Some basic FR measurements

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Last year I finished the first round of tweaking a new passive crossover for my three-way speakers (Audax HM210Z0 and HM100Z0 and Scan-Speak D2905-9300). The speakers sound very nice and very easy to listen to, but are perhaps not as transparent or immediate as they might be with my push-pull 300B monoblocks, and have what I perceive to be a dark tonal balance.

I did some electrical measurements of the crossovers, and after some substantial work on impedance correction, they are now an excellent match to 4th-order Linkwitz-Riley, as intended.

So to get an idea whether the warm bias of the sound was an issue with the crossovers themselves or a room problem, I thought I would do some simple acoustic measurements. I played 20 seconds of a white noise track from a test CD, and then used a Panasonic EM60 microphone into an EMU0202 interface to feed this into Audacity, which I used to calculate a power spectrum. The mic is on axis about 50cm from the respective midrange driver, but for the moment I'm not being obsessive about position.

The room is about 4-5 metres square, with a couple of metres square cut out of one corner for the doorway; the room is carpeted, with soft furnishings and Venetian blinds on the windows. The speakers are about a third and two-thirds of the way along a long wall, and are about 20cm from the wall with a slight toe-in.

Here are the responses of the two speakers, first the left then the right channel:

An externally hosted image should be here but it was not working when we last tested it.

An externally hosted image should be here but it was not working when we last tested it.


The responses are quite consistent between channels, which is reassuring, and also perhaps surprising, since the room is quite asymmetrically laid out.

I note the following:

- A raised response between 1.5kHz and 2kHz;
- Some peaking between 3kHz and 5kHz;
- The relative magnitude of the peaks at 2kHz and 4kHz is different in the two speakers;
- A rolloff above 6kHz. I wouldn't expect this from the Scan-Speak tweeters, as they have a very flat response up to 15kHz or more (though I am measuring slightly off the tweeter axis).

So - first of all, should I expect to get results from this technique which are at all useful? In other words, are the three features I mention above likely to be real? Secondly, using white noise at 50cm, how much would you expect room effects to influence the measured response?

The regular peaks at the HF end (separated by about 800Hz) look suspicious to me. Are they likely to be a measurement artefact?

Any other comments would be welcome!

Thanks,

Alex
 
Last edited:

Attachments

  • LH_50cm_response.gif
    LH_50cm_response.gif
    32.8 KB · Views: 106
  • RH_50cm_response.gif
    RH_50cm_response.gif
    32.7 KB · Views: 105
Hi Alex,

I think your measurements show some merit and should correlate to some degree with what you hear.

With regard to the periodic peaks above 2k, those are typical of having a strong early reflection. You noticed they have a regular spacing of 800 Hz and that confirms they are from a delayed reflection of about 1.25mSec delay. That would typically come from a reflecting point (say a cabinet edge or cabinet top that the tweeter "sees") that is about a foot or so distant from the tweeter. If you can track that down you should be able to find a way of treating it with damping foam or fiberglass.

Your low frequency curves look very smooth, devoid of any typical room effects. Is this some type of windowed or gating measurement system?

David S.
 
Hi Dave,

Yes, I wondered about reflections too, but there is certainly nothing that close to the speaker or to the microphone. I'll have another look at my setup this evening and think about it.

I am using the plain vanilla default settings in Audacity: I have a feeling that some windowing is selected, but digital signal processing is something I am relatively unfamiliar with. I will check this out later too and report back.

Alex
 
The speakers sound very nice and very easy to listen to, but are perhaps not as transparent or immediate as they might be with my push-pull 300B monoblocks, and have what I perceive to be a dark tonal balance.

The mic is on axis about 50cm from the respective midrange driver, but for the moment I'm not being obsessive about position.

- A raised response between 1.5kHz and 2kHz;
- Some peaking between 3kHz and 5kHz;
- The relative magnitude of the peaks at 2kHz and 4kHz is different in the two speakers;
- A rolloff above 6kHz. I wouldn't expect this from the Scan-Speak tweeters, as they have a very flat response up to 15kHz or more (though I am measuring slightly off the tweeter axis).

So - first of all, should I expect to get results from this technique which are at all useful? In other words, are the three features I mention above likely to be real? Secondly, using white noise at 50cm, how much would you expect room effects to influence the measured response?
Alex,
The response is representative of what the microphone is picking up at the location you placed it at, and the microphone's frequency response.
Without testing the microphone against a known "flat" mic you can't be 100% sure what you are seeing is what the actual response is.

