• 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 310A Review (Powered Monitor)

Francis Vaughan

Addicted to Fun and Learning
Forum Donor
Joined
Dec 6, 2018
Messages
933
Likes
4,697
Location
Adelaide Australia
Genelec also uses cheap Peerless stuff in their monitors. 100€ is actually a good spending, consider that this $30 woofer is more or less what's used in the 8030C and KH120A:
https://www.parts-express.com/peerless-by-tymphany-830860-5-1-4-ppb-cone-hds-woofer--264-1080
https://www.soundimports.eu/en/peerless-by-tymphany-hds-p830860.html

As always, implementation is key and can often compensate for the raw material's flaws.

Absolutely. Those of us that meddle in DIY speakers tend to lose sight of this a bit. There is a temptation to go for the best possible (aka most expensive) drivers simply because they must be better. Right? There might be some value in this in that you are probably more likely to get a reasonable result, because high end drivers are often more easy to deal with. But if you have the experience, design and test tools and the ability to customise the precise driver design, the apparently cheaper components are not a disadvantage.
It is pretty clear that the drivers in the KH 310A, whilst very likely of Tymphany origin, are not the same drivers that are commercially available to everyone else. Minimally the magnet on the bass driver looks to be larger, and the surround is different. There could easily be changes to the motor design, cone and spider. This makes it a completely different driver. Same for the mid-range. The overall speaker is a synergistic design, and an active speaker gives the designer a lot of freedom to move around the optimisations.
Going out and buying the visually similar part doner drivers and trying to build a clone is going to be a futile exercise.
 

Pepperjack

Member
Joined
May 9, 2020
Messages
90
Likes
64
Not exactly on topic but it came up in a few comments about hiss. I had a couple of rca to xlr cables from guitar center and the hissing was terrible. Switched to some of the best monoprice offering and two cables had no audible hiss and one had hiss. Replaced it and no audible hiss. Fast forward to getting some new bud+ earbuds for audiobooks etc and when using the setting where the microphones are used to pass ambient sound to the listener I discovered my air purifier was super loud. I turned it off and standing right next to the kh310 with the buds on max ambient there wasn’t any hiss. So, if you hear hissing, get a new cable.
 

Maiky76

Senior Member
Joined
May 28, 2020
Messages
440
Likes
3,704
Location
French, living in China
Hi,

I was reluctant at first glance to have a look at the EQ as the data was near perfect. However after looking more closely it appeared to me the the integration of the mid range was, maybe, not that “perfect”.
I understand the compromise to get the third way: increase LF bandwidth and max SPL capabilities but it does not change the fact that the DI clearly shows the directivity errors.
It might be inconsequential in near field in a mixing studio where the direct sound might be paramount but in a domestic environment, where the domain of validity of Olive score is supposed to reside, it will decrease the Predicted Preference.

Here are some thoughts I shared:
https://www.audiosciencereview.com/...7v-spinorama-and-eq-inside.17283/#post-560611

[...]Regarding the scoring method itself, models in general, not just in audio, are judged with two main criteria:
- Can it describe what we observe reasonably well?
- Do the predictions derived from the model hold against new observations?


The first criterium is covered in the original paper:
https://www.researchgate.net/public...e_Ratings_of_Around-Ear_and_On-Ear_Headphones


I agree with you we need more data for the second criterium to be more understood.
We have some data point from the Harman team comparing different speakers with different PIR shapes and EQing targets but this not as detailed as the first paper.[...]


One may not agree (whatever that means) with the Score but this is the best we’ve got.
Nothing is preventing anyone from coming up with something different but the validation will be hard to say the least…

In the case of the KH310, the error from flatness are higher than one may think at first glance.
See the spinorama with just bass EQ (to look like the Neumann data):
Neumann KH310 Just Bass EQ Spinorama.png


Quick comparison with the KH80
KH80 vs KH310 Just Bass EQ.png

NBDON = 0.34dB/20th octave on average
NBDPIR = 0.30dB/20th octave on average
SMPIR = 0.78 (closer to 1 is better, interpreted as smoothness of the PIR in the Olive paper)
Score: 6.35 / 7.76 with Sub

To be compared with the KH80 :
NBDON = 0.25dB/20th octave on average
NBDPIR = 0.22dB/20th octave on average
SMPIR = 0.86
Score: 6.04 / 8.4 with Sub
the higher score of the KH is due to the better LF response
The Score does not take into account the SPL headroom/THD.

KH80 vs KH310 Just Bass EQ radar.png


The small one is just smoother and the sub score shows, in my opinion, that pound for pound and within its SPL capabilities, the smaller one is a better speaker.
Adding more ways on the speaker improve some parameters but also bring some other compromises, more is not always 100% better…

For the pound for pound see there:
https://www.audiosciencereview.com/forum/index.php?threads/jbl-308p-mkii-studio-monitor-review.17338/page-13

[...]Here is the calculation I perform, which seems to match what others are doing:
PPR_LF = 12.69 - 2.49*NBD_ON - 2.99*NBD_PIR - 4.31*log10(14.5) + 2.32*SM_PIR
see there for some details:
https://www.audiosciencereview.com/...urements-community-project.14929/#post-467858


This assumes:
- Frequency response 14.5Hz @-6dB. 14.5Hz is chosen so the theoretical max score is 10 although nothing prevent a system from achieving better performances and therefore the score could exceed 10; this is the first issue.
- Perfect integration whatever that might mean, which is the second issue and where I don't quite adhere to the concept.


