Bass Response Correction for Klippel NFS Measurements

[MrPeabody]

A win win if also Klippel take this feedback and improve their product . That's peer review is so powerful.

Absolutely a win for Klippel to have these two hard-working committed professionals working out the hiccups for them. It seems to be a question of the adequateness of the instructions. What isn't yet certain, though, is whether a set of instructions could be written that would cover every conceivable kind of speaker. Looking at all of this from afar, I am more in awe than ever of just how technically difficult it is to measure loudspeakers.

 

[MrPeabody]

Question: will Amir and Erin recalculate and update all the 100+ speaker reviews they have combined? That is quite a burden...

I think the answer is that all of the measurements thus far taken for a very great many speakers are extremely useful and extremely accurate notwithstanding the limitation of the representation of bass extension with a fewe large tower speakers.

 

[Mnyb]

Absolutely a win for Klippel to have these two hard-working committed professionals working out the hiccups for them. It seems to be a question of the adequateness of the instructions. What isn't yet certain, though, is whether a set of instructions could be written that would cover every conceivable kind of speaker. Looking at all of this from afar, I am more in awe than ever of just how technically difficult it is to measure loudspeakers.

Suppose a speaker mfg works for years on a product and also confirms their spec in several ways with different techniques used to avoid a single point failure . Fully aware of the vagaries of acoustic measurements .

Amir and Erin does their best on a limited time budget , and have to juggle this seemingly complex system for their purposes .

 

[Deleted member 2944]

That's actually a good point. I don't know how the magnepan acts as a bass source... maybe its reference point is somewhere other than the axis Amir set for expansion? You guys would have to answer that. I'm just thinking you might be on to something here.

I doubt you'd see much of a difference in the LRS case......if applying this type of "correction." That speaker was mischaracterized in a different way.

Say you had a smallish two-way box speaker on your NFS and it exhibited a nicely leveled bass response and it sounded good during listening. But, then you placed it on the floor. Would you expect a different audible result? Maybe somewhat bass heavy?
Think about that situation in reverse and you have a simplified version of how the LRS measured.

Dave.

 

[Zvu]

You all probably remember how low frequency measurements look for NRC anechoic chamber that isn't anechoic bellow 100Hz. Klippel measures also in closed space trying to simulate what would it look like in anechoic. That is always problematic. To be able to get the most precise measurements of low frequencies in closed space (say anechoic chamber in NRC or NFS in room), you'd have to calibrate for every particular loudspeaker because number of woofers and ports don't couple in the same way depending on their exact number, position and cabinet size.

That's one of the main reasons why at Kef they do outdoor measurements. Axiom Audio made 30m high pole with the lift for loudspeaker cabinet. They turn the loudspeaker on its back and lift it 28m in the air. There is a microphone at top of the pole that measures the response down to 20Hz anechoic from 2m distance, distance more than enough to be able to see how ports and woofers interact in a cabinet of a certain height.

 

[amirm]

You all probably remember how low frequency measurements look for NRC anechoic chamber that isn't anechoic bellow 100Hz. Klippel measures also in closed space trying to simulate what would it look like in anechoic. That is always problematic. To be able to get the most precise measurements of low frequencies in closed space (say anechoic chamber in NRC or NFS in room), you'd have to calibrate for every particular loudspeaker because number of woofers and ports don't couple in the same way depending on their exact number, position and cabinet size.

This is not correct. The optimizations are there for the numerical methods to work. They are not intended to describe the speaker in any specific way.

 

[Zvu]

This is not correct. The optimizations are there for the numerical methods to work. They are not intended to describe the speaker in any specific way.

The only way to check the influence of that approach is to do a next complex (multiwoofer/multiport) loudspeaker measurement with NFS optimized this way and to confirm with outdoor ground plane measurement.

Here is Axiom Audio rig for precise low frequency measurements. They had exact same company build for them the exact copy of anechoic chamber in NRC. I tried to timestamp it but if not, it is at 15:40

 

[amirm]

The problem with anechoic chamber calibration is not at all what we are dealing with here. They have to worry about modal response of the room as induced by speaker. Nfs handles that without issue. As for ground plane it has its sources of errors.

 

[ctrl]

Now, the only question I still have is... how do you *know* your result is A-OK?

Don't want to be a pain, as I've suggested this a few times, but the easiest way to check would be to calculate the low frequency response from near field measurements of woofer and BR ports.
As @napilopez has proven with his measurements and explained in his howto, if the measurements are done correctly and then calculated, the accuracy is quite good.

