Neumann KH 80 DSP Speaker Measurements: Take Two

[napilopez]

We have that in the revel center speaker. Thats my reason to trust the NFS. I was slightly skeptical when the kali in-8 measured so poorly but with the revel speaker and that it turned out the kali sample was broken i trust Amirs measurements.

Again, no one is saying not to trust amir's measurements. We're just wondering what the limit of the NFS is. The Neumann example has brought some interesting insight.

These small differences are largely immaterial for broader comparisons of which speaker is better than the other, but it's to nice to know how close the measurements are to anechoic data and how they compare to well done DIY measurements. As Amir tests more and more speakers, there will be more top-performing models, and it will be interesting to compare measurements among them. But to do that we need to know how the system works and what its strengths, limitations, and outright differences are compared to traditional measurements. This thread has revealed some of that and for that I'm grateful.

 

[edechamps]

This all sounds like something one should ask Klippel about.

Yes. I guess the question I would ask Klippel is: how does the NFS deal with microphone directivity issues at high frequencies, given that some of the measurement points can have very steep incidence angles from the twitter to the microphone? How does the NFS avoid the systematic negative bias in high frequency response that would result?

As far as I can tell, the NFS whitepapers do not answer these questions.

 

[gr-e]

Hmm, looking at this video it seems that the tweeter point (expansion point) and the reference point can be set separately

Spoiler: Training 8 - Measurement of Loudspeaker Directivity

@9:18

 

[gr-e]

@amirm can you please show these windows for your kh80 measurements?

 

[edechamps]

Hmm, looking at this video it seems that the tweeter point (expansion point) and the reference point can be set separately

Yes, something like that was mentioned in the past and I asked for a spinorama generated that way (i.e. measured using the Klippel-recommended tweeter axis, but spinorama data plotted relative to manufacturer-specified reference axis - 5 cm down in the case of the KH80). I think that might be the best possible solution, but it seems @amirm has not run that particular experiment yet.

It looks like @BYRTT is trying to emulate this using the data we already have (see post #259) but I suspect offsetting the vertical angle is not quite the same thing as doing a proper spatial translation, especially in the case of far field data.

 

[MZKM]

Thousand thanks, back in post #167 shared VituixCAD can offset microphone so spin of the 76 txt files gets rotated and that makes your take 1 for KH 80 shine as seen in the mico overlay in left lower part of image, admit i didn't address you there but below is result for new take 4 where above 1kHz area smooth out when microphone point in space is lowered, it shall be noted than in VituixCAD there is a button called "CTA-2034-A" and when pushed listening distance setting dial in on a 2000mm number, i guess this behavour is because CTA2034 call for that but then again think its a little important Klippel software is set to compute acoustic spin for the 76 directionality steps to same 2 meter distance to base its Spinorama plot, well > 2 meters think doesn't mean so much but if its set any number < 2 meter then curves begin differ and we can't really make comparisons to Harman Spinorama's or other users of CTA2034 standard.

View attachment 49545
View attachment 49546

There is still the dip >13kHz in sample #2, which is likely in the speaker itself and not due to the measurement.

 

[BYRTT]

Well, we can't go by what it looks like. The filtering makes a huge difference as to how smooth that graph looks. Who can ask them in German?

There is still the dip >13kHz in sample #2, which is likely in the speaker itself and not due to the measurement.

Think we dial in on that Neaumann S&H curve if the -53mm microphone offset on Y axis happen can be approved, below plot is set to same X/Y axis ratio as S&H plot, upper black is ASR take 1 including the 53mm offset and yellow a copy including MK-255 calibration curve, offset 10dB lower are take 4 1000 points curves.

MZKM not shure agree, admit i'm not Klippel trained and have no idea of their software and interface other than its probably very complicated and long haired technical stuff to manage and dial in, about that upper octave notice take 4 was technical set to 1000 points (Killing 2 hours run) plus probably higher order expansion, now please notice lower image where the original measurement not a remeasure of Harbeth 30 was reloaded and asked recompute with higher order expansion gave also dip up there.

In general it could be the higher and higher optimal resolution amirm get system dialed in to perform and extract, the more we see a little roll off in say 20kHz area, say that because have noticed datasheet for MK-255 tell a 1/2" number and not a 1/4" as most and heck while its probably a microphone to die for would imagine there is probably a bandwidth bill to pay using mass of 1/2" verse a 1/4 incher.

 

[edechamps]

upper black is ASR take 1 including the 53mm offset

Unless I missed it, @amirm never published any data that includes any kind of "spatial offset" (i.e. data reference axis different from measurement axis). If what you did is just recenter the existing data an a different vertical angle, then please state the angle instead - using such terms as "53mm offset" could be misleading because it makes an implicit assumption on measurement distance (which in this case I suspect is wrong).

 

[BYRTT]

Yes sir but think maybe i really can't know because that robot was all over the holographic field visit about 500 unknown adresses in space, its origin was probaly set 5mm centered in front of tweeter, then it compute its thing and outputs the 36 hor and 36 ver directional steps, and it is these steps i can rotate X or Y in mm and to calculate in degree have to now at what distance and there will say at the moment i can't make my mind

EDIT can it help the minus 53mm number on Y axis is the difference from tweeter center axis to Neumann recommended on axis.

