Fostex T90A Supertweeter Filter

[simon]

I bought a little Dayton mic a few weeks ago - the same one Steve found. £20 from Amazon, can't say fairer than that. My main intention was to measure the filter crossover for the Supravox 285 bass helpers with the Fostex 208s when I eventually put them in the Quasars - something that seemed like a sensible thing to do during an email conversation with Mark a while back.

In the meantime I've used it with my back horns and it's been an enlightening experience . The first thing I realised was that I set the supertweeter level way too low with the L-pad - basically so I could only just hear it. Using the Dayton mic with a simple (free!) spectrum analyzer app on my tablet showed me the error of my ways.

But what did surprise me was when I looked at the individual contributions of the T90As and 208s - the T90A is working down to around 2kHz which I don't think it should. Here's the overall frequency response using a simple sine sweep generator 20Hz to 20kHz on my phone. I'm not making any claims for absolute accuracy but it should be good enough.

Ignore the green plot, that's just background noise.



Here's the 208 alone, which looks to me to roll off from somewhere in the region of 8kHz.


And here's the T90A alone, protected by a 2.7uF cap which should roll it off around 8kHz (simple 6dB/octave)


But there's clearly audible output below the filter point, and I don't think there should be. If there shouldn't be, what am I doing wrong??? Should I be using a smaller cap to raise the frequency roll off?

[Toppsy]

Those plots clearly show the reason for not buying a cheap measuring mike for designing speakers and then using it with a phone app. They bear little resemblance to the manufacturers published data sheets. However, it does clearly show you have some issues.

Because you are using a simple 1st order filter this has quite a wide crossover band width which is why you are hearing sound below the XO point, which is far too low for the T90A. You also have a potential mismatch of efficiencies. The 208 is rated as 97dB/W whereas the T90A is 106dB/W. The L-Pad needs to be set for a -9dB attenuation.

If you check the published datasheets and FR plot for the 208 this has a rising response up to 3.5KHz then a rapid 9dB fall to around 95dB at 5KHz. this further drops off to around 91dB at 10KHz. The datasheets for T90A recommend XO frequency more than 7KHz with a minimum 12dB/oct filter. The stated reproduction FR is 5KHz - 35KHz. You really need to design a proper filter with a 2nd order XO. There are a few free design packages on the web you can use.

[simon]

Food for thought Colin, thanks. I did say that I wasn't guaranteeing the plot accuracy .

This is an excerpt from the 208 datasheet. I really am no expert looking at these plots trying to average out the response - fitting a notional curve I'd say it rolls off anywhere between 5k and 8k? But a second blue curve might make it 4.5k? And actual reality might well be different anyway?


Perhaps I just shouldn't be looking at it in this way.

The recommended 12db/octave filter is a good spot - I've only ever used the simple example they show. I'll do some reading.

[simon]

2 caps and 2 inductors ordered, be interesting to see what difference they make. It'll be a few weeks till they arrive though.

[IslandPink]

I'm with Colin, you don't seem to be getting anything meaningful out of your analysis combo there - not enough to guide changes.
2.7uF as 1st-order HP shouldn't sound too bad if you aren't playing loud, from my experience with things like ribbon tweeters. At the moment I'd just listen by ear until you can get the mic etc sorted out. Bear in mind you might need tweeter polarity reversed depending on where you locate the two drivers , try both. Let's not get started on the many issues that affect where the acoustic centre of the 8" driver would be for predicting best polarity....

[cressy]

I'd re measure the t90 simon, its plot looks far too close to the combined plot. When i measured the medium metronomes with the same app it showed me where to look for the peaks i wanted rid of an gave me a starting point to tweak the crossovers.
Those plots tell me there is a dip in the response then a slight peak, so surely it tells enough to start. Im not sure what else a simple frequency sweep could tell you.
What's the point in spending big money on measurement equipment you may only use once?
If you are going to use it regularly, fair enough, but for a general overview of the in room response what's wrong with those plots?

If i had only seen the first plot I would have looked at bringing the t90 in lower ( how low will it actually go though?)as there is a clear dip where fostex is tailing off but the t90 hasn't come up.

As to why the t90 plot shows a fairly level output down to where it does, I'd triple check the measurements.

Personally i am more inclined to believe what an in room measurement says in this case than the manufacturer plots, having been in simons room listening to them myself recently, (pretty large with loads of reflective surfaces, wooden floor with rug,)the combined plot looks like what I heard.
And as these are his own and not a commercial speaker so optimising them for the room makes sense to me.

