• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Alan Shaw on Anechoic and Simulated Measurements

ahofer

Major Contributor
Forum Donor
Joined
Jun 3, 2019
Messages
4,952
Likes
8,698
Location
New York City
The thread is here, for members: https://www.harbeth.co.uk/usergroup/threads/measuring-the-bass-output-of-loudspeakers.80244/

After the first post, when Shaw said, among other things:

Here is another subject that you won't read about in audio forums, but is of paramount importance.

Once upon a time when engineers ruled rather than marketing people, and facts were respected as facts, engineering budgets were directed at the construction of anechoic chambers or dead rooms. These 'free-field' environments were for all practical purposes reflection (echo) free, and when a loudspeaker was measured in such chamber it was possible to say with a high degree of confidence that the frequency response measurement was that of the speaker alone, not a mix of the speaker and the environment in which it was being tested.

Sadly, the engineers lost the argument, and now magazines such as Stereophile and other have created their own methods for measuring loudspeakers 'on the kitchen table' or reviewer's back yard. The acoustic conditions there clearly are far from echo free, and hence, the measurements of tested speakers are necessarily corrupted by echoes from the environment. The human eye is not capable of 'removing' the echoes from a frequency response plot: there is no practical way of accurately removing the corrupting effect of the reflections from a wide-band measurement made in a reverberant space.

But speaker designers face the same problem as the technical reviewer does. I wrote to John Atkinson of Stereophile some years ago draw attention to my observation that most of his review frequency response curves showed a significant boost in the low frequencies which appeared to be consistent and therefore related to his methodology, and not really present on measurements made under truly free-field conditions. I suggested why it might be occurring. No reply was forthcoming, but this issue of excessive bass in Stereophile speaker measurements seems to be well known in the speaker industry. I didn't pursue it as I have the greatest sympathy for those trying to make accurate - or even repeatably inaccurate - technical measurements of loudspeakers without a handy anechoic chamber. It's an issue that dogs all objective speaker designers who must know the relative energy balance between bass. mid and top in their new designs. That's because the bass output of the speaker is the pivotal point in the audio spectrum against which the ear judges the overall correct balance.

Over the last twenty years, I have lost track of how many hours I've invested in this subject of making-do without a real anechoic chamber. I believe that there are just a handful in the world now, and it's not relevant that Microsoft Corp. or other large corporates have a huge ones of impressive performance as they're not available to the audio industry. Ditto those used by defence contractors and (I assume) those likely to be operated by universities for teaching and research, and even audio-centric businesses such as Dolby or Apple are not going to share their resources with speaker tinkerers. So what can we do?

Well, some thirty or so years ago it was suggested that the measurement of the entire bass, mid, top frequency response of a loudspeaker system could be broken down into a measurement of the bass output with the microphone very close to the woofer, and then an additional second measurement of the mid/HF with the mic 1m or so away. This would allow 'windowing' of the mid/high frequencies to catch them before their echoes arrived at the microphone, and by placing the mic close to the woofer, the long-wavelength, long period low frequencies could be measured with enough energy that the inevitable echoes would be swamped. So that opened up the possibility of measuring speakers in relatively ordinary non-anechoic rooms, just as Stereophile does in the back yard.

But - big but: different techniques for measuring the bass output of a speaker yield significantly different results when there can be only one "anechoic" truth.

I've been working on this subject again over the past few weeks, and here are a group of measurements made on the bass output of the same speaker within minutes, at 20 degrees C, using several different techniques, a lot of work. I've overlaid the many different measurements into one graph, below.

Which one do you believe? BTW: one of the curves was the output of a laser pointed at a shiny spot on the cone. Can we assume or not that the no-contact laser method must (surely) give the 'truest' results given the very high financial investment?

1620044060321.jpeg

Sorry but considering that making these measurement, post processing them and graphing them has taken around 15 hours, I have obscured the precise details for public viewing.

I asked, immediately, what he thought of the Klippel system. For some reason this question was killed by the moderator (or Shaw himself). Then:

All measurement systems, including the Klippel one that do not make use of a real, physical anechoic chamber have to mathematically manipulate two frequency response measurements: one made very close to the woofer (say, 5mm away from the dust cap - almost touching) and another made, say 1m away (the standard anechoic distance from speaker to mic).

