Neumann KH 80 DSP Speaker Measurements: Take Two

[amirm]

One possible error is that Audio Science Review’s CEA2034 calculation was done with the virtual measurement mic at 0.3 m (unless I’ve misunderstood the graph). This is then a near field measurement with a different frequency response compared to a measurement at 1.4 m (closer to far field). The effect is shown in the below graph, but this still doesn’t fully explain the earlier measurements on this forum.

Yes, you are misunderstanding the measurement distance. Klippel is a near-filed scanner. By definition, it will measure in the near field which in my case was around 0.3 meters. So that is the actual mic distance, not virtual. Klippel software however uses that data to solve the differential wave equations. Once there, it can compute the far-field response at any point away from the speaker. The CEA-2034 module which generates the frequency response measurements follows the standard:

As we see above, 2 meter is a compromise selection in the standard as a balance between correct far field response, and not needing too large of an anechoic chamber.

Klippel NFS follows the standard, projecting the soundfield to 2 meter, and then reporting at 1 meter. Alas, its computation of SPL is incorrect for active speakers and hence the manual notation I put in there with respect to loudness of the speaker under test.

Importantly, Klippel NFS, using its optional module that I use, removes the effects of the room/reflections. This is not the case in anechoic chamber where the drivers/ports and microphones could be in modal region. A simple calibration is not enough to remove this frequency dependent effect. As such, anechoic chamber measurements could be subject to error.

Your use of 1.4 meter distance is not in accordance with the standard so can have some effect due to above factor. If we had the raw, uncalibrated response of your measurements, it would help figure out what modal effects there are. Do you have this by chance?

Summary
Klippel NFS does properly use "far-field" response of 2 meter as defined in the standard. Measurements are made in near-field to achieve far better signal to noise ratio and allow for measurements in ordinary rooms. Removal of room reflections then provides free-field response better than what is achieved in just about any anechoic chamber.

 

[amirm]

Your transparency in providing detailed measurements both on the Neumann website and here is unrivaled, as far as I know.

While a lot of what is brought is appreciated, it is not the best I have seen. Best manufacturers come to me, provide another sample, and take back what I tested for their measurements. We then have a complete circle to resolve what the problem might be. This would have been a far faster process than the many months that this interaction has taken.

Ultimately we are not helped much here with them measuring a different speaker than the pair I tested. What do I change moving forward? How do we resolve the next dispute? Surely by not taking 5 months and testing another sample elsewhere.

The approach, which I understand from their point of view, is to disprove the data here. It is not to help advance a collaboration where we arrive at a mutual solution. It leaves me half full as it is.

 

[echopraxia]

While a lot of what is brought is appreciated, it is not the best I have seen. Best manufacturers come to me, provide another sample, and take back what I tested for their measurements. We then have a complete circle to resolve what the problem might be. This would have been a far faster process than the many months that this interaction has taken.

Ultimately we are not helped much here with them measuring a different speaker than the pair I tested. What do I change moving forward? How do we resolve the next dispute? Surely by not taking 5 months and testing another sample elsewhere.

The approach, which I understand from their point of view, is to disprove the data here. It is not to help advance a collaboration where we arrive at a mutual solution. It leaves me half full as it is.

This makes sense. Naturally, I was thinking from the perspective of a objectively-minded consumer shopping for speakers (where usually more data is better), rather than as a volunteer reviewer being criticized. What I mean here is: I don't know of any other speaker manufacturer who posts measurements as detailed as Neumann does on every single product page on their website.

I certainly hope they follow through to help understand the source of these discrepancies. If their only action here is to drop new measurements and then disappear, with no other context or assistance in resolving the discrepancy, then I would agree that will come across as mostly self-serving. From an outside observer’s perspective though, it’s not clear whether that has happened yet.

 

[napilopez]

Yes, this seems extremely important. I was under the impression that the Klippel software doesn’t have virtual mic distance as an adjustable parameter to the CEA2034 computations. I was also under the impression that CEA2034 requires an exact microphone distance of 2m (virtual or real being irrelevant here) with +6db added to scale SPL for easy comparability with 1m SPL specs.