[Cressy Snr]

Hmmm.....
I use a 3.3uF first order Xover with a Monacor RBT-95 ribbon tweeter, and modified Fostex FF225WK widebander, with zero problems. I used a pink noise generator on my Mac Mini, with the very same mic that Simon is using connected to my iPad, running an Audio in-room third octave spectrum analysis package. The results were excellent and the analyser output charts clearly showed where the improvements had been made to the midband performance of my Mets.
Anyone who has heard my Mets in my room would, I think, agree that they sound the biz, even when cranked up

Left - before, Right- after (much better across the inportant midband, though we can see that sensitivity has dropped a bit. Overall though a price worth paying for a far more even in room response) It shows up in the listening too.


I killed the top-end of my Fostex widebander quite deliberately, by the use of mechanical slugging ring of black-tak at the junction of the alloy dustcap and cone. The ribbon then fills in the top beautifully as it is not battling to compete with the output of the widebander. I'm not suggesting for one minute that Simon should slug his own Fostex 208EZ (perish the thought) but sometimess mechanical solutions can be better than electrical where "full range" drivers speakers are concerned. IME supertweeters with full rangers can be more trouble than they're worth, which is why I approached things from the other direction, by tailoring the full ranger response instead. It works a treat.

[andrew Ivimey]

Oh I don't mean to carp so I'm only imho when I say that '....when cranked up' speakers often sound better than when they are quiet. A quiet speaker that does the business is more difficult to achieve in general for two reasons .... well maybe just one really and that is because the human ear is most sensitive by far to the mid frequencies ( as defined by (me) as 'the range of human speech' - google 'phons curves' and you will see what I mean) and that whenever something gets louder (in this case music) the response by the human ear is to flatten out any curve until it gets painful i.e. too far.

A truly full range speaker of any sort is much harder to achieve under 60dBSPL than anything over 80-90dBSPL - at 110dBSPL if it isn't distorting it is easy to achieve a flat frequency response.

Me, I like a T90a with Sigma208s and I like using a peanut size amp with them.

OTOH, I love the splendor of everything else so long as it is splendid and there are many lovely ways of achieving such.

[Cressy Snr]

Bring back the "loudness" button I say

[andrew Ivimey]

Everything in its place and this doesn't have to be on a cheap Jap jobbie from the80s.

[simon]

If i had only seen the first plot I would have looked at bringing the t90 in lower ( how low will it actually go though?)as there is a clear dip where fostex is tailing off but the t90 hasn't come up.

That's exactly what I thought, and why I then thought I'd measure the T90A alone. Hence my surprise.

Overall I'd say I'm hearing what the overall plot shows, and it sounds pretty good too I think. I'm just not sure about the supertwatter though, and why it's not showing the expected roll off. I'll try to redo the measurements tomorrow night, but I heard what the plot shows. The mic was good enough to show me where the issue is .

[cressy]

As Colin says, the frequency response on the t90 is between 5 and 35khz so I'd expect to see something in the lower range, but the rolloff is what is confusing. I would have expected a much steeper rolloff from 8k
However, if it will go down to 5k and there is some sort of output that low, and it will actually stand it and not fry, why not try a lower crossover point?
It might fill in the hole at the current point. Might be better, might not, they sounded alright to me as they are

[simon]

I was really surprised to hear any output at 2k, let alone as much as there was. This just isn't adding up to me.

Anyway, I want to try 1uF just to see if I can produce a roll off. Assuming I can do that then I can play a little more.

[chris661]

Its worth noting that manfacturer's graphs will often be very smoothed.
See if you can swap to a Log frequency scale. Much more information there.

If I remember rightly, you're not too far from me (I'm just south of Sheffield). I've recently got a Beyerdynamic MM1, looking to try it out a bit on some more kit. Let me know if you fancy arranging something.

Chris

[SimonC]

Its worth noting that manfacturer's graphs will often be very smoothed.
See if you can swap to a Log frequency scale. Much more information there.

If I remember rightly, you're not too far from me (I'm just south of Sheffield). I've recently got a Beyerdynamic MM1, looking to try it out a bit on some more kit. Let me know if you fancy arranging something.

Chris

I can add an ECM8000 into the mix if we are going to run a 'mic-off'... a day of sine-wave sweeps and white noise - I'll pack some ear plugs

Simon C