Wednesday, May 03, 2017

Valentine's Day microphone test revisited

In the original Valentine's Day microphone comparison I tested three sets of mics in different scenarios. The files are still available, so you should visit that article if you haven't already.

Then, in Comments on field recording gear I showed how we match recorders to microphones, and demonstrated that for the three configurations I used, the recorders and mics are well-matched. This means that the pre-amps are adding no appreciable noise. Any noise we hear is the self-noise of the microphones. (Or comes from the sources we are recording.)

Now it's time to reveal which mics were which in the listening test.

The Big Reveal

A drum-roll please:

MIC 1 = DPA 4060 + Zoom F8
MIC 2 = AT 3032 + Zoom F8
MIC 3 = EM172 + Olympus LS-11

When I revealed this on Facebook I got some interesting responses. Some listeners were surprised by the amount of noise on the DPA mics, and assumed I had made some sort of a mistake. Nope! The results are in line with the published self-noise measurements of the microphones:

DPA 4060 = 23 dB
EM 172 = 14 dB
AT 3032 = 8 dB (Tested figure, not as specified.)

The difference between the three microphones is in no way insignificant. And these differences are reflected accurately in the recordings. This is largely due to laws of physics. A larger diaphragm can capture a bigger cross-section of a pressure wave in the air. The mic then simply has more energy to work with, and can transduce this more effectively. We usually pay a price for using smaller, handier mics.

Further Considerations

Consider that these specs are all A-weighted, which means that a particular weighting curve has been applied to each frequency band as tested. In my considered opinion as an audio engineer, this is an inappropriate curve to apply to noise measurements. Nonetheless, manufacturers persist in what is really a deception, since they can publish smaller numbers using the A curve than if they measure noise across the frequency band without applying a curve.

To their credit, DPA publish two numbers for equivalent noise level on their spec sheet. The full text for the first is: "A-weighted Typ. 23 dB(A) re. 20 ┬ÁPa (max. 26 dB(A))". And for the second: "ITU-R BS.468-4 Typ. 35 dB (max. 38 dB)". The ITU-R curve is a lot more appropriate from a technical standpoint.

But... 35 dB versus 23 dB. That's a pretty large difference. Which number sells more mics?

Regardless, not all noise is equal. Listen to the Site B recordings again.

With the AT 3032, the noise is evenly distributed across frequency, so it is much less likely to call attention to itself. The EM172 has very even mid to high frequency noise, with little noticeable noise in the lows. The DPA 4060 has significant low frequency noise in addition to what's there at other frequencies.

It seems from this, that even if we had ITU-R measurements for all three microphones, the AT3032 would still come out ahead, maybe even by a larger margin.

What Are Measurements Worth?

Some recordists regularly slam specifications as meaningless, preferring to believe only the evidence of their ears. I am aware of psychoacoustic effects, conditioning, and expectation. These affect me as much as the next person, so I prefer to challenge my biases. I use specifications with full knowledge of their limitations. And then confirm or deny these with listening tests.

Which I suppose was the whole point of this exercise.

If it's so noisy, why is the DPA 4060 so popular for field recording? I have heard hundreds (maybe thousands) of great recordings made with DPA 4060s. If you listen to the first set of recordings (Site A) again, you will hear why. This nature soundscape has objectionable noise in the lower frequencies, which is typical of real-world scenarios. Traffic rumbles and drones from distant machinery are undesirable to (almost) all listeners. The DPAs naturally attenuate these sounds.

Meanwhile, the primary content (bird calls) are in the mid and higher frequencies. The DPA renders these sounds well, managing to capture a good dynamic range, likely better than either of the other microphones. I'd say the 4060s have very good transient response. (I have no numbers to support what my ears hear on this point.)

It comes down to picking the best microphone for the application. Specifications can be used to guide you in this task. Recordists who disparage measurements usually do so because they do not understand how they should be applied. Yes, they have limits, but this cannot be taken to mean they have no use at all.

And then, after all else is said and done, we must use our ears.


No comments:

Post a Comment