Neumann KH 80 DSP Monitor Review

[MZKM]

It is very flat (no bass drop until 65 Hz where it falls sharply) and very close to the manufacturers measurements, are you confusing the blue lines which are some optional room filters?

JBL 305P MKII Preference Rating
SCORE: 4.4
SCORE w/ subwoofer: 6.4

They are very similar except for Smoothness of the Predicted In-Room; however, when looking at the smoothness graphs, they are both very smooth; the only way I can think of as to why it has a score thats ~1/2 as good for Smoothness is the tonal balance, the target slope wants a slope-down, whereas the JBL is almost flat. Now, Amir listened to these on a desk I believe, so that is likely the reason he preferred the wider soundstage; the preference rating is for a listening room. Or, my formula is borked somehow that I can't see (but my formulation matches the one that's generated via the graph). Also, it has ~1/2 the data points, so that effects it a bit.
https://docs.google.com/spreadsheets/d/1MDuE6hFfo9pofi2063s6IXqYcxtXMph8gqLLkL9X8pA/edit?usp=sharing


KH80 Radar:

305P MKII Radar

KH80 Smoothness:

305P MKII Smoothness

 

[sweetchaos]

Can't tell if this was a joke

I'm pointing out how the models are named. I don't see a joke here.

 

[Xyrium]

I'm pointing out how the models are named. I don't see a joke here.

Perfect, that makes two of us then.

 

[MZKM]

I can record the preference scores for each speaker as part of the planned review database.

Nice work @MZKM.

The formula is still getting kinks worked out, so score may be altered over time. I'd hold off.

 

[pozz]

The formula is still getting kinks worked out, so score may be altered over time. I'd hold off.

That's no issue. I'll keep track of your posts to make sure.

 

[BYRTT]

Seems "Informal Listening Tests/Conclusions" from review makes some noise here and there in objective response scales was not same and in process of being dialed in, then for JBL verse Neumann graphs combined the two on same CEA2034 scales in REW including Estimated In-Room Resp.

 

[maxxevv]

JBL 305P MKII Preference Rating
SCORE: 4.36
SCORE ignoring LFX: 6.43

They are very similar except for Smoothness of the Predicted In-Room; however, when looking at the smoothness graphs, they are both very smooth; the only way I can think of as to why it has a score thats ~1/2 as good for Smoothness is the tonal balance, the target slope wants a slope-down, whereas the JBL is almost flat. Now, Amir listened to these on a desk I believe, so that is likely the reason he preferred the wider soundstage; the preference rating is for a listening room. Or, my formula is borked somehow that I can't see (but my formulation matches the one that's generated via the graph).
https://docs.google.com/spreadsheets/d/1MDuE6hFfo9pofi2063s6IXqYcxtXMph8gqLLkL9X8pA/edit?usp=sharing


KH80 Radar:
View attachment 46860

305P MKII Radar
View attachment 46861

KH80 Smoothness:
View attachment 46862

305P MKII Smoothness
View attachment 46863

As for the preference rating, the way I look at it, considering Amir's listening impressions, it doesn't quite reflect how the speakers sound at varying distances.

Clearly, the 305P will do better at 2 meters than the KH80. And the predicted results of the KH80 based on Toole's criteria favour the KH80 too.

Is there any way to extract the data to reflect preferences at varying distances ? 1 meter, 2 meters, 5 meters ?

Pretty sure floor standers will not do well at closed in listening and yet better at beyond 3 meters compared to say a 4 inch driver bookshelf, regardless of how smooth and how well its directivity is implemented.

Also, I think 90 dB and 1 meter is pretty loud. I believe the monitoring "standard" that they adhere to in studios is something like 82 dB ( will need to look that up, as this is just purely from memory). It will be good if its close to what is practical / replicable.

 

[LeftCoastTim]

JBL 305P MKII Preference Rating
SCORE: 4.36
SCORE ignoring LFX: 6.43

They are very similar except for Smoothness of the Predicted In-Room; however, when looking at the smoothness graphs, they are both very smooth; the only way I can think of as to why it has a score thats ~1/2 as good for Smoothness is the tonal balance, the target slope wants a slope-down, whereas the JBL is almost flat. Now, Amir listened to these on a desk I believe, so that is likely the reason he preferred the wider soundstage; the preference rating is for a listening room. Or, my formula is borked somehow that I can't see (but my formulation matches the one that's generated via the graph).
https://docs.google.com/spreadsheets/d/1MDuE6hFfo9pofi2063s6IXqYcxtXMph8gqLLkL9X8pA/edit?usp=sharing

Do you plan to make a spreadsheet of all the reviewed speakers and their summary scores? Also, a glossary of the acronyms you use would be awesome! Thanks for your wonderful work!

