Could you post the CEA/PIR data so I can calculate the score difference (if any) for this as well?
-
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!
Neumann KH 80 DSP Speaker Measurements: Take Two
- Thread starter amirm
- Start date
OP
- Thread Starter
- #222
Here is the spin data for 1000 point and high order of expansion.
OP
- Thread Starter
- #223
We covert this already. Measurements must be at tweeter axis or the system will have more error. The software automatically computers what it thinks is the reference axis.
OP
- Thread Starter
- #224
The angle of incident is zero as well for the front measurements in NFS.
Err… okay. I thought the point of the experiment was to test the theory that, with more points, the system becomes more accurate at modelling the response precisely when the measurement axis is not the tweeter, because that's where problems were uncovered. Looks like I was confused and this was about investigating the fitting error in the tweeter axis measurement case instead.
I understand that there are problems when the measurement axis used is not the tweeter, that's quite evident from your first post. But I thought it would be nice to understand why that's the case. Anyway…
By "front" I assume you mean directly in front of the tweeter. Sure. The problem is, NFS doesn't just measure in front of the tweeter, it measures in other places as well with various angles of incidence, and (assuming I understand the process correctly), measurements taken at other angles could affect the on-axis, i.e. 0° response, because they participate in the overall field calculation. This is problematic, because in these cases, at high frequencies, the SPL measured by the microphone is not the true SPL at this point. This can in turn throw off NFS calculations and create errors in the response.
I'd assume the error will be reported as fitting error but it can't easily be fixed just by adding more measurement points, because these additional points will be wrong too. This is not a problem of not having enough data points for field calculations - the problem is that the data points themselves are wrong.
My hypothesis is: if the NFS is focused on the tweeter axis, the math works out in such a way (or there is special handling involved in the Klippel software, I don't know) that the measurements taken at odd angles from the tweeter do not affect for the on-axis response at high frequencies, thus avoiding the problem. (It might be that the Klippel software simply decides to not "trust" measurements at an angle at high frequencies for the purposes of computing on-axis response.) But if you tell the NFS to use a different measurement axis, then the whole thing breaks down and now you get problems with microphone directivity because the "trusted" axis is not in front of the tweeter anymore. Thus explaining the measurement you showed in the first post of this thread.
If my hypothesis is correct, then the solution is: for normal speakers with a well-defined tweeter, keep measuring using the tweeter axis, but post-process the measurement data using an offset to produce the spinorama along the correct reference axis (assuming that's possible - so far I've never seen you post any data using this technique). For exotic speakers that don't have a single, forward-firing tweeter, I wouldn't assume the problem is solvable even if you crank up the number of measurement points.
I am a bit surprised, I had expected that by reducing the fitting error the measurement would be much less rippled in the frequency range above 5khz.
The ripple in the frequency range 100Hz to 10kHz is 1.2dB according to the S&R measurement.
With the Klippel NFS, the old and new measurements arrive at a ripple of about 3dB.
The fitting error in this frequency range is <0.5dB (as Amir has pointed out, this is generously calculated and in reality usually below this value), so the measurements of S&R, Neumann and Amir should be closer together.
Here is a comparison of the old (from the first KH 80 thread) and the new measurement for vertical -10deg.
Comparison of the two measurements:
Comparison with normalization to the old measurement:
Comparison with normalization to the new measurement:
The ripple in the frequency range 100Hz to 10kHz is 1.2dB according to the S&R measurement.
With the Klippel NFS, the old and new measurements arrive at a ripple of about 3dB.
The fitting error in this frequency range is <0.5dB (as Amir has pointed out, this is generously calculated and in reality usually below this value), so the measurements of S&R, Neumann and Amir should be closer together.
Here is a comparison of the old (from the first KH 80 thread) and the new measurement for vertical -10deg.
Comparison of the two measurements:
Comparison with normalization to the old measurement:
Comparison with normalization to the new measurement:
I am not using symmetry optimization so it is measuring all around. Left to right symmetry is good with the KH80 so I could have done that:
View attachment 49433
Then again given the food fights we are having over small variations, it is better that I am not.
Haha I think you will find most of us have no trouble with symmetric measurements on symmetric speakers! The time savings are definitely worth it if you're running low on time or if you plan on measuring multiple speakers. It's pretty standard procedure.
I doubt anyone would complain if you chose to do this going forward or in specific cases where it would save you time.I think most of the food fighting has been over what aspects of sound we would expect to be consistent between measurement setups(i.e. low q deviations). But we mostly know the reasons now - the klippel is a new system so we're learning as we go. In some ways it's very different than traditional measurements, so we have to adjust our expectations. This isn't to say that's a bad thing, we're just learning.
Wouldn't have objected at all to speakers with a symmetrical design.
You act as if some users here in the forum are splitting hairs over every detail
It would halve the time needed for the measuring process.
Note, however, that the two dotted grey lines are only 1.2dB apart.
So the dip you marked should be in the range of 0.2dB.
As I think you mentioned or someone else mentioned before there was less compression on this sample than the previous Neumann sample due to the SPL level? This would be something a compression test could quickly put to rest, which I think Amir said he was working on.
