Erin's Audio Corner gets a Klippel NFS!

[MZKM]

Hi Erin - is there a reason for there to be multiple squiggles (technical term) in the 40-200Hz range in the EPDR curve when the phase and impedance grpahs look smoother ?

You mean those 3 peaks? Well, it’s based off the interaction of phase and impedance, any time the phase is 0° the EPDR is the same as the impedance, so as it transitions from positive to negative angles (or vice-versa), you will get a peak.

 

[hardisj]

^ better said than I would have said it.

TBH, the EPDR is new to me. But I do like it as a way to get a more accurate account of what the amplifier sees for the load. That is why I have included it in my graphics. Though, Dr. Jack did caution us that this calculation was created with the intention to be used for Class AB amps per the below:

Just a note of caution that the EPDR derivation is based on a class-B output stage so it's valid for typical class-AB amps but certainly not for class-A and probably has only marginal relevance for class-D amps (would love to hear from a class-D expert on this topic).

 

[MZKM]

^ better said than I would have said it.

TBH, the EPDR is new to me. But I do like it as a way to get a more accurate account of what the amplifier sees for the load. That is why I have included it in my graphics. Though, Dr. Jack did caution us that this calculation was created with the intention to be used for Class AB amps per the below:

Yeah, would love to know how it differs for Class-D.

You did that live stream with Purifi about speakers, maybe another could be done about Class-D amps

 

[hardisj]

Well, gentlemen, I was making pretty good progress until yesterday when we were gifted a new puppy. This is Rae. Seven weeks old German shepherd.

Really cool story how we got her... made short ... My wife had a patient whom she went above and beyond for. The patient unfortunately passed away. The patient’s family breeds dogs and they gave us this little girl as a token (for lack of a better word) of their appreciation for all my wife had done for their family. The name Rae was chosen in honor of that patient.

So, gonna be a couple weeks of madness around here. Good news is my golden (Rosie) has taken to her and the two have been playing all day.

 

[Kachda]

Well, gentlemen, I was making pretty good progress until yesterday when we were gifted a new puppy. This is Rae. Seven weeks old German shepherd.

Really cool story how we got her... made short ... My wife had a patient whom she went above and beyond for. The patient unfortunately passed away. The patient’s family breeds dogs and they gave us this little girl as a token (for lack of a better word) of their appreciation for all my wife had done for their family. The name Rae was chosen in honor of that patient.

So, gonna be a couple weeks of madness around here. Good news is my golden (Rosie) has taken to her and the two have been playing all day.

View attachment 113848

View attachment 113850

So many babies to play with, including the NFS !

 

[Steve Dallas]

A couple of weeks of madness? I'm thinking about 6 years. Germans are little nuclear energy plants for what seems like forever!

 

[Colonel7]

^ better said than I would have said it.

TBH, the EPDR is new to me. But I do like it as a way to get a more accurate account of what the amplifier sees for the load. That is why I have included it in my graphics. Though, Dr. Jack did caution us that this calculation was created with the intention to be used for Class AB amps per the below:

Maybe a short 2 or 3 sentence explanation of EPDR and the Dr. Jack caveat on its applicability to amplifier type, then it just becomes part of your written standard review template.

 

[hardisj]

In the interest of full disclosure, here is a long data-driven post that most probably won't care about but is important to the overall goal of learning the NFS and making sure my results make sense....




A few weeks ago @amirm measured the JBL 708P. Here is the link to his review. The owner (@Bugal1998) offered it to me for testing to help me compare my results to Amir's to make sure my setup was okay.

In my first test, I got a result that was about 2dB off from Amir's results in the midrange as shown below.
Mine = Red.
Amir's = Blue.

So, yea, this is no bueno. I contacted Amir and the owner to make sure there were no settings that were in play (eq, etc) when the unit was tested. The answer was "no", as expected. I informed Amir of my NFS setup and Amir suggested I move the microphone distance out a bit and re-test. I did. And my results much matched better. Still, not exactly the same (more on this below).

