• 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!

dominikz

Addicted to Fun and Learning
Forum Donor
Joined
Oct 10, 2020
Messages
803
Likes
2,626
Hello all,

This will be a summary of (many :)) measurements I did with the Neumann KH 120A that I recently purchased.
The pair were manufactured in 2013 and sport the older, IMO slightly industrial looking finish:
1632683838819.png


A few notes on measurement methodology:
  • The loudspeaker has been measured at ~130cm distance from any reflective surfaces, with the microphone at 50cm distance from the speaker (I also remeasured at 1m to confirm the results, as you will see below), so the first reflection was around 6ms after the main impulse.
  • The reflections were removed by temporal gating at ~6ms in REW to get some (low) resolution of quasi-anechoic measurement down to about 200-300Hz range (resolution is much better above ~1kHz).
  • LF response (<350Hz) is spliced from nearfield measurement and therefore not absolutely accurate (especially since one cannot directly access the woofer due to the non-removable grille). Take the absolute shape of the response there with a few grains of salt.
  • The loudspeaker front baffle was vertically aligned with center of speaker stand/turntable rotation
  • Loudspeaker stand has been covered with a blanket to reduce potential reflections and care was taken to angle the microphone boom to minimize boom reflections
  • Microphone was pointed to the manual specified acoustic axis (17cm from bottom and 9,1cm from side of loudspeaker), confirmed with laser level
  • REW, Cross-Spectrum labs calibrated Dayton EMM-6 microphone and RME Babyface interface was used to measure, and VituixCAD to generate the spin (and exported with 1/24 octave resolution)
I also built a (pretty simple) new turntable/stand to facilitate a bit more precision in my measurements:
1632684120321.png

The stand itself is 85cm tall, but I can extend it to about 115cm by putting my desktop monitor stand on top - which puts the speaker at more-or-less ideal height for measuring in my room (midpoint between floor and ceiling) - which is how I measured.

Basic measurements
Anyway, here's the spinorama for one of the units:
Neumann KH 120A - spinorama at 50cm from acoustic axis - unit 1.png

Preference rating as calculated by VituixCAD is:
  • Speaker alone: 6.1
  • Speaker with sub: 8.1
Driver components:
Neumann KH 120A - Driver Components Nearfield.png

We can see port resonances are well attenuated, but there is a visible, though low-in-level, resonance (~1,6kHz) around the crossover.

Horizontal directivity to 90 degrees:
Neumann KH 120A unit 1 Directivity (hor - line).png

Again we can see a bit of resonance coming up around 1,6kHz at very off-axis angles. Overall it looks really good, as off-axis response is very similar to on-axis in shape, with just the slope changing as we move more off-axis.

Horizontal directivity (normalized to on-axis response):
Neumann KH 120A unit 1 Directivity (hor).png

Vertical directivity (normalized to on-axis response; we can see here that crossover and acoustic response is indeed optimized for the specified acoustic axis):
Neumann KH 120A unit 1 Directivity (ver).png


Comparisons
Here's how the spin compares to a Klippel NFS spin published by Neumann here:
Neumann KH 120A - spinorama - Compared to Klippel spin.png



As mentioned, the nearfield measurement below 350Hz is not fully reliable and is only qualitative, and resolution until ~1kHz is relatively low due to the 6ms gate - but we still have a pretty good match.
Above 1kHz, where I expect a close match of the two methods, we also see a few small deviations, namely:
  • In my measurement there is a bump in most curves around 1,6kHz. The bump is especially visible in off-axis responses but also in the port NF measurement (see graphs above).
  • My measurements show a slightly larger directivity error around 2kHz
  • My ER curve has an ~1dB dip 2-6kHz
    [EDIT 2021-09-27]: The source of the ER difference has been identified as a calculation error in older Klippel SW by @napilopez in post #6 (fixed since then by Klippel), supported by the comparison graph of ER calculations vs Klippel in post #8.
Although all of the deviations are rather small and fall within ~1dB vs the NFS, I still remeasured several times (different units, setups and distances) to see if I missed something, but got similar results each time. Lastly I reached out to Neumann to check if they see any issues with these results and the applied methodology. Their view was that the results fall within the expected tolerance of different loudspeaker units and measurement setups - so I decided to leave well enough alone and drop further investigation :)
However, since I collected the data, here's a couple more comparisons:
Neumann KH 120A - spinorama - Two units compared.png

As you can see, unit-to-unit variance is pretty low, most of it being in the crossover area (but still within less than 1dB)!