If the various CEA2034 charts we’ve seen (around here and elsewhere) have varying measurement parameters like this, doesn’t that throw the whole notion of directly comparing different speaker’s spins and preference scores into question?

Fortunately, even if so, I believe it should be possible (in theory at least) to recompute CEA2034 charts using consistent virtual mic distances — since the whole point of the Klippel data is that it enables such things without re-measuring.

While technically yes, 2m is the recommended distance, I certainly wouldn't say the standard is stringent about requiring that as an 'exact' distance.

The standard also says "ideally measurements should be made in the far field" of the device, and that can vary dramatically. 2m is basically provided as a practical recommendation that works for most speakers. For some speakers, like the KH80, it's very likely overkill. Consider that 2m is provided as a universal recommendation for anything from desktop speakers to massive towers. 2m is obviously not necessarily suitable or a minimum requirement for all of them.

I only test bookshelf speakers, and I've yet to see one that shows a difference at 2m vs 1m (within the resolution of quasi-anechoic measurements, of course). It may be there, but it doesn't seem to be large. If it affects something, it'll most likely be in the vertical measurements. For bookshelves, I usually begin to see differences at about half a meter.

Of course, if suitable to the intended use of the speakers and if youre measurement system can handle it (e.g. not measuring speakers on your kitchen island) using 2m as a minimum makes sense.

 

[Vintage57]

Or they could of watched us in action and decided to respond at a time they thought appropriate. They have a lot more invested and can't assume the science here is any better than anywhere else on the web.

In the big picture we're not their market, pro audio is.

Glass is too big.

 

[amirm]

In the big picture we're not their market, pro audio is.

Our site gets 1 million visitors per month. People just search for reviews and go where they take them. And we are one of those destinations. We are third in-line in Google search:

And they said that they have gotten a lot of inquiries about testing here.

So no, it didn't make sense to wait so long just to contact me and send over a small speaker to test that they had measured. We would have been farther head four months ago.

We should be the best friend of the company seeing how we rely so much on measurements and objective data.

 

[Vintage57]

I think there’s an opportunity here where both sides gain. Possibly start a dialogue with more manufacturers. It may raise the bar for everyone.
If not their loss too.

 

[ernestcarl]

I do suspect they’ll come back around. Yes, it certainly took a while... Hopefully, it’s not going to take them too long to respond to Amir’s request for remeasuring the same sample now.

 

[ctrl]

Ultimately we are not helped much here with them measuring a different speaker than the pair I tested. What do I change moving forward? How do we resolve the next dispute? Surely by not taking 5 months and testing another sample elsewhere.

The approach, which I understand from their point of view, is to disprove the data here.

Can you blame the manufacturer?

After the first measurement of the KH80, many forum users tried to show you that your measurements could not be correct, since other independent measurement results were also very close to Neumann's measurements.

Your reaction back then was not really accommodating. You simply wanted to perform as many measurements as possible and not bother with "details".

https://www.audiosciencereview.com/...nn-kh-80-dsp-monitor-review.11018/post-312156

I don't trust any manufacturer or magazine measurements relative to capabilities we have. Using the identical measurement system allows us to much better compare one speaker to another.

https://www.audiosciencereview.com/...eaker-measurements-take-two.11323/post-323687

You still seem to think that their data is the bible and anything found different must be an error on our part, and not theirs. Company's own measurement data is useless as it has just a graph and nothing else but the words "interpolated" written on it. The third-part measurements are very well subject to errors per their own documentation and issues stated above.

I am leaving room for NFS system having errors but you are a long way away from justifying the aggressive stance you are taking.

Your attack vector needs to be to get Neumann to prove their measurements are correct and are not affected by their measurement techniques.

I can understand that the manufacturer chose this way and now proves with the help of Klippel that the ASR measurements are incorrect.

But even this brings us further, because the Neumann/Klippel measurement together with the low tolerances that Neumann specifies for the KH80, this loudspeaker can certainly serve as a rough measurement reference, at least in the 50-10000Hz range.

