Complaint thread about speaker measurements

[amirm]

There seems to be never ending set of protest posts, complains, personal insults, etc. regarding speaker measurements in the review thread. That is not what those threads are about. They are making it difficult for people who just want to read a review about a speaker to follow through. Seemingly the same argument is also dragged from on review thread into another.

From now on, if you have a beef about how or why speakers are measured, this is the thread for it. If you post it in the review thread, Thomas or I may delete or move it here depending on how we feel about you at the moment. I suggest not taking the chance and vent here.

I also like to remind the membership that this is a professional forum and highly values positive and constructive contributions.

 

[RayDunzl]

Not a complaint, but a request:

@amirm

Please include distortion measures with speaker tests.

While your background noise level will affect the measures for speakers with minimal distortion, it should still reveal those that border on gross distortions.

An on-axis sweep, and a few selected single frequencies would suit me. Maybe select two tones within the range of each mounted driver.

Level?

Needn't be terribly high. Let's say 90dB SPL at the mic to keep things within reason.

If the harmonics are present there, they likely will only increase at higher levels.


Here, a single tone at 10 feet.

JBL LSR 308 cone (woofer range)

ML electrostat (panel range)

While "audibility" may or may not be decided, we should be able to see which devices sing their own tune along with whatever the source provides.

I should think this is a fundamental aspect of speaker design, to omit it doesn't make sense to me, at least, I find it to be important to me.

You decide.

 

[amirm]

I can include them if no one asks me to prove they are right. Deal?

 

[Thomas savage]

There seems to be never ending set of protest posts, complains, personal insults, etc. regarding speaker measurements in the review thread. That is not what those threads are about. They are making it difficult for people who just want to read a review about a speaker to follow through. Seemingly the same argument is also dragged from on review thread into another.

From now on, if you have a beef about how or why speakers are measured, this is the thread for it. If you post it in the review thread, Thomas or I may delete or move it here depending on how we feel about you at the moment. I suggest not taking the chance and vent here.

I also like to remind the membership that this is a professional forum and highly values positive and constructive contributions.

Wow , my mum can finally be proud of me now I'm associated with something ' professional ' .

 

[Purité Audio]

You haven’t been with a ‘pro’ before?
Keith

 

[Ron Texas]

No complaint here although some of this stuff is hard to digest.

 

[dshreter]

Not a complaint, but a request:

@amirm
Level?

Needn't be terribly high. Let's say 90dB SPL at the mic to keep things within reason.

I don’t have a POV on appropriate level, but would like something that can survive from large full range speakers to small desktop monitors if possible, so we don’t end up with multiple bases for comparison.

If necessary, this could be paired with power handling, but I’m personally less interested to see that measured and satisfied to read off a manufacturer spec.

 

[amirm]

The only easy measurement for distortion is the one that is automatically generated as part of the measurements. This means the same level rest of measurements are performed at.

 

[sergeauckland]

The only easy measurement for distortion is the one that is automatically generated as part of the measurements. This means the same level rest of measurements are performed at.

90dB SPL seems to be the norm, if not actually a formal standard, for measuring distortion. A sweep based on 90dB SPL at 1kHz would be great, but otherwise three spot frequencies of 100Hz, 1kHz, 10kHz would suffice to characterise what the 'speaker does.

As you build up a corpus of measurements, distortion measurements like on and off axis responses will give a comparative measure, but it then becomes essential to maintain a standard methodology, as any differences in method render comparisons difficult if not meaningless.

Nevertheless, a great start! I only wish I could get my home-built 801s measured on the Kippel rig.

S

 

[Thomas_A]

What is the official specification (error range) that Klippel give ?

Since the measurement > 2 kHz is using gating, how are reflections from the setup equipment itself handled, e.g. microphone attachment/holder?

 

[hardisj]

The only easy measurement for distortion is the one that is automatically generated as part of the measurements. This means the same level rest of measurements are performed at.

Help me out here... So the distortion measurements are ran simultaneously with the NFS measurements? What module does it use? I'd like to look up the AN for it to learn a bit about it. When I was using the Klippel I used DIS. Now they have DIS3. Not sure what the difference there is. But with that and/or TRF you could measure distortion as you wished. Measurements were done nearfield (distance relative to driver diameter).

As for level, I would often measure at 90dB, 96dB, and 102dB to observe trends (nearfield; spl was adjusted). I did this with tweeters without passive or active protection but in some cases would limit the low frequency to 500hz or so. I occasionally measured very low sensitivity drivers and those were fun when trying to get them to 102dB @ 1m equivalent.
Off-axis won't yield any difference because it's a nearfield measurement (there is no off-axis this close). Unless you are measuring outside in a very quiet environment or an anechoic chamber, nearfield is the only way to get useful distortion in the low frequencies (same principle as deriving FR which the Klippel NFS uses).

 

[Thomas savage]

You haven’t been with a ‘pro’ before?
Keith

Well done ha ha

 

[Fluffy]

I would also like to see some kind of linearity test, to see how loud a speaker can go before it's noticeably compressed/distorted.

 

[sergeauckland]

I would also like to see some kind of linearity test, to see how loud a speaker can go before it's noticeably compressed/distorted.

