• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Speaker equalization - a real case study

Pio2001

Senior Member
Joined
May 15, 2018
Messages
315
Likes
477
Location
Neuville-sur-Saône, France
I have been trying to equalize my JBL 305P mk2 speakers in order to improve their sound and make it more neutral. According to measurements, they should be quite easy to equalize.
I also have got a pair of Neumann KH-120 monitors, that are among the most neutral speakers one can find.

The JBL retail at 250 euros, the Neumann at 1200 euros. Is it really possible that for 250 euros and an extra 110 euros for an external DSP, I can get the same performance as the Neumann KH-120 ?

Here, I am going to summarize my experiments. The answer to this question is no, I didn't manage to get the same sound with both speakers. However, I discovered many very interesting things during the process, and the result with the JBL, although not identical to the Neumann, is nonetheless excellent.

1 - Context

My setup is an ordinary installation in an ordinary room (6.5 x 3.5 x 2.5 meters) with no acoustic treatment. The speakers are 80 cm away from walls, and the listening position is away from walls.
The room has strong low frequency resonances that need heavy room equalization.
The listening distance is 2 meters, the 60 dB decay time measured between the speakers and the listening position is 0.45 seconds at 500 Hz.

The system includes a MiniDSP 2x4 that is in charge of room equalization below 700 Hz. All other frequencies are left uncorrected, except if the speaker itself needs to be corrected, according to the generally accepted principle that we can "hear through the room".

Three kind of measurements will be used:
-Spinorama of both speakers made with a Klippel device. The one posted by ASR for the JBL and the one posted by Neumann for the Neumann. They represent the actual performance of the speaker without any influence from the room.
-Moving microphone measurement realized with a pink periodic noise of 64 k samples, analyzed with REW RTA, 64 k FFT, 1/48th octave. The microphone is moved above one seat. The curves are displayed either with no smoothing, or with Var smoothing. They represent the result at the listening position.
-Gated measurements made pointing the microphone at the speaker, 70 cm away, halfway between the tweeter and woofer. The speaker being as far as possible from any object. They also represent the actual behaviour of the speaker without any influence of the room, but because I don't have a Klippel measurement device at home, they are limited to high frequencies only, and more and more smoothed towards lower frequencies, until there are no data anymore.

2 - Reference system

I will compare the results that I get with the JBL 305P to the results got with the Neumann KH-120, my reference system, that I have carefully setup and that give an extremely balanced sound.

Here are the measurements of the KH-120 reference system with and without equalization. The blue curve is the equalization applied with the MiniDSP :

00 System.png


Two things are important here :
-The frequency response is uneven above 700 Hz. This is because I have chosen to have the speaker's frequency response being flat, not the in-room response. Actually, it sounds very neutral.
-The target curve below 700 Hz is the result of subjective adjustments made by ear alone. It is said everywhere that room problems below the transition frequency should be corrected, but there is no rule saying how. The target level depends on the room, on the speaker, and on the listening distance, and its value is unknown.
The lack of bass below 100 Hz, for example, is unusual, but I found no other way for it to sound right. A straight line at 51 dB sounds boomy. Anyway, the topic here is to try to get as good a result with the JBL speakers.

Here is the spinorama of the Neumann KH-120, in addition to the measurements at the listening position:

00 SP.png


All curves are dispalyed at the same scale. The two directivity index curves have been inverted.

We can see that above 700 Hz, where the speakers are uncorrected, the sound at the listening position is dominated by the sound power curve of the speaker.
This is a surprise : I was expecting the result to be half way between the listening window curve (direct sound) and the sound power directivity index (indirect sound). But the variations at the listening position go actually beyond the variations of the sound power DI. They are identical to the sound power variations themselves !
In fact, with a perfectly neutral ox-axis curve, the sound power and sound power DI curves should have been identical.

What we can remember for the time being is that with this setup, the directivity index of the speaker is going to play an important role in the result.

3 - System under test

Now, what can we expect if we replace the KH-120 with the 305P ? Let's have a look at their respective spinoramas.

01 Neumann spino.png

02 JBL spino.png


The directivity indexes of both speakers are quite similar, with the same anomaly around 1800 Hz. In theory, we can expect very good results if we equalize properly the JBL speaker.
The Klippel measurements are very accurate. If a small problem looks identical on all 4 direct curves, it should be equalized. This is something very different from room equalization, where positive corrections are unadvised, and accurate corrections don't work above 1000 Hz because they depend too much on the listening position. Here, there is no limitation on positive corrections other than the maximum playback level, and no dependency on the listening position.

4 - Early attempts

The first trials, made in january 2020, very quite disappointing!
After having copied the above curve in REW in order to create a correction, it didn't sound right. A first problem was visible on the measurement:

03 MMM full eq.png


On the right side, at 8000 Hz, the result measured from the listening position follows exactly the correction. Which is wrong. It should behave like it does at 1700 Hz: a correction was made, and the result is good. At 8000 Hz, the correction is completely useless. The result was good before, and it is no more correct after.

The cause of the problem was discovered in the following weeks: the original spinorama included artifacts from the protection cage of the microphone. The 8000 Hz dip was one of these artifacts. The JBL 305P has no dip at this frequency.