Here's how the same unit measured at two different microphone distances compares:
Neumann KH 120A - spinorama - Same unit measured at 1m vs 50cm.png

The differences are marginal. There's a bit more difference in the 300-900Hz range - but this is mostly due to use of different temporal gates, mandated by the measurement distance. Resolution in this range is pretty low anyway.

Distortion
These THD measurements were done in-room at 50cm distance from tweeter, while SPL on the graphs is scaled for 1m distance (i.e. reduced by 6dB due to distance doubling).
I'm only showing 2nd, 3rd and 5th harmonic to make the graphs less busy (as these dominate the THD anyway) and total THD (2-9th harmonic).
[EDIT 2021-10-02] Additional distortion measurements added in post #14.
Neumann KH 120A - 74dB SPL at 1kHz at 1m.png

Neumann KH 120A - 84dB SPL at 1kHz at 1m.png

Neumann KH 120A - 84dB SPL at 1kHz at 1m (SPL).png

Unfortunately I didn't really measure any high-SPL, but up until ~84 dB SPL that I did measure the distortion seems pretty low!

Let's see how THD of both my units matches:
Neumann KH 120A - 84dB SPL at 1kHz at 1m - two units compared.png

Seems pretty close - again showing very good unit-to-unit matching - better than my previous (though much cheaper) JBL LSR305s.

In-room measurements
I use these on my desktop at about 70cm listening distance - so very nearfield. Therefore I'd expect the in-room response to track the LW pretty well in HF.
Neumann KH 120A - spin LW vs in-room MMM - without EQ.png

As we can see the match to quasi-anechoic LW is pretty impressive above ~1,2kHz! You can also see I have a very severe resonance at my listening position around 130Hz - this I'm normally compensating with the EQ within RME TotalMix SW - however all EQ was disabled for the above measurement.

Equalization
Next I designed a flat-LW-target EQ in REW based on the average of both unit's LW:
Filter Settings file

Room EQ V5.20.2
Dated: 26.09.2021. 11:41:01

Notes:Neumann KH 120A - spinorama LW EQ by dominikz 2021-06-26

Equaliser: Generic
Preamp: -2 dB
Filter 1: ON PK Fc 1578 Hz Gain 2.00 dB Q 1.693
Filter 2: ON PK Fc 1479 Hz Gain -1.70 dB Q 4.614
Filter 3: ON PK Fc 1703 Hz Gain -1.50 dB Q 6.164
Filter 4: ON PK Fc 2885 Hz Gain 0.60 dB Q 2.210
Filter 5: ON PK Fc 5284 Hz Gain 1.70 dB Q 1.929
Filter 6: ON PK Fc 14435 Hz Gain 2.00 dB Q 1.348
Filter 7: ON PK Fc 15664 Hz Gain -3.00 dB Q 3.000
This EQ improves the spin further (simulated in VituixCAD):
Neumann KH120A spin with EQ.gif

The preference score also rises to 6,6 (8,6 with sub).

If I now apply LW-based EQ above 300 Hz and MMM-based in-room EQ below 300 Hz, I get the following in-room response (the two pairs of responses are scaled for better visibility):
Neumann KH 120A - response before and after EQ.png

The peak around 430Hz and the dips before and after look problematic on paper, but I can't hear them with music - or even when trying to pin-point them with a slow-moving sine sweep. On the other hand the narrow (but deep null) around 206Hz is very audible with a sine sweep (but not with music).
Also, in the uncorrected response, the huge room resonance around 130Hz is very audible (with both music and sweeps).
Interesting thing is that there is 0,7dB difference in level between my channels when uncorrected - but this could be aligned with the input gain pot on the monitors or the RME balance control.