But, with my first test being so off I wanted to make sure that the second test was right. Sure, it matched Amir's better in the bass/midrange but blindly assuming that is the 'de-facto' is a bit presumptive, too. I mean, I got two different results using the same system. I needed to understand what was going on. So, with these differences still in mind, the logical thing was to take the speaker outside and measure quasi-anechoically; basically what I used to do before I had the NFS. I measured in the ground plane at 2m and then on a 6-foot ladder, illustrated in the photos below.

I then overlaid the on-axis results from these tests with that of my two NFS tests, shown below. My original NFS run in red and the updated in black. As noted in the legend, the positioning of the microphone on the original NFS was approximately 28cm and when I backed the mic out to be about 44cm the fitting error cleaned up and now matches (within reason) my two outdoor measurements mentioned above. Ground plane is in teal and ladder test is in blue. **NOTE: The two outdoor tests were designed to give me the information I needed on the bass and midrange areas. I wasn't really concerned with HF so making sure I had the speaker perfectly aligned wasn't top priority. Especially for the ladder test (I only have one ladder and the speaker was on it). This is why the HF is so different in these tests.**

The ground plane and quasi-anechoic more closely match the updated (black) NFS scan in the bass and mid-frequencies. The bass in my original scan (red) had an obvious error as illustrated by the rebound at ~28Hz. That was an indicator of high error in the bass, but since it was the midrange that was also off, that is where I keyed in on initially. So, bottom line, the quasi-anechoic and GP measurements indicated that my updated run and Amir's data were, indeed, accurate. Or, at least, more accurate than my initial run. And here is my final SPIN from the 708P:

And if you want to compare my updated NFS run against Amir's you can see how closely they match (zoomed y-scale to show differences).
Mine = Black.
Amir's = Blue.

The differences are practically only above 1kHz and can likely be explained by:

1. My microphone is less influenced by the boom as has already been discussed.
2. According to Amir and @MZKM's data the DUT was not quite centered on the stand. You can see MZMK's post:

   So a the mic was off center by a bit to the right and a bit too low.

   Yeh, the horn makes it hard to position the mic correctly as does the curved sides on the enclosure.

   Now, when I received the speaker there were pads under it (I assume to keep from scratching a desk up). I made a video for the owner when I set it up so he would understand why I removed the one in the center, because as you can see the center pad caused the speaker to tilt down and not sit flat on the NFS stand. I had a feeling this would show up in measurements and it looks like I am right. This is likely the reason why Amir's data shows the HF droop compared to mine.

So, thanks to Amir and @Bugal1998 for giving me the opportunity to help get my setup more correct and learn something about the NFS that I will use going forward.



I think at this point I am now ready to continue forward with "real" measurements and start the reviews using the NFS.

 

[hardisj]

Maybe a short 2 or 3 sentence explanation of EPDR and the Dr. Jack caveat on its applicability to amplifier type, then it just becomes part of your written standard review template.

Yea, I was actually thinking about including that in the title of each plot. That way it stays with the data in case someone plucks it from a review and the context is not given.

 

[napilopez]

In the interest of full disclosure, here is a long data-driven post that most probably won't care about but is important to the overall goal of learning the NFS and making sure my results make sense....




A few weeks ago @amirm measured the JBL 708P. Here is the link to his review. The owner (@Bugal1998) offered it to me for testing to help me compare my results to Amir's to make sure my setup was okay.

In my first test, I got a result that was about 2dB off from Amir's results in the midrange as shown below.
Mine = Red.
Amir's = Blue.

View attachment 113978



So, yea, this is no bueno. I contacted Amir and the owner to make sure there were no settings that were in play (eq, etc) when the unit was tested. The answer was "no", as expected. I informed Amir of my NFS setup and Amir suggested I move the microphone distance out a bit and re-test. I did. And my results much matched better. Still, not exactly the same (more on this below).

But, with my first test being so off I wanted to make sure that the second test was right. Sure, it matched Amir's better in the bass/midrange but blindly assuming that is the 'de-facto' is a bit presumptive, too. I mean, I got two different results using the same system. I needed to understand what was going on. So, with these differences still in mind, the logical thing was to take the speaker outside and measure quasi-anechoically; basically what I used to do before I had the NFS. I measured in the ground plane at 2m and then on a 6-foot ladder, illustrated in the photos below.