The idea, I guess, is to compare speakers if the LF extension is literally taken out of the equation, akin the pound-for-pound rating.
To me it would make more sense to use it this way rather than thinking "if I buy a sub then I'll get the astonishing system predicted by the sub score". That is just not going to happen: sub or not the room will still determine the system LF response.[...]


Here is my take on the EQ nonetheless:
I think unless the user has a decent understanding on measurements and their interpretation it is probably better not to play with the EQ or just FYI. The final user should measure his/her own units, at least the LW.
Room integration EQ is highly recommended thought.

The raw data with corrected ER and PIR:
Score no EQ: 6.19
With Sub: 7.6

Spinorama with no EQ:
  • LF response, is it temperature related or otherwise? I EQed it so that it to resemble the data published by Neumann to get it out of the way.
  • HF it seems that there are tonal control that do pretty much the same job that the EQ I added…
  • The midrange integration maybe related to production tolerance?
Neumann KH310 No EQ Spinorama.png


Directivity:
Better stay at tweeter within +/-10 deg of the mid point Tweeter/mid both vertically and horizontally.
Neumann KH310 2D surface Directivity Contour Only Data.png

Neumann KH310 LW better data.png

EQ design:
I have generated two EQs. The APO config files are attached.
  • The first one, labelled, LW is targeted at making the LW flat
  • The second, labelled Score, starts with the first one and adds the score as an optimization variable.
  • The EQs are designed in the context of regular stereo use i.e. domestic environment, no warranty is provided for a near field use in a studio environment although the LW might be better suited for this purpose
Score EQ LW: 6.82
with sub: 8.22

Score EQ Score: 6.98
with sub: 8.39

Code:
Code:
Neumann KH310 APO EQ LW 96000Hz
November232020-121217

Preamp: -2.3 dB

Filter 1: ON HPQ Fc 29.1 Hz Gain 0 dB Q 0.9
Filter 2: ON PK Fc 89 Hz Gain 1.63 dB Q 2.21
Filter 3: ON PK Fc 707 Hz Gain 1.18 dB Q 4.87
Filter 4: ON PK Fc 1548 Hz Gain 2.32 dB Q 6.71
Filter 5: ON PK Fc 4779 Hz Gain -0.88 dB Q 8.74

Neumann KH310 APO EQ Score 96000Hz
November232020-110222

Preamp: -2.3 dB

Filter 1: ON HPQ Fc 29.4 Hz Gain 0 dB Q 0.9
Filter 2: ON PK Fc 88.4 Hz Gain 1.63 dB Q 2.21
Filter 3: ON PK Fc 700.8 Hz Gain 1.06 dB Q 4.56
Filter 4: ON PK Fc 1525 Hz Gain 2.25 dB Q 10
Filter 5: ON PK Fc 4714 Hz Gain -1.31 dB Q 5.85
Filter 6: ON PK Fc 12878 Hz Gain -0.69 dB Q 1.04
Neumann KH310 EQ Design.png

Spinorama EQ LW
Neumann KH310 EQ LW Spinorama.png

Spinorama EQ Score
Neumann KH310 EQ Score Spinorama.png

Zoom PIR-LW-ON
Zoom PIR-LW-ON.png

Regression - Tonal the EQ score makes the On flat
Neumann KH310 Regression-Tonal.png

Radar no EQ vs EQ score
Some improvements
Neumann KH310 Radar.png

The rest of the plots are attached in particular Horizontal directivity with positive and negative angles separated.
Neumann KH310 Horizontal 3D Directivity data.png
 

Attachments

  • Neumann KH310 APO EQ LW 96000Hz.txt
    302 bytes · Views: 101
  • Neumann KH310 Vertical 3D Directivity data.png
    Neumann KH310 Vertical 3D Directivity data.png
    572.4 KB · Views: 104
  • Neumann KH310 Normalized Directivity data.png
    Neumann KH310 Normalized Directivity data.png
    455.1 KB · Views: 108
  • Neumann KH310 Raw Directivity data.png
    Neumann KH310 Raw Directivity data.png
    793 KB · Views: 113
  • Neumann KH310 Reflexion data.png
    Neumann KH310 Reflexion data.png
    213.4 KB · Views: 127
  • Neumann KH310 LW data.png
    Neumann KH310 LW data.png
    286.1 KB · Views: 127
  • Neumann KH310 APO EQ Score 96000Hz.txt
    356 bytes · Views: 107
  • Neumann KH310 2D surface Directivity Contour Data.png
    Neumann KH310 2D surface Directivity Contour Data.png
    308.9 KB · Views: 117
  • Neumann KH310 3D surface Vertical Directivity Data.png
    Neumann KH310 3D surface Vertical Directivity Data.png
    488.2 KB · Views: 116
  • Neumann KH310 3D surface Horizontal Directivity Data.png
    Neumann KH310 3D surface Horizontal Directivity Data.png
    482.9 KB · Views: 120
Last edited:

Pepperjack

Member
Joined
May 9, 2020
Messages
90
Likes
64
Hi,

I was reluctant at first glance to have a look at the EQ as the data was near perfect. However after looking more closely it appeared to me the the integration of the mid range was, maybe, not that “perfect”.
I understand the compromise to get the third way: increase LF bandwidth and max SPL capabilities but it does not change the fact that the DI clearly shows the directivity errors.
It might be inconsequential in near field in a mixing studio where the direct sound might be paramount but in a domestic environment, where the domain of validity of Olive score is supposed to reside, it will decrease the Predicted Preference.