The high frequency roll off probably looks to be worse than actual because the mic is off axis from the tweeter. Placing it midway between the midrange and HF would give a better impression of the crossover region and HF rolloff. Testing outdoors at quite some distance from the ground will determine if the early reflections are cabinet or room related.

Multiple mic locations in room will also give you a better idea of what the cabinet response compared to room response is.

10 dB variations equate to sounding half as loud or twice as loud referenced to 1000 Hz.

You say you used white noise (which sounds like the hissing you hear between radio stations) as a source, but the response looks like pink noise (equal volume per octave) was used.
 
Suggest removing the tweeter connection, test, then remove the woofer connection, test and you can see what the drivers are doing separately. Assuming parallel xover.

If you don't see the same ripple separately, then you have interference going on...

Also it is good to look at just the impulse. Well, I like to.

_-_bear
 
Theres quite a few of dedicated f/r software, try to switch to one of these.
As pointed out, you need to ensure your mic/preamp is calibrated.
Then you can measure each drivers output to ensure that your xover & driver combination results in the desired acoustic LR4 output.
 
Just another Moderator
Joined 2003
Paid Member
I use holmimpulse HOLM Acoustics for gated measurements and REW REW - Room EQ Wizard Home Page for doing pink noise (with RTA) in room measurements (also great for measuring impedance curves)

I like your thinking with using audacity, but I think that using REW with pink noise and some averaging in the RTA you will probably get a better in room representation.

Tony.
 
Thanks for the helpful suggestions.

I still can't see any obvious source for reflection at that distance. I wondered whether it might be a front-back reflection inside the mid-treble cabinet, since the boxes are about 35cm front to back, and the ultra-lightweight HDA midrange cone is notorious for passing the reflected backwave. However the tweeter is closed at the back, and the crossover is at about 3.2 kHz, so you wouldn't expect any substantial internal echoes at higher frequencies than that. I will try measuring with the speakers further away from the walls, just to see.

Yes, I see that "size" in Audacity defaults to 512 - a quick test reveals that increasing this extends the bass response accordingly, but I haven't repeated my measurements with this setting.

I am definitely feeding my amp white noise: it is track 65 of the Hi-Fi News Test CD, marked "white noise". Pink noise sounds like a waterfall, instead of the "hiss" I hear with this track.

It had indeed occurred to me to measure the drivers individually with and without the crossover, to test whether they (the mid in particular) are working consistently with the published specs.

I will also measure with a different (solid state) amplifier to eliminate my 300B amp as a possible cause of the HF rolloff.

Finally I will investigate other software packages. Holm appears not to be available for a Mac, but I will try REW.

Alex
 
Last edited:
There are definitely strong reflections going on, as evidenced by the periodic ripple. The lower curve is worse and the 6dB peak to trough indicates that the reflection strength is about 1/3rd the direct sound strength. It could also be a reflection at the microphone end. What is your mic support/stand like?

White noise is correct for this type of analyzer.

David S.
 
I've now installed REW on my MacBook, and started to have a look through the comprehensive manual and FAQ. There are instructions for using an SPL meter to calibrate a microphone, which reminded me that I have an old RS analogue SPL meter (modded as per Eric Walin's instructions), which I dusted off after ten years in storage and put new batteries in. I had forgotten that it also has a fixing for a camera tripod, which I had never used, but might well be handy.

I'll experiment and report back!

Alex
 
I think I have solved the reflection issue. I re-did the measurements on one of the speakers using the RS analogue SPL meter instead of my homebrew microphone, and used a higher data length for the FFT in Audacity, and got the result below, which looks much more encouraging: I don't see the ~800Hz repeat at all.

The mic I was using before consisted of the Panasonic capsule in the end of a 30cm plastic tube, which was wrapped in tape to suppress vibrations: my feeling is that the reflections occurred along the interior length of the tube (possibly through the cable), and would have been relatively undamped at high frequencies.

The overall response is much smoother through the midband too. I will experiment with REW over the next few days to try to improve the measurements.

Alex
 

Attachments

  • LH channel new SPL meter.gif
    LH channel new SPL meter.gif
    35.6 KB · Views: 52
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.