View attachment 113968View attachment 113969



I then overlaid the on-axis results from these tests with that of my two NFS tests, shown below. My original NFS run in red and the updated in black. As noted in the legend, the positioning of the microphone on the original NFS was approximately 28cm and when I backed the mic out to be about 44cm the fitting error cleaned up and now matches (within reason) my two outdoor measurements mentioned above. Ground plane is in teal and ladder test is in blue. **NOTE: The two outdoor tests were designed to give me the information I needed on the bass and midrange areas. I wasn't really concerned with HF so making sure I had the speaker perfectly aligned wasn't top priority. Especially for the ladder test (I only have one ladder and the speaker was on it). This is why the HF is so different in these tests.**


View attachment 113972


The ground plane and quasi-anechoic more closely match the updated (black) NFS scan in the bass and mid-frequencies. The bass in my original scan (red) had an obvious error as illustrated by the rebound at ~28Hz. That was an indicator of high error in the bass, but since it was the midrange that was also off, that is where I keyed in on initially. So, bottom line, the quasi-anechoic and GP measurements indicated that my updated run and Amir's data were, indeed, accurate. Or, at least, more accurate than my initial run. And here is my final SPIN from the 708P:


View attachment 113973







And if you want to compare my updated NFS run against Amir's you can see how closely they match (zoomed y-scale to show differences).
Mine = Black.
Amir's = Blue.

View attachment 113979



The differences are practically only above 1kHz and can likely be explained by:

1. My microphone is less influenced by the boom as has already been discussed.
2. According to Amir and @MZKM's data the DUT was not quite centered on the stand. You can see MZMK's post: Now, when I received the speaker there were pads under it (I assume to keep from scratching a desk up). I made a video for the owner when I set it up so he would understand why I removed the one in the center, because as you can see the center pad caused the speaker to tilt down and not sit flat on the NFS stand. I had a feeling this would show up in measurements and it looks like I am right. This is likely the reason why Amir's data shows the HF droop compared to mine.

So, thanks to Amir and @Bugal1998 for giving me the opportunity to help get my setup more correct and learn something about the NFS that I will use going forward.



I think at this point I am now ready to continue forward with "real" measurements and start the reviews using the NFS.

Thank you for this, I definitely appreciate the thoroughness here!

As opposed to there being a positioning difference, are we also sure that your mic doesn't just run a touch brighter than Amir's in the treble? In even the last run of the R3 measurements, your on-axis treble was a teensy bit hotter than Amir's. I think my mic is a little hotter than Amir's too .

Not that it really matters, just noting that might be partly the cause rather than a difference in on-axis centering.

 

[MZKM]

Thank you for this, I definitely appreciate the thoroughness here!

As opposed to there being a positioning difference, are we also sure that your mic doesn't just run a touch brighter than Amir's in the treble? In even the last run of the R3 measurements, your on-axis treble was a teensy bit hotter than Amir's. I think my mic is a little hotter than Amir's too .

Not that it really matters, just noting that might be the cause rather than a difference in on-axis centering.

How much money does one have to spend to get a guaranteed super flat mic?

But then, who measures the measurer?

Measuring weight with speakers:

 

[hardisj]

Thank you for this, I definitely appreciate the thoroughness here!

As opposed to there being a positioning difference, are we also sure that your mic doesn't just run a touch brighter than Amir's in the treble? In even the last run of the R3 measurements, your on-axis treble was a teensy bit hotter than Amir's. I think my mic is a little hotter than Amir's too .

Not that it really matters, just noting that might be the cause rather than a difference in on-axis centering.

Possibly. But the response deviates above the 1kHz and not just above 8-10kHz which is where calibration is effective.

I do want to contact the manufacturer to inquire about a calibration file but if all else fails I’ve been looking at purchasing a calibrated mic.

