Neumann KH 80 DSP Speaker Measurements: Take Two

[jhaider]

The Harbeth is really good, but is warm (treble deemphasis).

Harbeth has a glaring dispersion disruption at the crossover, which is going to lead to different midrange timbre in the reflections compared to the direct sound. Compare Neumann and Harbeth horizontal polar maps.

Essentially perfect. You cannot determine the mid-tweeter crossover frquency from the polar map.

This one reminds me a lot of a Monitor Audio speaker I reviewed. Maybe years of using coaxes and waveguides has over sensitized me to such problems compared to typical listeners. No doubt the 6.5"-7" 2-way with flat waveguide is the de facto audiophile standard.

 

[napilopez]

Harbeth has a glaring dispersion disruption at the crossover, which is going to lead to different midrange timbre in the reflections compared to the direct sound. Compare Neumann and Harbeth horizontal polar maps.

Essentially perfect. You cannot determine the mid-tweeter crossover frquency from the polar map.

This one reminds me a lot of a Monitor Audio speaker I reviewed. Maybe years of using coaxes and waveguides has over sensitized me to such problems compared to typical listeners. No doubt the 6.5"-7" 2-way with flat waveguide is the de facto audiophile standard.

The Olive paper does allow for directivity problems to be corrected for with the on-axis sounds, and I do think this is a valid, if not ideal way of designing speakers. That bump in the Harbeth's on-axis seems to very cleanly account for the crossover directivity dip, hence a PIR curve that's smoother than the Neumann's. One would presume it wouldn't sound as good in a nearfield setup though.

The way I see it, the very best speakers will have both smooth on axis and smooth directivity, as this will net you the highest possible scores. Most good speakers will do this too. But short of that highest tier, balancing out crossover dips with on-axis seems serviceable approach, though it will probably mean that the speaker's sound will be more variable with different listening spaces (i.e. you probably don't want to treat those sidewall reflection). And the data won't look at pretty =]

 

[NTK]

@amirm, let me first (and belatedly) express my gratitude for your enormous sacrifice both in time, money and grief to bring us all these information we otherwise wouldn't be able to get. Your effort is helping us all in advancing our knowledge and understanding of speaker performance and measurement, and I am very grateful for what you've done and are doing.

I'm think out loud here ...

CTA-2034 specifies a measuring distance of 2 m, and then "normalize" the measurements to 1 m for reporting. Are the numbers reported by the Klippel NFS reconstructed at 1 m? If it is the case, then the number from the Klippel NFS should be equivalent to anechoic measurements taken at 1 m, not at 2 m then normalized to 1 m. If 1 m is not sufficiently far-field for a speaker, the 1 m measurement may very well be different from a 2 m measurement normalized to 1 m.

Can you please verify? Thank you very much!

 

[thewas]

The Olive paper does allow for directivity problems to be corrected for with the on-axis sounds, and I do think this is a valid, if not ideal way of designing speakers. That bump in the Harbeth's on-axis seems to very cleanly account for the crossover directivity dip, hence a PIR curve that's smoother than the Neumann's. One would presume it wouldn't sound as good in a nearfield setup though.

Not only in a nearfield setup, according to Toole even in higher distances above the transition frequency the direct sound dominates the tonal perception, that's why he also doesn't recommend EQ in that region from speakers with not known anechoic data just based on listening position measurements which is what actually the Harbeth does and contradicts a bit that Olive metric. In this topic (room EQ above transition) there is more contradiction on both reseachers and imho although the Olive metric is valuable and the only one we have for now, its far from perfect and the example "Neumann vs. Harbeth" really shows it.

 

[MZKM]

If this is the case, and this dip is due to speaker unit variation, then the average of the preference scores for the two units measured should be used in your scoring charts, as this would give a more accurate prediction of the performance of any particular unit a customer might receive. So the score for the Neumann should be 5.29, which makes it pretty much joint best speaker so far with the Harbeth (5.31).

Probably.

 

[MZKM]

Harbeth has a glaring dispersion disruption at the crossover, which is going to lead to different midrange timbre in the reflections compared to the direct sound. Compare Neumann and Harbeth horizontal polar maps.

Essentially perfect. You cannot determine the mid-tweeter crossover frquency from the polar map.

This one reminds me a lot of a Monitor Audio speaker I reviewed. Maybe years of using coaxes and waveguides has over sensitized me to such problems compared to typical listeners. No doubt the 6.5"-7" 2-way with flat waveguide is the de facto audiophile standard.

Hence the boost on-axis for that same region, and how the in-room response in that region is fine (it’s even a little too emphasized, the boost is a tad much).

The Neumann is a better designed speaker in terms of directivity control, but in terms of the sound hitting your ears in the far-field, it’s not much different. The in-room for the Neumann actually has deemphasized response at 2kHz (so does the Harbeth). The slope of the in-room is also not as steep as the ideal (-1.15 vs -1.75), so it will sound a bit bright in the far-field, and the bass droop along with being linear in the bass from 150Hz-300Hz in the in-room doesn’t help, that latter aspect being one most people won’t pay attention to at first glance; look at the Harbeth‘s in-room and how the slope is mostly continuous, it’s a bit too steep though at -2.17 (but is closer to the target).

 

[MZKM]

Not only in a nearfield setup, according to Toole even in higher distances above the transition frequency the direct sound dominates the tonal perception,

I think he says the on-axis dominates above ~10kHz, so higher than what we are talking about. At 1kHz-5kHz the early reflections have a lot of impact.

