• 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!

M-Audio BX5 D3 | Active Monitor | Measurements, Teardown & Subjective Impressions

Weeb Labs

Addicted to Fun and Learning
Joined
Jun 24, 2020
Messages
604
Likes
1,417
Location
Ireland
The M-Audio BX5 D3 is rather ubiquitous among budget studio monitors, so I thought it only fitting that I purchase a unit and publish the first complete set of measurements.

This monitor typically retails for approximately €80 but I paid only €60 for this particular unit, so it has the potential to represent exceptional value for money. The front baffle design is quite unassuming and the small white power LED also functions as a listening axis indicator. When viewed off-axis, its intensity is attenuated.


A1pdVoueIkL.png

The rear panel offers XLR and TRS inputs but no RCA. There is also an input gain control without detents or a unity point marker and a three position switch for LF attenuation. As we will soon discover, that switch is best left in the -4dB position.


bx5d3_back_hires-backtoback.png

These monitors make use of class AB amplifiers and so the rear plate tends to reach rather high temperatures when left idling. Thermal imaging reveals temperatures approaching 60°C (140°F) on the portion of the plate which sits directly behind the heatsink. For an exposed surface, I consider this to be excessively hot and would not feel comfortable leaving the monitor to idle unsupervised overnight.


1641061465130.png


As always, we begin with the on-axis and off-axis measurements. These utilize a composite of ground plane, nearfield and gated methods, which should be reasonably comparable to measurements acquired using a Klippel NFS or anechoic chamber. For these measurements, the LF attenuation switch was in the "Flat" position.

1640882913869.png


And the in-room response.

1640883949877.png


The on-axis response is very reasonable, although we can see a very large upper bass hump. In-room response tilt is quite steep as a result, which will sound slightly muffled. Directivity is smooth and with only a few simple PEQ filters, there exists the potential for an excellent response.

This is an example of what can be achieved using ten PEQ filters. For specifics, please see Maiky76's excellent post.

BX5 D3 Power+DI.png



Next, we have nearfield measurements of all driver components. Port resonances are kept to very low levels, which is always nice to see. The tweeter bump at 14.6KHz appears to serve as correction for the waveguide rolloff, as it does not appear in farfield measurements. We can also see a slight crossover gap followed by a rise in tweeter response, which explains the 1.2KHz dip and subsequent bump that appears in the farfield results.

driver comp2.png



Next are polar and line charts for directivity. Horizontal directivity is lovely, with almost 50 degrees of freedom.

1640736379334.png



As is often the case with non-coaxial designs, vertical directivity is not so lovely. You will definitely want to remain within about 25 degrees on the vertical axis. If you must listen off-axis, then prefer above to below.

1640736605969.png


1640736650615.png


1640736661284.png



Finally, we have harmonic distortion. This data was captured as part of the ground plane measurement and should be reasonably accurate.

THD.png



Subjective impressions were excellent but I immediately identified the large bump within the upper bass region. Moving the LF attenuation switch to the -4dB position effectively engages the baffle step compensation and fixes this problem completely. You can see the rather dramatic effect below. This measurement is only valid below 13KHz, so please disregard the high frequency rolloff.

switchpos.png


A little bit of vocal emphasis remained but after placing a wide PEQ filter at 1.8KHz, the result was an immensely enjoyable experience as I shuffled through my Apple Music library. Hanaregumi's Dokuji no Life was a particularly pleasant track on these monitors. For silly fun, I then paired my single BX5 unit with a Mackie MR524 and generated an Audiolense correction for the odd pair with surprisingly good results!



For €60-80 per unit, this is rather incredible performance and I would highly recommend this monitor to anybody with a limited budget.

Thank you for reading and I hope this thread has been of help to you.

Please find attached the full spin data. Teardown will be added very shortly.
 

Attachments

  • M-Audio BX5 D3 Full Spinorama.zip
    112.5 KB · Views: 320
Last edited:

sweetchaos

Major Contributor
The Curator
Joined
Nov 29, 2019
Messages
3,872
Likes
11,549
Location
BC, Canada
Also, I suggest you use a 50db vertical scale, like Amir and Erin does, for their spinorama.
Your scale is 40db, which stretches the data.

