measuring speaker frequency response? (1 Viewer)

[Steve_AS]

For tedious reasons, I want to measure the outputs each of my five satellites , across the frequency range, for comparison purposes. Among the possible tools I have at hand are a Rat Shack analog SPL meter, two Stereophile Test Discs (#s 1 and 2), Avia and Sound&Vision test DVDs, an an analog volt/ohm meter.

What is the msot practical way for me to set up this measurement (positioning of the meter wrt the speaker being measured, what calibration setting to use, etc). ?

It needn't be 'anaechoic chamber' quality, I just want to get an idea of how well-matched my speakers are across the audible frequency range. From subjective listening I'm getting the impression that one of them may be intrinsically 'off' compared to the others (even though they're identical models) -- not due to room placement.

 

[MingL]

With what you have, you could measure them SPL levels using sine tones at different frequencies at regular intervals. Eg. 125hz, 250hz, 500hz, 1000hz, 2000hz............. finer intervals = finer resolution = very tedious.

Or you could connect your SPL meter to a computer via line-in and use spectraplus or trueRTA to do a FFT to get the freq response.

Would this be what you intend to measure?

 

[Kevin C Brown]

Yeah, I would suggest the Radio Shack meter, a PC, and:

http://www.etfacoustic.com/

I have done the "plot discrete CD test tones on a spreadsheet" thing, and it is very tedious. Much easier to use a PC.

 

[Mike Dzurko]

If you're trying to make sure two speakers are the same, make sure you do your measurements with them in the exact same position, and with the mic placed exactly the same in relationship to the speakers.

 

[Marc_J]

I'm not sure if the response of the rat shack meter will be flat enough to get a really accurate measure. The guys at rivesaudiocom have a test CD that has tones that compensate for the irregularity. I've personally not used it, but it seems like a good idea.

jm2c

 

[Steve_AS]

With what you have, you could measure them SPL levels using sine tones at different frequencies at regular intervals. Eg. 125hz, 250hz, 500hz, 1000hz, 2000hz............. finer intervals = finer resolution = very tedious.

Or you could connect your SPL meter to a computer via line-in and use spectraplus or trueRTA to do a FFT to get the freq response.

Would this be what you intend to measure?

Yup. I'm not interested in the system response, or the room response, I want to compare as best I can how each of the five supposedly identical speakers performs over the audible frequency range. This means measuring each speaker under the same conditions; I'm curious to know what the best conditions would be, this side of building an anaechoic chamber -- how far to place the mic from the speaker, shoudl the speaker be on a stand or on the floor, etc. I already know about the Rat Shack metere needing correction factors. I haven't tried any PC-based measurment software.

 

[Wayne A. Pflughaupt]

I really can’t see the point of doing separate measurements on identical speakers – except to see how the room affects their response at the various locations.

That said, accurate measuring of speaker response is tricky. It will take an instrument or program with extremely fine resolution for both “horizontal” and “vertical” readings – say, 1/24-octave or finer for frequency response (“horizontal”) and dB calibrations of 1/2dB or so (“vertical”).

In addition, it will be impossible to eliminate the room as a factor with any consumer-grade equipment. Professional testers use equipment that takes instantaneous measurements as soon as the sound wave hits the mic, then shuts off before the room reflections arrive.

If you think taking the readings will be tedious, wait until you start analyzing and comparing them. I think you will find that while subtle differences are easy to detect by ear, it will be difficult to “see” or find where those differences show up on a response chart. For instance, a deviation of 1/2dB or so over a limited range will be audible, but so close in measurement to the reference speaker that exactly locating where the deviation appears will be difficult to determine.

Actually, I think you will find that even 1/24-octave resolution will prove inadequate. As the frequency cycles go higher, there are hundreds and even thousands frequencies between each octave. As you can see, 1/24-octave resolution, while excellent for taking room measurements, is pretty coarse for measuring actual speaker response. If those 1/2dB deviations I mentioned above happen to fall between the 24 sampled points, you’ll hear them but you’ll never see them.

Regards,
Wayne A. Pflughaupt

 

[MingL]

I personally prefer to measure the speakers as how they are heard; including room and system response.

Calibrating by computer using ETF is one of the better ways to measure freq response.

To closely approximate anechoic, try measuring 1m away from the speaker front. As always, there are flaws with this sort of of close mic measurement, thus I tend to prefer meas that include system+room response to correct for flaws in the entirety.

 

[BruceD]

Wayne,

In addition, it will be impossible to eliminate the room as a factor with any consumer-grade equipment. Professional testers use equipment that takes instantaneous measurements as soon as the sound wave hits the mic, then shuts off before the room reflections arrive.

Actually the ETF software mentioned by Kevin and Ming allows you to perform this exact function. You can adjust the capture time (it's called windowing) to only display the frequency response from the initial impulse signal without any room reflections.

But I agree with you and Ming, combined room/speaker response is more useful.

 

[Kevin C Brown]

You can also input the correction curve into ETF too. Improves the accuracy of the Radio Shack meter.