 

[thewas]

No, see here, from 400-500 Hz is the region where he says that direct sound dominates and EQ based on just listener position measurement (which also includes reflections) is not recommended.

 

[hardisj]

No, see here, from 400-500 Hz is the region where he says that direct sound dominates and EQ based on just listener position measurement (which also includes reflections) is not recommended.

I'm surprised there is no mention of Schroeder frequency.
https://www.soundandvision.com/content/schroeder-frequency-show-and-tell-part-1

Essentially, the working thought wrt EQ is that it is best applied below the Schroeder frequency and the loudspeaker itself should be left to "do its thing" above that transition point. (this is another reason why car audio is so hard; the whole car can act like a baffle and there is no separation of the "room" and the "speaker" because the room itself is practically the "speaker", unless the drive unit is well above the beaming frequency)

 

[thewas]

About two octaves above the Schroeder frequency is the transition zone, so up to 400-500Hz EQ use can be quite reasonable too.

 

[Juhazi]

Above Schröder tone control type very low Q adjustment for tonality is also "allowed". This is actually a part of diy loudspeaker design, "tuning to your taste". Modern measurement techniques and dsp-eq make this very easy , but use 1 octave smoothing!

Regarding in-room listening position measurements and eq, we should always look at decay and RT60 too. Reflectiveness of walls is not even through spectrum. Wildest differences are however always in bass, below Schrder.

 

[jhaider]

Actually, this is not a joke. It is very standard industry practise to have one or two references so you have something to verify system accuracy/repeatability from time to time.

The Klippel shipped with a JBL speaker, right? The calibration speaker need not be good, just stable. I use an inexpensive KEF egg as my calibration speaker, mostly because it has a durable metal shell. It is not a bad speaker for its size, but it is no world beater either.

The Neumann is a better designed speaker in terms of directivity control, but in terms of the sound hitting your ears in the far-field, it’s not much different.

I don’t buy that. Others have referenced studies showing perception of timbre correlates with direct sound. Those results are consistent with my (sighted) experience.

Is anyone aware of a blind preference study comparing a carefully engineered speaker (flat and smooth on axis, smooth off axis with no horizontal dispersion disruption at the tweeter crossover) with a standard speaker of similar extension and output voiced to mimic the in room response of the carefully engineered speaker? I am not but it would be interesting to see. My intuition is band aids such work to an extent, but not completely.

 

[pozz]

@MZKM Would you mind posting the updated 1000 point spin graphs and preference scores (in the same manner as you usually do) here and in the Harbeth thread for archival purposes? That way I can capture them and point to a source post.

 

[MZKM]

@MZKM Would you mind posting the updated 1000 point spin graphs and preference scores (in the same manner as you usually do) here and in the Harbeth thread for archival purposes? That way I can capture them and point to a source post.

Sure; I’m not at my computer, but I’ll do it when I am.

 

[BYRTT]

Comparison the two txt-file spins amirm shared for KH 80 DSP.

 

[MZKM]

1000pt data, @pozz
Preference Rating
SCORE: 5.2
SCORE w/ subwoofer: 7.6

 

[bobbooo]

@pozz in case you weren't aware, the above score is specifically for this measurement of the second Neumann unit. The final score in MZKM's ranking tables is, as it should be, the average of this score and the score for the first unit tested, in order to account for the unit variation seen (especially notable in the high treble), and so give a better prediction of the sound quality someone would expect from any given unit they receive.

By the way, I've noticed some of the other scores in your interactive speaker table also aren't quite right - there were some corrections to the calculations at some point. But if you just take the values from MZKM's table, they'll be the latest correct scores.

 

[carlob]

in order to account for the unit variation seen

Perhaps in order to account for the different methodology used as one was measured on the tweeter axis and the other on the acoustic center or whatever, I haven't seen any unit variation. Not even sure if averaging "take one" and "take two" measurements is correct.

 

[bobbooo]

Perhaps in order to account for the different methodology used as one was measured on the tweeter axis and the other on the acoustic center or whatever, I haven't seen any unit variation. Not even sure if averaging "take one" and "take two" measurements is correct.

Several measurements of the second unit were made at different reference axes. The measurements and score posted above are for the measurement of the second unit using the tweeter axis, just like the measurement of the first unit, so they are comparable measurements and an average of the final preference ratings makes sense due to the use of two different units, giving a better indication of the average performance someone could expect from their particular unit.

 

[pozz]

@pozz in case you weren't aware, the above score is specifically for this measurement of the second Neumann unit. The final score in MZKM's ranking tables is, as it should be, the average of this score and the score for the first unit tested, in order to account for the unit variation seen (especially notable in the high treble), and so give a better prediction of the sound quality someone would expect from any given unit they receive.

By the way, I've noticed some of the other scores in your interactive speaker table also aren't quite right - there were some corrections to the calculations at some point. But if you just take the values from MZKM's table, they'll be the latest correct scores.

Taking the average seems reasonable in this instance, though for the Harbeth I think the updated information is more primary.

The way I've dealt with this before was to have links in the speaker table to the precise post in the review thread where @MZKM gives this data. I think to avoid having him do more manual work than necessary, like the repost I asked for in this instance, I'll take out those links from now on and refer to his tables only like you suggested.

Just remember that I won't know of updates unless I'm notified or I happen to read the thread.