Here are some thoughts I shared:
https://www.audiosciencereview.com/...7v-spinorama-and-eq-inside.17283/#post-560611

[...]Regarding the scoring method itself, models in general, not just in audio, are judged with two main criteria:
- Can it describe what we observe reasonably well?
- Do the predictions derived from the model hold against new observations?


The first criterium is covered in the original paper:
https://www.researchgate.net/public...e_Ratings_of_Around-Ear_and_On-Ear_Headphones


I agree with you we need more data for the second criterium to be more understood.
We have some data point from the Harman team comparing different speakers with different PIR shapes and EQing targets but this not as detailed as the first paper.[...]


One may not agree (whatever that means) with the Score but this is the best we’ve got.
Nothing is preventing anyone from coming up with something different but the validation will be hard to say the least…

In the case of the KH310, the error from flatness are higher than one may think at first glance.
See the spinorama with just bass EQ (to look like the Neumann data):
View attachment 95191

Quick comparison with the KH80
View attachment 95215
NBDON = 0.34dB/20th octave on average
NBDPIR = 0.30dB/20th octave on average
SMPIR = 0.78 (closer to 1 is better, interpreted as smoothness of the PIR in the Olive paper)
Score: 6.35 / 7.76 with Sub

To be compared with the KH80 :
NBDON = 0.25dB/20th octave on average
NBDPIR = 0.22dB/20th octave on average
SMPIR = 0.86
Score: 6.04 / 8.4 with Sub
the higher score of the KH is due to the better LF response
The Score does not take into account the SPL headroom/THD.

View attachment 95214

The small one is just smoother and the sub score shows, in my opinion, that pound for pound and within its SPL capabilities, the smaller one is a better speaker.
Adding more ways on the speaker improve some parameters but also bring some other compromises, more is not always 100% better…

For the pound for pound see there:
https://www.audiosciencereview.com/forum/index.php?threads/jbl-308p-mkii-studio-monitor-review.17338/page-13

[...]Here is the calculation I perform, which seems to match what others are doing:
PPR_LF = 12.69 - 2.49*NBD_ON - 2.99*NBD_PIR - 4.31*log10(14.5) + 2.32*SM_PIR
see there for some details:
https://www.audiosciencereview.com/...urements-community-project.14929/#post-467858


This assumes:
- Frequency response 14.5Hz @-6dB. 14.5Hz is chosen so the theoretical max score is 10 although nothing prevent a system from achieving better performances and therefore the score could exceed 10; this is the first issue.
- Perfect integration whatever that might mean, which is the second issue and where I don't quite adhere to the concept.


The idea, I guess, is to compare speakers if the LF extension is literally taken out of the equation, akin the pound-for-pound rating.
To me it would make more sense to use it this way rather than thinking "if I buy a sub then I'll get the astonishing system predicted by the sub score". That is just not going to happen: sub or not the room will still determine the system LF response.[...]


Here is my take on the EQ nonetheless:
I think unless the user has a decent understanding on measurements and their interpretation it is probably better not to play with the EQ or just FYI. The final user should measure his/her own units, at least the LW.
Room integration EQ is highly recommended thought.

The raw data with corrected ER and PIR:
Score no EQ: 6.19
With Sub: 7.6

Spinorama with no EQ:
  • LF response, is it temperature related or otherwise? I EQed it so that it to resemble the data published by Neumann to get it out of the way.
  • HF it seems that there are tonal control that do pretty much the same job that the EQ I added…
  • The midrange integration maybe related to production tolerance?
View attachment 95204

Directivity:
Better stay at tweeter within +/-10 deg of the mid point Tweeter/mid both vertically and horizontally.
View attachment 95208
View attachment 95193
EQ design:
I have generated two EQs. The APO config files are attached.
  • The first one, labelled, LW is targeted at making the LW flat
  • The second, labelled Score, starts with the first one and adds the score as an optimization variable.
  • The EQs are designed in the context of regular stereo use i.e. domestic environment, no warranty is provided for a near field use in a studio environment although the LW might be better suited for this purpose
Score EQ LW: 6.82
with sub: 8.22

Score EQ Score: 6.98
with sub: 8.39

Code:
[

I think you may be underestimating the impact of temperature. While you clearly know way way more than I do overall I notice that you pulled the final test of the kh80 where the temperature was corrected. Based on what we know it seems possible the speaker was at 14c or below (if left in the garage overnight perhaps the speaker itself had not yet actually heated up to 14?) regardless the room was heated to 68 but it seems that is when the test began? The data we see here looks exactly like the temperature impact we saw in the fridge test which showed that it took quite a while for the internals of the speaker to warm up sufficiently-based on the test that were taken every 3 minutes. Given the larger size of everything here it seems reasonable it would take even longer. To my untrained eye it looks like this speaker was warmer than the original kh80 test, but not up to fully intended room temperature-this both the dip and the score are falling inbeteeen the two.

