Human Ear Polar Patterns

[MRC01]

As we measure rooms to improve response, we should ensure the microphone hears what our ears do. For example, when I measure my room, the UMIK-1 mic shows a bass null that doesn't appear with Rode NT1As. I think this is because that null is caused by sound reflecting off the back wall, which the NT1As don't pick up because they're facing forward and this reflection comes from behind.

This leads to the question: how sensitive is our hearing to sound waves coming from different directions? It seems we should measure the polar sensitivity pattern of human hearing and use microphones that match this pattern. I found this interesting article in which the human hearing pattern looks asymmetric, in which case an omni mic might not be the best way to tune the room. Is this a topic covered here or elsewhere, that I can read more about?

PS: that doc says human hearing is not sensitive to azimuth at low frequencies - e.g. it's like an omni pattern in the bass, and becomes more cardiod-like in the midrange & treble. Since most of our room tuning is at low frequencies, perhaps we should use omni mics to tune the bass and cardiods for higher frequencies?

 

[Kal Rubinson]

As we measure rooms to improve response, we should ensure the microphone hears what our ears do. For example, when I measure my room, the UMIK-1 mic shows a bass null that doesn't appear with Rode NT1As. I think this is because that null is caused by sound reflecting off the back wall, which the NT1As don't pick up because they're facing forward and this reflection comes from behind.

This doesn't make sense to me. The null exists as a site of low energy at that particular frequency and, unless the microphone is so large as to approximate a significant portion of a wavelength, it should detect that low energy as null regardless of its polar properties.

 

[A800]

There are also binaural microphones.

 

[MRC01]

This doesn't make sense to me. The null exists as a site of low energy at that particular frequency and, unless the microphone is so large as to approximate a significant portion of a wavelength, it should detect that low energy as null regardless of its polar properties.

It's a conundrum that I'd like to better understand. I sit with a reflective wall about 44" behind me. When I measure response with NT1A mics, bass response is smooth. When I measure it with UMIK-1, it shows a null at 74 Hz. It just so happens that 44" is 1/4 wavelength of 74 Hz, so the wave that hits the wall behind me and reflects forward has traveled 1/2 wavelength back to my head so it's 180* out of phase with the incoming wave. The only explanation I can think of is that the UMIK-1 is omni while the NT1A is a large diaphragm condenser with a cardiod pattern. The UMIK-1 "hears" waves coming from behind it (which, being out of phase, causes a null) while the NT1A does not. Since the NT1A can't detect waves coming from behind, there's no cancellation thus no null.

Yet if we imagine the sound wave as pressure pulse trains, this seems impossible. The pulses bouncing off the back wall arrive at my head just as the incoming sound is in between pulses. At the position of my head, the pulses are interleaved so the peaks of one fill the troughs of the other, creating a localized zone of neutral ambient pressure - a null. From this pressure perspective, the direction of the wave wouldn't seem to matter. And the mic is so much smaller than the wavelength of this low frequency, diffraction masking shouldn't be an issue.

From the doc I found above, the polar sensitivity pattern of human ears is frequency dependent: omni at low freqs and cardiod at high freqs. Are mics the same way? Does a cardiod mic become omni at low frequencies? This image for the NT1A suggests not:

If there's another explanation for this observation, I'd like to learn what it is.

 

[Kal Rubinson]

If there's another explanation for this observation, I'd like to learn what it is.

Is the mic on a stand or are you holding it in front of you?

 

[Blumlein 88]

It's a conundrum that I'd like to better understand. I sit with a reflective wall about 44" behind me. When I measure response with NT1A mics, bass response is smooth. When I measure it with UMIK-1, it shows a null at 74 Hz. It just so happens that 44" is 1/4 wavelength of 74 Hz, so the wave that hits the wall behind me and reflects forward has traveled 1/2 wavelength back to my head so it's 180* out of phase with the incoming wave. The only explanation I can think of is that the UMIK-1 is omni while the NT1A is a large diaphragm condenser with a cardiod pattern. The UMIK-1 "hears" waves coming from behind it (which, being out of phase, causes a null) while the NT1A does not. Since the NT1A can't detect waves coming from behind, there's no cancellation thus no null.

