Pio2001
Senior Member
Hi,
After reading the review of the JBL 305P mkII speaker (that I own), I decided to try and equalize them according to the measurements that were published.
The two conditions for the equalization to succeed were that the directivity indices are smooth, and that my own pair sounds exactly the same as the one measured by Amir.
We can see in the review that the first condition is met, and that these speakers are good candidates for equalization.
The second condition seems to be fulfilled according to my own measurements (see my post here with the curves : https://www.audiosciencereview.com/...pro-monitors-review.10811/page-19#post-302946 )
I thus generated a set of eq filters in the REW software that should compensate for the JBL's frequency response, as measured by Amir (on-axis and listening window curves).
Here is the equalization curve that I managed to generate. It is displayed upside down, and I have overlaid it on Amir's graph. It is the light blue curve. I am very happy with the accuracy that I got while playing manually with all the PK coefficients, just looking at Amir's picture.
I didn't set any correction above 14 kHz for the time being, as I am completely deaf above 13.5 kHz anyway.
Once applied to the speakers, it sounds... a bit weird.
I then measured the frequency response before and after the correction from the listening position. I use the moving microphone method for this purpose, drawing the cumulated RTA of a Pink periodic noise.
Unfortunately, I can't measure the windowed response of the speaker alone, as I can't apply the correction to the REW signal. I can only apply it to music played by the Foobar2000 software player, with its convolver plugin. What we see here is the room curve in addition to the speaker curve :
The left part of the curve, up to 500 Hz, can be ignored. It represents the room's response rather than the speaker's response, with a correction setup for the Neumann KH-120 monitors, that are front vented, while the JBL are rear-vented.
The right part, from 1000 Hz to 20000 Hz, is supposed to decrease gently as the frequency goes up, as expected in any room.
Oops ! My correction obviously created some problems that were not there to begin with. Let's compare this curve with the correction itself :
The part around 500 Hz is difficult to interpret. On one hand, it looks like the correction is too strong, but on the other hand, this frequency range is supposed to by higher than the treble frequencies, measured from the listening position. Let's put it aside. The real problem is obvious : from 4000 Hz and up, the correction is useless. The original curve was neutral and didn't need any correction.
Let's drop this part from the eq and keep only the useful corrections, up to 4000 Hz.
Making a "half correction" is something very difficult, because there is no reference level for the part that is not corrected. In my first try, I just naively drew a straight line from 4000 Hz upwards. But the result didn't sound good. The treble were muffled. I realized that I had not calculated any target level for the part that is above 4000 Hz.
So I went back to Amir's graph, and tried to evaluate the average level measured above 4000 Hz. I restricted my reading to the 5000 to 15000 Hz range, ignoring what's happening above. I tried to draw a straight line so that half of the curve is above, and half is below. Then I noticed that this level was exactly the same as the 150 Hz peak. That would be my reference point.
Reading now my REW correction overlaid on Amir's graph, I noticed that it was accurately following it at 150 Hz. I therefore set the shelf level of my correction above 4000 Hz exactly identical to my 150 Hz peak (I mean 150 dip, since the curve is displayed upside down).
I generated the impulse response of the new correction and loaded it in Foobar's convolver and... wow ! Now, that sounds really great !!!
Here is how it measures from the listening position :
When I say it sounds great, I'm talking about medium and high frequencies. The low frequencies need re-equalization, as the bass reflex of the JBL is on the back of the speaker, and the MiniDSP room correction (still active from 35 to 600 Hz) was setup for the Neumann.
The measurement shows yet a little problem around 1000 Hz that was not there before correction. Here is the response without eq, the final correction, and REW's prediction of the result.
We can see that the level is unduly decreasing from 1000 to 1400 Hz, while it was not the case before correction.
Comparing with the first picture with Amir's graph, we can see that the correction is supposed to be ok. Since I'm not sure what's happening here, I didn't try any other adjustment.
The forum doesn't allow audio nor binary attachments, so here are links to my correction :
44100 Hz impulse response for convolvers : JBL 305P mk2 equalization.wav
Eq filters for REW : JBL 305P mkII eq filters.req
Important :
The correction is valid for JBL 305P mkII speakers, not for regular LSR 305 speakers.
The impulse response works at 44100 Hz only. The convolver must be preceded by a 44100 Hz resampler, or have a built-in resampler. Otherwise, if a file with a sampling frequency different from 44100 Hz is played, the correction will be wrong.