The EQ-ed response sounds pretty good to me - all in all! Though I have to say that normally I just use the 3-band RME TotalMix EQ to correct the few larger resonances below ~300Hz and call it a day - I wouldn't say heavy DSP correction we see in graphs above is really needed to enjoy speakers as good as these.

Conclusion
I'm pretty impressed by these speakers - technically they are a clear step-up from the budget monitors I had before. Unit-to-unit consistency is amazing, and match to vendor-provided data is very good - both huge pluses for Neumann in my book!
They of course also sound great, but as I said before, IMHO diminishing returns set-in pretty early in audio these days and one can get pretty similar sound with nice-measuring budget monitors (at least at lower SPLs). Still, there is something to said for owning something as well made and engineered as the KH120As - I'm definitely enjoying mine :D

In case @MZKM and/or @pierre (or others) might be interested to do their magic, CTA-2034-A exports from VCAD are attached as well.

[EDIT 2021-10-05] Added step response plot comparison in post #15.
 

Attachments

  • Neumann KH120A CTA-2034-A by dominikz.zip
    210.3 KB · Views: 158
Last edited:

Matias

Master Contributor
Forum Donor
Joined
Jan 1, 2019
Messages
5,029
Likes
10,796
Location
São Paulo, Brazil
Awesome job, congrats!
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
I use these on my desktop at about 70cm listening distance - so very nearfield.

At such close distance, do you still prefer to leave the HF flat as-is?

I have mine shelved (not the internal EQ, but my own applied digitally) to lessen fatigue with long-term use.
 
OP
dominikz

dominikz

Addicted to Fun and Learning
Forum Donor
Joined
Oct 10, 2020
Messages
803
Likes
2,626
At such close distance, do you still prefer to leave the HF flat as-is?

I have mine shelved (not the internal EQ, but my own applied digitally) to lessen fatigue with long-term use.
I've not 100% decided yet. Initially I felt the HF a bit much for my untreated room so I used the HF -1dB switch at the back. Recently however I've had them set to flat and haven't had any fatigue.
 

napilopez

Major Contributor
Forum Donor
Joined
Oct 17, 2018
Messages
2,109
Likes
8,420
Location
NYC
Amazing thorough work @dominikz! Would you mind sharing how you built the stand? I've been meaning to come with something a little less reflection-prone than my current setup =]

Once again shows the consistency of Neumann speakers.

One note: I'd be willing to bet the difference in the ER around 2-6kHz is not due to an error in your measurement or deviation in your unit, it's because the Klippel measurement has an incorrect early reflections calculation due to the ambiguity in the CTA-2034A document.

If try you calculate your ER using the old incorrect method(which I believe has since been corrected in the latest NFS update) you'd probably see even closer agreement. If I remember correctly, the old way the NFS calculated it was a single average of the following 26 measurements:

Horizontal: 0, ±10 to ±90, 180
Vertical: -20, -30, -40, +40, +50, +60

It places less weight on the vertical component, so in general calculating the ER using that method tends to obscure issues around the crossover, hence less of a crossover dip. The correct method used in VCAD is an average of 5 averages (each of the ER components), and including the full horizontal rear response

Maybe give the incorrect single average a try just to confirm =]

Edit: Given the bump at 1.6kHz is escaping from the port, is present in both of your speakers, and it's not in Neumann's measurement, I wouldn't be surprised if they minimized this resonance in newer models somehow.
 
Last edited:
OP
dominikz

dominikz

Addicted to Fun and Learning
Forum Donor
Joined
Oct 10, 2020
Messages
803
Likes
2,626
Amazing thorough work @dominikz!
Thanks a lot!

Would you mind sharing how you built the stand? I've been meaning to come with something a little less reflection-prone than my current setup =]
Sure! The basis is the Ikea Snudda lazy susan, which I disassembled first. On the smaller circular board I bolted three 80cm long metal table feet with M10 screws (the feet had an M10 thread at the bottom side, normally used for table feet adjustment):
1632723365404.png

On the other side of the table feet I screwed-in a 60cm long wooden shelf with 4 short wood screws per leg - pretty simple stuff, though I took a lot of time to measure and remeasure to try and be as precise as possible. I also used a table drill stand to make sure holes were perfectly vertical and precise:
1632723697236.png

Lastly I made a small modification to the lazy susan to improve stability - ideally one would add small wheels or ball bearings to make rotation easier, but instead I used what I had on hand - the plastic inlay I removed from the lazy susan and some pins (height adjusted with washers to fit the gap between the two circular boards):
1632724986800.png

1632724996252.png

1632725005817.png

This helped a lot with stability, while still not being too hard to rotate.