I then removed all corrections above 4000 Hz, getting this resulting curve for the JBL 305P:

04 Correction v1 .png


According to REW's predictions, that correction should give me this result at the listening position:

05 prediction v1.png


And this one for the speaker itself, measured with a 5ms window (because of a bug in REW, the pink curve had no microphone calibration):

06 Predicted.png


But the audible results were not as good as expected. The human voices and acoustic instruments were definitely more natural with the Neumann and its jaggy curve at the listening position, than with the JBL and its beautiful curve !

This result is consistent with the hypothesis that this is the frequency response curve of direct sound that counts. Not the one at the listening position.
The Neumann KH-120 would have a direct sound even more neutral than the corrected JBL.
But this is not the end of the story !

5 - New improvements

Recently, I decided to give a second try and to analyze more accurately what's going on. So I went and brought back the JBL 305P in my listening room.

I reloaded the above correction and setup a proper room correction for them. Since the bass reflex port is behind the speaker, the room interactions are not the same as with the Neumann, whose port is in front of the speaker. The deviations in the amount of bass are enough to spoil the general balance.

After few listening sessions I identified a problem with human voices.
I tried to reproduce the problem with a graphic equalizer in order to find the frequency of the problem. I found that if I push the 1700 Hz frequency band of the graphic equalizer up 10 dB, the problem sounds the same. Therefore there must be something wrong around 1700 Hz.
This frequency is outside the room correction frequencies, so I need to look at the anechoic response in order to find the cause:

07 Medium problem.png


And indeed, we can see that there is too much energy at 1700 Hz.

This was not visible in the january measurements posted above. And it also contradicts Klippel measurements done by Amir.
The variation on the 1000 Hz side of the 1700 Hz bump is not surprising: I messed with these frequencies in order to adjust the overall target of the new room correction. But the shallow dip on the other side, around 3000 Hz, was not there in january, in the pink curve posted above. I don't know why. Maybe the microphone was not positioned exactly in the same axis during both measurements. And it is possible that my copy of JBL 305P is not identical to the one measured by Amir. After all, this is a 110 euros speaker.
Anyway, I changed the correction.

It was also obvious that the JBL had much more treble than the Neumann. I had forgotten about the treble trim ! You know, the little switch on the back of the speaker, that allows to adjust the level of bass and treble. Neumann advises to set it at -1 dB for listening in a room with live acoustics, and that was the case from the beginning... but not on the JBL.
This correction starts around 8000 Hz, but, given the measurement above, I decided to start it at 5000 Hz in order to further improve the anechoic response of the JBL.

Last, I raised the level below 300 Hz in order to get a warmer sound.

08 New correction.png


Now, my anechoic curve looks flat from 1000 to 6000 Hz ! I don't pay much attention to what's happening above, as it strongly depends on the exact measurement axis (in front of the tweeter, mid-way between the teeter and woofer etc.)

I also checked the result with a 2 ms window, because 5 ms in not strictly anechoic with my measurement setup:

09 5 ms Window.png


We can see that the 1700 Hz correction is still needed. The problem was not caused by the 5 ms windowing.
We can also see that the treble variations above 6000 Hz are real, they don't show up in the Neumann measurement.
The inclusion of small reflections that are ocurring between 2 and 5 ms doesn't change the overall balance of the curve, but introduces narrow oscillations.
The rise on the extreme left of the 2 ms curve, from 700 to 500 Hz, doesn't appear in the 5 ms curves, which is consistent with the hypothesis that it is just an image of the window itself and must not be taken into account. It suggests that in the 5 ms curves, the rise from 400 to 200 Hz is an artifact of the same kind and doesn't exist in reality.

After loading this new correction in the MiniDSP, here is how it measures from the listening position. The Spinorama is here only for information, since the anechoic response of the speaker is no more the same. The two (inverted) directivity indexes curves, however, are unchanged.

11 SP.png


We can see that the in-room response is rising from 1700 Hz to 4000 Hz !
This is strange, but the experiment requires the speaker to be equalized flat above 700 Hz, so I can't change this.
A possible explanation is that, according to REW, in my room, the measured RT rises from 500 to 5000 Hz (from 0.45 to 0.50 seconds IIRC), which is in itself unusual. This is probably related to the absence of carpet or curtains.

We can also see that the treble it perfectly smooth above 4000 Hz, while it was not the case in the anechoic measurements. Here, the Neumann and JBL behave completely differently. Had I tried to correct the anechoic response of the JBL above 6000 Hz, I would have spoiled this result.


6 - How does it sound ?

Well, now it begins to sound really good !
I've been listening to this setup for several days. Here is how it compares with the Neumann KH-120.

The first difference is treble. The Neumann treble sounds weak, too far away. The JBL treble, on the opposite, is strong.
On human voices, according to the recording, the treble ranges from "just right" to "too bright". The voices treble always sounds right, but not the "s" sounds, that, on some recordings, stand out too much. This never happens with the Neumann.
On the opposite, with the Neumann, the treble ranges from "just right" to "too weak".
Besides the level of treble, there is also a difference in treble texture. The JBL treble always sound crystalline, while the treble of the Neumann always sound dull. Even if I increase the treble on the Neumann and decrease it on the JBL, their textures remain unchanged.
I have experimented with the variations above 6000 Hz, trying to reproduce the crystalline sound of the JBL on the Neumann by coying the anechoic oscillations measured above, and there was a bit of change. It sounded less dull. So maybe this difference in texture is caused by these minor variations around 10000 Hz on the JBL.
I have also measured in one point at the listening position, in order to see if the treble anomalies of the JBL's response were visible here without moving the microphone, but I didn't see any obvious difference between the forest of peaks produced by one speaker or the other in one point.