That said, assuming the overall level was not too loud(do we know the actual SPL @amirm?) I'm inclined to agree with the NFS. Nearfield bass summation, which both S&R and I use, is always a bit of a guessing game. On the KH80, it's a also bit harder than usual because the fixed grille doesn't let you get as close to the woofer as you should. I'm not at my REW computer, but between placing the microphones at the right distance (which, again, you can't really do on the KH80 because of the fixed grilles), summing the port and wooofers correctly(which is also more difficult because of the fixed grille, plus picking the exact right effective piston diameter), picking the right splice point, and properly modelling baffle loss, I always assume some deviations in nearfield summations. Well-done ground plane measurements might be better in this regard though they tend to be noisier.
I previously used this comparison to show the consistency in my HF measurements of the Neumann (using two different microphones, speakers, and setups), but it can also be used to show the (slight) inconsistency in my nearfield bass summations.
A slightly bigger difference is in the on-axis measurement in this image, where I'd picked a different splice point when it was made:
Obviously small differences, but just an example of potential deviations in nearfield summations even when done by the same person. It's also worth noting I did these summations after seeing Neumann's data. While that shouldn't affect them in terms of procedure, it did mean I automatically assumed the data was correct because it closely tracks neumanns. WhatBias always finds a way
On the other hand, to the point about smoothing, I do not think S&R is using very much at all. My measurements above are 1/24 octave smoothed and are not averaged except for the listening window. Even taking into account the reduced resolution of gated measurements, above 3K or so the klippel and my DIY measurements should have similar resolution after smoothing. the HF deviations are smaller than those captured by the klippel, despite being recorded in a living room and my graph using a taller aspect ratio than Amir's(relative to dB).
So here I assume either:
-The bumps are caused by reflections in the setup ( I take extra care to avoid these by using a long boom angled such that the mic doesn't "see" it's stand)
-The klippel is somehow capturing something about HF soundfield not captured by a single mic.
To be clear these bumps do not personally concern me much though other than as an academic curiosity. I was more concerned with low-q deviations which we now mostly have the answers for.
Last edited:
- Joined
- Apr 15, 2019
- Messages
- 375
- Likes
- 758
I think Amir should buy his own KH80 so that he can re-measure it every month. Why? To keep forum traffic up!Let’s see other speakers now
Let’s see other speakers now
It's not really about the KH80 anymore =]
We're just trying to understand the strengths and limitations of the NFS, and the KH80 is probably the best way since multiple members have one to measure, there are two sets of anechoic data, and it might be one of the flattest speakers on the market. But Amir certainly doesn't need to spend more time on this.I think Amir should buy his own KH80 so that he can re-measure it every month. Why? To keep forum traffic up!
kH80 remeasurements will pay for the klippel within a year or two, I assume!
In the normalized view in Post#227 the compression can be seen very well due to the high SPL level.
Off axis for a measurement mic is going to be a well understood question. You have a small diaphragm omni. There is a very very tiny directional effect due to the diameter of the diaphragm. What there isn’t is some ill defined weirdness that can’t be managed that throws things off. If the particular mic in use had enough of an issue it would be reasonably easy to incorporate a correction during processing. One would assume Klippel have worked through this question.
Checkout the performance of a well known measurement mic here. https://earthworksaudio.com/wp-content/uploads/2017/07/M50-Data-Sheet-2017.pdf
Checkout the performance of a well known measurement mic here. https://earthworksaudio.com/wp-content/uploads/2017/07/M50-Data-Sheet-2017.pdf
Did I get it right, the KH 80 measured now comes from a different member than the one measured in the first thread?
The frequency response deviations of both measurements (old from the first thread and the new high resolution measurement with another KH 80) are less than +-0.5dB in the range 100Hz to 10kHz at vertical -10deg.
This shows, on the one hand, that Amir's measurements have a very good reproducibility, and that Neumann's statements regarding the low serial dispersion of the loudspeakers were not exaggerated.
The frequency response deviations of both measurements (old from the first thread and the new high resolution measurement with another KH 80) are less than +-0.5dB in the range 100Hz to 10kHz at vertical -10deg.
This shows, on the one hand, that Amir's measurements have a very good reproducibility, and that Neumann's statements regarding the low serial dispersion of the loudspeakers were not exaggerated.
Neumann change:Years? I thought I was bad with my review pace.
5.34 to 5.24; the new dip >13kHz hurts it.
Harbeth change:
5.08 to 5.31; everything except LFX was improved slightly.
Do you think I should change the scores I have in my spreadsheet, or keep them the same to match the same # of points the other speakers were measured against (regardless if some had less/more fitting errors?
OP
- Thread Starter
- #236
Correct.
OP
- Thread Starter
- #237
Since I am going to try for lower fitting error from here on, I think it makes sense to update these. We will have a lot more speakers measured in the future than the past.
Cool. (this now puts the Harbeth as the best speaker measured so far though).
Last edited:
Right, but one has duplicate the amp the speaker cables and cable bridges if used.
The Neumann is active and therefore free of matchy matchy
- Joined
- Jun 5, 2016
- Messages
- 2,878
- Likes
- 4,699
There has to be something wrong with the metric if a speaker with such a large dispersion disruption at the crossover gets a higher score than a speaker with very smoothly rising directivity throughout the audible band, merely because the better speaker has a dip at 13kHz.
Similar threads
