Questions about taking relevant speaker measurements / spinorama data

[edc]

Hi all, this is my second thread here, and I would like to know more about doing correct speaker measurements/spinoramas.

I have a few pairs of vintage speakers that I would like to do spinorama measurements of and post them here on ASR. I have a few questions about how to take accurate and relevant speaker measurements.

Here is a JBL L15 which I would like to measure correctly.


Some months ago I built a spinorama platform, that's shown on this post. The platform holds the speaker at a 2-meter height, and there is 2m sphere of clearance in all directions, the same applies to the microphone.

I've read the excellent guide written by napilopez (how-to-make-quasi-anechoic-speaker-measurements-spinoramas), and I've made my first spinoramas using REW and VituixCAD.

---

Gear I use:
Microphone: Dayton Audio EMM-6 + calibration file
48V mic interface: iRig Pre 2 + calibration together with the ADC & DAC
ADC & DAC: 2x Apple 3.5mm adapters, one used as input, and the second used as output (they both work at the same time if forced in headphone/headset mode)
Amplification: Fosi Audio V3 + 48V PSU
SPL meter, voltage/ohm meter, unfortunately, I don't have any way to measure impedance response.

---

Questions I have:

1. How should the speaker be positioned on the spin platform? Should the center of rotation be the center of the drivers or the center of the cabinet? In the case of the JBL L15, the drivers are centered to each other, but they are offset on the baffle. Same question for the vertical position: should the center of rotation be the cabinet or the point between the tweeter & woofer?
2. Should the front baffle of the speaker be aligned with the rotational point, or should the averaged voice-coil position be at the center of rotation? In the case of a coaxial / horn driver that has the driver deeper inside the cabinet compared to a dome tweeter mounted on a flat baffle.
3. How should the microphone height position be decided relative to the driver arrangement? I'm assuming the mic should always be at the center of rotation, but what about the height, as this seems to influence the results the most? Is it ok to find the height (within reason / between drivers) that provides the smoothest response and go with that or is that biased?
4. During sweeps, is it ok to power the speakers using 2.83V measured from the amp (400Hz), or is it better to have a standard SPL value for all speakers, eg 86dB?
5. Is the 1-meter distance between the mic and speaker enough or should there be a 2-meter distance during the spin measurements?
6. Considering a 5ms impulse window width in REW, is 64k sweep length enough for spins, and 256k enough for distortion measurements? What about for ARTA burst decay?
7. Is it possible to measure "Instantaneous Compression" using regular equipment, or is there a need for a physical dedicated generator + analyzer? Same question for multitone distortion.
8. What are other measurements that offer useful information about performance? Step response, group delay,...?

Thanks!

 

[NTK]

Many of your questions can be found in ANSI/CTA-2034 "Standard Method of Measurement for In-Home Loudspeakers", which is available for free.

ANSI/CTA-2034-B

This standard describes how to determine the frequency response, directivity, and maximum output capability of a residential loudspeaker. It is intended to determine the audio performance of a loudspeaker, not the loudspeaker's ability to survive a given input signal. This standard applies only...

shop.cta.tech

1. Below are the figures showing how the speaker should be positioned during the test (spin).
2. See above
3. The mic should be aligned to the "reference axis" of the speaker, defined in section 7.3.
4. Sections 6.3/9.1 specifies 2.83 V as the standard for passive speakers. For powered speakers, the speaker should be tested with averaged SPL from 500 Hz to 2 kHz of 79 dBSPL at 2 m distance (85 dBSPL when corrected to 1 m).
5. Standard measurement distance is 2 m (section 7.4). For small speakers, 1 m is probably fine.
6. If measuring with sampling rate 96 kHz, 64 k, a sweep will be 0.683 seconds long. It should be enough for gated FR measurements. The Klippel NFS uses multiple quick sweeps to test, see video:
   https://www.facebook.com/video.php?v=609534989795383
7. I don't see why "regular equipment" won't work. Of course they need to be suitable for the job, e.g. an amplifier capable of supplying unclipped signal to the speaker, which you will need to verify for signals with high crest factors such as pink noise and multitone. You may also need some means of automation as some tests can be timing sensitive. For example, you may not want to wait too long between compression tests to keep the speaker voice-coil temperature from changing too much (also for consistency between tests).
8. You can get a lot of information such as step response, group delay, and etc. from the frequency response measurements.

 

[edc]

The lower part of the image answered the questions. Very informative.

Thank you for your response NTK!

 

[edc]

Since the CTA-2034 doesn't specifically mention the stimulus type and length required during measurements, I spent a few hours testing different configurations in Arta to see what effect they have on results.

I used Arta's "Spatial impulse response group record" to help me with the directivity measurements. In Arta it's possible to use as stimulus multiple signal types: Periodic Noise (PN), Sweeps or MLS (maximum length sequence). Another advantage of Arta is the ability to choose a very short stimulus length of 16k, which in REW is limited to 64k as minimum (for good reasons). A short sequence length will speed up directivity measurements considerably, but it lacks resolution for other measurement types such as distortion.

I've taken 6 individual 360deg measurements using the JBL L15 on horizontal axis (position didn't change between measurements) with the following settings to observe the results: MLS 16k, Periodic Noise 16k, sweep 16k, sweep 64k, sweep 256k, sweep 16k with 8.6 V @ 1 m.

Measurements were done on spin platform, 1m distance, tweeter center as the reference axis, 10ms IR gate.
I used a self-leveling laser at two positions to align the microphone to the reference axis.


Contour plot:


Waterfall:


The highest deviation I can notice is the 16k sweep where the power was increased from 2.83 V to 8.6 V, otherwise there doesn't seem to be major differences between stimulus type and sequence length. All within margin of error, since the rotation angles were set by hand at 10deg intervals, which could have a small impact.

I'm surprised of these results, which indicates that a sweep of 16k length has enough resolution to show the speaker directivity performance. Please let me know if I'm wrong

Raw Arta files of all 216 measurements (.pir, .dpf): Google Drive 50MB