 

[Edward J M]

If you are getting the impression that one of five supposedly identical speakers is "intrinsically off", I would imagine it might be defective in some manner and an audible anomaly should show up on a sweep, even one as "rough" as 1/24 octave.

As for the room, if you measure each of the five speakers in the same exact location, in the exact same manner each time, the room itself will cease to become a variable between speakers, even though it will clearly influence the frequency response of the speaker.

While I don't think it's needed in this case, if you really wanted to eliminate the room, just run a ground plane sweep on each speaker.

As for equipment, you can run from mild to wild, but a decent measurement mic might run $40-$50. A decent mic mixer with phantom power will run about the same. Then you have the software investment of say $100 for True RTA with the 1/24 octave capability. Then you have to procure (or have built) custom interconnects to go between the mic mixer, the sound card, and the AVR or pro/pro. I'd say you are looking at about a $300 investment at that point. While not exactly "pro grade", it will be FAR more accurate than using the RS meter, and it will give a total system response of 20-20k +/- 1 dB after True RTA corrects for sound card and mic response anomalies.

 

[RichardHOS]

Steve,

There are two options available to you. One if you want to use your RatShack SPL meter, then you should take nearfield measurements of the speaker. Set the SPL meter on books or somesuch so that it is roughly at a vertical height that is equal to the midpoint between tweeter and woofer of your surrounds (I assume that these are two-way designs?).

The speaker can be on a floor, a stand, or whatever. So long as the conditions are the same for each speaker, you will still have a vaild comparison (which is all it seems you're after... not absolute frequency response).

Place the microphone within a couple of inches from the front baffle to eliminate most of the room effect. At this close distance, you will likely measure peaks and dips in the response as a result of comb filtering in the nearfield, but again... for relative comparisons this is fine so long as for each speaker both the SLP meter and speaker are in the exact same position.


The other method is using a software package such as ETF, which was mentioned above. You'd still place each speaker and the microphone in identical positions for each speaker being measured, but with ETF you wouldn't need to take extreme nearfield measurements (note that you can use the RatShack SPL meter in combination with ETF, so if you have an adequate sound card, you shouldn't need to purchase anything other than the software itself). A mic distance of a couple of feet will allow a better picture of the combined absolute tweeter and woofer response, as the comb filtering will be much less severe at that distance. ETF uses a windowed maximum length sequence and transform analysis to measure the response of the speaker in the time window before most room reflections arrive, providing a "semi-anechoic" response down to some lowest frequency that is determined by the window length used. This method is more accurate in absolute terms, but either will probably work for relative comparisons.

I would point out though that frequency response abberations might only show up in a very narrow bandwidth, so making measurements with the SPL meter you'd need pretty find graduations of test tones, and that's going to be very tedious for multiple speakers and take quite a bit of time to get anything like accurate results.

 

[BruceD]

(note that you can use the RatShack SPL meter in combination with ETF, so if you have an adequate sound card, you shouldn't need to purchase anything other than the software itself).

While I do advocate that you buy the software (I started with version 4 and find ETF indispensable), you can download the software for a free trial run for use with the RS SPL meter.

 

[AndyDC]

I thought the Rives disk would be nice for quick-and-dirty checks (I know, not what this person wanted, but I thought I'd mention it). If the system is reasonably well set-up, you can just run through the tracks without adjusting the settings on the meter too much. It's good for quickly checking a system in a house you're visiting, or a store.

But it's only 1/3 octave tones, so it's no use at all for doing eq or any serious analysis.

 

[RichardHOS]

Ahh, thanks for pointing that out BruceD, I had forgotten about their trial version. That's what the trial version is for, after all... try it out on a "real world" problem and see if it is useful to you.

Then again, doesn't the trial version prevent you from running and saving multiple sweeps/measurements?

Erm... you can always make a quick screenshot of the response graph for the purpose of comparing one speaker's response to another.

 

[BruceD]

Richard,

Then again, doesn't the trial version prevent you from running and saving multiple sweeps/measurements?

Erm... you can always make a quick screenshot of the response graph for the purpose of comparing one speaker's response to another.

You are correct, the trial version doesn't allow multiple measurements, saving, or printing. But like you said, you can save a screen shot.

 

[Kevin C Brown]

Actually, the last trial version of ETF I used allowed 2 sweeps. Then you had to exit the program and start it up again.

 

[Steve_AS]

I really can’t see the point of doing separate measurements on identical speakers – except to see how the room affects their response at the various locations.

To check if they really are identical, despite being the same model. In my case they're from different production runs (separated by years). And rough measurement seemed to indicate that at least one of them is outputting a few dB louder than the others at some frequecy..also appears to be a resonance rattle in one of them that's not in the others.

I accept that with my equipment it'll be iompossible to eliminate room effects, but if they're kept the same for each measurment, then any differences between the speakers should still be detectable. (the 'all other things being equal' approach.)

 

[Steve_AS]

You guys rock, btw. Thanks for all the input.