 

[hardisj]

OT a bit but I have to admit, when I saw @amirm complain about how long the test took at 2 hours I snickered. Amir, if you only knew how good you have it. When I test loudspeakers it takes that long and that's with me doing all the work. Setting it up, sweeping, going back to the speaker to turn it, sweep, and then all the post-processing of data... I haven't started my process of developing CEA2034 tests yet but I can bet it's going to take me considerably longer. And it won't be automated to where I can "set it and forget it" and go grab dinner or play with my kid. So, while 2 hours may seem like a very long time to you, just think of how much worse it could be.

 

[LeftCoastTim]

E
You paid 50,00 USD
to Audio Science Review LLC

Actually, it might be more profitable to ASR if Amir tested a different KH series speaker every few months, each of which is "slightly damaged" in some way (Oops, it fell off the stand again ; ) to keep interest (and donations!) up.

Seriously though, I think people here has waaaay too much confidence in manufacturer's self reported numbers and what's actually observed in the real world, especially for non-solid state, mechanical products, like speakers.

 

[Laserjock]

Man, I thought that said 50 thousand; then I remembered a lot of non-US countries switch the comma and period, and then I noticed the # of 0’s.

That was a hard lesson in a recent training I had in Europe. A Cataract laser’s precision is very important.

 

[Sancus]

Seriously though, I think people here has waaaay too much confidence in manufacturer's self reported numbers and what's actually observed in the real world, especially for non-solid state, mechanical products, like speakers.

I'd agree with this if it was just the on-axis Neumann graph we were referring to, especially since it's very low resolution and says "interpolated", whatever that means, with no information on smoothing level or anything else. It's not good.

However we also have Sound and Recording's data. It is reasonable to question significant differences between these two sets of third party measurements, especially if there is a good reason for those differences as exposing that reason educates everyone on the measurement process and on how much uncertainty and variance it involves.

 

[williamwally]

Just wondering if anyone could help me get up to speed and clarify the issues people are seeing in the measurements now?
Are the current issues in the high frequency measurements because the sound-field expansion measurements (angles) and subsequent calculations cannot be separated from the regular gated measurements that it runs, or is my understanding way off base?

 

[amirm]

I am disappointed to continue to see people wanting to get to published anechoic measurements as if they are the bible. None of the scrutiny that is applied to NFS is being applied to their measurement. Any issue with their data is brushed aside with conjecture.

So here is some hard data on a speaker that was measured in two anechoic chambers and with NFS:

Look at the same dip above 10 kHz (this is a bookshelf speaker). Look at how one anechoic chamber (blue) showed it to be deeper than the NFS and the other chamber. The same chamber is showing a higher peak at 5 kHz.

The above graph was performed as part of the due diligence by that company who then proceed to purchase the NFS system.

You guys want to get Neumann to come here and defend their measurements and explain how their speaker is free of baffle diffraction that causes such dips on-axis, I am game. Otherwise we are back down the unhealthy path of obsessing over detail that is not material. And exists in every type of measurements.

 

[LeftCoastTim]

I'd agree with this if it was just the on-axis Neumann graph we were referring to, especially since it's very low resolution and says "interpolated", whatever that means, with no information on smoothing level or anything else. It's not good.

However we also have Sound and Recording's data. It is reasonable to question significant differences between these two sets of third party measurements, especially if there is a good reason for those differences as exposing that reason educates everyone on the measurement process and on how much uncertainty and variance it involves.

I was thinking of individual unit differences when I said "real world". I doubt Sound und Recording bought a few units off the shelf somewhere and did their measurements. And if they found something odd, they probably reached out to Neumann, who might have provided a "pristine" sample for a well known publication. Who knows.

All I'm saying is that when you buy something from the store, you're rolling the dice. I lost out on my KH80, it seems, even with Neumann's outstanding QC control otherwise.

 

[napilopez]

Actually, it might be more profitable to ASR if Amir tested a different KH series speaker every few months, each of which is "slightly damaged" in some way (Oops, it fell off the stand again ; ) to keep interest (and donations!) up.

Seriously though, I think people here has waaaay too much confidence in manufacturer's self reported numbers and what's actually observed in the real world, especially for non-solid state, mechanical products, like speakers.

And I think you greatly underestimate the repeatability of DIY measurements, especially for high frequencies . No one cares about neumann's data in isolation (aside from the fact their previous measurements for other speakers have also been verified by third parties). It's because it's been verified by S&R and me that people want to know the reasons for the differences. Well, no one else cares about my amateur DIY measurements probably, but mind you I did measure two different KH80s.

You know, you could actually help out here, since it's your KH80. Grab a cheap Umik-1 or whatever calibrated measurement mic you have. I'd be more than happy to send you one myself if you don't have one, I've got three. Measure the KH80 on a stand as far from any wall as possible, at a distance of 1m or greater. Shouldn't be too hard to get enough gating for reliable measurements above 1K or so. You only need to do one measurement, since we're only concerned with on-axis data, maybe two if you want to compare the tweeter with the reference axis. Shouldn't take more than 5-10 minutes. Gate out the reflections and it's essentially the same as an anechoic measurement.