Your Mackie MR524 spin needs fixing as well.
 

dfuller

Major Contributor
Joined
Apr 26, 2020
Messages
3,335
Likes
5,050
Re: that bass hump, are you sure that isn't an artifact of your measurement technique? I seem to recall Stereophile's measurements (which are taken largely in the same way) always have a bass hump right around there.
 
OP
Weeb Labs

Weeb Labs

Addicted to Fun and Learning
Joined
Jun 24, 2020
Messages
604
Likes
1,417
Location
Ireland
Also, I suggest you use a 50db vertical scale, like Amir and Erin does, for their spinorama.
Your scale is 40db, which stretches the data.

Your Mackie MR524 spin needs fixing as well.
Thank you! Not sure how I missed that. Both have now been fixed.
 
Last edited:
OP
Weeb Labs

Weeb Labs

Addicted to Fun and Learning
Joined
Jun 24, 2020
Messages
604
Likes
1,417
Location
Ireland
Re: that bass hump, are you sure that isn't an artifact of your measurement technique? I seem to recall Stereophile's measurements (which are taken largely in the same way) always have a bass hump right around there.
It is definitely real and was the first thing that I noticed upon listening to them, prior to capturing any measurements. It also appears in ungated in-room measurements such as this one.

To my knowledge, Stereophile does not typically make use of ground plane measurements. They often encounter floor bounce artifacts as a result.

unknown.png
 
Last edited:
  • Like
Reactions: czt

Maiky76

Senior Member
Joined
May 28, 2020
Messages
440
Likes
3,703
Location
French, living in China
The M-Audio BX5 D3 is rather ubiquitous among budget studio monitors, so I thought it only fitting that I purchase a unit and publish the first complete set of measurements.

This monitor typically retails for approximately €80 but I paid only €60 for this particular unit, so it has the potential to represent exceptional value for money. The front baffle design is quite unassuming and the small white power LED also functions as a listening axis indicator. When viewed off-axis, its intensity is attenuated.


The rear panel offers XLR and TRS inputs but no RCA. There is also an input gain control without detents or a unity point marker and a three position switch for LF attenuation. As we will soon discover, that switch is best left in the -4dB position.



As always, we begin with the on-axis and off-axis measurements. These are a composite of ground plane, nearfield and gated methods, which should be reasonably comparable to measurements acquired using a Klippel NFS or anechoic chamber.

For these measurements, the LF attenuation switch was in the "Flat" position.

View attachment 175221

The on-axis response is very reasonable, although we can see clear evidence of insufficient or nonexistent baffle step correction. Directivity is smooth and with only a few simple PEQ filters, there exists the potential for an excellent response.


Next, we have nearfield measurements of all driver components. Port resonances are kept to very low levels, which is always nice to see. There seems to be a tweeter resonance around 14.6KHz but it does not appear in any farfield measurements. We can see a slight crossover gap followed by a rise in tweeter response, which explains the 1.2KHz dip and subsequent bump that appears in the farfield results.

View attachment 175203


Next are polar and line charts for directivity. Horizontal directivity is lovely, with almost 50 degrees of freedom.

View attachment 175204


As is often the case with non-coaxial designs, vertical directivity is not so lovely. You will definitely want to remain within about 25 degrees on the vertical axis. If you must listen off-axis, then prefer above to below.

View attachment 175206

View attachment 175207

View attachment 175208


Finally, we have harmonic distortion. This data was captured as part of the ground plane measurement and should be reasonably accurate.

View attachment 175209


Subjective impressions were excellent but I immediately identified the large bump within the upper bass region. Moving the LF attenuation switch to the -4dB position effectively engages the baffle step compensation and fixes this problem completely. You can see the rather dramatic effect below. This measurement is only valid below 13KHz, so please disregard the high frequency rolloff.

View attachment 175211

A little bit of vocal emphasis remained but after placing a wide PEQ filter at 1.8KHz, the result was an immensely enjoyable experience as I shuffled through my Apple Music library. Hanaregumi's Dokuji no Life was a particularly pleasant track on these monitors. For silly fun, I then paired my single BX5 unit with a Mackie MR524 and generated an Audiolense correction for the odd pair with surprisingly good results!



For €60-80 per unit, this is rather incredible performance and I would highly recommend this monitor to anybody with a limited budget.

Thank you for reading and I hope this thread has been of help to you.