Yes, compression would be a good test. Good 'speakers don't do more that 1dB or so at 100dB SPL, but I have no idea how smaller 'speakers perform. I'd would guess that some, like the LS3/5a won't even get to 100dB, at least not at 2m.

S

 

[Blumlein 88]

My as well put it out there. You could avoid virtually all of the questions you don't like by remeasuring the kh 80. Do I understand right that you can change the axis without remeasuring? The other thing was level of the bass.

Your review was timely and helpful as I was about to buy some 2nd hand. The lack of output in the low end makes it a no go. So your review helped me.

Whether you see it or not a redo on the KH80 is a benefit to you, the forum and questions about your methodology.

 

[Cahudson42]

Needn't be terribly high. Let's say 90dB SPL at the mic to keep things within reason.

If the harmonics are present there, they likely will only increase at higher levels.


Here, a single tone at 10 feet.

JBL LSR 308 cone (woofer range)

@RayDunzl Do we also need a simultaneous multi-tone for IMD?

 

[napilopez]

Originally thought it'd be best to keep concerns on the related speaker, but given they don't seem to be going away, crossing onto other threads, and that some of it has devolved into personal insults, it's probably best to put them in one place.

Obligatory axis comment:

And scaling. I'm guessing the klippel software doesn't provide much control over scaling the measurements so I know you're resizing the graphs manually. Just it's a bit odd to see all the different graph sizes, making some speakers look better or worse, and I know. Maybe some feedback you can give klippel to let you manually set an aspect ratio?

(Though to be fair Harman isn't great about being consistent nor following CTA-2034-A either. I'm actually not even sure the language in CTA-2034-A is accurate to their images either lol)

I would also like to see some kind of linearity test, to see how loud a speaker can go before it's noticeably compressed/distorted.

This is especially important for active speakers, as they will be the ones most affected by it in real-world listening situations, as the compression is a hard limit. It would be useful to know how much headroom they have.

 

[Costas EAR]

Please don't remeasure the Neumann 80's.

Please measure the level of self generated noise on active speakers and everything coming out of the bass reflex tube.

I would like to see a max spl graph for every speaker at 1&3% THD, like the Neumann graphs:

This graph is really what i want for every speaker.

Thanks!

 

[NTK]

I would like to point out a lesser discussed part of CTA-2034-A — the on-axis maximum usable SPL test (section 8).

The test has 2 parts, the first is the maximum usable continuous output (SPLmuco). It is tested based on the ability of the loudspeaker to maintain its frequency response using CTA-2034 weighted pink noise. A nominal input signal level (1 V for passives) is run first to establish the baseline FR. Then the input is progressively increased until the FR deviates from the baseline by more than 3 dB (the actual procedure is a little more complicated that this). Every step of the test will need to be maintained for 1 minute minimum to make sure the driver voice coils are close enough to their steady state temperatures. The broadband SPL is the SPLmuco. [Edit - added] SPLmuco is basically a measure of when compression becomes excessive.

The peak SPL is measured using tone bursts. SPLpeak is determined using a rather complicated formula based on harmonic distortions. [Edit - added] SPLpeak is basically a measure of when HD becomes excessive.

The maximum SPL of the loudspeaker is then summarize to one single number (referenced to 1 m):

if SPLpeak - SPLmuco ≥ 9 dB, then max SPL = SPLmuco;​

if SPLpeak - SPLmuco < 9 dB, then max SPL = SPLpeak - 9 dB.​

As the loudspeaker will be tested at very high SPL for long periods of time, I don't know if this is something Amir's family will be willing to suffer through, multiple days a week. But this is how CTA-2034-A distills the harmonic distortions and compression of a loudspeaker into a single number that is intended to be used by the general public.

[Edit] Add notes on compression and harmonic distortion.

 

[dshreter]

I would like to point out a lesser discussed part of CTA-2034-A — the on-axis maximum usable SPL test (section 8).

The test has 2 parts, the first is the maximum usable continuous output (SPLmuco). It is tested based on the ability of the loudspeaker to maintain its frequency response using CTA-2034 weighted pink noise. A nominal input signal level (1 V for passives) is run first to establish the baseline FR. Then the input is progressively increased until the FR deviates from the baseline by more than 3 dB (the actual procedure is a little more complicated that this). Every step of the test will need to be maintained for 1 minute minimum to make sure the speaker voice coil is close enough to the steady state temperature. The broadband SPL is the SPLmuco.

The peak SPL is measured using tone bursts. SPLpeak is determined using a rather complicated formula based on harmonic distortions.

The maximum SPL of the loudspeaker is then summarize to one single number (referenced to 1 m):

if SPLpeak - SPLmuco ≥ 9 dB, then max SPL = SPLmuco;​

if SPLpeak - SPLmuco < 9 dB, then max SPL = SPLpeak - 9 dB.​

As the loudspeaker will be tested at very high SPL for long periods of time, I don't know if this is something Amir's family will be willing to suffer through, multiple days a week. But this is how CTA-2034-A distills the harmonic distortions of a loudspeaker into a single number that is intended to be used by the general public.

Given many of the speakers will be borrowed, stress tests may also be less feasible.