Once again shows the consistency of Neumann speakers.
Agreed!

One note: I'd be willing to bet the difference in the ER around 2-6kHz is not due to an error in your measurement or deviation in your unit, it's because the Klippel measurement has an incorrect early reflections calculation due to the ambiguity in the CTA-2034A document.

If try you calculate your ER using the old incorrect method(which I believe has since been corrected in the latest NFS update) you'd probably see even closer agreement. If I remember correctly, the old way the NFS calculated it was a single average of the following 26 measurements:

Horizontal: 0, ±10 to ±90, 180
Vertical: -20, -30, -40, +40, +50, +60

It places less weight on the vertical component, so in general calculating the ER using that method tends to obscure issues around the crossover, hence less of a crossover dip. The correct method used in VCAD is an average of 5 averages (each of the ER components), and including the full horizontal rear response

Maybe give the incorrect single average a try just to confirm =]
Wow, that's amazing insight - and IMHO hits the nail right on the head:
Neumann KH 120A - Early Reflections - incorrect and correct calculation method vs Klippel NFS ER.png

The above dotted line is just the single RMS average calculated in REW from the specified 26 measurements (Horizontal: 0, ±10 to ±90, 180; Vertical: -20, -30, -40, +40, +50, +60) compared to published Klippel NFS ER (red) and the ER calculated by VituixCAD.
Seems that indeed the difference comes from incorrect ER calculation in older Klippel NFS SW revision!

Edit: Given the bump at 1.6kHz is escaping from the port, is present in both of your speakers, and it's not in Neumann's measurement, I wouldn't be surprised if they minimized this resonance in newer models somehow.
Might be - would be interesting to know more!
 
Last edited:

napilopez

Major Contributor
Forum Donor
Joined
Oct 17, 2018
Messages
2,109
Likes
8,420
Location
NYC
Thanks a lot!


Sure! The basis is the Ikea Snudda lazy susan, which I disassembled first. On the smaller circular board I bolted three 80cm long metal table feet with M10 screws (the feet had an M10 thread at the bottom side, normally used for table feet adjustment):
View attachment 155844
On the other side of the table feet I screwed-in a 60cm long wooden shelf with 4 short wood screws - pretty simple stuff, though I took a lot of time to measure and remeasure to try and be as precise as possible. I also used a table drill stand to make sure holes were perfectly vertical and precise:
View attachment 155845
Lastly I made a small modification to the lazy susan to improve stability - ideally one would add small wheels or ball bearings to make rotation easier, but instead I used what I had on hand - the plastic inlay I removed from the lazy susan and some pins (height adjusted with washers to fit the gap between the two circular boards):
View attachment 155852
View attachment 155853
View attachment 155854
This helped a lot with stability, while still not being too hard to rotate.


Agreed!


Wow, that's amazing insight - and IMHO hit's the nail right on the head:
View attachment 155842
The above dotted line is just the single RMS average calculated in REW from the specified 26 measurements (Horizontal: 0, ±10 to ±90, 180; Vertical: -20, -30, -40, +40, +50, +60) compared to published Klippel NFS ER (red) and the ER calculated by VituixCAD.
Seems that indeed the difference comes from incorrect ER calculation in older Klippel NFS SW revision!


Might be - would be interesting to know more!

Glad that calculation worked out! I saw it happen several times that there's be a significant difference between the NFS ER and the proper ER, so eventually I went about importing most of Amir's measurements into VCAD for the correct thing lol. Naturally this means the PIR is also affected, albeit to a lesser extent. But yeah in the proper calculation those six vertical angles end up contributing a lot more.