After reading the review of the JBL 305P mkII speaker (that I own), I decided to try and equalize them according to the measurements that were published.
The two conditions for the equalization to succeed were that the directivity indices are smooth, and that my own pair sounds exactly the same as the one measured by Amir.
We can see in the review that the first condition is met, and that these speakers are good candidates for equalization.
The second condition seems to be fulfilled according to my own measurements (see my post here with the curves : https://www.audiosciencereview.com/...pro-monitors-review.10811/page-19#post-302946 )
I thus generated a set of eq filters in the REW software that should compensate for the JBL's frequency response, as measured by Amir (on-axis and listening window curves).
Here is the equalization curve that I managed to generate. It is displayed upside down, and I have overlaid it on Amir's graph. It is the light blue curve. I am very happy with the accuracy that I got while playing manually with all the PK coefficients, just looking at Amir's picture.
I didn't set any correction above 14 kHz for the time being, as I am completely deaf above 13.5 kHz anyway.
Once applied to the speakers, it sounds... a bit weird.
I then measured the frequency response before and after the correction from the listening position. I use the moving microphone method for this purpose, drawing the cumulated RTA of a Pink periodic noise.
Unfortunately, I can't measure the windowed response of the speaker alone, as I can't apply the correction to the REW signal. I can only apply it to music played by the Foobar2000 software player, with its convolver plugin. What we see here is the room curve in addition to the speaker curve :
The left part of the curve, up to 500 Hz, can be ignored. It represents the room's response rather than the speaker's response, with a correction setup for the Neumann KH-120 monitors, that are front vented, while the JBL are rear-vented.
The right part, from 1000 Hz to 20000 Hz, is supposed to decrease gently as the frequency goes up, as expected in any room.
Oops ! My correction obviously created some problems that were not there to begin with. Let's compare this curve with the correction itself :
The part around 500 Hz is difficult to interpret. On one hand, it looks like the correction is too strong, but on the other hand, this frequency range is supposed to by higher than the treble frequencies, measured from the listening position. Let's put it aside. The real problem is obvious : from 4000 Hz and up, the correction is useless. The original curve was neutral and didn't need any correction.
Let's drop this part from the eq and keep only the useful corrections, up to 4000 Hz.
Making a "half correction" is something very difficult, because there is no reference level for the part that is not corrected. In my first try, I just naively drew a straight line from 4000 Hz upwards. But the result didn't sound good. The treble were muffled. I realized that I had not calculated any target level for the part that is above 4000 Hz.
So I went back to Amir's graph, and tried to evaluate the average level measured above 4000 Hz. I restricted my reading to the 5000 to 15000 Hz range, ignoring what's happening above. I tried to draw a straight line so that half of the curve is above, and half is below. Then I noticed that this level was exactly the same as the 150 Hz peak. That would be my reference point.
Reading now my REW correction overlaid on Amir's graph, I noticed that it was accurately following it at 150 Hz. I therefore set the shelf level of my correction above 4000 Hz exactly identical to my 150 Hz peak (I mean 150 dip, since the curve is displayed upside down).
I generated the impulse response of the new correction and loaded it in Foobar's convolver and... wow ! Now, that sounds really great !!!
Here is how it measures from the listening position :
When I say it sounds great, I'm talking about medium and high frequencies. The low frequencies need re-equalization, as the bass reflex of the JBL is on the back of the speaker, and the MiniDSP room correction (still active from 35 to 600 Hz) was setup for the Neumann.
The measurement shows yet a little problem around 1000 Hz that was not there before correction. Here is the response without eq, the final correction, and REW's prediction of the result.
We can see that the level is unduly decreasing from 1000 to 1400 Hz, while it was not the case before correction.
Comparing with the first picture with Amir's graph, we can see that the correction is supposed to be ok. Since I'm not sure what's happening here, I didn't try any other adjustment.
The forum doesn't allow audio nor binary attachments, so here are links to my correction :
44100 Hz impulse response for convolvers : JBL 305P mk2 equalization.wav
Eq filters for REW : JBL 305P mkII eq filters.req
Important :
The correction is valid for JBL 305P mkII speakers, not for regular LSR 305 speakers.
The impulse response works at 44100 Hz only. The convolver must be preceded by a 44100 Hz resampler, or have a built-in resampler. Otherwise, if a file with a sampling frequency different from 44100 Hz is played, the correction will be wrong.