I believe it is reasonable to see the dip you are focusing on as potentially a result of the garage temp influencing the unit again, no? Here are some screen shots from the earlier post:
 

Attachments

  • 97625D79-89A7-4499-BD80-47EB3FE28412.png
    97625D79-89A7-4499-BD80-47EB3FE28412.png
    644.2 KB · Views: 111
  • 630C9CA4-2BDB-4CDC-9AC1-A28B8674D43F.png
    630C9CA4-2BDB-4CDC-9AC1-A28B8674D43F.png
    539.7 KB · Views: 99
  • 4A686D94-F0A4-413C-B623-9EAA09BBAED5.jpeg
    4A686D94-F0A4-413C-B623-9EAA09BBAED5.jpeg
    247 KB · Views: 109
  • DFE271A1-5C2F-4397-9946-D543BF80A062.jpeg
    DFE271A1-5C2F-4397-9946-D543BF80A062.jpeg
    158.9 KB · Views: 103

YSC

Major Contributor
Joined
Dec 31, 2019
Messages
3,194
Likes
2,570
I think you may be underestimating the impact of temperature. While you clearly know way way more than I do overall I notice that you pulled the final test of the kh80 where the temperature was corrected. Based on what we know it seems possible the speaker was at 14c or below (if left in the garage overnight perhaps the speaker itself had not yet actually heated up to 14?) regardless the room was heated to 68 but it seems that is when the test began? The data we see here looks exactly like the temperature impact we saw in the fridge test which showed that it took quite a while for the internals of the speaker to warm up sufficiently-based on the test that were taken every 3 minutes. Given the larger size of everything here it seems reasonable it would take even longer. To my untrained eye it looks like this speaker was warmer than the original kh80 test, but not up to fully intended room temperature-this both the dip and the score are falling inbeteeen the two.

I believe it is reasonable to see the dip you are focusing on as potentially a result of the garage temp influencing the unit again, no? Here are some screen shots from the earlier post:
It seems interesting, btw anyone have idea of did Genelecs pose similar temperature dependency? As I recall they use peerless drivers also so for material response I don’t think it will behave differently. Though it’s of no concern to me as HK is always a warm to hot place where the drivers should always be in their optimal temperature
 

Wombat

Master Contributor
Joined
Nov 5, 2017
Messages
6,722
Likes
6,459
Location
Australia
Not exactly on topic but it came up in a few comments about hiss. I had a couple of rca to xlr cables from guitar center and the hissing was terrible. Switched to some of the best monoprice offering and two cables had no audible hiss and one had hiss. Replaced it and no audible hiss. Fast forward to getting some new bud+ earbuds for audiobooks etc and when using the setting where the microphones are used to pass ambient sound to the listener I discovered my air purifier was super loud. I turned it off and standing right next to the kh310 with the buds on max ambient there wasn’t any hiss. So, if you hear hissing, get a new cable.

Maybe that is why my JBL LSR 305P MKIIs do not hiss. Guys on this forum won't believe me.
headbang.gif
 

YSC

Major Contributor
Joined
Dec 31, 2019
Messages
3,194
Likes
2,570
Maybe that is why my JBL LSR 305P MKIIs do not hiss. Guys on this forum won't believe me. View attachment 95228
I kind of agree on cables did do something on hissing, maybe the 305P originally have some higher noise floor for hissing, but that noise floor can be from cable interference
 

Robbo99999

Master Contributor
Forum Donor
Joined
Jan 23, 2020
Messages
6,874
Likes
6,672
Location
UK
Nope, it's not that simple. As I understand current research the FR of the first wave front (on axis) defines SQ and smooth FR of the reflected sound (off axis) is required for excellent SQ. An on axis FR with a dip which is filled by a peak in the off axis FR may give smooth in room FR but does not sound as good as a flat on axis FR and smooth off axis FR giving similar smooth in room FR.
(Well I'm talking in relation to distortion measurments)
 

Robbo99999

Master Contributor
Forum Donor
Joined
Jan 23, 2020
Messages
6,874
Likes
6,672
Location
UK
Hi,

I was reluctant at first glance to have a look at the EQ as the data was near perfect. However after looking more closely it appeared to me the the integration of the mid range was, maybe, not that “perfect”.
I understand the compromise to get the third way: increase LF bandwidth and max SPL capabilities but it does not change the fact that the DI clearly shows the directivity errors.
It might be inconsequential in near field in a mixing studio where the direct sound might be paramount but in a domestic environment, where the domain of validity of Olive score is supposed to reside, it will decrease the Predicted Preference.

Here are some thoughts I shared:
https://www.audiosciencereview.com/...7v-spinorama-and-eq-inside.17283/#post-560611

[...]Regarding the scoring method itself, models in general, not just in audio, are judged with two main criteria:
- Can it describe what we observe reasonably well?
- Do the predictions derived from the model hold against new observations?


The first criterium is covered in the original paper:
https://www.researchgate.net/public...e_Ratings_of_Around-Ear_and_On-Ear_Headphones


I agree with you we need more data for the second criterium to be more understood.
We have some data point from the Harman team comparing different speakers with different PIR shapes and EQing targets but this not as detailed as the first paper.[...]


One may not agree (whatever that means) with the Score but this is the best we’ve got.
Nothing is preventing anyone from coming up with something different but the validation will be hard to say the least…

In the case of the KH310, the error from flatness are higher than one may think at first glance.
See the spinorama with just bass EQ (to look like the Neumann data):
View attachment 95191

Quick comparison with the KH80
View attachment 95215
NBDON = 0.34dB/20th octave on average
NBDPIR = 0.30dB/20th octave on average
SMPIR = 0.78 (closer to 1 is better, interpreted as smoothness of the PIR in the Olive paper)
Score: 6.35 / 7.76 with Sub

To be compared with the KH80 :
NBDON = 0.25dB/20th octave on average
NBDPIR = 0.22dB/20th octave on average
SMPIR = 0.86
Score: 6.04 / 8.4 with Sub
the higher score of the KH is due to the better LF response
The Score does not take into account the SPL headroom/THD.