The second audible difference is in the medium frequencies: the JBL medium sounds overall a bit more balanced than the Neumann, however less natural.
Let me explain. With the Neumann, nothing ever sounds artificial, but there is a general feeling of sweetness. A soprano forte, in the real world, hurts the ears in a particular way (personally experienced at the opera) , but the Neumann never seem to be able to hurt one's ears. The JBL "harshness" in soprano voices seem closer to reality. Maybe this is related to the 2000 Hz dip in the resulting in-room frequency response of the Neumann. It complies with the measurements at the listening position, that look better with the JBL.
On the other hand, human voices don't always sound as "realistic" as the Neumann. Listening with the Neumann, the quality of the speakers is never questioned except maybe for the lack of treble. With the JBL, there is still something that sounds sometimes unnatural. You can tell that you're listening to a recording. Sometimes with the Neumann, this feeling disappears and we can hear "through the recording", as if the real voices were there.
I wonder if this performance is permitted by the extremely accurate neutrality of the Neumann's anechoic frequency response, while the slight lack of balance would be caused by the accident in their sound power curve, that plainly affects the result at the listening position.
Unfortunately, the gated anechoic measurements that I can perform at home are not accurate enough to investigate further, by improving the neutrality of my JBL.

The third difference is simple : the JBL have slightly deeper bass. In my setup, they reach 30 Hz at -5 dB, while the Neumann reach only 35 Hz.

Here are the compared measurements at the listening position, with VAR smoothing.

10 Result 2.png


7 - Speaker directivity

In the above measurements, there is obviously something very different between these two speakers. The treble of the JBL being stronger is not an illusion !
The two curves are aligned in level from 400 to 1000 Hz. Above 1000 Hz, all room correction has stopped. Therefore the sudden variation after 1500 Hz is caused by the directivity of the speakers. The JBL look here to have a much wider dispersion than the Neumann in high frequencies.

Let's see if this correlates with the spinoramas of both speakers.

159_JBL_Neumann_DI.png


Here, I have overlaid the two spinoramas at the same scale, and changed the colours of the Neumann (replacing red with blue). Looking at the directivity indexes at the bottom, we can see that the JBL have weaker early reflections at all frequencies (in this inverted representation, the higher the curve, the weaker the sound ! ). It tells nothing for us, since we are looking for a difference above and below 1500 Hz, while this effect is the same above and below 1500 Hz. The sound power directivity index is also similar.

But wait, what if we realign these DI curves around 1000 Hz, where they locally diverge ?

160_JBL_Neumann_DI.png


Now we have exactly the same behaviour in the spinorama and in reality : the JBL is stronger than the Neumann from 1700 Hz and above (remember, the DI curves are inverted in the spinorama).

But in order to get this result, we had to offset the DI curves. Physically, the real directivity indexes are the ones plotted in the first diagram ! The waveguide around the dome tweeter of the Neumann spreads exactly the same amount of energy around the room as the waveguide around the dome of the JBL. How is it that we get more treble with the JBL than with the Neumann ?

The answer is a bit complicated, and related to the concept of partial room correction.
Remember that making the room correction for both speakers, I set the target curve between 30 and 700 Hz by ear, listening to a lot of human voices. The real target curve is always unknown. Doing this, what obviously happened is that I used the audible level around 1000 Hz as my subjective reference, thus artificially aligning the sound power of both speakers around 1000 Hz. That's always a problem when correcting only a limited part of the frequency response: we need a reference level. Here, without even paying attention, when I defined all the filters for room eq, I found myself using the 700 - 1700 Hz range as the reference level, rather than the 1700 - 10000 Hz range. That's how it sounded natural to me with human voices.

8 - Future improvements

As I said in the listening results, the Neumann sound a bit weak in treble, and the JBL a bit strong.

Of course, I already tried to correct this difference, always without success. The result sounded worse, leading me to believe that the JBL had naturally more treble than the Neumann in their power response (which, as we saw, is wrong).
The problem was that I was trying to correct the curves measured at the listening position, using a high shelf at 1700 Hz, without paying attention to the anechoic response, nor to the directivity index.
But now I have more information : according to the measurements, this difference in treble is caused by a sudden divergence in the sound power directivity of both speakers between 700 and 2000 Hz, and to the fact that the reference level for room correction was chosen in this range.

Knowing this, I should be able to try a different approach, for example starting the high shelf closer to 700 Hz than to 2000, in order to preserve the anechoic response of the speakers around 1700 Hz, that was spoiled when I was trying to correct of the curves measured at the listening position.
 
OP
Pio2001

Pio2001

Senior Member
Joined
May 15, 2018
Messages
315
Likes
477
Location
Neuville-sur-Saône, France
Another possible progress, for the Neumann KH-120, is the listening window curve of the spinorama. It is not completely flat. We can see that it is 1 dB below the 1000 Hz level from 1700 to 7000 Hz (from 1200 Hz actually).
Correcting this, more than half of the difference between the JBL and the Neumann curves at the listening position should disappear.