Yet if we imagine the sound wave as pressure pulse trains, this seems impossible. The pulses bouncing off the back wall arrive at my head just as the incoming sound is in between pulses. At the position of my head, the pulses are interleaved so the peaks of one fill the troughs of the other, creating a localized zone of neutral ambient pressure - a null. From this pressure perspective, the direction of the wave wouldn't seem to matter. And the mic is so much smaller than the wavelength of this low frequency, diffraction masking shouldn't be an issue.

From the doc I found above, the polar sensitivity pattern of human ears is frequency dependent: omni at low freqs and cardiod at high freqs. Are mics the same way? Does a cardiod mic become omni at low frequencies? This image for the NT1A suggests not:
View attachment 62080


If there's another explanation for this observation, I'd like to learn what it is.

Here is the polar chart for the KSM32 cardioid to 125 hz. You can find other similar plots. Yes, at lower frequencies they tend to get near an omni.

 

[MRC01]

Is the mic on a stand or are you holding it in front of you?

In both cases, UMIK-1 and NT1A, the mics are on a stand sticking up from the couch where I sit (where my head is when listening). Nothing obstructs the back of the mics.
Blumlein's diagrams show that while a cardiod pattern gets more omni at low freqs, it's still significantly attenuated (-15 dB or so) from the back. That's a lot.

 

[Blumlein 88]

In both cases, UMIK-1 and NT1A, the mics are on a stand sticking up from the couch where I sit. Nothing obstructs the back.

Did you ever try turning the NT1A around 180 degrees?

 

[MRC01]

Did you ever try turning the NT1A around 180 degrees?

Not yet, because first I'm trying to decide what to expect and how to interpret whatever result I get. I was thinking, put one facing forward, the other facing back at the same position, play a test sweep, see if REW can show their responses with the phase difference between them?

 

[Blumlein 88]

Not yet, because first I'm trying to decide what to expect and how to interpret whatever result I get. I was thinking, put one facing forward, the other facing back at the same position, play a test sweep, see if REW can show their responses with the phase difference between them?

If you align the diaphragms internally carefully, they not show a difference that way. That is actually how omni's for larger condenser's work. They have a double sided diaphragm. That is where I suggested pairing them up pointing both ways and combining the plots.

Two ways to approach unexpected results. One is to figure it out in theory, and then test that theory. Often however, you just have to mess around with it different ways and you'll figure it out. I've not tested the supposition that cards are that directional especially as around 74 hz they are at least halfway between omni and cardioid. I can't see it being deaf enough at 74 hz to be smooth one way and sharply nulled at the other. But try some of this and the truth will become apparent whatever it is.

If I get the chance and can get an omni to null like yours I'll try it out. I've got a microphone that lets me choose the pattern after the fact. It records both sides of a double diaphragm and I can later add and alter phases for any pattern I wish.

BTW, what is the wall made of behind you? To work that well in the bass as a reflector it likely needs to be something substantial like concrete or stone or to be below ground.

 

[Kvalsvoll]

This is a result of the properites of the sound field at the position where you measure in combination with the microphone directivity. The sound field has a pressure null, but there is still particle velocity and direction in the sound. The omni umik is pressure sensitive, it does not care about direction, the other mic with directivity will also measure pressure, but it will also measure velocity, so you get a signal wven when the pressure component is zero.

 

[Blumlein 88]

Okay did some measurement quick and dirty. I had a speaker almost against the wall firing across 19.5 ft room. Microphone 43 inches from the distant opposite wall. Sheetrock over studs backed by a brick wall. A dip might be expected at 78 hz or so, there maybe is one at 77 hz.

Microphone was a Shure KSM44A which is multi-pattern. So once positioning the mic I only needed to flip a switch to change patterns.

There are much deeper nulls at 135 and 167 hz. Red is omni, and green is cardioid. I did use cal files, but at these lower frequencies there isn't really a need. BTW, these are raw unsmoothed files.

I also switched to figure 8 or bidirectional mode. This would be a pressure gradient mic or similar to a velocity mike.
Here is that compared to the omni.
The 77 hz and 135 hz dips are gone, but the 167 hz dip is much deeper.
Red is omni and blue is bidirectional.