 

[ctrl]

Is there no device to secure the speaker on the measurement platform?
A simple lashing strap would be sufficient and should hardly influence the measurement - or does it only look so unsafe in the pictures?

 

[amirm]

A simple lashing strap would be sufficient and should hardly influence the measurement - or does it only look so unsafe in the pictures?

I have built two auxiliary platforms to measure larger speakers. For my recent Genelec review, I initially wanted to use a strap but it still felt unsafe so I used one of my platforms. It is tricky to build these in that you don't want them to poke out in front and create reflections. That then forces you to have the added platform be cantilevered. So good bit of thought is needed to make ones that are stable, can support a speaker well, and don't cause artifacts of their own.

 

[ctrl]

I do want to contact the manufacturer to inquire about a calibration file but if all else fails I’ve been looking at purchasing a calibrated mic.

Without a calibration file, measuring even with a Class 1 microphone is a pure gamble at frequencies above 6kHz, because the standard allows deviations of +2/-3dB at 10kHz.

My measurement microphone is already pretty good, but still the correction in the range above 6kHz is from -0.33dB to 1.1dB.

6300    -0.33   
7100    -0.21   
8000    0.01   
9000    0.13   
10000    0.27   
11200    0.42   
12500    0.73   
14000    1.02   
16000    1.12   
18000    0.94   
20000    0.55

 

[hardisj]

I have built two auxiliary platforms to measure larger speakers. For my recent Genelec review, I initially wanted to use a strap but it still felt unsafe so I used one of my platforms. It is tricky to build these in that you don't want them to poke out in front and create reflections. That then forces you to have the added platform be cantilevered. So good bit of thought is needed to make ones that are stable, can support a speaker well, and don't cause artifacts of their own.

If you want, I can forward you the CAD drawing. I mentioned to Klippel that I was going to have a machine shop to make me an adapter plate for large speakers and they sent the drawing to me. Just PM me if you want and I'll email it to you.

 

[hardisj]

I went to the packaging and pulled out the spec sheet provided for the mic I am using. I just ran it through SPLTrace and found it is +0.664dB @14.4kHz and -0.892dB at 20kHz. So, not bad at all. Still, I made the calibration file from that. Good to go.

 

[TimVG]

+0.664dB @14.4kHz

Looks closer to 1dB. Did you mean +0.964?

 

[hardisj]

Looks closer to 1dB. Did you mean +0.964?

You are correct.

Still, it’s within +/-1dB without calibration. That’s pretty damn good.

 

[NTK]

As I understand it, the mathematics behind sound field separation works like this:

The room in which the measurements take place is separated into 3 zones -- the interior zone, source-free zone, and exterior zone. All the sound sources of the speaker under test must be inside the interior zone (should include the entire cabinet if it isn't absolutely inert and diffraction free). Measurements are taken in the source-free zone. All the reflection and exterior noise sources are in the exterior zone.

By measuring within the source-free zone, the math can separate the sound generated in the interior zone from those in the exterior zone.

The interior zone is a sphere centered at the acoustic (expansion) center, typically on the tweeter axis for a 2-way and slightly behind the diaphragm. The diameter of the sphere is the minimum diameter to enclose all the sound radiating sources of the speaker (including diffraction edges).

The exterior zone is defined by a sphere concentric to the interior zone sphere, where all the external sound sources are located. It is typically defined by the closest wall, ceiling, or floor.

The source-free zone, where measurements are made, is the spherical shell between the interior and exterior zones.

For the NFS, which measures in cylinders, measurements must be taken outside of the green cylinder as shown. When Erin said that his first scan was taken 28 cm (measurement cylinder radius?), I wonder if in this case the bottom port is outside the interior zone (The 708P is 44.1 cm tall). If it is the case, and if my understanding of the math is correct, some of the bass output from the bottom port may have been excluded by the math, causing the lower measured bass extension.

I also wonder if this is also the reason of the poor bass extension in Amir's Revel F328Be measurement. The F328Be has 2 rear ports, and the lower port is pretty far from the expansion center, causing the interior zone to be pretty large, and would require much larger measurement distances.