Funny how 1 dB looks meaningless in the spinorama, but very important on the overlaid listening position measurements.
 

dougi

Addicted to Fun and Learning
Joined
May 31, 2020
Messages
530
Likes
410
Location
ACT, Australia
Another possible progress, for the Neumann KH-120, is the listening window curve of the spinorama. It is not completely flat. We can see that it is 1 dB below the 1000 Hz level from 1700 to 7000 Hz (from 1200 Hz actually).
Correcting this, more than half of the difference between the JBL and the Neumann curves at the listening position should disappear.

Funny how 1 dB looks meaningless in the spinorama, but very important on the overlaid listening position measurements.
Yeah I had something similar with Proac Response D2s. My room seemed to exaggerate treble variations observed at 1m. I either corrected with RoomPerfect automatically or tweak EQ manually a bit.
 
OP
Pio2001

Pio2001

Senior Member
Joined
May 15, 2018
Messages
315
Likes
477
Location
Neuville-sur-Saône, France
I have tried to improve the above results.
First I applied a few corrections to the Neumann according to their exact frequency response (listening window) according to the spinorama:

20 Neumann corr1.png


I'm not sure that this correction is ok, because it doesn't match what my own windowed measurement says, but given the small deviations and the difficulty to accurately measure the speaker, I rather trust the Klippel measurement.

But the result doesn't sound better. The treble sounds "thin".
I also tried a weaker correction :

21 Neumann corr2.png


This one sounds better.
But in the meantime, I also improved the result with the JBL. I found that the voices still had a boxy sound, around 750 Hz, this time. So I removed the peak at this frequency :

22 JBL correction.png


And now they sound really great ! Better than the KH-120, actually. At least for human voice. The treble is a tie : quite recessed in the Neumann, quite bright on the JBL.

The JBL have a small problem at 750 Hz, that this correction helps to solve... or does it ? They were already supposed to be equalized. How is it that this extra correction was needed ? It should have been correct from the beginning.
To check what's happening, let's look at the speaker correction alone vs the new correction:

23 JBL corrections.png


This new correction definitely belongs to the room correction part, and is only justified because the curve measured from the listening position looks better.

Now if we compare the final result with the JBL and the Neumann (without the last correction above), we have this:

24 Result v8.png


Obviously, looking at the 600 - 2000 Hz part, we can see why the JBL sounds better now.

The problem is that since the last correction was purely a room correction, there's no reason not to apply it also to the Neumann. Otherwise, the comparison is unfair. The JBL benefits from a more accurate room correction.

I'm not sure anymore what I'm doing here : I'm dealing with frequencies that are supposed to be well over room correction, and I begin to make subjective assessments.

The only thing that seems clear to me is that the sound power frequency response of a speaker plays an important role in the final result.
We use to say that, looking at a spinorama, what's important in the sound power directivity index is that it is smooth, without accidents, but that's forgetting another very important data : the slope of sound power directivity index.
A steep sound power directivity index means that, in an average room, the speaker will sound muffled, while a flat sound power directivity index means that the speaker will sound bright.
 
Last edited:

thewas

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
3,375
Likes
7,110
I'm not sure anymore what I'm doing here : I'm dealing with frequencies that are supposed to be well over room correction, and I begin to make subjective assessments.
I think the problem with the KH120 are not the higher frequencies but the ones below 1 kHz which need reduction in the sound power in usual listening positioning as otherwise it sounds too warm.
In my desktop monitoring case I didn't need to correct above 1 kHz but more below than with other loudspeakers:

1606347311351.png


The only thing that seems clear to me is that the sound power frequency response of a speaker
plays an important role in the final result.
Exactly.
 

STRA

New Member
Joined
Nov 26, 2020
Messages
1
Likes
1
I think the problem with the KH120 are not the higher frequencies but the ones below 1 kHz which need reduction in the sound power in usual listening positioning as otherwise it sounds too warm.
In my desktop monitoring case I didn't need to correct above 1 kHz but more below than with other loudspeakers:

View attachment 95880


Exactly.

+1
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
1,635
Likes
1,140
Location
Canada
Since early on, I've always thought of the KH120's as dark sounding speakers. Which is why I still I prefer the much brighter sounding (even way before EQ!) Sceptre S8s more often than not despite being way more uneven in FR.

1606533708149.png

KH120s have virtually zero EQ above 300Hz. Dip at 450Hz is from desk -- it comes and goes as one shifts around the listening position. Only really obvious if using single test tones. HF response of the Sceptres in the graph is "uber EQ'd" to the max -- despite that, I am very, very, very happy with what I'm hearing post EQ. o_O

My own attempts at increasing the KH120's treble response from a simple shelf to more targeted PEQs I've all but scrapped out eventually -- perhaps there's really little room for improvement(?). I've just not come up with a decision whether the overall effect in totality is for the better. For the Sceptres, despite my extensive "EQ abuse", the result is a much more simple/definitive yes! :D
 

AnalogSteph

Major Contributor
Joined
Nov 6, 2018
Messages
1,796
Likes
1,666
Location
.de
My 2 cents -

This is a good demonstration for why basic room acoustics still matter:
  1. A vastly underdamped room gives trouble well into the kHz. In a room like the OP's, 2 m is too far for something like the KH120s IMO. I'd really want to address the most problematic reflection points at least.
  2. An ~80 cm wall distance (from back of speaker?) is in the "not recommended zone" and likely to cause the dip a little under 100 Hz (λ/4 ~= 90 cm). You may be able to nuke the ~53 Hz peak with that by placing them about 1.5 m out, fine tuning with measurements. Otherwise try a position as close to the wall as possible and EQ out the boundary gain, it could only help level handling.
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
1,635
Likes
1,140
Location
Canada
Since early on, I've always thought of the KH120's as dark sounding speakers. Which is why I still I prefer the much brighter sounding (even way before EQ!) Sceptre S8s more often than not despite being way more uneven in FR.