 

[amirm]

I can understand that the manufacturer chose this way and now proves with the help of Klippel that the ASR measurements are incorrect.

Not at all. What Neumann has proven is that the Klippel NFS is able to measure their speaker the same as they have with anechoic chamber. They took a sample speaker, measured it themselves, and then sent it to Klippel and got the same results. So all the arguments when I started testing that the Klippel NFS must be doing things wrong, that you can't measure near-field, etc., etc. were completely false as I stated then.

The key here is that Neumann did NOT measure the speakers that I measured. You have no idea if the sample used was especially good and had a flat response relative to the ones I tested. It is not like they went to a store and bought a random sample and had it tested. I have no reason to doubt them. Nor have a reason to trust them.

As I noted earlier, the proper way to show my measurements were in error was to contact me and ask for the sample. Or offer me theirs. Remember the Kali measurements? They didn't reach out to me but went and found the returned speaker and discovered that the woofer was faulty. A repeat measurement with another sample showed much better performance to their and our satisfaction. And they offered to send me a speaker (although at the end did not and I had to buy my own sample). And they hung around here while we had the discussions to get everything sorted out.

Even if this is not a golden sample issue, we still have unresolved protocol differences. I have and continue to use tweeter axis as reference. One good test would be for them to show tweeter axis measurement. If it matches mine, then your conclusion is wrong again as that would prove my measurements are correct.

 

[amirm]

Can you blame the manufacturer?

Blame? No, this is not an emotional war for me that it is for you. I have tons and tons of speakers to test and life goes on for me unchanged.

The opportunity for Neumann would have been to build a solid relationship with us moving forward. This could have highlighted design and production issues which would have helped them. Or find issues with my testing which would have helped future reviews of their products.

I was in their shoes for decades producing products that got reviewed by others. I always, always tried to build a good relationship and goodwill with reviewers no matter what.

Clearly they are choosing a different path and I am here to say it is not a good one. But not here to force them otherwise. As with electronics, we will find companies that realize that we are the place to highlight good engineering. To the extent they want to ignore us and only deal with us in passing as we see here, is hopefully something that will change in the future, lest they are not dedicated to this principal.

 

[amirm]

But even this brings us further, because the Neumann/Klippel measurement together with the low tolerances that Neumann specifies for the KH80, this loudspeaker can certainly serve as a rough measurement reference, at least in the 50-10000Hz range.

I don't know what this means. Rough measurement for what? We already have a proper references in the form of Genelec 8341A:

And company measurements matching mine almost to the letter (on-axis):

This, is a reference. And across the full frequency response.

 

[TimVG]

I was able to match Amir's KH80 listening window measurement above ~600Hz pretty spot on in a decent size space. I have some low range gain due to the column I am using to lift the speaker up to a decent height.

Interestingly enough I also tested a Genelec G2 and in the averaged power response measurement It shows more energy around the 100Hz region when compared to the KH80 despite both manufacturers claiming flat(ness) in that region.

 

[ernestcarl]

Guys... It’s only been a day since their own measurements were posted in this thread (yeah, still, better late than never), so I don’t see why Neumann would not respond back positively for a retest of the exact same sample that was used here. I’d give them at least a few days or a week before doubting/second-guessing the intention.

 

[ctrl]

Not at all. What Neumann has proven is that the Klippel NFS is able to measure their speaker the same as they have with anechoic chamber. They took a sample speaker, measured it themselves, and then sent it to Klippel and got the same results. So all the arguments when I started testing that the Klippel NFS must be doing things wrong, that you can't measure near-field, etc., etc. were completely false as I stated then.

The key here is that Neumann did NOT measure the speakers that I measured. You have no idea if the sample used was especially good and had a flat response relative to the ones I tested. It is not like they went to a store and bought a random sample and had it tested. I have no reason to doubt them. Nor have a reason to trust them.

As I noted earlier, the proper way to show my measurements were in error was to contact me and ask for the sample.