The 167 hz or so dip might make sense. The wavelength is 81 inches. It would travel past the mic, bounce off the wall (which reverses phase of it) and at one wavelength with reversed phase it would cancel out. 81 inches isn't quite 43 inches x 2 however.

 

[MRC01]

Here's how the UMIK1 and NT1A compare, each at the listener position, each using factory mic calibration. Note: this Y scale is 2 dB / division where the above graphs are 5 dB / division.

 

[Blumlein 88]

I probably should mention I've compared the KSM44 in Omni to the Umik1. Below 1000 hz there's no real difference. Above 1 khz there are very small differences. The larger diaphragm of the 44 makes it a little directional above 8 khz.

 

[MRC01]

Listening to some slow low frequency sweeps, and 70 Hz & 140 Hz tones, the bass response sounds more even or "flat" when not EQed. If I apply an EQ to partially offset the 70 Hz null that the UMIK-1 shows, the low frequency sweep perceptually sounds less even with an audible bump. EQing the corrections for each separately and comparing them, I've found that subjectively, my perceptual response seems closer to what the cardoid measures, than what the omni measures.

PS: I do hear a dip a 70 Hz where the omni mic shows it. But the dip I hear is, perceptually, not the 14 dB that the mic shows. Playing with the volume to test & compare, it sounds more like 2-3 dB.

Comparing omni & cardiod responses has been an interesting experiment. It definitely sheds more light on what's going on with the room response and how we might perceive it.

 

[RayDunzl]

Listening to some slow low frequency sweeps, and 70 Hz & 140 Hz tones, the bass response sounds more even or "flat" when not EQed.

You made me think about that a little, since I have a black hole around 48Hz, that I didn't notice, until measuring..

I might try some bass sweeps with overtone levels typical of a bass instrument and see how that sounds.

Musical tones are seldom if ever naked sines.

 

[MRC01]

You made me think about that a little, since I have a black hole around 48Hz, that I didn't notice, until measuring..

I see a couple of psychoacoustic factors at play here.

First, physical. Our hearing isn't perfectly omni, sounds coming from the rear are attenuated, so we don't hear exactly what an omni mic picks up. Of course, we don't necessarily hear exactly what a cardio picks up either. Our perception might be somewhere in between.

Second, psychoacoustic. Even if the eardrum does detect a response dip, we won't necessarily perceive it like that. When we hear the harmonics of a bass note, most people "hear" the fundamental too even when the fundamental isn't there. The brain fills in the missing fundamental because it has learned through experience that the harmonic structure says the fundamental should be there.

 

[Blumlein 88]

I've still had this issue on my mind the last couple days and haven't decided what I think. I had suggested you turn your cardioid around 180 degrees and remeasure. Did you ever do that? BTW, combining those two results, 0 degrees and 180 degrees should give you an omni result. Use the (a+b)/2 function in REW. See if it shows the same dip as using the UMIK.

I've also been pondering if using figure-8 responses might allow us to point to which room modes are causing dips or peaks we don't want. I'm thinking measuring perpindicular to the room in one direction and rotating 90 degrees and remeasuring. As the figure 8 pattern ignores 90 degrees off axis that might work.

 

[MRC01]

I had suggested you turn your cardioid around 180 degrees and remeasure. Did you ever do that? BTW, combining those two results, 0 degrees and 180 degrees should give you an omni result. Use the (a+b)/2 function in REW. See if it shows the same dip as using the UMIK. ...

Yep. Here it is. Red = facing forward, Blue = facing backward, Green = (A+B)/2. Does Room EQ Wizard account for phase when it sums them? In the SPL/phase diagrams, phase looks the same, which is suspicious.

 

[RayDunzl]

Does Room EQ Wizard account for phase when it sums them? In the SPL/phase diagrams, phase looks the same, which is suspicious.

"Average The Responses calculates an average of the dB SPL values of those traces which are selected when the button is pressed. Phase is not taken into account, measurements are treated as incoherent. "

"Vector average, which averages the currently selected traces taking into account both magnitude and phase. It can only be applied to measurements that have an impulse response. "

https://www.roomeqwizard.com/help/help_en-GB/html/graph_allspl.html