View attachment 96247
KH120s have virtually zero EQ above 300Hz. Dip at 450Hz is from desk -- it comes and goes as one shifts around the listening position. Only really obvious if using single test tones. HF response of the Sceptres in the graph is "uber EQ'd" to the max -- despite that, I am very, very, very happy with what I'm hearing post EQ. o_O

My own attempts at increasing the KH120's treble response from a simple shelf to more targeted PEQs I've all but scrapped out eventually -- perhaps there's really little room for improvement(?). I've just not come up with a decision whether the overall effect in totality is for the better. For the Sceptres, despite my extensive "EQ abuse", the result is a much more simple/definitive yes! :D

Okay, I was wrong about the KH120s having zero EQ. The monitor's own -1dB shelf filter had been switched on -- sorry, I always seem to forget about it.

*BUT, still, even if the +1dB HF shelf were turned on in the KH120s, the Sceptres out of the box still sound brighter (even when activating their internal -1.5dB HS filter as well).
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
1,635
Likes
1,140
Location
Canada
Otherwise try a position as close to the wall as possible and EQ out the boundary gain, it could only help level

I went for the opposite approach and moved the speakers quite well away from boundaries -- other than the ceiling and standing desk. I've lost a fair bit of bass boost from boundary gain, but have a sub to augment for that loss anyway.
 

Robbo99999

Major Contributor
Forum Donor
Joined
Jan 23, 2020
Messages
3,179
Likes
2,692
Location
UK
I have been trying to equalize my JBL 305P mk2 speakers in order to improve their sound and make it more neutral. According to measurements, they should be quite easy to equalize.
I also have got a pair of Neumann KH-120 monitors, that are among the most neutral speakers one can find.

The JBL retail at 250 euros, the Neumann at 1200 euros. Is it really possible that for 250 euros and an extra 110 euros for an external DSP, I can get the same performance as the Neumann KH-120 ?

Here, I am going to summarize my experiments. The answer to this question is no, I didn't manage to get the same sound with both speakers. However, I discovered many very interesting things during the process, and the result with the JBL, although not identical to the Neumann, is nonetheless excellent.

1 - Context

My setup is an ordinary installation in an ordinary room (6.5 x 3.5 x 2.5 meters) with no acoustic treatment. The speakers are 80 cm away from walls, and the listening position is away from walls.
The room has strong low frequency resonances that need heavy room equalization.
The listening distance is 2 meters, the 60 dB decay time measured between the speakers and the listening position is 0.45 seconds at 500 Hz.

The system includes a MiniDSP 2x4 that is in charge of room equalization below 700 Hz. All other frequencies are left uncorrected, except if the speaker itself needs to be corrected, according to the generally accepted principle that we can "hear through the room".

Three kind of measurements will be used:
-Spinorama of both speakers made with a Klippel device. The one posted by ASR for the JBL and the one posted by Neumann for the Neumann. They represent the actual performance of the speaker without any influence from the room.
-Moving microphone measurement realized with a pink periodic noise of 64 k samples, analyzed with REW RTA, 64 k FFT, 1/48th octave. The microphone is moved above one seat. The curves are displayed either with no smoothing, or with Var smoothing. They represent the result at the listening position.
-Gated measurements made pointing the microphone at the speaker, 70 cm away, halfway between the tweeter and woofer. The speaker being as far as possible from any object. They also represent the actual behaviour of the speaker without any influence of the room, but because I don't have a Klippel measurement device at home, they are limited to high frequencies only, and more and more smoothed towards lower frequencies, until there are no data anymore.

2 - Reference system

I will compare the results that I get with the JBL 305P to the results got with the Neumann KH-120, my reference system, that I have carefully setup and that give an extremely balanced sound.

Here are the measurements of the KH-120 reference system with and without equalization. The blue curve is the equalization applied with the MiniDSP :

View attachment 95023

Two things are important here :
-The frequency response is uneven above 700 Hz. This is because I have chosen to have the speaker's frequency response being flat, not the in-room response. Actually, it sounds very neutral.
-The target curve below 700 Hz is the result of subjective adjustments made by ear alone. It is said everywhere that room problems below the transition frequency should be corrected, but there is no rule saying how. The target level depends on the room, on the speaker, and on the listening distance, and its value is unknown.
The lack of bass below 100 Hz, for example, is unusual, but I found no other way for it to sound right. A straight line at 51 dB sounds boomy. Anyway, the topic here is to try to get as good a result with the JBL speakers.