 

[MZKM]

As for the preference rating, the way I look at it, considering Amir's listening impressions, it doesn't quite reflect how the speakers sound at varying distances.

Clearly, the 305P will do better at 2 meters than the KH80. And the predicted results of the KH80 based on Toole's criteria favour the KH80 too.

Is there any way to extract the data to reflect preferences at varying distances ? 1 meter, 2 meters, 5 meters ?

Pretty sure floor standers will not do well at closed in listening and yet better at beyond 3 meters compared to say a 4 inch driver bookshelf, regardless of how smooth and how well its directivity is implemented.

Also, I think 90 dB and 1 meter is pretty loud. I believe the monitoring "standard" that they adhere to in studios is something like 82 dB ( will need to look that up, as this is just purely from memory). It will be good if its close to what is practical / replicable.

It will be an alteration to the formula (ignoring LFX just disregards bass extension).
However, before I typed this, I was going to see if I could alter it for curiosity's sake by trying to ignore the target slope.

 

[maxxevv]

Do you plan to make a spreadsheet of all the reviewed speakers and their summary scores? Also, a glossary of the acronyms you use would be awesome! Thanks for your wonderful work!

Believe @pozz will be working on compiling that for ASR.

 

[bobbooo]

JBL 305P MKII Preference Rating
SCORE: 4.36
SCORE ignoring LFX: 6.43

They are very similar except for Smoothness of the Predicted In-Room; however, when looking at the smoothness graphs, they are both very smooth; the only way I can think of as to why it has a score thats ~1/2 as good for Smoothness is the tonal balance, the target slope wants a slope-down, whereas the JBL is almost flat. Now, Amir listened to these on a desk I believe, so that is likely the reason he preferred the wider soundstage; the preference rating is for a listening room. Or, my formula is borked somehow that I can't see (but my formulation matches the one that's generated via the graph).
https://docs.google.com/spreadsheets/d/1MDuE6hFfo9pofi2063s6IXqYcxtXMph8gqLLkL9X8pA/edit?usp=sharing


KH80 Radar:
View attachment 46860

305P MKII Radar
View attachment 46861

KH80 Smoothness:
View attachment 46862

305P MKII Smoothness
View attachment 46863

We also need to consider that the JBL data has only 10 points per octave, compared to the Neumann which has the required 20 points per octave to accurately use Olive's preference formula. As you've shown for the NHT Pro M-00, this can affect the final score. (The JBL measurements also only have 7 data points in the first octave for some reason, but since they're below the LFX point, this shouldn't matter. Just a bit strange as all measurements for the other speakers have equal numbers of data points per octave.)

 

[maxxevv]

Think @amirm was using the 305P as a test run.
He could simply do a re-run with it using his better fine-tuned test protocol now.

I honestly didn't even know about that difference!

 

[MZKM]

We also need to consider that the JBL data has only 10 points per octave, compared to the Neumann which has the required 20 points per octave to accurately use Olive's preference formula. As you've shown for the NHT Pro M-00, this can affect the final score. (The JBL measurements also only have 7 data points in the first octave for some reason, but since they're below the LFX point, this shouldn't matter. Just a bit strange as all measurements for the other speakers have equal numbers of data points per octave.)

Yeah, but with this, the smoothness is drastically lower because it doesn't follow the target slope, it's almost flat for the in-room.
I'm trying to see if I can get smoothness to only look at smoothness regardless of slope, but I'm not having luck. Any alterations to target slope have no effect on the final score.

 

[bobbooo]

I also upped the number of points to 20/octave. This unfortunately made the analysis phase take over half hour. So I like to go back to 10 points or something like it.

Thanks for the review! I know it's a pain, but I really hope you continue with 20 points per octave, as then we'll know for sure that we have the highest precision data we can get to use with Sean Olive's preference rating formula, which may produce unpredictably inaccurate ratings at lower precision - we have no way of knowing it won't for all speakers, even if it doesn't change the score massively for a few we calculate.

It would also be a shame not to use a state of the art measuring system like the the Klippel NFS to its full potential. I think most people would much prefer to slightly sacrifice your very high rate of reviews and measurements being posted, for the highest precision measurements and preference ratings produced for each speaker. Using this higher precision also future-proofs the data against any improvements Olive might make to his preference ratings that would change the scores even more drastically based on precision, or other proposed preference ratings / data analysis in the future. If you stick to 20 points per octave, you have a chance here to create a database of some of the highest precision measurements of speakers outside places like Harman and Klippel themselves, which could hugely benefit consumers and even academic research for years to come

You said you tried to recompute the Yamaha HS5 raw data with 1/20-octave smoothing instead of 1/10, did that work? If so, could you do the same for the JBL 305P? And any idea why there are only 7 data points for the first octave in the JBL measurements? (Luckily the LFX point is in the second octave so isn't affected.)