Please find attached the full spin data. Teardown will be added very shortly.
Hi,

Great effort again!
Score probably not comparable to the data gathered using the NFS due to the smoothing effect of the windowing and I don't know if the PIR/ER is correctly calculated by Vituixcad. In addition these speakers are probably intended for near filed in which case the score might not be accurate.

Here is my take on the EQ based on what I believe is the Flat position of the LF Switch.

The following EQs are “anechoic” EQs to get the speaker right before room integration. If you able to implement these EQs you must add EQ at LF for room integration, that is usually not optional… see hints there: https://www.audiosciencereview.com/...helf-speaker-review.11144/page-26#post-800725

The raw data with corrected ER and PIR:

Score no EQ: 4.9
With Sub: 7.5

Spinorama with no EQ:

M-Audio BX5 D3 by Weeb Labs No EQ Spinorama.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.
  • Quite a few sharp boost that may not translate well with other units and would definitively warant listening tests...

Score EQ LW: 5.8
with sub: 8.4

Score EQ Score: 6.2
with sub: 8.8

Code:
M-Audio BX5 D3 by Weeb Labs APO Score LW 96000Hz
December292021-111919

Preamp: -2.5 dB

Filter 1: ON HPQ Fc 53.33,    0.00,    1.29
Filter 2: ON PK Fc 132.74,    -5.27,    0.78
Filter 3: ON PK Fc 459.00,    2.32,    4.36
Filter 4: ON PK Fc 885.43,    -0.75,    3.38
Filter 5: ON PK Fc 1222.33,    2.56,    6.91
Filter 6: ON PK Fc 1771.00,    -1.45,    2.27
Filter 7: ON PK Fc 3164.49,    2.26,    1.70
Filter 8: ON PK Fc 6533.77,    1.51,    2.38
Filter 9: ON PK Fc 12142.26,    -1.31,    3.74
Filter 10: ON PK Fc 17068.65,    2.55,    1.76

M-Audio BX5 D3 by Weeb Labs APO Score EQ 96000Hz
December292021-111829

Preamp: -2.4 dB

Filter 1: ON HPQ Fc 51.33,    0.00,    1.29
Filter 2: ON PK Fc 124.83,    -5.27,    0.66
Filter 3: ON PK Fc 461.00,    2.45,    4.86
Filter 4: ON PK Fc 865.93,    -0.75,    4.88
Filter 5: ON PK Fc 1220.33,    2.56,    6.16
Filter 6: ON PK Fc 1605.33,    -1.16,    1.55
Filter 7: ON PK Fc 2821.90,    2.53,    3.20
Filter 8: ON PK Fc 5874.77,    1.06,    1.63
Filter 9: ON PK Fc 12827.15,    -1.68,    2.10
Filter 10: ON PK Fc 22244.08,    3.05,    1.36

M-Audio BX5 D3 by Weeb Labs EQ Design.png


Spinorama EQ LW
M-Audio BX5 D3 by Weeb Labs LW EQ Spinorama.png


Spinorama EQ Score
M-Audio BX5 D3 by Weeb Labs Score EQ Spinorama.png


Zoom PIR-LW-ON
M-Audio BX5 D3 by Weeb Labs Zoom.png


Regression - Tonal
ON Flat with EQ Score
M-Audio BX5 D3 by Weeb Labs Regression-Tonal.png


Radar no EQ vs EQ score
Nice improvements
M-Audio BX5 D3 by Weeb Labs Radar.png

Impact of the LF trim:
The measurements seem to have been taken at different times so the accuracy is not guaranteed.
I focussed on the range under 300Hz, error still better than 0.5dB.
Switch.png

-2dB@LF
Score -2dB@LF: 4.7
with sub: 7.5
M-Audio BX5 D3 by Weeb Labs -2dB@LF Trim Spinorama.png


-4dB@LF
Score -4dB@LF: 4.5
with sub: 7.3
M-Audio BX5 D3 by Weeb Labs -4dB@LF Trim Spinorama.png
 

Attachments

  • M-Audio BX5 D3 by Weeb Labs APO Score EQ 96000Hz.txt
    542 bytes · Views: 202
  • M-Audio BX5 D3 by Weeb Labs APO Score LW 96000Hz.txt
    542 bytes · Views: 162

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,741
Likes
16,174
Very nice discovery in that budget range, so thank you very much for those measurements!