Here is the spinorama of the Neumann KH-120, in addition to the measurements at the listening position:

View attachment 95022

All curves are dispalyed at the same scale. The two directivity index curves have been inverted.

We can see that above 700 Hz, where the speakers are uncorrected, the sound at the listening position is dominated by the sound power curve of the speaker.
This is a surprise : I was expecting the result to be half way between the listening window curve (direct sound) and the sound power directivity index (indirect sound). But the variations at the listening position go actually beyond the variations of the sound power DI. They are identical to the sound power variations themselves !
In fact, with a perfectly neutral ox-axis curve, the sound power and sound power DI curves should have been identical.

What we can remember for the time being is that with this setup, the directivity index of the speaker is going to play an important role in the result.

3 - System under test

Now, what can we expect if we replace the KH-120 with the 305P ? Let's have a look at their respective spinoramas.

View attachment 95030
View attachment 95031

The directivity indexes of both speakers are quite similar, with the same anomaly around 1800 Hz. In theory, we can expect very good results if we equalize properly the JBL speaker.
The Klippel measurements are very accurate. If a small problem looks identical on all 4 direct curves, it should be equalized. This is something very different from room equalization, where positive corrections are unadvised, and accurate corrections don't work above 1000 Hz because they depend too much on the listening position. Here, there is no limitation on positive corrections other than the maximum playback level, and no dependency on the listening position.

4 - Early attempts

The first trials, made in january 2020, very quite disappointing!
After having copied the above curve in REW in order to create a correction, it didn't sound right. A first problem was visible on the measurement:

View attachment 95032

On the right side, at 8000 Hz, the result measured from the listening position follows exactly the correction. Which is wrong. It should behave like it does at 1700 Hz: a correction was made, and the result is good. At 8000 Hz, the correction is completely useless. The result was good before, and it is no more correct after.

The cause of the problem was discovered in the following weeks: the original spinorama included artifacts from the protection cage of the microphone. The 8000 Hz dip was one of these artifacts. The JBL 305P has no dip at this frequency.

I then removed all corrections above 4000 Hz, getting this resulting curve for the JBL 305P:

View attachment 95036

According to REW's predictions, that correction should give me this result at the listening position:

View attachment 95037

And this one for the speaker itself, measured with a 5ms window (because of a bug in REW, the pink curve had no microphone calibration):

View attachment 95038

But the audible results were not as good as expected. The human voices and acoustic instruments were definitely more natural with the Neumann and its jaggy curve at the listening position, than with the JBL and its beautiful curve !

This result is consistent with the hypothesis that this is the frequency response curve of direct sound that counts. Not the one at the listening position.
The Neumann KH-120 would have a direct sound even more neutral than the corrected JBL.
But this is not the end of the story !

5 - New improvements

Recently, I decided to give a second try and to analyze more accurately what's going on. So I went and brought back the JBL 305P in my listening room.

I reloaded the above correction and setup a proper room correction for them. Since the bass reflex port is behind the speaker, the room interactions are not the same as with the Neumann, whose port is in front of the speaker. The deviations in the amount of bass are enough to spoil the general balance.

After few listening sessions I identified a problem with human voices.
I tried to reproduce the problem with a graphic equalizer in order to find the frequency of the problem. I found that if I push the 1700 Hz frequency band of the graphic equalizer up 10 dB, the problem sounds the same. Therefore there must be something wrong around 1700 Hz.
This frequency is outside the room correction frequencies, so I need to look at the anechoic response in order to find the cause:

View attachment 95048

And indeed, we can see that there is too much energy at 1700 Hz.

This was not visible in the january measurements posted above. And it also contradicts Klippel measurements done by Amir.
The variation on the 1000 Hz side of the 1700 Hz bump is not surprising: I messed with these frequencies in order to adjust the overall target of the new room correction. But the shallow dip on the other side, around 3000 Hz, was not there in january, in the pink curve posted above. I don't know why. Maybe the microphone was not positioned exactly in the same axis during both measurements. And it is possible that my copy of JBL 305P is not identical to the one measured by Amir. After all, this is a 110 euros speaker.
Anyway, I changed the correction.

It was also obvious that the JBL had much more treble than the Neumann. I had forgotten about the treble trim ! You know, the little switch on the back of the speaker, that allows to adjust the level of bass and treble. Neumann advises to set it at -1 dB for listening in a room with live acoustics, and that was the case from the beginning... but not on the JBL.
This correction starts around 8000 Hz, but, given the measurement above, I decided to start it at 5000 Hz in order to further improve the anechoic response of the JBL.

Last, I raised the level below 300 Hz in order to get a warmer sound.

View attachment 95049

Now, my anechoic curve looks flat from 1000 to 6000 Hz ! I don't pay much attention to what's happening above, as it strongly depends on the exact measurement axis (in front of the tweeter, mid-way between the teeter and woofer etc.)

I also checked the result with a 2 ms window, because 5 ms in not strictly anechoic with my measurement setup:

View attachment 95050

We can see that the 1700 Hz correction is still needed. The problem was not caused by the 5 ms windowing.
We can also see that the treble variations above 6000 Hz are real, they don't show up in the Neumann measurement.
The inclusion of small reflections that are ocurring between 2 and 5 ms doesn't change the overall balance of the curve, but introduces narrow oscillations.
The rise on the extreme left of the 2 ms curve, from 700 to 500 Hz, doesn't appear in the 5 ms curves, which is consistent with the hypothesis that it is just an image of the window itself and must not be taken into account. It suggests that in the 5 ms curves, the rise from 400 to 200 Hz is an artifact of the same kind and doesn't exist in reality.