 

[amirm]

You said you tried to recompute the Yamaha HS5 raw data with 1/20-octave smoothing instead of 1/10, did that work?

I think it did. I was not watching it since it took so long.

 

[spacevector]

Levels were matched using pink noise.

Amir, if not too much trouble, please include SPL level used for matching speakers and during listening (do you mess with volume while listening?).

It would also be good to see a snapshot of the SUT's (speaker under test) frequency response at the LP compared to the reference LSR305(say). This would help 'see' what the room is doing to the bass.

 

[Vintage57]

I’m pleased to learn that Neumann is on the list of manufacturers who provide repeatable data. That’s big in this Hi-Fi world.

As for preferring the JBL sound I think there is an east coast west coast sound preference. There was over 40 years ago when I first auditioned JBL L100 (west coast) I went with the AR (east coast). Many many speakers later and more recently I auditioned in my home the JBL 4367, known as the passive version of the M2 that measures flat. I also auditioned the Neumann KH 420 likewise a flat measuring speaker.
I kept the Neumann and am very happy with my system. The JBL gave me listening fatigue, I couldn’t tell you why only that it was there for me.

Kinda reminds me of wine tasting, I can’t tell you much about grapes of origin or vintage but I can tell you if I like it or not.

Let’s not understate the personal taste quotient.

Good work Amir

I would send a KH 420 however international borders would make it a chore. Maybe someone closer can volunteer

 

[amirm]

Amir, if not too much trouble, please include SPL level used for matching speakers and during listening (do you mess with volume while listening?).

The tests are informal so I don't want to spec them too much to make them otherwise. That said, it was about 75 dB. I do turn the master volume up and down during the test.

 

[napilopez]

I made some t/s measurements on an sb acoustic driver (they manufacture drivers for revel and are a serious manufacturer). Measurements before break in, after break in and then again after they where in storage for about a year, they. They all where all slightly different, not a big deal but still different. They add a disclaimer on every data-sheet that t/s parameters are measured on drivers that are broken in.
In tymphany data sheets they add a 15% tolarance on their fs specification. That is not because they are bad drivers its because their spec sheets are honest.

Would they still have that variation when actually played as part of a system though? (just asking - i've not built a speaker before). I'd still find such a change a bit concerning, but I guess we can't know without knowing the age of speakers.

I'm being persnickety in part because I specifically chose the KH80 as my 'reference speakers for measurements. According to the previously existing anechoic measurements, they're were the flattest speakers on the market I could find short of maybe the JBL M2's listening window. I figured if I could match my quasi-anechoic measurements to the existing fully anechoic data of the KH80s, I'd be confident in my results going forward. The KH80 seemed so flat, in fact, that my original Umik-1 was not flat enough to capture its performance, and I had to get a specially calibrated one from CSL to get closer to the truth. Indeed, the speaker measured flatter with a better microphone.

Now I'm less confident in those results.

To illustrate the differences more clearly, here's the KH80's on-axis performance as measured by the Klippel, Neumann, Sound&Recording, and me (quasi anechoically), scaled to match aspet ratio. It kind of concerns me that the Klippel is the biggest outlier in terms of broad FR shape. We know the reference axis was a bit off, but I'd consider these differences of significant amplitude given the precision we're aiming for here.

 

[maxxevv]

Would they still have that variation when actually played as part of a system though? (just asking - i've not built a speaker before). I'd still find such a change a bit concerning, but I guess we can't know without knowing the age of speakers.

I'm being persnickety in part because I specifically chose the KH80 as my 'reference speakers for measurements. According to the previously existing anechoic measurements, they're were the flattest speakers on the market I could find short of maybe the JBL M2's listening window. I figured if I could match my quasi-anechoic measurements to the existing fully anechoic data of the KH80s, I'd be confident in my results going forward. The KH80 seemed so flat, in fact, that my original Umik-1 was not flat enough to capture its performance, and I had to get a specially calibrated one from CSL to get closer to the truth. Indeed, the speaker measured flatter with a better microphone.

Now I'm less confident in those results.

To illustrate the differences more clearly, here's the KH80's on-axis performance as measured by the Klippel, Neumann, Sound&Recording, and me (quasi anechoically), scaled to match aspet ratio. It kind of concerns me that the Klippel is the biggest outlier in terms of broad FR shape. We know the reference axis was a bit off, but I'd consider these differences of significant amplitude given the precision we're aiming for here.

View attachment 46873

Jokes aside, yours certainly seem like an average of the Klippel , Neumann and S&R curves !!