One little remark though:
The on-axis response is very reasonable, although we can see clear evidence of insufficient or nonexistent baffle step correction.
If there would be no baffle step correction, the response would rather have a mid hump https://www.micka.de/en/bafflestep.php , this response looks more like an over-correction of the bass.
 

bunkbail

Addicted to Fun and Learning
Forum Donor
Joined
Feb 27, 2018
Messages
520
Likes
662

Gabe33

Member
Joined
Jan 5, 2021
Messages
16
Likes
13
I own a pair of BX D2s , they have the same problem in the lower frequencies, this causes a sharp peak around 128hz in all of the rooms i measured them in , im glad m audio added a low frequency switch to compensate for that, it would have been great if they re-tuned the speaker completely. The Bx5 D2s i own have degraded in sound quality after 9 years of daily studio use . But they still play louder than my Adam T7s . It would be great if you guys tested the bigger BX8 D3
 

pierre

Addicted to Fun and Learning
Forum Donor
Joined
Jul 1, 2017
Messages
962
Likes
3,046
Location
Switzerland
The M-Audio BX5 D3 is rather ubiquitous among budget studio monitors, so I thought it only fitting that I purchase a unit and publish the first complete set of measurements.

This monitor typically retails for approximately €80 but I paid only €60 for this particular unit, so it has the potential to represent exceptional value for money. The front baffle design is quite unassuming and the small white power LED also functions as a listening axis indicator. When viewed off-axis, its intensity is attenuated.


The rear panel offers XLR and TRS inputs but no RCA. There is also an input gain control without detents or a unity point marker and a three position switch for LF attenuation. As we will soon discover, that switch is best left in the -4dB position.



As always, we begin with the on-axis and off-axis measurements. These are a composite of ground plane, nearfield and gated methods, which should be reasonably comparable to measurements acquired using a Klippel NFS or anechoic chamber.

For these measurements, the LF attenuation switch was in the "Flat" position.

View attachment 175221

The on-axis response is very reasonable, although we can see clear evidence of insufficient or nonexistent baffle step correction. Directivity is smooth and with only a few simple PEQ filters, there exists the potential for an excellent response.


Next, we have nearfield measurements of all driver components. Port resonances are kept to very low levels, which is always nice to see. There seems to be a tweeter resonance around 14.6KHz but it does not appear in any farfield measurements. We can see a slight crossover gap followed by a rise in tweeter response, which explains the 1.2KHz dip and subsequent bump that appears in the farfield results.

View attachment 175203


Next are polar and line charts for directivity. Horizontal directivity is lovely, with almost 50 degrees of freedom.

View attachment 175204


As is often the case with non-coaxial designs, vertical directivity is not so lovely. You will definitely want to remain within about 25 degrees on the vertical axis. If you must listen off-axis, then prefer above to below.

View attachment 175206

View attachment 175207

View attachment 175208


Finally, we have harmonic distortion. This data was captured as part of the ground plane measurement and should be reasonably accurate.

View attachment 175209


Subjective impressions were excellent but I immediately identified the large bump within the upper bass region. Moving the LF attenuation switch to the -4dB position effectively engages the baffle step compensation and fixes this problem completely. You can see the rather dramatic effect below. This measurement is only valid below 13KHz, so please disregard the high frequency rolloff.

View attachment 175211

A little bit of vocal emphasis remained but after placing a wide PEQ filter at 1.8KHz, the result was an immensely enjoyable experience as I shuffled through my Apple Music library. Hanaregumi's Dokuji no Life was a particularly pleasant track on these monitors. For silly fun, I then paired my single BX5 unit with a Mackie MR524 and generated an Audiolense correction for the odd pair with surprisingly good results!



For €60-80 per unit, this is rather incredible performance and I would highly recommend this monitor to anybody with a limited budget.

Thank you for reading and I hope this thread has been of help to you.

Please find attached the full spin data. Teardown will be added very shortly.
@Weeb Labs Thanks for the data.
Can you include the full data? all SPL H and V? That makes nicer graphs :)
 

DSJR

Major Contributor
Joined
Jan 27, 2020
Messages
3,312
Likes
4,425
Location
Suffolk Coastal, UK
£79 each on Amazon UK.
 

