Which microphone to use for speaker analysis?

[DBMandrake]

Originally Posted by Ron E

Measurement mics need to be omnidirectional. A directional (unidirectional, cardioid, anything but omni) mic changes frequency response depending on distance from the source.

Erm, can you please explain how microphone directivity would cause the measured frequency response to change with distance from a sound source differently than it would with an omni ? I can't see any mechanism where this would be the case.

Now, if you're talking about measuring an un-gated steady-state response in a room then yes a directional mic will give a different result than an omni at high frequencies because it's collecting less of the room response, but neither result is a useful or valid measurement of a speaker, nor of how the room response will be perceived at high frequencies.

An on axis gated measurement to a single driver will work just as well with a directional mic as an omni, with no difference in response (between omni and directional) at different distances.

On a large speaker measured too close the more directional mic might show a slight difference due to being slightly off axis to some of the drivers, but even with an omni mic in that situation you still have significant errors due to being off-axis from the drivers.

There's nothing inherently wrong with some microphone directivity for speaker measurements, the real reason for wanting an omni mic for most measurements is to get the capsule size as small as possible to push resonances up out of the audible range and to minimize diffraction - most cardioid and directional mic's are physically large and therefore have problems with in band resonances and diffraction.

[CopperTop]

Is it just a case of correcting the mic's frequency response? Are they essentially perfect apart from a few bumps in the frequency response?

I'm very excited by the idea that it might be possible to go further than correcting a speaker's frequency response and effectively convolve the signal with a correcting FIR filter or similar to completely linearise it. Are these mics up to the job of providing the measurements? (given suitably anechoic surroundings etc.)

Beyond that, it might also be possible to create self-learning multidimensional FIRs that correct for non-linear nasties as well...

I have in mind the idea of being able to build a speaker using any drivers in any box and once the thing has been physically tweaked to the best it can be, the DSP can refine it to as close to 'transparent' as is possible. I would never be bored again.

[spot]

In the past I have wanted to do live compensation for speakers response, but now its a more humble measure and equalize. Things like the 'Convolve' filter for ffdshow sound amazing, I can't wait to try it, but first I'm going to build great speakers to start with.

The microphone I ended up getting is a FM-6B from www.hy-q.com.au (found this in the forum). $6 ea + postage in Australia and flat supposedly to about 1db from 20-20K.

Some Ideas I had in the past-
1) Use a microphone directly in front of the woofer for closed loop control of the sound. This may be possible with sealed boxes only. I have read an article on a design a long time ago, but it had a loop gain of 4 which is woefully low - not much control at all.

2) Use a microphone as part of a control loop continually adjusting a FIR filter. Essentially doing live adjustments to the equalisation based on microphone input while you are using the stereo. This is very similar to the way noise cancelling headphones/speakers work, except what is being cancelled is the difference between your speakers behaviour and a perfect sound. (Only acts on linear effects though, you need the previous method for non-linear).

[mickeymoose]

spot: now that we are off-thread google "Motional Feedback". Philips did that in the 70s! E

[simon7000]

OK, so does anyone understand how to measure the frequency response of a pair of loudspeakers WITHOUT using a microphone?

[krivium]

, can you please explain how microphone directivity would cause the measured frequency response to change with distance from a sound source differently than it would with an omni ?

Proximity effect: Proximity effect (audio) - Wikipedia, the free encyclopedia

Every unidirectional pattern microphone (figure of 8, hyper,super,cardio,subcardio, hypocardio) exhibit this behavior. This is the reason for use of omni mic for analysis.

The best i've heard are Bruel & Kjaer (now DPA) and some Earthworks but they are clearly off budget... The little behringer is ok as long you don't go over 90/94 DB Spl. Beyond that level you can't trust them.

[CopperTop]

Originally Posted by simon7000

OK, so does anyone understand how to measure the frequency response of a pair of loudspeakers WITHOUT using a microphone?

Could you do something like bolt the speakers together, facing each other, driving one with a test signal while measuring the output of the other?

I haven't quite worked out what it would tell you. Maybe nothing!

[simon7000]

Originally Posted by CopperTop

Could you do something like bolt the speakers together, facing each other, driving one with a test signal while measuring the output of the other?

I haven't quite worked out what it would tell you. Maybe nothing!

It is called reciprocity. You don't need to bolt the speakers, just face them at each other about 1 meter apart. Drive one and use the other as a microphone, reverse the hook up and do it again.

That is how microphones get calibrated!

One of these days I will do an article on it!

[CopperTop]

Originally Posted by simon7000

It is called reciprocity. You don't need to bolt the speakers, just face them at each other about 1 meter apart. Drive one and use the other as a microphone, reverse the hook up and do it again.

That is how microphones get calibrated!

One of these days I will do an article on it!

This is very interesting!

Are you saying that if a speaker has, say, a particular group delay for a range of frequencies, that by using a matching speaker as the 'microphone' this will be doubled? Similarly, I imagine that resonances in the playback speaker would also enhance resonances in the 'microphone'.

Not quite sure about the significance of the 1m... Are you then introducing a need for an anechoic chamber etc.? Or do you need a gap for a more representative free field response as opposed to the pressure response anyway?

[Ron E]

Originally Posted by simon7000

It is called reciprocity. You don't need to bolt the speakers, just face them at each other about 1 meter apart. Drive one and use the other as a microphone, reverse the hook up and do it again.

Actually you need an auxiliary (third) transducer. You can use two speakers and your microphone.

You can do self reciprocity with a single microphone and a perfectly reflective wall. But it is tricky and mics don't make very strong signals.

It is difficult to get good data with good signal to noise. I did a trial reciprocity calibration using my behringer and an omni dynamic mic and a tweeter. I got plausible results, but I can't know for sure if they were correct.