Hello, DIY community.
A very short list of magazines / reviewers use maximum SPL measurements, such as compression and/or distortion limited SPL, to evaluate speakers. The most prominent is the German magazine Production Partner, and its sister magazines Fidelity Online and Sound & Recording, with Prof. Dr. Anselm Goertz at the helm.
A few years ago, I started to replicate their testing methods & compared my results, with the aim to gain the opportunity to directly compare the performance of my own speakers, or third party speakers I'm interested in and could get a hold of and measure, to the wide variety of speakers that those magazines had already tested.
For those interested in the topic, here's documentation of the process.
Specific Requirements:
Types of Max. SPL measurements:
- Max. SPL @ THD: sine bursts that increase in level until a pre-determined total harmonic distortion limit (THD) is reached; then switching to the next higher frequency (in 1/3 to 1/12 oct steps) and repeating the process. Usual THD limits are 1% (for HiFi speakers), 3%, or 10% (subwoofers, PA speakers). Here's a demo of mine that shows the process on a subwoofer (note: low frequency = large wavelengths only, no flat surface required) -> ARTA / STEPS distortion limited SPL measurement - THAM 15 w/ 15LB075-UW4
- multitone distortion: full range multi sine signal that is increased in level until a pre-determined total distortion limit (intermodular distortion (IMD) & harmonic distortion (HD)) is reached
Comparisons on commercial speakers both the Production Partner and I have tested:
TW Audio T24N:
Frequency response 1m 4pi; Production Partner orange curve, me black curve:
Max. SPL @ 3% THD; Production Partner blue curve, me black curve (note: my measurement intentionally limited @ ~130 dB):
QSC CP8:
Frequency response 1m 4pi; Production Partner red curve, me black curve:
Max. SPL @ 10% THD:
Multitone distortion, 6 tones / oct, EIA-426B output filter, 110 dB RMS; left Production Partner, right me:
The congruence turned out as best as can be expected from commercial speakers with at least a minor amount of series deviation & difference in break-in, difference in environment variables (temperature, pressure, humidity), and known calibration tolerance of the hardware. A 100% match isn't possible - but, as we can see, a ~80-90% match was accomplished, which is more than good enough for most comparisons, where major, not minor differences, are the decider.
That's it. I hope the documentation proved interesting.
Regards
Stoneeh
A very short list of magazines / reviewers use maximum SPL measurements, such as compression and/or distortion limited SPL, to evaluate speakers. The most prominent is the German magazine Production Partner, and its sister magazines Fidelity Online and Sound & Recording, with Prof. Dr. Anselm Goertz at the helm.
A few years ago, I started to replicate their testing methods & compared my results, with the aim to gain the opportunity to directly compare the performance of my own speakers, or third party speakers I'm interested in and could get a hold of and measure, to the wide variety of speakers that those magazines had already tested.
For those interested in the topic, here's documentation of the process.
Specific Requirements:
- frequency calibrated microphone w/ low distortion at high SPL; my choice: Isemcon EMX-7150 (135 dB @ 1% THD)
- SPL Calibrator; my choice: Isemcon SC-1 (Class 2, temperature & supply-voltage compensated; Accuracy +/- 0,5 dB)
- software with Max. SPL measurement routines; my choice: ARTA, STEPS
- reflexion free, low noise measurement environment; my choice: GPM (ref.: Gander, AES; Erin) on smooth concrete strip in rural outdoor environment on wind free days
Types of Max. SPL measurements:
- Max. SPL @ THD: sine bursts that increase in level until a pre-determined total harmonic distortion limit (THD) is reached; then switching to the next higher frequency (in 1/3 to 1/12 oct steps) and repeating the process. Usual THD limits are 1% (for HiFi speakers), 3%, or 10% (subwoofers, PA speakers). Here's a demo of mine that shows the process on a subwoofer (note: low frequency = large wavelengths only, no flat surface required) -> ARTA / STEPS distortion limited SPL measurement - THAM 15 w/ 15LB075-UW4
- multitone distortion: full range multi sine signal that is increased in level until a pre-determined total distortion limit (intermodular distortion (IMD) & harmonic distortion (HD)) is reached
Comparisons on commercial speakers both the Production Partner and I have tested:
TW Audio T24N:
Frequency response 1m 4pi; Production Partner orange curve, me black curve:
Max. SPL @ 3% THD; Production Partner blue curve, me black curve (note: my measurement intentionally limited @ ~130 dB):
QSC CP8:
Frequency response 1m 4pi; Production Partner red curve, me black curve:
Max. SPL @ 10% THD:
Multitone distortion, 6 tones / oct, EIA-426B output filter, 110 dB RMS; left Production Partner, right me:
The congruence turned out as best as can be expected from commercial speakers with at least a minor amount of series deviation & difference in break-in, difference in environment variables (temperature, pressure, humidity), and known calibration tolerance of the hardware. A 100% match isn't possible - but, as we can see, a ~80-90% match was accomplished, which is more than good enough for most comparisons, where major, not minor differences, are the decider.
That's it. I hope the documentation proved interesting.
Regards
Stoneeh
Last edited:
Using a using a perceptual based weighting in your distortion measurement, for example as in CEA 2010, can be a useful extension of this.
CEA-2010 & its modernisations, up to CTA-2010C, is probably the most meaningful max. SPL routine for subwoofers - but, for subwoofers only.