View attachment 95214

The small one is just smoother and the sub score shows, in my opinion, that pound for pound and within its SPL capabilities, the smaller one is a better speaker.
Adding more ways on the speaker improve some parameters but also bring some other compromises, more is not always 100% better…

For the pound for pound see there:
https://www.audiosciencereview.com/forum/index.php?threads/jbl-308p-mkii-studio-monitor-review.17338/page-13

[...]Here is the calculation I perform, which seems to match what others are doing:
PPR_LF = 12.69 - 2.49*NBD_ON - 2.99*NBD_PIR - 4.31*log10(14.5) + 2.32*SM_PIR
see there for some details:
https://www.audiosciencereview.com/...urements-community-project.14929/#post-467858


This assumes:
- Frequency response 14.5Hz @-6dB. 14.5Hz is chosen so the theoretical max score is 10 although nothing prevent a system from achieving better performances and therefore the score could exceed 10; this is the first issue.
- Perfect integration whatever that might mean, which is the second issue and where I don't quite adhere to the concept.


The idea, I guess, is to compare speakers if the LF extension is literally taken out of the equation, akin the pound-for-pound rating.
To me it would make more sense to use it this way rather than thinking "if I buy a sub then I'll get the astonishing system predicted by the sub score". That is just not going to happen: sub or not the room will still determine the system LF response.[...]


Here is my take on the EQ nonetheless:
I think unless the user has a decent understanding on measurements and their interpretation it is probably better not to play with the EQ or just FYI. The final user should measure his/her own units, at least the LW.
Room integration EQ is highly recommended thought.

The raw data with corrected ER and PIR:
Score no EQ: 6.19
With Sub: 7.6

Spinorama with no EQ:
  • LF response, is it temperature related or otherwise? I EQed it so that it to resemble the data published by Neumann to get it out of the way.
  • HF it seems that there are tonal control that do pretty much the same job that the EQ I added…
  • The midrange integration maybe related to production tolerance?
View attachment 95204

Directivity:
Better stay at tweeter within +/-10 deg of the mid point Tweeter/mid both vertically and horizontally.
View attachment 95208
View attachment 95193
EQ design:
I have generated two EQs. The APO config files are attached.
  • The first one, labelled, LW is targeted at making the LW flat
  • The second, labelled Score, starts with the first one and adds the score as an optimization variable.
  • The EQs are designed in the context of regular stereo use i.e. domestic environment, no warranty is provided for a near field use in a studio environment although the LW might be better suited for this purpose
Score EQ LW: 6.82
with sub: 8.22

Score EQ Score: 6.98
with sub: 8.39

Code:
Code:
Neumann KH310 APO EQ LW 96000Hz
November232020-121217

Preamp: -2.3 dB

Filter 1: ON HPQ Fc 29.1 Hz Gain 0 dB Q 0.9
Filter 2: ON PK Fc 89 Hz Gain 1.63 dB Q 2.21
Filter 3: ON PK Fc 707 Hz Gain 1.18 dB Q 4.87
Filter 4: ON PK Fc 1548 Hz Gain 2.32 dB Q 6.71
Filter 5: ON PK Fc 4779 Hz Gain -0.88 dB Q 8.74

Neumann KH310 APO EQ Score 96000Hz
November232020-110222

Preamp: -2.3 dB

Filter 1: ON HPQ Fc 29.4 Hz Gain 0 dB Q 0.9
Filter 2: ON PK Fc 88.4 Hz Gain 1.63 dB Q 2.21
Filter 3: ON PK Fc 700.8 Hz Gain 1.06 dB Q 4.56
Filter 4: ON PK Fc 1525 Hz Gain 2.25 dB Q 10
Filter 5: ON PK Fc 4714 Hz Gain -1.31 dB Q 5.85
Filter 6: ON PK Fc 12878 Hz Gain -0.69 dB Q 1.04
View attachment 95203
Spinorama EQ LW
View attachment 95205
Spinorama EQ Score
View attachment 95206
Zoom PIR-LW-ON
View attachment 95207
Regression - Tonal the EQ score makes the On flat
View attachment 95200
Radar no EQ vs EQ score
Some improvements
View attachment 95201
The rest of the plots are attached in particular Horizontal directivity with positive and negative angles separated.
View attachment 95195
I think the tonal regressions you show are very telling/useful, and it's interesting the Listening Window EQ seems to be always having less downwards slope than the Score EQ.....I know I found a Listening Window EQ a tad too bright in my room when I did it for my JBL 308p Mkii speakers (unless that's down to unit to unit variation), so the EQ I'm using now is closer overall to your Score EQ even though I use elements of the Listening Window measurement to EQ portions of the speaker.
 
Last edited:

TimVG

Major Contributor
Joined
Sep 16, 2019
Messages
1,181
Likes
2,573
Hi,

I was reluctant at first glance to have a look at the EQ as the data was near perfect. However after looking more closely it appeared to me the the integration of the mid range was, maybe, not that “perfect”.
I understand the compromise to get the third way: increase LF bandwidth and max SPL capabilities but it does not change the fact that the DI clearly shows the directivity errors.
It might be inconsequential in near field in a mixing studio where the direct sound might be paramount but in a domestic environment, where the domain of validity of Olive score is supposed to reside, it will decrease the Predicted Preference.

Here are some thoughts I shared:
https://www.audiosciencereview.com/...7v-spinorama-and-eq-inside.17283/#post-560611

[...]Regarding the scoring method itself, models in general, not just in audio, are judged with two main criteria:
- Can it describe what we observe reasonably well?
- Do the predictions derived from the model hold against new observations?


The first criterium is covered in the original paper:
https://www.researchgate.net/public...e_Ratings_of_Around-Ear_and_On-Ear_Headphones


I agree with you we need more data for the second criterium to be more understood.
We have some data point from the Harman team comparing different speakers with different PIR shapes and EQing targets but this not as detailed as the first paper.[...]