With a little practice, the calculation is done in less than 15 minutes, even with complex speakers - depending on how accurate you are with the delay of the individual sound sources and their baffle step correction - I'm sure @napilopez can confirm this.

How accurate the calculated frequency response is (see turquoise curve below) can be seen nicely when comparing the different measurements. Please keep in mind that as a rough rule of thumb, the calculated FR is only valid up to about 200-300Hz (it's depending on the chassis and cabinet size), in this frequency range the sound pressure level comparison should be made with the NFS measurement.

The calculated frequency response is a good indicator to determine outliers in the low frequency measurements of the NFS.

At 30Hz, the deviation of the old NFS measurement from the calculated FR is 9dB. The new NFS measurement, on the other hand, shows no deviation from the calculated FR at 30Hz.

The calculated low frequency response at 30Hz is between the Audioholic's measurement (ground plane) and the manufacturer's measurement.
Which is another indication that the frequency response calculated from the near-field measurements is quite reliable - which is not surprising, since at the large wavelengths simplified physical models still achieve very good results.

In order for the near-field measurements to be as accurate as possible, a few conditions must be met.

- When measuring the sound sources of a loudspeaker, the measurement distances should always be the same - to ensure that the sound pressure level summation is correct

- The measurement distance depends on the sound source dimension. Since the cabinet influence is calculated later (baffle step correction), it is sufficient to use driver and BR port dimensions:

Measuring distance < 0.11m x source dimensions → error <1dB

Source: Arta Handbook
The measuring distance for an 8'' woofer would thus be <2cm (<<1'').
Since this is not possible with the NFS measuring device, it is best to do this manually with an extra microphone cable (using the Klippel software).

The Klippel software, as far as I know, is also controllable via a scripting language. It would make sense to write a script that, for example, starts with a twenty second delay, does not move the robot arm, and takes measurements every twenty or thirty seconds, for a total of five.
For the HDI-3800, for example, this would complete the near-field measurements in two minutes.

If the best possible accuracy is not required, you can probably even hold the measurement microphone by hand.

 

[hardisj]

Don't want to be a pain, as I've suggested this a few times, but the easiest way to check would be to calculate the low frequency response from near field measurements of woofer and BR ports.

In my case, the easiest way is a ground plane measurement. Weather notwithstanding, I take the speaker out to my driveway, run some cables and get a simple GP measurement within about 15 minutes from setup to takedown.

But, yes, the method that @napilopez has laid out would work well for Amir if he wants to pursue that or in lieu of a GP measurement. That post by @napilopez is pure gold.

 

[hardisj]

The measuring distance for an 8'' woofer would thus be <2cm (<<1'').
Since this is not possible with the NFS measuring device, it is best to do this manually with an extra microphone cable (using the Klippel software).

To be clear, you can set up the NFS to position the mic at any distance you want from <0mm (beyond the center of the NFS depending on how long the boom is) to (the max length of the r-axis+mic). In my case, with the r-axis extension I *think* that puts the mic at a max of about 1m from the center of the NFS. Then, depending on the depth of the DUT, the mic would be placed no further than this. For example, when I backed out the NFS mic to get as far as I could from the Focal Twin6 Be I measured last week, the distance was 0.871 meters from the tweeter to the mic.

On the flip side, I also position the mic about 15cm from the DUT for my NF testing. But you can go as close as you want.

So, there is no limit to how close you can get to the DUT with the NFS. Just a limit on how far away you can get.

 

[hardisj]

Don't want to be a pain, as I've suggested this a few times, but the easiest way to check would be to calculate the low frequency response from near field measurements of woofer and BR ports.

Alternatively, Amir already provides nearfield measurements. I don't want to speak for him but - since his plate is full - maybe he could provide the results in a text file and tell you guys the distance he measured from and one of you who is well-versed in that method can use that to generate the bass profile. At least for a few up front to gain confidence in the setup when there isn't other data (like anechoic or groundplane) available.

In fact, you may be able to do that now using SPLtrace, and assuming the distance of the mic to the DUT is the same for each driver/port.

Just an idea.

 

[napilopez]

Alternatively, Amir already provides nearfield measurements. I don't want to speak for him but - since his plate is full - maybe he could provide the results in a text file and tell you guys the distance he measured from and one of you who is well-versed in that method can use that to generate the bass profile. At least for a few up front to gain confidence in the setup when there isn't other data (like anechoic or groundplane) available.