All these twin-measurement approached rely on combining these two very different measurements into one wideband 20Hz - 20/40kHz conventional anechoic frequency response measurement. The maths is simple - too simple - and in the many published papers it relies on finding a single frequency where the two curves can be cut and past together, typically around 200-300Hz. But as we'll see, that is a gross simplification of a much more complex issue.

In my long experience with this matter, the outcome for the cut and paste can be whatever you want it to be according to the frequency you chose to make the junction between the measurements. If you have a wideband anechoic-chamber measurement of the same speaker (at 1m) you can literally printout and slide your two DIY measurement curves up and down in level and frequency until you have picked a transition frequency which visually matches the overall real anechoic curve. But if you didn't have that anechoic curve as a guide, should you give ten speaker engineers the near field and the far field measurements and asked them to cut and paste them into a wideband measurement, you'd get ten very different takes on at what frequency and what level to join them together. It's a fuzzy solution to a hard acoustic problem.

And that is a nonsense.

Thoughts?
 

Koeitje

Major Contributor
Joined
Oct 10, 2019
Messages
2,292
Likes
3,880
Thoughts? He doesn't understand the maths behind the Klippel NFS.
 

Zvu

Addicted to Fun and Learning
Joined
May 1, 2020
Messages
830
Likes
1,416
Location
Serbia
BS

Only better thing from anechoic chamber is no chamber aka outdoor measurement for low frequencies.

Instead of trying to justify poor decisions and by that measurements of Harbeth loudspeakers, wouldn't it be easier if he just did his f*in job right ?

Luckily, era of murky water is almost gone.
 
Last edited:

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,747
Likes
16,180
AS has unfortunately in many of his business related topics a state of knowledge still from the "good old 70s BBC research times" which can be seen on such posts and also of course on his loudspeaker engineering and thus he hasn't understood how the Klippel NFS works. Heck, for bass measurements you don't even need a NFS or a huge anechoic chamber, you can even do them for example outside with the ground-plane method.
 

Ericglo

Senior Member
Joined
Apr 21, 2020
Messages
452
Likes
323
Harman has an anechoic chamber. When Erin did his measurements of their speakers, his lined up with Harmans.

Alan has a point about lower frequencies, but I think most people put a disclaimer on their measurements to account for this.
 

hardisj

Major Contributor
Reviewer
Joined
Jul 18, 2019
Messages
2,907
Likes
13,908
Location
North Alabama
Thoughts? He doesn't understand the maths behind the Klippel NFS.



This quote shows he has no understanding of the Klippel. Hell, you can't even get 1 meter away from the DUT in most cases. Even with the R-axis extension.

All measurement systems, including the Klippel one that do not make use of a real, physical anechoic chamber have to mathematically manipulate two frequency response measurements: one made very close to the woofer (say, 5mm away from the dust cap - almost touching) and another made, say 1m away (the standard anechoic distance from speaker to mic).


I wish people would not talk about things they obviously know nothing of. In this case, he has not even the slightest idea of how the NFS works.
 
Last edited:

hardisj

Major Contributor
Reviewer
Joined
Jul 18, 2019
Messages
2,907
Likes
13,908
Location
North Alabama
On the topic, I have shown repeatedly - via real data - that:
1) One can gain quasi-anechoic measurements that are well within reasonable accuracy of a true anechoic measurement. (±1dB per the CTA spec))
2) The NFS is better than an anechoic measurement because anechoic are typically only anechoic >80-100Hz. A room correction curve is only useful for the speaker (or similar type) used to create the curve. A curve generated with a sealed speaker will not work with a vented enclosure.
3) A ground plane measurement is better for LF accuracy than anechoic. See #2 for reasons.
4) An outdoor measurement is worse than the NFS because it is incredibly impractical not only to get a speaker 100 feet off the ground but also to perform SPIN measurements ... and wind... and .... well, you get the idea.

