AR-90 frequency response measurements

[Kuja]

Hello again!

After starting a thread regarding replacement dustcaps and foam surrounds:

http://www.classicspeakerpages.net/IP.Board/index.php?showtopic=6392

Carl gave me an idea to measure each driver separately from 0.5 cim distance.

So I did:

Here are the results:

Green = woofer

Blue = lower midrange driver

Red = upper midrange driver

Violet = tweeter

Vertical thicker blue lines are crossover frequencies found in AR-90 literature:

They are quite different from what I have measured!

Crossover frequency between the UMR and the tweeter should be 7000Hz, but it measured around 5000Hz.

According to the AR literature, crossover slope on the tweeter should be 18dB/octave at 7000Hz and the graph looks far from it.

Crossovers are all original except for the old aging capacitors, that were replaced with new polys of the EXACTLY same values.

What are your thoughts?

PS

Both speakers had the same measured results, difference was less than or around +/- 1dB.

So I guess both crossovers and all drivers are OK, and they are functioning properly.

.

[Carlspeak]

I have more questions than thoughts at this point.

Did you do NF on all the drivers? or, just the woofer? If all, I suggest you pull the mic back to about 1 or 2 meters from the baffle board and take another measurement. Be sure to gate it.

What level of smoothing did you use?

Could you change the display so the vertical axis is higher and more detailed?

[Kuja]

Yes, I separately measured all drivers from the 0.5cm distance,

so that I can better see how they behave regarding their crossover points.

Here is the measurement I made from 1m distance, measuring mic pointing at the upper midrange driver (the way that AR measured these speakers):

[Kuja]

PS

I used 1/6 octave smoothing on the graphs, in order to visually avoid effects of room reflections.

Here is the bigger image:

[Carlspeak]

Yes, I separately measured all drivers from the 0.5cm distance,

so that I can better see how they behave regarding their crossover points.

Here is the measurement I made from 1m distance, measuring mic pointing at the upper midrange driver (the way that AR measured these speakers):

That looks pretty darn good!

Enjoy the music.

[Kuja]

Do you have any idea why measured crossover frequencies appear to be different when compared to the AR literature?

[Carlspeak]

Do you have any idea why measured crossover frequencies appear to be different when compared to the AR literature?

No I don't. Just a guess, they may have changed the crossover point and the literature you have does not match?

Does it really matter anyway if all sounds well?

[Steve F]

No I don't. Just a guess, they may have changed the crossover point and the literature you have does not match?

Does it really matter anyway if all sounds well?

A possible explanation of the crossover variance:

There are really two different ways to characterize the so-called "crossover" of a multi-driver system. One is electrical and one is acoustic/mechanical.

If you just take the crossover network itself and run the voltage curves through it, you'll get an exact electrical measurement of where that network is directing the various frequencies. It's not really open to too much question, for that specific network sample, anyway. Another sample may well show a different result, depending on the tolerance of the components and how those tolerances stack up. If your network is using, say, 10% tolerance parts and there are three components in that path, it's conceivable that you could have a .9 x .9 x .9 situation, or only .729 (or 1.33 if all three are +10%) of the design "value" in that signal path, resulting in a not-insignificant crossover frequency variance.

Yet, technically, the network is "in spec."

The other thing to consider is the acoustic/mechanical "crossover" of the drivers themselves. Even though the electrical network may be "telling" the drivers where to cross over, drivers have mass, they have resonances, they have inertia, they have imperfect frequency response, all kinds of things. It's not like flicking a switch, it's not an 'on-off' situation. Drivers in a system often have considerable intentional overlap.

So drivers don't always listen to the "orders" of the network that well. If you plot the electrical crossover of the network and you plot the actual acoustic crossover of the drivers through the finished system, they're often--usually--not the same.

Crossover frequencies are a guess, an approximation, at best, for these and other reasons. Ask any speaker engineer. They've all undoubtedly had this conversation with Marketing: "Where do you want to say it crosses over?"

Steve F.

[deek]

The only thing I see that's concerning is the gap left between the LMR and UMR.

The distance could be having an effect as well. Moving the mic away from one driver will reduce it's contribution to the curve and add the contribution of the other drivers. That alone could shift your "crossover" more than a few hz.

Back off the speaker about a meter and put a composite curve over the top of your individual curves. I bet you see those gaps between drivers disappear.

-Deek

[Carlspeak]

Back off the speaker about a meter and put a composite curve over the top of your individual curves. I bet you see those gaps between drivers disappear.

-Deek

Hi did. Check post #3