One may not agree (whatever that means) with the Score but this is the best we’ve got.
Nothing is preventing anyone from coming up with something different but the validation will be hard to say the least…

In the case of the KH310, the error from flatness are higher than one may think at first glance.
See the spinorama with just bass EQ (to look like the Neumann data):
View attachment 95191

Quick comparison with the KH80
View attachment 95215
NBDON = 0.34dB/20th octave on average
NBDPIR = 0.30dB/20th octave on average
SMPIR = 0.78 (closer to 1 is better, interpreted as smoothness of the PIR in the Olive paper)
Score: 6.35 / 7.76 with Sub

To be compared with the KH80 :
NBDON = 0.25dB/20th octave on average
NBDPIR = 0.22dB/20th octave on average
SMPIR = 0.86
Score: 6.04 / 8.4 with Sub
the higher score of the KH is due to the better LF response
The Score does not take into account the SPL headroom/THD.

View attachment 95214

The small one is just smoother and the sub score shows, in my opinion, that pound for pound and within its SPL capabilities, the smaller one is a better speaker.
Adding more ways on the speaker improve some parameters but also bring some other compromises, more is not always 100% better…

For the pound for pound see there:
https://www.audiosciencereview.com/forum/index.php?threads/jbl-308p-mkii-studio-monitor-review.17338/page-13

[...]Here is the calculation I perform, which seems to match what others are doing:
PPR_LF = 12.69 - 2.49*NBD_ON - 2.99*NBD_PIR - 4.31*log10(14.5) + 2.32*SM_PIR
see there for some details:
https://www.audiosciencereview.com/...urements-community-project.14929/#post-467858


This assumes:
- Frequency response 14.5Hz @-6dB. 14.5Hz is chosen so the theoretical max score is 10 although nothing prevent a system from achieving better performances and therefore the score could exceed 10; this is the first issue.
- Perfect integration whatever that might mean, which is the second issue and where I don't quite adhere to the concept.


The idea, I guess, is to compare speakers if the LF extension is literally taken out of the equation, akin the pound-for-pound rating.
To me it would make more sense to use it this way rather than thinking "if I buy a sub then I'll get the astonishing system predicted by the sub score". That is just not going to happen: sub or not the room will still determine the system LF response.[...]


Here is my take on the EQ nonetheless:
I think unless the user has a decent understanding on measurements and their interpretation it is probably better not to play with the EQ or just FYI. The final user should measure his/her own units, at least the LW.
Room integration EQ is highly recommended thought.

The raw data with corrected ER and PIR:
Score no EQ: 6.19
With Sub: 7.6

Spinorama with no EQ:
  • LF response, is it temperature related or otherwise? I EQed it so that it to resemble the data published by Neumann to get it out of the way.
  • HF it seems that there are tonal control that do pretty much the same job that the EQ I added…
  • The midrange integration maybe related to production tolerance?
View attachment 95204

Directivity:
Better stay at tweeter within +/-10 deg of the mid point Tweeter/mid both vertically and horizontally.
View attachment 95208
View attachment 95193
EQ design:
I have generated two EQs. The APO config files are attached.
  • The first one, labelled, LW is targeted at making the LW flat
  • The second, labelled Score, starts with the first one and adds the score as an optimization variable.
  • The EQs are designed in the context of regular stereo use i.e. domestic environment, no warranty is provided for a near field use in a studio environment although the LW might be better suited for this purpose
Score EQ LW: 6.82
with sub: 8.22

Score EQ Score: 6.98
with sub: 8.39

Code:
Code:
Neumann KH310 APO EQ LW 96000Hz
November232020-121217

Preamp: -2.3 dB

Filter 1: ON HPQ Fc 29.1 Hz Gain 0 dB Q 0.9
Filter 2: ON PK Fc 89 Hz Gain 1.63 dB Q 2.21
Filter 3: ON PK Fc 707 Hz Gain 1.18 dB Q 4.87
Filter 4: ON PK Fc 1548 Hz Gain 2.32 dB Q 6.71
Filter 5: ON PK Fc 4779 Hz Gain -0.88 dB Q 8.74

Neumann KH310 APO EQ Score 96000Hz
November232020-110222

Preamp: -2.3 dB

Filter 1: ON HPQ Fc 29.4 Hz Gain 0 dB Q 0.9
Filter 2: ON PK Fc 88.4 Hz Gain 1.63 dB Q 2.21
Filter 3: ON PK Fc 700.8 Hz Gain 1.06 dB Q 4.56
Filter 4: ON PK Fc 1525 Hz Gain 2.25 dB Q 10
Filter 5: ON PK Fc 4714 Hz Gain -1.31 dB Q 5.85
Filter 6: ON PK Fc 12878 Hz Gain -0.69 dB Q 1.04
View attachment 95203
Spinorama EQ LW
View attachment 95205
Spinorama EQ Score
View attachment 95206
Zoom PIR-LW-ON
View attachment 95207
Regression - Tonal the EQ score makes the On flat
View attachment 95200
Radar no EQ vs EQ score
Some improvements
View attachment 95201
The rest of the plots are attached in particular Horizontal directivity with positive and negative angles separated.
View attachment 95195


Great job as usual.
Only thing I would personally change in your model is when doing listening window EQ, afterwards shift the ON regression to flat :). In all of my blind testing I've preferred this universally across loudspeakers.
 