This stuff isn't rocket science to people who have actually used these methods. I would know... I literally have a degree in rocket science and I have used all the above methods (maybe not 100 feet, but 7 feet is hard enough). :D
(that is the most egotistical thing I have ever said but every now and again I like to pat myself on the back)
 
Last edited:

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
I wish people would not talk about things they obviously know nothing of

This would be great, an ideal situation. Forum discussions would be then reduced to 5%. And it was the case some 20 - 25 years ago, when the internet was not available to everyone.
 

hardisj

Major Contributor
Reviewer
Joined
Jul 18, 2019
Messages
2,907
Likes
13,908
Location
North Alabama
This would be great, an ideal situation. Forum discussions would be then reduced to 5%. And it was the case some 20 - 25 years ago, when the internet was not available to everyone.

Well, to be clear, I'm all for discussion. But people speaking from a position of authority about something they then show obvious lack of knowledge in... those are the people I wish would stop.
 

musicapristina

Member
Audio Company
Forum Donor
Joined
Apr 22, 2020
Messages
21
Likes
8
Location
Raleigh, NC
Does anyone have data they can share showing measurements in an anechoic chamber compared with those coming from models? I would like to see the degree of correlation.
 

napilopez

Major Contributor
Forum Donor
Joined
Oct 17, 2018
Messages
2,109
Likes
8,420
Location
NYC
The thread is here, for members: https://www.harbeth.co.uk/usergroup/threads/measuring-the-bass-output-of-loudspeakers.80244/

After the first post, when Shaw said, among other things:



I asked, immediately, what he thought of the Klippel system. For some reason this question was killed by the moderator (or Shaw himself). Then:



Thoughts?

On the topic, I have shown repeatedly - via real data - that:
1) One can gain quasi-anechoic measurements that are well within reasonable accuracy of a true anechoic measurement. (±1dB per the CTA spec))
2) The NFS is better than an anechoic measurement.
3) A ground plane measurement is better for LF accuracy than anechoic because anechoic are typically only anechoic >80-100Hz. A room correction curve is only useful for the speaker (or similar type) used to create the curve. A curve generated with a sealed speaker will not work with a vented enclosure.
4) An outdoor measurement is worse than the NFS because it is incredibly impractical not only to get a speak 100 feet off the ground but also to perform SPIN measurements ... and wind... and .... well, you get the idea.

This stuff isn't rocket science to people who have actually used these methods. I would know... I literally have a degree in rocket science and I have used all the above methods (maybe not 100 feet, but 7 feet is hard enough). :D
(that is the most egotistical thing I have ever said but every now and again I like to pat myself on the back)


This was honestly all a little disheartening to read. I know that when you're running a speaker business and actually making speakers you might not be aware of all the same things we are as a forum -- the NFS is fairly new, after all -- but I agree with everything @hardisj said. It's possible Shaw is thinking of some other klippel device or software rather than the NFS. There's no splicing going on with that, and the system has shown its reliability over and over, even if there have been some calibration/software issues that now appear to be worked out.

By now Erin and Amir have repeatedly shown remarkable correlation between the NFS and anechoic measurements, and in many cases it is obvious the NFS is the more accurate measurement in the bass, because there are 'jaggies' from anechoic measurements becoming less anechoic at lower frequencies. Erin and I have also shown how close one can get to anechoic measurements with the splicing technique.

I'd also add that I'm not sure why he makes out stereophile's measurements to be a mystery. Atkinson clearly states in every one of his recent reviews (that I've seen) something along the lines of:

"The boost in the response in the upper bass will in part be due to the nearfield measurement technique, which assumes that the radiators are mounted in a baffle that extends to infinity in both vertical and horizontal planes."

(Taken from here)

Perhaps some of the older reviews don't have this disclaimer, but I could find them as far back as 2005.

That's not to say the issues Shaw talks about aren't real ones (aside from the lack of understanding of the NFS). The most important of these is:

"In my long experience with this matter, the outcome for the cut and paste can be whatever you want it to be according to the frequency you chose to make the junction between the measurements. If you have a wideband anechoic-chamber measurement of the same speaker (at 1m) you can literally printout and slide your two DIY measurement curves up and down in level and frequency until you have picked a transition frequency which visually matches the overall real anechoic curve. But if you didn't have that anechoic curve as a guide, should you give ten speaker engineers the near field and the far field measurements and asked them to cut and paste them into a wideband measurement, you'd get ten very different takes on at what frequency and what level to join them together. It's a fuzzy solution to a hard acoustic problem."[/quoute]

This is true. When I splice nearfield measurements to farfield gated ones, I can choose where to joing the graphs, and depending on where I do so, the bass can appear greater or lower than it is. This is indeed crucial for properly presenting the speakers tonal balance, because we do seem to reference it off of the 200-400 region (or something like that).

