Advise on crossover for tweeters

[Nicola]

Here is a simple question. I'm driving two tweeter (1x Fostex T90a and T825) with one amp which is driven by an active xcover. I thought to make an high pass first order butterworth, 6db/ octave at 12kHz in order to separate the two drivers. The T825 is driven directly by the amps (6300hz - pass) and the T90A would have been 12khz to pass. For that as the T90A is a 8ohm i inserted between the T825 and the T90A a MMK cap of 1.64uF. I expected i would see a 6db decline below 12khz for the T90A. If you look at the picture which shows the T90A filtered and not, you'll find a dip below 12khz with a bump a 10hkz and dip at 20khz followed with a bump.... What i've i done wrong ? Thanks in advance.

[picowallspeaker]

My suspicious mind wants to know...
You say that you use two tweeters . Are you using two equal amplifiers ? Oh! What a pity !
Measurements are taken with the same amp ! So let's blame it on the cross and other near circuitry , which may produce that attenuation ...
The bump may be explained, as you said you used a gentle roll-off,so the curves of the two tweeters sum together at that frequency .
Then , why that high frequency for crossing the first tweeter ? I imagine if choosing two tweeters for the top frequency reproduction ,you are allowed to think 1 KHz ,if the tweeter permits , as a good starting point . 6 KHz is insane !
Then also have to be taken in account the different geometry of the tweeters and the baffle they are mounted on : sure that peaks and dips are not caused by phase redundancy etc etc ?
Next time I want a complete examination ,completed with photos , measurement off-axis,and all that goodies that skilled DIYers use to post for full understanding,OK ?!!

[Cal Weldon]

Can you draw us a little schematic of how you've wired the tweeters. Somethings not right here.

[PaleRider]

[QUOTE=picowallspeaker;2198663]
Then , why that high frequency for crossing the first tweeter ? I imagine if choosing two tweeters for the top frequency reproduction ,you are allowed to think 1 KHz ,if the tweeter permits , as a good starting point . 6 KHz is insane !
QUOTE]

Wrong. 6khz is a good start, 1khz is insane. The clue is just to fing the right driver to work up to 6khz.

[Nicola]

Thanks for the feedback. Here is more :
a) single amp drives the two drivers, so each 8ohms, ends up 4ohms because there are in parallel
b) because the T90a is 106db and the T825 is 102db is use a lpad to lower the t90a (hence the mox resistor 1R5 et 33R)
c) the cap 1.6uf is after the T825 and before the lpad on the + of the T90A

I would have expected
a) to see a 6db/octave drop below 12khz, hence to be around 80db at 6khz
b) i don't understand the dip at 20khz
c) i noticed the filter being 1st order does a 90° phase shift which i decided to compensate by wiring the T90a reversed, meaning the + of the T90a is wired on - of the T825.

The read is the T90A without filter and lpad and the magenta sort of is T90A filtered at 12khz and lpad.

Regards

[Charles Darwin]

With 6dB roll off at 12k and an efficiency of 102dB I would expect a response of 93dB (taking into account that the crossover frequency itself is already the -3dB point)at 6k as that is one octave below 12k. To achieve a 6k response of 80dB you would need a 24dB roll off.
Going up one octave means doubling the frequency. The interval between 16Hz and 32Hz is one octave, the interval 3kHz and 6kHz is also one octave.

You can not compensate a 90deg phase shift by inverting the tweeter as inverting shifts the the phase by 180deg. you are still 90deg out, just in the opposite direction.

[picowallspeaker]

single amp drives the two drivers, so each 8ohms, ends up 4ohms because there are in parallel

Not true ,since the T90 is another circuit, separated by the condenser and attenuated by the resistor

Wrong. 6khz is a good start, 1khz is insane. The clue is just to fing the right driver to work up to 6khz.

Unless Nicola won't give the exact picture of the system ,how can we guess ?
I know the T90 is a supertweeter , so I supposed it would have helped something near a dome-midrange or a fat tweeter.
Now that I've made a quick search , I discovered the T825 being a strange kind of tweeter ,so never mind.

What I wanted to say is that we need a very clean source and amp to take advantage of such configuration ,since the treble region might be a minefield

[speaker dave]

I'm not seeing much evidence of a first order roll off, although it is hard to see with such a compressed dB scale. As to the dip at 20k, it is unrelated to your network. In fact both curves have a small dip, the curve behind has it somewhat masked by the curve in front. I would guess that that is a slight reflection from cabinet edge or grille. The system probably moved between curves so the dip frequency shifted slightly. Of no consequence.

Any chance you can feed the electrical output of your network into the analyzer? Calculated networks frequently give different results than expected, especially if the impedeance curve of the driver hasn't been taken into consideration.

David

[Nicola]

Sorry should have given the full picture. I'm running a 4ways active crossover with one amp for each frequency range. The high range which drives the tweeter cover the frequencies from 6300hz to above. The amp drive the two fostex supertweeter, one T825 and one T90A. Why the T90A + T825 ? The T825 is very spacial, open but does not cover well the hamonics beyond 15khz what the T90a which is a bit more directional but since it cover 6300 - 35khz, the harmonic definition above 14-15khz are more precise. I thought using a high pass for the T90a below 12khz would reduce the impact. But the measure as shown don't seem to show any roll off below 12khz

Question to picowallspeaker : you can find the "schematic", so what load does the amp sees ? 8 or 4 ohms ?

The T825 covers frequencies above 6300hz and the T90A - i thought with the cap - would cover frequencies above 12khz.