In fact, you may be able to do that now using SPLtrace, and assuming the distance of the mic to the DUT is the same for each driver/port.

Just an idea.

You need phase data to do it right unfortunately, which SPL trace doesn't provide. How did you go about it, @ctrl ?

 

[richard12511]

I'd be curious to see if giant sealed towers show similar issues.

 

[richard12511]

That's actually a good point. I don't know how the magnepan acts as a bass source... maybe its reference point is somewhere other than the axis Amir set for expansion? You guys would have to answer that. I'm just thinking you might be on to something here.

Yeah, that's one of the ones I would really love to see recalculated(to see if it matters).

 

[richard12511]

It is pretty remarkable to think back on the huge difference in the availability of available measurements over the last year or two. We had soundstage Network, Audioholics, Stereophile, Audioxpress for the English sites. But now we have such a wealth of data with Amir and Erin (and a little bit from me) and the spinorama is becoming increasingly common. When Dr Toole wrote the latest edition of his book, it was pretty much only Harman doing spins... now it's showing up more and more places

Probably the best couple years ever for objective minded loudspeaker consumers.

Just few years ago, Toole was lamenting the fact that consumers were incredibly limited with choices of speakers for which detailed measurements exist. Y'all are fixing that problem. I bet @Floyd Toole and @Sean Olive are loving this movement.

 

[NTK]

The only way to check the influence of that approach is to do a next complex (multiwoofer/multiport) loudspeaker measurement with NFS optimized this way and to confirm with outdoor ground plane measurement.

Here is Axiom Audio rig for precise low frequency measurements. They had exact same company build for them the exact copy of anechoic chamber in NRC. I tried to timestamp it but if not, it is at 15:40

This is Axiom's method of testing subwoofers (link)

 

[ctrl]

In my case, the easiest way is a ground plane measurement. Weather notwithstanding, I take the speaker out to my driveway, run some cables and get a simple GP measurement within about 15 minutes from setup to takedown.

Sure, with nice weather and your very good local conditions, the calculation via the near field measurements would not bring any advantage (ground plane measurements, are better/ more realistic with optimal conditions). Except that with ground plane measurements you have to move the large, heavy loudspeakers additionally.

On the flip side, I also position the mic about 15cm from the DUT for my NF testing. But you can go as close as you want.
So, there is no limit to how close you can get to the DUT with the NFS. Just a limit on how far away you can get.

Yes, but the problem for Amir is that he wants to avoid damaging the DUT at all costs (which I can well understand) and therefore keeps a safe distance.

In this case, one could test whether it is not easier and faster, to perform the near field measurements by hand with a separate microphone stand or even held by hand (might work for low frequency range, for mid and high frequencies a microphone stand is mandatory to avoid reflections).

Of course, only for large and complex loudspeakers the low frequency response needs to be calculated from the near-field measurements (and even there only if there are discrepancies with other measurements), in all other cases the NFS works reliably - especially after Amir figured out how to further optimize the low-frequency measurements.

You need phase data to do it right unfortunately, which SPL trace doesn't provide. How did you go about it, @ctrl ?

Yep, you will loose all excess phase information when only the frequency response is available.

The loss of excess phase is not so problematic when measuring the woofers (behaves approximately like a minimum phase system). That is, with a sealed speaker, the error would be minimal, even with multiple drivers.

But with a TL or BR loudspeaker the matter looks quite different, since the ports do not represent a minimum phase system (from the frequency response alone, can no longer be concluded to the phase frequency response).
With a BR port, we know that the sound radiated backwards by a driver is inverted, i.e. offset by 180° in the phase frequency response. Therefore, we invert the minimum phase determined from the frequency response for a BR port - for a rough approximation of the low frequency response this is usually sufficient.

But at the latest with a TL loudspeaker with multiple woofers, this approach no longer works at all. You really have to rely on the complete impulse response (and the resulting frequency response and phase frequency response) of the near-field measurements to create the low frequency response of the speaker.

 

[hardisj]

Yes, but the problem for Amir is that he wants to avoid damaging the DUT at all costs (which I can well understand) and therefore keeps a safe distance.

What?... You think I’m an outlaw cowboy going rogue on any speaker sent to me?


Kidding aside, setting up the Klippel for nearfield measurements is zero risk because you can manually move the mic arm/axes with the entire NFS unplugged from the wall. Or you can keep it powered on and use the remote/software if you want.

 

[Zvu]

This is Axiom's method of testing subwoofers (link)

Not just subs, all speakers. Nothing must stand in the way of ultimate precision