy at Fs (4.8kHz). The tweeter remains clean above this with a very fast decay time across it's usable bandwidth.



Distortion​

Looking at harmonic for the 85dB 1m test signal, we see H2 dominate however still moderate in level at only 0.42% at 10kHz. H3 and H4 remains well behaved.

The 95dB test signal shows H2 rise in a linear fashion as expected across the frequency range. There is nothing too concerning here.



Intermodulation Distortion

The below test was done with a 6th order high pass filter at 8kHz. This was done within the Hypex FA123 filter software.

Using a 12 band per octave test signal we see a subtle rise in IMD at 10kHz dropping to very low levels in the 15kHz region. Distortion is lower than many amplifiers I've measured in this region, so pair carefully!



Raising the the test signal from 85dB to 95dB 1m we see IMD rise in a linear fashion with 10kHz showing 55dB dynamic range improving to 74dB at 15kHz.



General Considerations About Bullet Style Super Tweeters

The Fostex T900A is an attractive solution for covering upper treble. Care should be taken to high pass filter with the correct frequency and slope. This will help keep intermodulation distortion at bay in the audible band. The 85dB and 95dB test signal is likely higher than normal use case listening levels, so the projected distortion represented in my test data is likely much higher than in real life use. Considering this, amplifier choice becomes paramount in order to ensure low distortion performance, otherwise the tweeter will likely take on a somewhat harsh sound character. It is suggested to use a low gain amplifier since the tweeter will only require around 0.1v in order to provide optimal listening levels. If you are using a high gain amplifier, you should attenuate the tweeter using an L-Pad circuit or Autoformer. This ensures you are out of the noise floor of the amplifier.

Conclusion

The Fostex T900A is a well executed and beautiful tweeter with very good objective test data, particularly in the time domain and intermodulation distortion results for the upper treble region.
 
..pretty, but far to expensive for a ring radiator IMO.

Maybe something can be *done with this ring radiator at a more reasonable cost:

https://www.bmsspeakers.com/index.php-339.html?id=4526nd_overview

*Maybe even a different profile and damping for the phase plug “cone” inside the driver to extend it out further (removing the protective bug screen), like ”capping” a larger different angle cone (3d printed) on top of the plug to extend out into a custom horn flair (maybe even trying diffraction types to improve upper freq. dispersion).

-the Fostex super tweeter that I miss most is the 96 with its excellent upper freq. dispersion.
 
Last edited:
glad that you love them but have you made mesures of your system respons with and without them ? i always found a that' it is very difficult to add a tweeter on top of a horn system and try to make it disappear in 6db or even 12db passive crossover . playing with frequency cut off ,phase and time delays with a DSP help a lot ..
 
Yes, I did. My system is completely aligned for my listening position via impulse response measurements. My in room response is within -+3db from 20hz to 20khz. It should be buried in some topic here as well. It has been more than a few years though...
 
Maybe something can be *done with this ring radiator at a more reasonable cost:

https://www.bmsspeakers.com/index.php-339.html?id=4526nd_overview

Though this is the next driver I think you should take a look at (Troy):

https://solen.ca/en/products/cmp65_vpa-8

https://www.kartesian-acoustic.com/copie-de-cmp35-vpa

-for an integrated metal surround (not poly) the non-linear distortion on the low-end (along with the extension) is spectacular for the money (especially considering post-pandemic pricing), while also having an extended upper freq. response.

Low Mmass vs Sd + High strength magnet (21,500 Gauss) + Integrated Metal Surround usually leads to increased clarity (..but also the metal surround usually leads to much higher non-linear distortion near Fs, though again: not here at least as shown on their website).
 
  • Like
Reactions: EarlK and mp9
Re: A very usable crossover for these sort of tweeters;

Here's how I do it ( this approach gives a clean @ 12db rolloff once the LCR has sculpted the first part of the stop-band.

( Grey trace is the LCR notch taken out of circuit )

Fostex 945N HiPass_1P+LCR_Fr+Circuit_.png



Sharp eyes should notice this circuit has some gain above 12K >>> due to setting up a PEQ boosting filter ( a series LCR ) that uses the tweeters L ( coil inductance ) directly connected in series with the single (C) capacitor.

And no, I don't hear the "ringing///resonance" >> since I can barely hear 12K. 😉

🙂
 
Last edited:
  • Like
Reactions: ScottG
Overall, are the generally lower (?) cost Fostex horn tweeter models-those still available https://www.madisoundspeakerstore.com/horn-tweeters/
less problematic and better performing than most ribbons, AMTs and/or beryllium tweeters, at least with the appropriate lens?

Soundstage size, dispersion pattern, smoothness, clarity, ease, total lack of grain, hardness, brightness, “hashyness” sounding, et al.


Or have some heard other tweeters that have more advantages, regardless of price?

https://www.madisoundspeakerstore.com/amt/

https://www.madisoundspeakerstore.com/ribbon-tweeters/




 
If the specifications were published without all the beautiful pictures no one would call that a super tweeter or even a high fidelity tweeter. There are many tweeters that have 1/4 the distortion at 85 dB. That is a great service posting these test results. The size of these prevents mounting them close enough to a midrange to allow for good off axis response through the crossover region. I can't think of any advantage to using an Alnico magnet with so many modern magnet options available. This caters to the audio jewelry collectors as it is indeed very beautiful.