(Btw for those who haven't been following along this particular quirk, MZKM and Pierre have been using the correct calculation for the scores well before the Klippel fix, so you don't have to worry about the scores being wildly off lol. )

Naturally, using this calculation for the ER means your ERDI will also match the Neumann/Klippel one much better. So other than the slightly more prominent bass peak (maybe port was summed a little high? Obviously not a big deal) and the difference at 1.6khz, your data matches impeccably.

One thing I appreciate about this forum and gr is that it's shown just how repeatable measurements can and should be. Even with units made several years apart!
 
OP
dominikz

dominikz

Addicted to Fun and Learning
Forum Donor
Joined
Oct 10, 2020
Messages
803
Likes
2,626
One thing I appreciate about this forum and gr is that it's shown just how repeatable measurements can and should be. Even with units made several years apart!
Indeed. Before finding ASR I never would have expected this level of repeatability of acoustic measurements!
In addition, I find the sense of community here is amazing - one really can learn a lot from other fellow members! Case in point: the ER calculation difference you directed me to :)
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
I've not 100% decided yet. Initially I felt the HF a bit much for my untreated room so I used the HF -1dB switch at the back. Recently however I've had them set to flat and haven't had any fatigue.

If you ever feel like the highs are becoming fatiguing (very loud listening volumes), you could try the ff. "simpler" or gentler PEQs :

1632773134973.png



1632772744901.png

EQ 1 = "mild"
EQ 2 = "milder"

Upper traces are yours (1m). Lower half is from Neumann's own Klippel graphs.

I used rePhase to make the time domain linear in mine, but it's not at all necessary.
 
OP
dominikz

dominikz

Addicted to Fun and Learning
Forum Donor
Joined
Oct 10, 2020
Messages
803
Likes
2,626
So other than the slightly more prominent bass peak (maybe port was summed a little high? Obviously not a big deal)
It is very much possible that I have summed the NF measurements less than ideally for this loudspeaker. Specifically, the issue here was that the non-removable woofer grille made it impossible to put the microphone close enough to the woofer to get a proper NF measurement, so calculating the baffle step and summing the NF responses of the woofer with the two ports was a bit of a guessing game in this case. :confused: It still matches the expectation pretty well, but I assume it would be even closer if I could measure the woofer closer to the cone/dust-cap.

If you ever feel like the highs are becoming fatiguing (very loud listening volumes), you could try the ff. "simpler" or gentler PEQs :

View attachment 155951


View attachment 155950
EQ 1 = "mild"
EQ 2 = "milder"

Upper traces are yours (1m). Lower half is from Neumann's own Klippel graphs.

I used rePhase to make the time domain linear in mine, but it's not at all necessary.
Thanks for the suggestion!
 
OP
dominikz

dominikz

Addicted to Fun and Learning
Forum Donor
Joined
Oct 10, 2020
Messages
803
Likes
2,626
Distortion
These THD measurements were done in-room at 50cm distance from tweeter, while SPL on the graphs is scaled for 1m distance (i.e. reduced by 6dB due to distance doubling).
I'm only showing 2nd, 3rd and 5th harmonic to make the graphs less busy (as these dominate the THD anyway) and total THD (2-9th harmonic).
Neumann KH 120A - 74dB SPL at 1kHz at 1m.png

Neumann KH 120A - 84dB SPL at 1kHz at 1m.png

Neumann KH 120A - 84dB SPL at 1kHz at 1m (SPL).png

Unfortunately I didn't really measure any high-SPL, but up until ~84 dB SPL that I did measure the distortion seems pretty low!

Let's see how THD of both my units matches:
Neumann KH 120A - 84dB SPL at 1kHz at 1m - two units compared.png

Seems pretty close - again showing very good unit-to-unit matching - better than my previous (though much cheaper) JBL LSR305s.
Today I extended a bit on the distortion measurements, measuring at 1m (in-room) at 70, 80 and 90dB SPL (above that I'm no longer comfortable, fearing neighbour retaliation :p).