After loading this new correction in the MiniDSP, here is how it measures from the listening position. The Spinorama is here only for information, since the anechoic response of the speaker is no more the same. The two (inverted) directivity indexes curves, however, are unchanged.

View attachment 95061

We can see that the in-room response is rising from 1700 Hz to 4000 Hz !
This is strange, but the experiment requires the speaker to be equalized flat above 700 Hz, so I can't change this.
A possible explanation is that, according to REW, in my room, the measured RT rises from 500 to 5000 Hz (from 0.45 to 0.50 seconds IIRC), which is in itself unusual. This is probably related to the absence of carpet or curtains.

We can also see that the treble it perfectly smooth above 4000 Hz, while it was not the case in the anechoic measurements. Here, the Neumann and JBL behave completely differently. Had I tried to correct the anechoic response of the JBL above 6000 Hz, I would have spoiled this result.


6 - How does it sound ?

Well, now it begins to sound really good !
I've been listening to this setup for several days. Here is how it compares with the Neumann KH-120.

The first difference is treble. The Neumann treble sounds weak, too far away. The JBL treble, on the opposite, is strong.
On human voices, according to the recording, the treble ranges from "just right" to "too bright". The voices treble always sounds right, but not the "s" sounds, that, on some recordings, stand out too much. This never happens with the Neumann.
On the opposite, with the Neumann, the treble ranges from "just right" to "too weak".
Besides the level of treble, there is also a difference in treble texture. The JBL treble always sound crystalline, while the treble of the Neumann always sound dull. Even if I increase the treble on the Neumann and decrease it on the JBL, their textures remain unchanged.
I have experimented with the variations above 6000 Hz, trying to reproduce the crystalline sound of the JBL on the Neumann by coying the anechoic oscillations measured above, and there was a bit of change. It sounded less dull. So maybe this difference in texture is caused by these minor variations around 10000 Hz on the JBL.
I have also measured in one point at the listening position, in order to see if the treble anomalies of the JBL's response were visible here without moving the microphone, but I didn't see any obvious difference between the forest of peaks produced by one speaker or the other in one point.

The second audible difference is in the medium frequencies: the JBL medium sounds overall a bit more balanced than the Neumann, however less natural.
Let me explain. With the Neumann, nothing ever sounds artificial, but there is a general feeling of sweetness. A soprano forte, in the real world, hurts the ears in a particular way (personally experienced at the opera) , but the Neumann never seem to be able to hurt one's ears. The JBL "harshness" in soprano voices seem closer to reality. Maybe this is related to the 2000 Hz dip in the resulting in-room frequency response of the Neumann. It complies with the measurements at the listening position, that look better with the JBL.
On the other hand, human voices don't always sound as "realistic" as the Neumann. Listening with the Neumann, the quality of the speakers is never questioned except maybe for the lack of treble. With the JBL, there is still something that sounds sometimes unnatural. You can tell that you're listening to a recording. Sometimes with the Neumann, this feeling disappears and we can hear "through the recording", as if the real voices were there.
I wonder if this performance is permitted by the extremely accurate neutrality of the Neumann's anechoic frequency response, while the slight lack of balance would be caused by the accident in their sound power curve, that plainly affects the result at the listening position.
Unfortunately, the gated anechoic measurements that I can perform at home are not accurate enough to investigate further, by improving the neutrality of my JBL.

The third difference is simple : the JBL have slightly deeper bass. In my setup, they reach 30 Hz at -5 dB, while the Neumann reach only 35 Hz.

Here are the compared measurements at the listening position, with VAR smoothing.

View attachment 95060

7 - Speaker directivity

In the above measurements, there is obviously something very different between these two speakers. The treble of the JBL being stronger is not an illusion !
The two curves are aligned in level from 400 to 1000 Hz. Above 1000 Hz, all room correction has stopped. Therefore the sudden variation after 1500 Hz is caused by the directivity of the speakers. The JBL look here to have a much wider dispersion than the Neumann in high frequencies.

Let's see if this correlates with the spinoramas of both speakers.

View attachment 95065

Here, I have overlaid the two spinoramas at the same scale, and changed the colours of the Neumann (replacing red with blue). Looking at the directivity indexes at the bottom, we can see that the JBL have weaker early reflections at all frequencies (in this inverted representation, the higher the curve, the weaker the sound ! ). It tells nothing for us, since we are looking for a difference above and below 1500 Hz, while this effect is the same above and below 1500 Hz. The sound power directivity index is also similar.

But wait, what if we realign these DI curves around 1000 Hz, where they locally diverge ?

View attachment 95066

Now we have exactly the same behaviour in the spinorama and in reality : the JBL is stronger than the Neumann from 1700 Hz and above (remember, the DI curves are inverted in the spinorama).

But in order to get this result, we had to offset the DI curves. Physically, the real directivity indexes are the ones plotted in the first diagram ! The waveguide around the dome tweeter of the Neumann spreads exactly the same amount of energy around the room as the waveguide around the dome of the JBL. How is it that we get more treble with the JBL than with the Neumann ?