Could be interesting to see what your results are. I bet you they'd look closer to S&R and mine than the NFS'. But hey what do I know.

Again, please understand I'm well aware these differences are mostly immaterial for broader comparative purposes, the reason why Amir bought the NFS. That data is much more important than differences of a db here and there largely being discussed here. But some of us still want to know what the NFS does well and what it might be a bit weaker at, or at least how it might be different than existing measurement methods. If the NFS is capturing something for single, on-axis measurements that can't be captured by a normal anechoic chamber or DIY methods, I want to know too. It'll come to play more as more speakers are tested.

 

[edechamps]

I am disappointed to continue to see people wanting to get to published anechoic measurements as if they are the bible. None of the scrutiny that is applied to NFS is being applied to their measurement. Any issue with their data is brushed aside with conjecture.

To be clear, as far as the bass issue (dip around 80 Hz) is concerned, I am not saying your measurement is wrong (since you fixed the level issue). I am not saying third party measurements are wrong, either. I'm just saying the discrepancy is unexplained. It would be nice to solve that mystery and learn something in the process, improving the state of speaker measurement for everyone. I would suggest sending an email to Klippel and Neumann and ask them about this. Klippel in particular is used to measuring Neumann speakers themselves so they might have something to say about this issue.

As for the high frequency issues, I think there's still a bit of "low hanging fruit" for improving accuracy somewhat. You looked into increasing the number of points, which didn't have much effect, but it was still worth trying. I suggested adjusting the reference axis that the plots are generated at (not the measurement axis) to match the manufacturer-specified reference axis. If that works, then it seems like a quick and easy way to fix some of the weirdness around treble tilts and the like.

So here is some hard data on a speaker that was measured in two anechoic chambers and with NFS:

Do we know the model of that speaker, and the axis for which the frequency response is plotted? The reason I'm asking is because we know, from past discussions, that some speakers can be easier for the NFS to measure than others. For example, I would expect coaxial speakers to be a very easy task for the NFS because the sound field is the simplest (single source). Likewise, speakers where the reference axis is the same as the tweeter axis will not raise the question of how to define these axes in the NFS. In such easy cases, the comparison graph between anechoic and NFS will look good, but that would not mean much when it comes to more "interesting" speakers.

 

[LeftCoastTim]

And I think you greatly underestimate the repeatability of DIY measurements, especially for high frequencies . No one cares about neumann's data in isolation (aside from the fact their previous measurements for other speakers have also been verified by third parties). It's because it's been verified by S&R and me that people want to know the reasons for the differences. Well, no one else cares about my amateur DIY measurements probably, but mind you I did measure two different KH80s.

You know, you could actually help out here, since it's your KH80. Grab a cheap Umik-1 or whatever calibrated measurement mic you have. I'd be more than happy to send you one myself if you don't have one, I've got three. Measure the KH80 on a stand as far from any wall as possible, at a distance of 1m or greater. Shouldn't be too hard to get enough gating for reliable measurements above 1K or so. You only need to do one measurement, since we're only concerned with on-axis data, maybe two if you want to compare the tweeter with the reference axis. Shouldn't take more than 5-10 minutes. Gate out the reflections and it's essentially the same as an anechoic measurement.

Could be interesting to see what your results are. I bet you they'd look closer to S&R and mine than the NFS'. But hey what do I know.

Again, please understand I'm well aware these differences are mostly immaterial for broader comparative purposes, the reason why Amir bought the NFS. That data is much more important than differences of a db here and there largely being discussed here. But some of us still want to know what the NFS does well and what it might be a bit weaker at, or at least how it might be different than existing measurement methods. If the NFS is capturing something for single, on-axis measurements that can't be captured by a normal anechoic chamber or DIY methods, I want to know too. It'll come to play more as more speakers are tested.

I can do a gated measurement when the speaker comes back to me. I could do one to my other speaker this week as time allows.

 

[Sancus]

All I'm saying is that when you buy something from the store, you're rolling the dice. I lost out on my KH80, it seems, even with Neumann's outstanding QC control otherwise.

I agree S&R's measurements have no particular authority over the NFS, but I think the comparison has revealed some useful things.

I'm guessing yours is the one being measured? I'm not sure we know enough to say it's out of spec, really, with the first unit's measurements being inaccurate due to the field expansion issue discovery. So maybe they all have that 13khz dip. Even the in-room measurement has it, although much diminished.

My own in-room measurements of the KH80 (NOT gated) do also indicate a dip there, measured on axis pointed directly at the speaker from 2m. But I'm not running out to return the speakers, or anything. It's unclear to me how audible it should be, but given it seems reduced in magnitude in every in-room measurement, it doesn't really bother me.

If we find a similarly sized, even better measuring active monitor I might swap them out one day when I feel the upgrade bug, but that's about it.