Robbo99999

Master Contributor
Forum Donor
Joined
Jan 23, 2020
Messages
6,874
Likes
6,672
Location
UK
Great job as usual.
Only thing I would personally change in your model is when doing listening window EQ, afterwards shift the ON regression to flat :). In all of my blind testing I've preferred this universally across loudspeakers.
(That would probably also match my EQ experience that I outlined in my previous post....unless I have unit variation from Amir's sample)
 

tuga

Major Contributor
Joined
Feb 5, 2020
Messages
3,984
Likes
4,281
Location
Oxford, England
Hi,

I was reluctant at first glance to have a look at the EQ as the data was near perfect. However after looking more closely it appeared to me the the integration of the mid range was, maybe, not that “perfect”.
I understand the compromise to get the third way: increase LF bandwidth and max SPL capabilities but it does not change the fact that the DI clearly shows the directivity errors.
It might be inconsequential in near field in a mixing studio where the direct sound might be paramount but in a domestic environment, where the domain of validity of Olive score is supposed to reside, it will decrease the Predicted Preference.

Here are some thoughts I shared:
https://www.audiosciencereview.com/...7v-spinorama-and-eq-inside.17283/#post-560611

[...]Regarding the scoring method itself, models in general, not just in audio, are judged with two main criteria:
- Can it describe what we observe reasonably well?
- Do the predictions derived from the model hold against new observations?


The first criterium is covered in the original paper:
https://www.researchgate.net/public...e_Ratings_of_Around-Ear_and_On-Ear_Headphones


I agree with you we need more data for the second criterium to be more understood.
We have some data point from the Harman team comparing different speakers with different PIR shapes and EQing targets but this not as detailed as the first paper.[...]


One may not agree (whatever that means) with the Score but this is the best we’ve got.
Nothing is preventing anyone from coming up with something different but the validation will be hard to say the least…

In the case of the KH310, the error from flatness are higher than one may think at first glance.
See the spinorama with just bass EQ (to look like the Neumann data):
View attachment 95191

Quick comparison with the KH80
View attachment 95215
NBDON = 0.34dB/20th octave on average
NBDPIR = 0.30dB/20th octave on average
SMPIR = 0.78 (closer to 1 is better, interpreted as smoothness of the PIR in the Olive paper)
Score: 6.35 / 7.76 with Sub

To be compared with the KH80 :
NBDON = 0.25dB/20th octave on average
NBDPIR = 0.22dB/20th octave on average
SMPIR = 0.86
Score: 6.04 / 8.4 with Sub
the higher score of the KH is due to the better LF response
The Score does not take into account the SPL headroom/THD.

View attachment 95214

The small one is just smoother and the sub score shows, in my opinion, that pound for pound and within its SPL capabilities, the smaller one is a better speaker.
Adding more ways on the speaker improve some parameters but also bring some other compromises, more is not always 100% better…

For the pound for pound see there:
https://www.audiosciencereview.com/forum/index.php?threads/jbl-308p-mkii-studio-monitor-review.17338/page-13

[...]Here is the calculation I perform, which seems to match what others are doing:
PPR_LF = 12.69 - 2.49*NBD_ON - 2.99*NBD_PIR - 4.31*log10(14.5) + 2.32*SM_PIR
see there for some details:
https://www.audiosciencereview.com/...urements-community-project.14929/#post-467858


This assumes:
- Frequency response 14.5Hz @-6dB. 14.5Hz is chosen so the theoretical max score is 10 although nothing prevent a system from achieving better performances and therefore the score could exceed 10; this is the first issue.
- Perfect integration whatever that might mean, which is the second issue and where I don't quite adhere to the concept.


The idea, I guess, is to compare speakers if the LF extension is literally taken out of the equation, akin the pound-for-pound rating.
To me it would make more sense to use it this way rather than thinking "if I buy a sub then I'll get the astonishing system predicted by the sub score". That is just not going to happen: sub or not the room will still determine the system LF response.[...]


Here is my take on the EQ nonetheless:
I think unless the user has a decent understanding on measurements and their interpretation it is probably better not to play with the EQ or just FYI. The final user should measure his/her own units, at least the LW.
Room integration EQ is highly recommended thought.

The raw data with corrected ER and PIR:
Score no EQ: 6.19
With Sub: 7.6

Spinorama with no EQ:
  • LF response, is it temperature related or otherwise? I EQed it so that it to resemble the data published by Neumann to get it out of the way.
  • HF it seems that there are tonal control that do pretty much the same job that the EQ I added…
  • The midrange integration maybe related to production tolerance?
View attachment 95204

Directivity:
Better stay at tweeter within +/-10 deg of the mid point Tweeter/mid both vertically and horizontally.
View attachment 95208
View attachment 95193
EQ design:
I have generated two EQs. The APO config files are attached.
  • The first one, labelled, LW is targeted at making the LW flat
  • The second, labelled Score, starts with the first one and adds the score as an optimization variable.
  • The EQs are designed in the context of regular stereo use i.e. domestic environment, no warranty is provided for a near field use in a studio environment although the LW might be better suited for this purpose
Score EQ LW: 6.82
with sub: 8.22

Score EQ Score: 6.98
with sub: 8.39

Code:
Code:
Neumann KH310 APO EQ LW 96000Hz
November232020-121217

Preamp: -2.3 dB

Filter 1: ON HPQ Fc 29.1 Hz Gain 0 dB Q 0.9
Filter 2: ON PK Fc 89 Hz Gain 1.63 dB Q 2.21
Filter 3: ON PK Fc 707 Hz Gain 1.18 dB Q 4.87
Filter 4: ON PK Fc 1548 Hz Gain 2.32 dB Q 6.71
Filter 5: ON PK Fc 4779 Hz Gain -0.88 dB Q 8.74