The answer is a bit complicated, and related to the concept of partial room correction.
Remember that making the room correction for both speakers, I set the target curve between 30 and 700 Hz by ear, listening to a lot of human voices. The real target curve is always unknown. Doing this, what obviously happened is that I used the audible level around 1000 Hz as my subjective reference, thus artificially aligning the sound power of both speakers around 1000 Hz. That's always a problem when correcting only a limited part of the frequency response: we need a reference level. Here, without even paying attention, when I defined all the filters for room eq, I found myself using the 700 - 1700 Hz range as the reference level, rather than the 1700 - 10000 Hz range. That's how it sounded natural to me with human voices.

8 - Future improvements

As I said in the listening results, the Neumann sound a bit weak in treble, and the JBL a bit strong.

Of course, I already tried to correct this difference, always without success. The result sounded worse, leading me to believe that the JBL had naturally more treble than the Neumann in their power response (which, as we saw, is wrong).
The problem was that I was trying to correct the curves measured at the listening position, using a high shelf at 1700 Hz, without paying attention to the anechoic response, nor to the directivity index.
But now I have more information : according to the measurements, this difference in treble is caused by a sudden divergence in the sound power directivity of both speakers between 700 and 2000 Hz, and to the fact that the reference level for room correction was chosen in this range.

Knowing this, I should be able to try a different approach, for example starting the high shelf closer to 700 Hz than to 2000, in order to preserve the anechoic response of the speakers around 1700 Hz, that was spoiled when I was trying to correct of the curves measured at the listening position.
Wow, well done on your huge amount of work & analysis you've done on this & writing up too!

How confident are you in your listening tests of comparing your results between the two speakers? I've found through EQ'ing my different headphones that my philosophy has recently changed - I've so far concluded that it's very difficult to compare "sound" in listening tests between headphones for a few reasons. Firstly it takes too long to change headphones over and listen to the same portion of the track so you kinda forget the subtle differences. Secondly, my brain gets burned in very quickly to a sound signature of a headphone, so putting on a different headphone always sounds a bit strange at first. Therefore it's really hard for me to assess "proper" tonality between headphones, in terms of selecting one as a reference standard. However, I have learned that it's very easy to assess EQ differences within one headphone - ie comparing EQ's at the flip of a switch in EqualiserAPO, so you can do that quickly and not forget the intricacies of the difference between the EQ changes on the same short passage of your music track. Having said that I can tell the tonal differences between headphones but only in relation to last pair I had on my head, but not in relation to which one is the most correct or reference standard. I even have "flat reference measured" speakers to compare my headphones against, but the same problem exists re quite long time between testing.

I think I've come to the conclusion that you can use measurements to verify how close you are a proposed standard, but I think it's very hard to listen test compare different speakers & headphones due to the time taken to switch between the different speakers or headphones and when combined with the phenomenon of "brain burn in" (and also level matching which I forgot to mention). However, I do think you can use EQ to affect measurements so that you can approach a proposed standard, and I believe you can make & assess EQ changes within one speaker or headphone to improve various aspects of it's sound, but I think it's extremely difficult to compare seperate speakers & headphones. (I do think I remember the spatial abilities of different headphones though, I remember that a lot more accurately than the tonality).
 
OP
Pio2001

Pio2001

Senior Member
Joined
May 15, 2018
Messages
315
Likes
477
Location
Neuville-sur-Saône, France
Wow, well done on your huge amount of work & analysis you've done on this & writing up too!

How confident are you in your listening tests of comparing your results between the two speakers?

Thanks !

There are two different things at play.

Hearing the difference between two setups is easier when the switching delay is several seconds rather than one or two minutes.
One or two minutes is the time it takes to replace the Neumann with the JBL.
In order to switch between two corrections, I use Foobar's convolver. It takes about 5 seconds.

But things are different when the matter is to tell which setup is more neutral. Short-term impressions must be avoided because, as you say, the ear keeps in memory the previous sound and remains used to it.
If we listen to a bad setup for some time, we get used to it, and when switching to a neutral one, the neutral sound will sound bad for some time, until our auditory system adapts itself to the new sound.

Objectively, when I say that the direct sound eq sounds better to me than the in-room eq, I couldn't swear that it is actually more neutral.

I can clearly hear the differences between two setups, and what I hear matches what is measured.
I can also hear that the Neumann and JBL are more neutral than the Kef R300, as shown in their measurements. When there is such a difference, I can tell which is more neutral.
I also hear that the tweeter of my old Dynaudio Gemini kit is too loud without measuring it (it is indeed setup +1.5 dB above 2000 Hz).

On the other hand, the fact that I find low frequencies balanced while the measurement shows that they are too weak is strange.

I have listened to a reference system once (Igor Kirkwood's system, setup by Jean-Luc Ohl), and I found it balanced. That's a good thing.

I also compared the church organ of my town to a recording of it played on my system (Neumann KH-120, flat direct sound), 15 minutes after the concert (it was another recording, but made in the same church). Medium and treble sounded the same as in reality, but there, I could hear that my system had less bass than reality, as the measurements show.

I rarely go to acoustic concerts, so I don't have a deep knowledge of "real" sound. But it already occurred to me, so I'm not completely without any reference.
 
Top Bottom