3125b

Major Contributor
Joined
May 18, 2020
Messages
1,352
Likes
2,204
Location
Germany
That‘s a lot better than I thought, always had my doubts about these being the cheapest of the 5“ monitors.
Surprising though to see how much less bass extension these have vs. the MR524 given that they are the same size.

Regular retail price for these is 88€ - that is not a lot of money for an active speaker, the Ayima chip amp alone isn‘t much cheaper.

@Weeb Labs
Thanks for the review, this is one of the most interesting kind of speakers since they are interesting for non-audio people unlike multi thousand dollar offerings.
How would the hiss on these compare to the Mackie?
 

abdo123

Master Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
7,423
Likes
7,940
Location
Brussels, Belgium
Very nice discovery in that budget range, so thank you very much for those measurements!

One little remark though:

If there would be no baffle step correction, the response would rather have a mid hump https://www.micka.de/en/bafflestep.php , this response looks more like an over-correction of the bass.
it might be intentional voicing since the -4dB is perfectly flat. but yeah there is obviously baffle step correction in this design.
 

abdo123

Master Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
7,423
Likes
7,940
Location
Brussels, Belgium
Hi,

Great effort again!
Score probably not comparable to the data gathered using the NFS due to the smoothing effect of the windowing and I don't know if the PIR/ER is correctly calculated by Vituixcad. In addition these speakers are probably intended for near filed in which case the score might not be accurate.

Here is my take on the EQ based on what I believe is the Flat position of the LF Switch.

The following EQs are “anechoic” EQs to get the speaker right before room integration. If you able to implement these EQs you must add EQ at LF for room integration, that is usually not optional… see hints there: https://www.audiosciencereview.com/...helf-speaker-review.11144/page-26#post-800725

The raw data with corrected ER and PIR:

Score no EQ: 4.9
With Sub: 7.5

Spinorama with no EQ:

View attachment 175239

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.
  • Quite a few sharp boost that may not translate well with other units and would definitively warant listening tests...

Score EQ LW: 5.8
with sub: 8.4

Score EQ Score: 6.2
with sub: 8.8

Code:
M-Audio BX5 D3 by Weeb Labs APO Score LW 96000Hz
December292021-111919

Preamp: -2.5 dB

Filter 1: ON HPQ Fc 53.33,    0.00,    1.29
Filter 2: ON PK Fc 132.74,    -5.27,    0.78
Filter 3: ON PK Fc 459.00,    2.32,    4.36
Filter 4: ON PK Fc 885.43,    -0.75,    3.38
Filter 5: ON PK Fc 1222.33,    2.56,    6.91
Filter 6: ON PK Fc 1771.00,    -1.45,    2.27
Filter 7: ON PK Fc 3164.49,    2.26,    1.70
Filter 8: ON PK Fc 6533.77,    1.51,    2.38
Filter 9: ON PK Fc 12142.26,    -1.31,    3.74
Filter 10: ON PK Fc 17068.65,    2.55,    1.76

M-Audio BX5 D3 by Weeb Labs APO Score EQ 96000Hz
December292021-111829

Preamp: -2.4 dB

Filter 1: ON HPQ Fc 51.33,    0.00,    1.29
Filter 2: ON PK Fc 124.83,    -5.27,    0.66
Filter 3: ON PK Fc 461.00,    2.45,    4.86
Filter 4: ON PK Fc 865.93,    -0.75,    4.88
Filter 5: ON PK Fc 1220.33,    2.56,    6.16
Filter 6: ON PK Fc 1605.33,    -1.16,    1.55
Filter 7: ON PK Fc 2821.90,    2.53,    3.20
Filter 8: ON PK Fc 5874.77,    1.06,    1.63
Filter 9: ON PK Fc 12827.15,    -1.68,    2.10
Filter 10: ON PK Fc 22244.08,    3.05,    1.36

View attachment 175234

Spinorama EQ LW
View attachment 175238

Spinorama EQ Score
View attachment 175237

Zoom PIR-LW-ON
View attachment 175235

Regression - Tonal
ON Flat with EQ Score
View attachment 175236

Radar no EQ vs EQ score
Nice improvements
View attachment 175232
Impact of the LF trim:
The measurements seem to have been taken at different times so the accuracy is not guaranteed.
I focussed on the range under 300Hz, error still better than 0.5dB.
View attachment 175230
-2dB@LF
Score -2dB@LF: 4.7
with sub: 7.5
View attachment 175240

-4dB@LF
Score -4dB@LF: 4.5
with sub: 7.3
View attachment 175241
Any idea why the -2dB trim scores higher when the -4dB trim is the flat one?
 