Neumann KH310 APO EQ Score 96000Hz
November232020-110222

Preamp: -2.3 dB

Filter 1: ON HPQ Fc 29.4 Hz Gain 0 dB Q 0.9
Filter 2: ON PK Fc 88.4 Hz Gain 1.63 dB Q 2.21
Filter 3: ON PK Fc 700.8 Hz Gain 1.06 dB Q 4.56
Filter 4: ON PK Fc 1525 Hz Gain 2.25 dB Q 10
Filter 5: ON PK Fc 4714 Hz Gain -1.31 dB Q 5.85
Filter 6: ON PK Fc 12878 Hz Gain -0.69 dB Q 1.04
View attachment 95203
Spinorama EQ LW
View attachment 95205
Spinorama EQ Score
View attachment 95206
Zoom PIR-LW-ON
View attachment 95207
Regression - Tonal the EQ score makes the On flat
View attachment 95200
Radar no EQ vs EQ score
Some improvements
View attachment 95201
The rest of the plots are attached in particular Horizontal directivity with positive and negative angles separated.
View attachment 95195

The KH310 has an edge when it comes to LF extension and HD and IMD.
Would adding a pair of subs to the KH80 bring them in line with the larger model in those aspects and at what cost?

The KH310s may be too "small" for use as mains in a large US sitting room but they're about perfect for a UK one, or most European homes.
 

Maiky76

Senior Member
Joined
May 28, 2020
Messages
440
Likes
3,704
Location
French, living in China
Great job as usual.
Only thing I would personally change in your model is when doing listening window EQ, afterwards shift the ON regression to flat :). In all of my blind testing I've preferred this universally across loudspeakers.

Sorry, I am not quite sure about the meaning of your sentence.
Do you mean: start with flat LW EQ then adjust it to get flat ON regression?
If not, could you explicit your method further? Sorry if you already did it somewhere...

From memory you own objectively good speakers, Genlec 8030c, KH80, a tribe Revels maybe more.
Your method may break down (or not) with lesser designs.
I do believe that some speakers must be voiced (designed with deviation from flat) to be optimized, the Boston recently reviewed is one of them.
My optimizer is specifically designed for that, finding the flattest response that maximizes the preference score in the *general* case.
Does it achieve this goal not sure... some designs remain problematic, I suspect this has more to do with with the score limits than the optimizer itself. These edge cases are the most interesting of course but modifying the score is a huge task.

What I do, which is not the end of it all:
1. The LW EQ is using a fixed target aimed at making the LW flat regardless of any other concerns.
I do this because the LW flatness is highly correlated with preference in the Olive study and this is easy to understand.
This would normally result in a flat tonal regression for the LW and probably slightly ascending ON regression.
That is what most, if not all, the other optimizers are doing: set a target and fit it be it for PIR, ON, LW, in-room, etc.

It is not sufficient *generally* as some aberration only found ON/near axis may result in weird EQs for example Kali IN80 with the coaxial tweeter.
On the other hand, If one were to EQ on score alone as I tried the results may look also quite funny and yourself though these EQ were suboptimal.
These EQs are similar to EQing to a pure slope PIR and to unreasonably boosting LF if left unchecked .


2. The Score EQ starts from the LW EQ and adds the score into the optimization process so that, theoretically, the directivity is taken on board and the EQ does not try to correct things that should not be or at least tries to reasonably correct them. There is no fixed target anymore.
Many speakers will end up with a (close to) flat ON regression but some won't because of their directivity and/or because of the "aggressiveness" of the EQ. The "aggressiveness" level is set by the LW EQ: if the EQ LW complexity can deliver a reasonably flat LW, then it should be a decent starting point.The LW usually ends up having a gentle negative slope similar to the ones observed is the Olive study.

In all cases this not a target that I set but I'd say that speakers with reasonable directivity will tend to end up close to flat ON though.
Sometimes the results are easy to understand but sometimes the deviation form flatness is so large (or so small) that the interpretation becomes more and more difficult without listening trials.


@Pepperjack
From my initial post:

Spinorama with no EQ:
  • LF response, is it temperature related or otherwise? I EQed it so that it to resemble the data published by Neumann to get it out of the way.
The peculiarities of the LF response may or may not be caused by the temperature, they probably are, but I also tend to trust the Neumann data.
That's why I just EQed the LF to the the Neumann data and showed it independently to the rest of the the EQ design as I did not want to repeat the KH80 thread...
 

TimVG

Major Contributor
Joined
Sep 16, 2019
Messages
1,181
Likes
2,573
Do you mean: start with flat LW EQ then adjust it to get flat ON regression?

Yes, that's exactly right.
Your method may break down (or not) with lesser designs.

I'm in the process of producing spinoramas for a couple of speakers like I did recently with the Genelec 1032A. I have a few Behringer models, a KRK model, also a Yamaha HS8 I believe. They would be good test subjects. I even have some RCF, Bose and Yamaha PA stuff available. It will take me a good while to process all of this though, months probably.
 

thefsb

Addicted to Fun and Learning
Joined
Nov 2, 2019
Messages
796
Likes
657
I cranked it way up and then I could detect a bit of distortion but if you were not looking for it, you would be plenty satisfied.
When SPL is high enough, are you sure distortions you hear are coming from the speaker?

When an emergency vehicle with dual tone siren gets close I enjoy listening to the hetrodyne tone that I believe comes from a distrotion introduced by my ear.
 
Top Bottom