Problems measuring two speakers in room using RTA...

[fjhuerta]

I have a stupid question...

If I measure one speaker in room with an RTA, everything measures fine.

If I measure the other speaker in room, it measures almost the same.

If I measure both speakers at the same time, I get destructive interference from around 900 to 2.3 KHz. Strange.

I've already tried the following:

1. Check the crossover. Fine.
2. Check speaker impedance. Both are identical.
3. Check individual speaker FR. Both are identical.
4. Connect both speakers to the same channel of my amp. Same result.
5. Move speakers closer to one another. This does help. The closer they are, the less interference there is, until finally, it disappears and measurements are flat again.

I'm thinking that measuring both speakers using an RTA is creating destructive interference from 1 KHz on up. Reason being, one needs to place mids and tweeters as close as possible to the woofers for them to sum up, so, can I assume that the distance between the speakers mean it's impossible to measure both speakers simultaneously with an RTA?

Or, is it something else? Can stereo speakers be measured with an RTA? Should I still keep looking? for a solution?

[Kal Rubinson]

5. is the clue. There is interaction, constructive and destructive, due to room reflections and the different positions of the speakers. However, this should be instructive as you probably are hearing at the mic position what the mic is measuring .

Kal

[minkuni]

Keep in mind that the microphone is a "dumb" monaural pressure sensor. What you see in a graph is not necessarily what you hear. Try googling around for the terms binaural decoloration and binaural dereverberation.

The brain is a remarkable processor of sensory input

[Dennis H]

Yeah, what minkuni said. Or put another way, at high frequencies, your head partially blocks the sound from the right speaker reaching your left ear and vice versa. So, it doesn't sound as bad as it measures with a mic that 'hears' sounds from all directions about equally.

[Kal Rubinson]

Keep in mind that the microphone is a "dumb" monaural pressure sensor. What you see in a graph is not necessarily what you hear. Try googling around for the terms binaural decoloration and binaural dereverberation.

The brain is a remarkable processor of sensory input

All that helps but the FR aberration is still there and the system would sound better without it.

Kal

[minkuni]

Better get out the head-vise and bolt it to the floor at a location exactly equidistant to both loudspeakers (I assume the room is perfectly symmetrical about the midline as well with regards to walls, furniture, acoustic treatment etc.).

If both speakers are reproducing exactly the same signal (mono) then there will be interference if the distances from the microphone to each loudspeaker are not exactly equal. How much depends on the difference in distances and the contribution from the room reflections.

edit:
In other words, if the measurements are being done for system adjustment - concentrate on one channel at a time (two if using a subwoofer to handle the bass). There will be more than enough interference and suckouts in the measurements from the room contribution, no need to complicate matters

[brianpowers27]

It seems to me the problem you see is comb filtering. If you play the exact same s mono signal from two loudspeakers they are likely to cancel in a comb like structure.
#5 reduces the problem but could never solve it until the two speakers are closer than their physical enclosure allows.

For a fun test, play the rta signal while moving the mic toward/away from the speakers. The peaks/dips should change in frequency.

[fjhuerta]

Thank you all!

I thought at first it was comb filtering, but I wasn't really sure.

It's interesting to think that the FR we eventually hear is the one that the "summed up" signal on the RTA looks like. If this is so, then... could it be that, by pure chance, a badly designed speaker summed up flatter than a perfectly designed one on a real world room?

[Kal Rubinson]

Thank you all!

I thought at first it was comb filtering, but I wasn't really sure.

It's interesting to think that the FR we eventually hear is the one that the "summed up" signal on the RTA looks like. If this is so, then... could it be that, by pure chance, a badly designed speaker summed up flatter than a perfectly designed one on a real world room?

Nearly any combination and permutation of speakers, placement and room can lead to Nirvana or disaster. What science and good practice does is to reduce the odds and make success more likely. Remember, too, that your head is always moving and your brain is sampling the sound over a larger area than does the mic/RTA, even when you think you are sitting still.

Kal

[fjhuerta]

Thanks Kal.. it's great to see someone from Stereophile here, since my next question was going to be...

I've learned a lot from John Atkinson, and I find his measurement features incredibly insightful. Is this the reason why he takes 30 RTA measurements around the listener's head, and averages them?

[Dennis H]

Thanks Kal.. it's great to see someone from Stereophile here, since my next question was going to be...

I've learned a lot from John Atkinson, and I find his measurement features incredibly insightful. Is this the reason why he takes 30 RTA measurements around the listener's head, and averages them?

A minor point, I don't think JA is taking RTA measurements. He's doing his best to filter out the room contribution by taking gated HF measurements and splicing in nearfield LF measurements.

[Kal Rubinson]

A minor point, I don't think JA is taking RTA measurements. He's doing his best to filter out the room contribution by taking gated HF measurements and splicing in nearfield LF measurements.

Right.

Kal

[fjhuerta]

I've seen Atkinson's gated measurements, but sometimes he does RTA measirements at 1/6 octave resolution, And he takes 30 samples. He doesnt do it all the time, though.

Example:

http://www.stereophile.com/standloudspeakers/208fre/index4.html

There's the gated measurement, and.. "Fig.7 shows the spatially averaged response of the Escalantes, calculated by taking 18 1/6-octave-smoothed responses for each speaker individually in a 48"-by-18" vertical rectangular grid centered on the position of Larry's ears in his listening chair. (I used SMUG Software's Fuzzmeasure program running on my Apple PowerBook, in conjunction with a Metric Halo ULN-2 and a calibrated EarthWorks QTC-40 microphone.)"