Freq response charts vs real world experience

[bgavin]

Another measurement approach is 100w at 10m. This gives the same relative chart as 1w/1m, but shows the effects of power compression (if any).

 

[fdeck]

I would want the test to be done at a reasonable multiple, like 5x, of the overall dimension of the driver array. Thus something like an 8x10 should probably be done at 5 meters. Anything can be scaled back to a 1-meter rating by assuming the inverse square law. Getting too close will result in a sensitivity figure that overestimates the efficiency of the system.

 

[will33]

Although it should technically be the same aside from distortion, I like the idea of upping the power and the distance of the mic placement as long as consistency wasn't affected. Would really mirror more of a real world playing situation. Seems to make more sense when talking about 810's, 412's or a crossed-over cab with different size drivers, 1 meter doesn't seem far enough to get the effect of the entire system. If your measuring 1 driver only, 1 meter should be just fine. Doesn't Bill F. measure his at 10 meters with 28.3v instead of 1 meter at 2.83v or something like that?

 

[bgavin]

Tom Danley uses 100w/10m for measuring his horn products.

 

[seamonkey]

I'm thinking the 100w at 10m makes sense.
This seems like a reasonable audience distance.

I'm not sure how a voltage (2.83 vs 28.3) would work, seems like a lower iimpedance would have higher wattage applied to it.

So now my next question is on sweeps.
An impedance sweep is an important measurement.
But should you sweep for the other measurements or do a white noise? For sweeps the impedance of the speaker changes, so you couldn't keep a constant wattage flowing. How would you determine 100w if you were sweeping?

 

[alexclaber]

I'm not sure how a voltage (2.83 vs 28.3) would work, seems like a lower iimpedance would have higher wattage applied to it.

SPL tests have to be done with constant voltage not power and then referenced to the power into the nominal impedance.

Alex

 

[A9X]

Tom Danley uses 100w/10m for measuring his horn products.

That's aminly because of the very high SPL's they're capable of and the LF nature of many of the products.

I'm thinking the 100w at 10m makes sense.
This seems like a reasonable audience distance.

I'm not sure how a voltage (2.83 vs 28.3) would work, seems like a lower iimpedance would have higher wattage applied to it.

So now my next question is on sweeps.
An impedance sweep is an important measurement.
But should you sweep for the other measurements or do a white noise? For sweeps the impedance of the speaker changes, so you couldn't keep a constant wattage flowing. How would you determine 100w if you were sweeping?

It's the nature of spekers that they're impedance varies. An amplifier that's behaving as a voltage source, ie very low output Z won't affect frequency response.

 

[A9X]

SPL tests have to be done with constant voltage not power and then referenced to the power into the nominal impedance.

Alex

Exactly. That's why I suggested 400Hz as an example of what to use as 'nominal Z'. It's above any Z tuning peaks, below xovers and inductance shouldn't make much contribution this low.

 

[fdeck]

I'm thinking the 100w at 10m makes sense.
This seems like a reasonable audience distance.

I'm not sure how a voltage (2.83 vs 28.3) would work, seems like a lower iimpedance would have higher wattage applied to it.

So now my next question is on sweeps.
An impedance sweep is an important measurement.
But should you sweep for the other measurements or do a white noise? For sweeps the impedance of the speaker changes, so you couldn't keep a constant wattage flowing. How would you determine 100w if you were sweeping?

To the extent that a speaker is a linear system, sweeps and white noise should be equivalent. Impedance is the ratio of voltage to current. You could imagine mixing two sinewaves in the generator, and then extracting the voltage and current amplitudes (and phases) at the two frequencies in the output waveforms. Doing it with a white noise source and FFT is just an extension of this concept. I imagine the difficulty with a white noise based impedance measurement would be extracting accurate phase information at each frequency without extensive averaging.

Like alexclaber says, the measurement is done at constant voltage. The "Watt" of sensitivity ratings is a funny Watt. It is the square of the RMS voltage divided by the nominal impedance. Thus, 28.3 V is treated as 100 W into 8 Ohms, regardless of the actual impedance.

In my view, Watts are just a convenient way of comparing systems, but design and analysis are done with voltage and current.

About the impedance sweep, you could get away with doing only a small number of frequencies, since the impedance curve is dictated by the electromechanical behavior of the system. The theory is quite accurate at low frequencies, and at higher frequencies, exact impedance is less of a concern because there is typically less power in the upper harmonics. It might not be quite so straightforward with bass horns.

 

[billfitzmaurice]

In my view, Watts are just a convenient way of comparing systems, but design and analysis are done with voltage and current.

Exactly. The power draw of any speaker varies with impedance, and impedance varies with frequency, so the entire notion of sensitivity per watt input is intrinsically flawed. Speakers are measured with a constant input voltage, not constant power. How it ever went from SPL/volt to SPL/watt is lost in the archives of a century ago, but I suspect it was one of the earliest cases of the engineering department being over-ruled by the marketing department.
BTW, the easiest and most logical way to measure is to use one of the free software programs widely available. You also need specific hardware, but that's the case whether one uses a software program or not.