In your reply you completely ignore the narrow tolerances for the frequency response on axis that Neumann specifies.
Reproduction accuracy between 100 Hz and 10 kHz:
100% of loudspeakers produced: ±0.26 dB
80% of loudspeakers produced: ±0.23 dB
50% of loudspeakers produced: ±0.17 dB

Other measurements of various KH80 production samples have shown how reliable Neumann's tolerance specifications are - which correspond very well with Neumann's measurements.

As showed in post#433, the deviations of Neumann's measurement in the range 40-11000Hz compared to Klippel's are just +-0.5dB (apart from a few outliers) - which means that these are also very reliable:

In addition, we know today that the ASR measurements of that time are faulty because
- the protective cage around the microphone has caused a comb filter effect at frequencies above 5kHz
- if the evaluation of the NFS measurements is not carried out for "small" loudspeakers at a distance of 2m as recommended, phase shifts/phase errors around the crossover frequency and the addition of BR port and woofer will occur when evaluating at a distance of e.g. 0.3m

Furthermore, the low frequency range is so often conspicuous in ASR measurements that another previously undiscovered error could lurk there - see below.

This, is a reference. And across the full frequency response.

Okay, if the Genelec 8341A is to be the reference, then we compare the ASR measurement with the manufacturer's specifications. At that time the problem with the protective cage on the microphone was already solved, therefore the interferences above 5kHz are missing in the ASR measurement. The other problems were still present.

If we look at the deviations from the manufacturer's measurement, the following diagram results:

In the 200-13000Hz range, the measurements of ASR and the manufacturer, within the scope of the measurement accuracy, are quite well matched.

But below 200Hz or below 60Hz at the latest, the deviations are, as with many other measurements of loudspeakers, much too high - when using your self-designated reference.

 

[echopraxia]

If anything, it seems to me that Klippel should have a strong vested interest in investigating and resolving these discrepancies — both between ASR measurements and their own, and between their own and Neumann.

But especially, I would expect Klippel would want to resolve differences between their own measurements and ASR. Because it’s the same machine, it might not reflect so well on Klippel if discrepancies keep appearing that might indicate calibration quality issues, or setup validation issues: the best test systems are self-validating to an extent, in realization of how many ways things can (and will) go wrong or vary “in the field”. If I bought one, I would want to know that the results I’m getting are extremely reliable, and that the system setup is self-validating to some extent so that you can’t just blame potential user error (except in exceptionally rare of cases).

Of course, resolving this is not made easy if Neumann doesn’t share the same sample with ASR that they gave to Klippel.

But as others have said: Hasn’t it only been a day since they posted the measurements?

Unless I am missing some information, perhaps we should wait and see how they follow up before jumping to judgements of their intentions. If I were in their shoes and I read some parts of this thread, I might be a little hesitant or fearful of direct interactions in a hostile environment. I understand frustrations, but I just hope we can keep the tone here friendly and professional enough that they are not driven away.

 

[QMuse]

But especially, I would expect Klippel would want to resolve differences between their own measurements and ASR. Because it’s the same machine,

It is the same machine, but it wasn't the same speaker. Until the same speaker is measured with both machines it is hard not to speculate about inter-specimen differences.

 

[Vintage57]

On the subject of time. Most non essential businesses, (Neumann being one) in Europe and North America have been closed or down to skeletal staff the past 3 months.

Give them time then lash out if you still feel a need to.

 

[Mawclaw]

I think they will eventually send the sample. I notice in my job that a lot of the German and Japanese manufacturers are not very "nimble" especially when it comes to social media. I bet it will just take a few weeks and likely many meetings for some unfortunate employees to get it moving.

 

[Deleted member 16502]

Measurements should be repeatable, especially when you use a commercial measurement rig that's not exactly cheap. Then again, I think it's remarkable that a major company responds to third party measurements with additional and comparable data at all. I'll be patient.

PS: I would however be interested in a comment on why the FR shown on the Neumann page is flatter (compare >10khz) that the one posted here (both Neumann and Klippel measurements).