OP
Weeb Labs

Weeb Labs

Addicted to Fun and Learning
Joined
Jun 24, 2020
Messages
604
Likes
1,417
Location
Ireland
That‘s a lot better than I thought, always had my doubts about these being the cheapest of the 5“ monitors.
Surprising though to see how much less bass extension these have vs. the MR524 given that they are the same size.

Regular retail price for these is 88€ - that is not a lot of money for an active speaker, the Ayima chip amp alone isn‘t much cheaper.

@Weeb Labs
Thanks for the review, this is one of the most interesting kind of speakers since they are interesting for non-audio people unlike multi thousand dollar offerings.
How would the hiss on these compare to the Mackie?
These are actually significantly smaller than the MR524s but almost a kilogram heavier. Amplifier noise floors are similar but the MR524 is very slightly quieter.
 

3125b

Major Contributor
Joined
May 18, 2020
Messages
1,352
Likes
2,204
Location
Germany
You're right, I assumed they would be the same size, sure look similar in the pictures, but actually aren't. MR524 are 11,2l, BX5 only 8.8l, making them one of the smallest 5" nearfield monitors**. That would explain the bass extension, probably why they decided to go for that mid-bass hump. The MR524 are already rather stuffed with the transformer, electronics, heat sink and damping material, I wonder how they fit even more weight into these. Probably a larger transformer since they claim double the amplifier power rating.

**Size comparison: https://comparesizes.com/comparison...5V-vs-Swissonic-ASM5-Thomann-vs/1640793518485
There are quite a few. Simply delete those that don't interest you, the others will be rearranged automatically.
Width range: 170-190mm (6.7-7.5in)
Height range: 234-300mm (9.2-11,8in)
Depth range: 197-297mm (7.8-11.7in)
The width is similar, only 2cm difference, but height and especially depth varies.

As far as self-noise goes: I ask because I question the S&R measurements. I've just gotten my new Alpha 65 Evo wich, according to S&R, only have 0.9dB lower self-noise than the old Alpha 65. But there is no way that is true, it's a huge difference, they are not even half as loud and actually quieter than the MR524. That is a welcome surprise, but it makes me doubt their measurements. That probably has to do with gain, but in that case they need to be set to a comparable level for a fair comparison.
 
Last edited:

Maiky76

Senior Member
Joined
May 28, 2020
Messages
440
Likes
3,703
Location
French, living in China
Any idea why the -2dB trim scores higher when the -4dB trim is the flat one?
M-Audio BX5 D3 by Weeb Labs -4dB@LF Trim Radar.png

Two factors:

1. The -4dB trim @LF changes the -6dB point of the speaker form 58.6 to 64.5, this parameters has a heavy weight on the metric
2. The slope of the PIR, the steeper the higher the SM PIR, with similar NBD PIR, which is also a heavy weight on the metric.

Looking at the sub score 7.5 to 7.3 the SMPIR contributes to a loss of 0.2 points while the Loss of LF also contributes for about the same loss.
The improvement in smoothness NBD ON and NBD PIR do not compensate for these two parameters.
The metric favors steep slopes... this is why if one ONLY optimizes for the score the EQ will always end up with huge boost at LF and decrease at HF (depending on the directivity).
Here after is an example where the additional variables (on top of the metric itself) of my algorithm are unchecked but with identical structure and complexity. EQing for the score only is equivalent more or less to increase the LF output and fit a strong slope for the PIR.
M-Audio BX5 D3 by Weeb Labs Unchecked EQ.png

The LF is unreasonably boosted and the HF is trimmed down to have a steeper slope and better regression matching.
The NBD parameters are kept very similar; that's the beauty of true optimization, the best compromised is found.
BTW, the theoretical score becomes 6.6/9.0.

M-Audio BX5 D3 by Weeb Labs Unchecked EQ radar.png
 
Last edited:
Top Bottom