Loudspeaker and microphone position was sub-optimal this time so I had a pretty big null ~260 Hz (as well as some comb filtering), significantly impacting the normalized plots - so I will only show the normalized plot for the 90 dB SPL measurement. Still, at least the trends are visible. :)

For better visibility I'm only showing the noise floor, 2nd and 3rd harmonic, and total HD (2nd-9th), as contribution of the other individual harmonics is much smaller.

Here are the results:

70 dB SPL:
Neumann KH 120A - In-room sweep distortion measurement - 70 dB SPL at 1m.png


80 dB SPL:
Neumann KH 120A - In-room sweep distortion measurement - 80 dB SPL at 1m.png


90 dB SPL:
Neumann KH 120A - In-room sweep distortion measurement - 90 dB SPL at 1m.png
Neumann KH 120A - In-room sweep distortion measurement (normalized) - 90 dB SPL at 1m.png
 
OP
dominikz

dominikz

Addicted to Fun and Learning
Forum Donor
Joined
Oct 10, 2020
Messages
803
Likes
2,626
One thing I forgot to mention in my original post - KH 120A electronics seem to invert the absolute signal polarity, here's a comparison to another active (JBL LSR305) and a passive (Revel M16) loudspeaker:
Neumann KH 120A vs JBL LSR305 vs Revel M16 - Step response and absolute phase comparison.png
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
One thing I forgot to mention in my original post - KH 120A electronics seem to invert the absolute signal polarity, here's a comparison to another active (JBL LSR305) and a passive (Revel M16) loudspeaker:
View attachment 157383

Last time I checked it's not inverted if you look at the impulse graphs (not step response) -- only the mid-driver looks inverted.

*Closing the ports seems to flip it back up:
1633508967068.png

Red trace is with the KH120 phase linearized.
 
Last edited:

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
Last time I checked it's not inverted if you look at the impulse graphs (not step response) -- only the mid-driver looks inverted.

*Closing the ports seems to flip it back up:
View attachment 157472
Red trace is with the KH120 phase linearized.

Just to let you know: I happened to be using a different DAC today and the output looks inverted now. Hmmmn.... wonder why... but it seems to me the DAC is inverting the final output so depends on which device I use. Doesn't really matter in the end as one could always invert it back up again if so desired -- just have to check the final acoustic output.
 
OP
dominikz

dominikz

Addicted to Fun and Learning
Forum Donor
Joined
Oct 10, 2020
Messages
803
Likes
2,626
Just to let you know: I happened to be using a different DAC today and the output looks inverted now. Hmmmn.... wonder why... but it seems to me the DAC is inverting the final output so depends on which device I use. Doesn't really matter in the end as one could always invert it back up again if so desired -- just have to check the final acoustic output.
Thanks! Yeah, I agree it doesn't probably matter that much; and likely there is good reason why the woofer phase is inverted in the first place - I assume probably to optimize the crossover behavior.
Just thought it was interesting!
 

nc535

Member
Joined
Dec 30, 2021
Messages
52
Likes
62
I'm exploring the idea that 3D printing enclosures that add a pair of small, delayed inverted side speakers to left and right sides of KH--120A, I could have a Kii-like cardioid response for a fraction of their price. Simulating this in Vituix using generic 5.25" midwoofer directivity is encouraging. I am about to trace your horizontal directivity chart but I don't see one for the verticals. Do you have one you could post or better yet post your own Vituix directivity files?
Thanks!
 
OP
dominikz

dominikz

Addicted to Fun and Learning
Forum Donor
Joined
Oct 10, 2020
Messages
803
Likes
2,626
I'm exploring the idea that 3D printing enclosures that add a pair of small, delayed inverted side speakers to left and right sides of KH--120A, I could have a Kii-like cardioid response for a fraction of their price. Simulating this in Vituix using generic 5.25" midwoofer directivity is encouraging. I am about to trace your horizontal directivity chart but I don't see one for the verticals. Do you have one you could post or better yet post your own Vituix directivity files?
Thanks!
Sure, full spin data is attached.
Your project sounds very interesting, would be great to see more - hope you will post about it! :)
 

Attachments

  • Neumann KH120A 1m by dominikz - full spin data - 2021-12-31.zip
    337.7 KB · Views: 105
Top Bottom