Only perception weighted routine used for full frequency spectrum I'm aware of is the GedLee Metric, which, warranted or not, has not been widely adopted. My view is that multitone distortion testing performed by the Production Partner & replicated by me, which is at least output filtered to represent a music signal, is probably fine for that purpose.
Only perception weighted routine used for full frequency spectrum I'm aware of is the GedLee Metric, which, warranted or not, has not been widely adopted. My view is that multitone distortion testing performed by the Production Partner & replicated by me, which is at least output filtered to represent a music signal, is probably fine for that purpose.
Thank you that is great information!
I do wonder with MD Anselm Goertz did measure -17 dB Distortion below 1 kHz while you measure -30 dB Distortion. In his measurement 12 dB crest factor is noted. Maybe 6 tones per octave isn't enough.
Do you know how Klippel Measures MD? @bikinpunk what do you use to measure MD? Would be very curious since you are the only person that does MD measurements of hifi speakers.
I do wonder with MD Anselm Goertz did measure -17 dB Distortion below 1 kHz while you measure -30 dB Distortion. In his measurement 12 dB crest factor is noted. Maybe 6 tones per octave isn't enough.
Do you know how Klippel Measures MD? @bikinpunk what do you use to measure MD? Would be very curious since you are the only person that does MD measurements of hifi speakers.
Last edited:
I did measure the arta output with audacity for 10 seconds, Amplified the signal to 0 dB peak. Measured the rms value for 12, 9 and 6 tons per octave. Why is the crestfactor the same?
I've been receiving some PMs, basically on "how to?". Most to all questions will be answered by reading these essential, or at least highly recommended documents. No, that's not light and short reading, and if you can't afford a large amount of time to learn these measurement methods in theory and practice, as well as bring to the table a decent amount of knowledge of acoustics and loudspeaker mechanics, I don't recommend starting (not seeking to discourage, just being realistic):
ARTA User Manual
ARTA Tutorial (additional, practice oriented tutorial)
STEPS User Manual
STEPS Tutorial (esp. Pt. 3.5; again, many valuable tipps from practical experience)
Re-reading the point "Specific requirements" in the OP, and the links within, could also prove beneficial; especially since the measurements can't be gated, and thus the only suitable measurement setup is GPM, free field, anyway. (In Situ correction would be possible in principle, but ARTA doesn't support it.)
Important: a reflection free environment is not only necessary for representative measurement results, but in the Max. SPL @ THD measurement also for speaker protection. Reflections will impact the harmonics different than the fundamental, so the THD reading will be wrong (too high or too low, depending on freq. & wavelength). THD increases with membrane excursion, so a THD limit in the measurement routine is essentially an excursion limiter for the speaker. However, if the program reads the THD lower than it actually is, it will try to increase power to the speakers until a much higher THD, meaning much higher excursion. A voltage limit should be imposed on the DUT anyway, as the tutorials note, but such a thermal-electrical limit will not protect every speaker mechanically under all circumstances.
Also, attached is my EIA-426B filter as ASCII, importable in ARTA:
ARTA User Manual
ARTA Tutorial (additional, practice oriented tutorial)
STEPS User Manual
STEPS Tutorial (esp. Pt. 3.5; again, many valuable tipps from practical experience)
Re-reading the point "Specific requirements" in the OP, and the links within, could also prove beneficial; especially since the measurements can't be gated, and thus the only suitable measurement setup is GPM, free field, anyway. (In Situ correction would be possible in principle, but ARTA doesn't support it.)
Important: a reflection free environment is not only necessary for representative measurement results, but in the Max. SPL @ THD measurement also for speaker protection. Reflections will impact the harmonics different than the fundamental, so the THD reading will be wrong (too high or too low, depending on freq. & wavelength). THD increases with membrane excursion, so a THD limit in the measurement routine is essentially an excursion limiter for the speaker. However, if the program reads the THD lower than it actually is, it will try to increase power to the speakers until a much higher THD, meaning much higher excursion. A voltage limit should be imposed on the DUT anyway, as the tutorials note, but such a thermal-electrical limit will not protect every speaker mechanically under all circumstances.
Also, attached is my EIA-426B filter as ASCII, importable in ARTA:
Attachments
On page 50, the ARTA manual states:
"Multitone signal contains sine signals spaced 1/R octave in the user defined range from 5 to 20000 Hz. Crest factor is 13 +/- 1 dB. Resolution R can be user defined from 1 to 12 tones per octave."
Your Audacity levels are within .25 dB of a 13dB crest factor, verifying the ARTA output is producing the correct output level.
ARTA's three types of periodic noise, White PN, Pink PN and Speech PN, all have a crest factor of 12-13dB. That said, page 45 says:
"By further processing, the crest factor can be lowered to values below 6dB.
In ARTA, the periodic noise generation is implemented with a crest factor lower than 10 dB."
ARTA has menu commands to show the RMS value and crest factor:
So you can check the value of whatever signal you are generating before applying it.
Fixed or swept sine has only a 3dB crest factor, so heats voice coils rapidly by comparison to the other signals.
Art
I was wondering how it is possible that the crest factor is not dependent on the number of tones per octave used. Arta automatically reduces the Amplitude with increased number of tones to keep the rms value constant and additionally phase shifts the tones to maintain the same peak value.
AES75 is another way of measuring maximum linear SPL that might be of interest. Seems to now be the standard in proaudio. https://www.aes.org/standards/AES75/
Eclipse Audio offers freeware to run it. https://eclipseaudio.com/speakermeasure/
Eclipse Audio offers freeware to run it. https://eclipseaudio.com/speakermeasure/