This can be illustrated in the example I used in my quasi-anechoic measurement guide:

1620058498192.png


I can choose to line up the bass lower or higher, and that will affect the appearance of the overall bass level.

But I'm not sure it's quite as big of a problem as he paints it out to be. When I started making measurements, I would indeed practice by lining up the splice to match anechoic or other reliable measurements when I could find them. But it quickly became apparent that having a reference wasn't necessary. He mentions splicing at 200-300, but I usually splice higher, 400-600Hz, and by then it should be rather apparent how much the measurements line up. The above is actually one of the 'worst' correlations between the nearfield and gated data I've had, but it was still pretty obvious where to splice.

Even then, nearfield has its issues, so I do understand where he's coming from. BUT if you're using ground plane like @hardisj used to do, you should have much more accuracy and be able to splice high enough that lining up the measurements isn't a big issue. Neumann uses an anechoic chamber, but still uses ground plane technique for its bass measurements because it is more accurate. And the bass has correlated pretty much perfectly with the NFS (after we figured out the temperature issue). Same for soundandrecording.de.

The other reason he might mention bass variation with splicing is that in some older splicing setups, it would be a hard splice. So if there was a small peak at, say, 250 Hz, and you aligned the measurement with the top of the peak, your bass would appear exaggerated. But both REW and VituixCAD have options now to blend the frequency response over a small range of frequencies, and this virtually eliminates this problem.
 
Last edited:

musicapristina

Member
Audio Company
Forum Donor
Joined
Apr 22, 2020
Messages
21
Likes
8
Location
Raleigh, NC
@napilopez Not sure if you are speaking to me: "This was honestly all a little disheartening to read." I'm not building speakers.

"By now Erin and Amir have repeatedly shown remarkable correlation between the NFS and anechoic measurements", can you point me to those data? Still trying to find my way around here.
 

Katji

Major Contributor
Joined
Sep 26, 2017
Messages
2,990
Likes
2,273
Welll...

AS has unfortunately in many of his business related topics a state of knowledge still from the "good old 70s BBC research times" which can be seen on such posts and also of course on his loudspeaker engineering and thus he hasn't understood how the Klippel NFS works.

On the other hand, I like what he says about speaker cables...and that talking about how amplifiers sound is not allowed on the forum. :) The wording of the rule is quite amusing.
 

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,747
Likes
16,180
Welll...
On the other hand, I like what he says about speaker cables...and that talking about how amplifiers sound is not allowed on the forum. :) The wording of the rule is quite amusing.
True, me too, on the other hand unfortunately he does not also allow really reasonable discussions on loudspeaker measurements in his forum, which leaves a selective bitter aftertaste.
 

Soniclife

Major Contributor
Forum Donor
Joined
Apr 13, 2017
Messages
4,500
Likes
5,417
Location
UK
True, me too, on the other hand unfortunately he does not also allow really reasonable discussions on loudspeaker measurements in his forum, which leaves a selective bitter aftertaste.
I see him as running his own cult, a lot of what he says is good, but he will not tolerate freedom of thought on his forum, and absolutely will not allow being questioned. He's got the avuncular thing down as well so anyways sounds calm and reassuring.
 

617

Major Contributor
Forum Donor
Joined
Mar 18, 2019
Messages
2,403
Likes
5,296
Location
Somerville, MA
What is the value of a Harbeth speaker? They don't compete on technical excellence, which is fine. They are sort of nice looking, and sound a certain way which their customers like and expect. It is far easier to differentiate a speaker based on appearance, marketing and 'sound' than it is to differentiate based on technical performance.

Who cares if he's running a cult? Every brand encourages cultishness.

Now, if I was running a content creation company, and someone decided to buy Harbeth's for monitoring, I might be annoyed, but it's not like recording studios earnestly seek accuracy in their monitoring anyway.
 
Top Bottom