JBL 306P MK II Analysis

[roci_big_ear]

Amir measured a dead spot at the cross over frequency right above the tweeter axis. This is readily apparent when playing a ~1.5 kHz tone and results in a musical reproduction that lacks presence: subjectively sound comes from around the speaker, not coherently from within the speaker. Generally, I listen to speakers slightly above the tweeter axis, so am I supposed to put the speaker upside down ?!

A / B testing with my older LSR305 makes the latter preferred since it provides more robust and confident sound.

I was determined to figure out why this is. I suspected it is driver phase misalignment.

Physically, the speaker has the tweeter and the base ring of the woofer exactly aligned, at 3 cm from the front plane. I thought the mechanical designers did a remarkable job:

Internally the speaker has 2 bridged class D power amplifiers. I measured the output of the half bridges of the woofer and tweeter. At the cross-over frequency of 1.425 kHz, the electrical waveforms need to be exactly aligned for constructive interference in front of the speaker.

Sadly, this is not the case:

The woofer is lagging the tweeter by 128 micro seconds at 1.425 kHz. Or in other words, the tweeter needs to be 128 us * 340 m/s = 4.4 cm behind the woofer for sound to constructively add.

In another thread on ASR, I fixed the issues with the Behringer B2031A and made it sound good. However, sadly the 306P uses a DSP and the program / filter coefficients are stored in flash, making it unsuitable for DIY without very significant reverse engineering and complete firmware rewrite. What I will try doing is to 3D print a bracket to shift the tweeter back by 4 cm. Hopefully, this will not affect the wave guide functionality by much. Let me know in the comments if you think this can work please.

Since, I had the speaker opened, I thought it will be interesting to measure the pre-equalization the DSP does. Here is the output of the amps:

In purple, I doubled the tweeter output, to compensate for the different driver sensitivity - with that the cross over point is at about 1.425 kHz, like the JBL spec.

Let's compare the above to the speaker response Amir measured:

We can see that the bump at 55 Hz was needed, to keep the bass flat. However, the bump at 1.1 kHz is questionable. If they had proper phase alignment between the drivers, it likely would have been unnecessary.

 

[roci_big_ear]

Fun fact: the datasheet specs the flash retention time at 20 years. Don't plan on leaving these kind of electronics for your grand kids, folks!

 

[Sancus]

I mean all normal layout 2-way speakers have vertical nulls, some worse than others. But I don't know of any that don't have one, I think the best case scenario is the Neumann KH80. But that has a very small woofer so directivity matching is easier.

To me the solution is to buy a good coaxial. I also find them annoying so that's the solution I went with.

 

[fluid]

I was determined to figure out why this is. I suspected it is driver phase misalignment.

Phase misalignment is the root cause of the null, but in speakers with vertically separated drivers they are almost inevitable. This is generally referred to as lobing.

The shape of the directivity can be changed and the direction the main lobe points in can be changed through different crossover settings this diagram shows something similar to what you want to do.

You cannot put the tweeter back 4cm with a 3D printed adapter without a having a pretty serious effect on the waveguide response.

What you can do if you have some form of processing available on the input is to use an allpass filter or FIR equivalent to phase delay the tweeter frequencies to get the same effect and tilt the lobe up. It won't remove the null but it should move it to somewhere that doesn't bother you so much.

Analog active version from Linkwitz
https://www.linkwitzlab.com/filters.htm#4

Rephase can create all pass filter as well as "phase Equalization" through a bank of filters, you could use this to test (if you have a way of including convolution filters in your audio path)
https://rephase.org/

 

[roci_big_ear]

To me the solution is to buy a good coaxial.

Hmm, the tweeter waveguide changes shapes depending on what bass you play...

You cannot put the tweeter back 4cm with a 3D printed adapter without a having a pretty serious effect on the waveguide response.

It turns out, I can. The pipe had less than 2 dB impact on the off axis measurements across the entire response band. I suspect the difference is from not exactly the same positioning. I tried 30 degrees horizontally and 15 degrees vertically. The main issue with it is that it resonates a bit at ~14kHz since the diameter is one wavelength.

What you can do if you have some form of processing available on the input is to use an allpass filter or FIR equivalent to phase delay the tweeter frequencies to get the same effect and tilt the lobe up.

The processing needs to be past the cross-over, like your middle picture, and not in the common signal path like you describe.

 

[Vini darko]

Point the speaker up by a few degrees?

 

[roci_big_ear]

The main issue with offsetting the tweeter is that it fixed the cross over dip, but now I have +5 dB bump at 1.5-3.5 kHz.

 

[fluid]

It turns out, I can. The pipe had less than 2 dB impact on the off axis measurements across the entire response band. I suspect the difference is from not exactly the same positioning. I tried 30 degrees horizontally and 15 degrees vertically. The main issue with it is that it resonates a bit at ~14kHz since the diameter is one wavelength.

A picture and the measurements would be helpful in seeing what you did and how it affected the response.

The processing needs to be past the cross-over, like your middle picture, and not in the common signal path like you describe.

Certainly there is greater flexibility if it is done within the crossover branches but with the right filter the response can be changed to tilt the lobe up without it. The same way a crossover can be made linear phase with the right preconditioning filter.

 

[alex-z]

IMO, the time alignment mismatch is so slight that it makes more sense to just use a small amount of EQ to fill the 1500Hz dip. Any physical change you make will result in a compromise off-axis, you are just moving the response lobe around. And yes, of course you could just run the speakers upside down, there is no harm in that.

The bigger priority would be fixing the port and cabinet resonances. There is an especially prominent one from the port at 1100Hz. Maybe with some clever 3d printing and application of absorption material that could be avoided. Rebuilding the entire cabinet seems a waste on such cheap monitors.

 

[roci_big_ear]

Adapter
2.5 cm extra offset

Outside

Measurements
In room at approximately the same placement, but not exact:

The tube resonance is apparent at 14 kHz, thankfully not much of an issue for my older ears.

Factory pre-equalization is now becoming a problem after the drivers are phase-aligned better. +5-6 dB around 2kHz.

More, off axis:

Conclusion

The mod is a mixed bag. Definitely sound coming from the speaker is more coherent and authoritative. However, tube resonance, bump around 2 kHz and the worse waveguide are downsides. Thankfully, now that 2kHz bump can be pre-equalized nicely, unlike before.

Overall, I summarize the exploration as interesting, but likely not worth it: 2 steps forward, 1 step backwards. Time spent: 2 days.

 

[YSC]

Adapter
2.5 cm extra offset
View attachment 222476

Outside
View attachment 222477

Measurements
In room at approximately the same placement, but not exact:

View attachment 222478
The tube resonance is apparent at 14 kHz, thankfully not much of an issue for my older ears.

Factory pre-equalization is now becoming a problem after the drivers are phase-aligned better. +5-6 dB around 2kHz.

More, off axis:

View attachment 222479View attachment 222480

Conclusion

The mod is a mixed bag. Definitely sound coming from the speaker is more coherent and authoritative. However, tube resonance, bump around 2 kHz and the worse waveguide are downsides. Thankfully, now that 2kHz bump can be pre-equalized nicely, unlike before.

Overall, I summarize the exploration as interesting, but likely not worth it: 2 steps forward, 1 step backwards. Time spent: 2 days.

I feel that it actually make things worse than that.. the messed up waveguide would likely mess up directivity also, which is a bad thing for even slightly off axis listening (like any practical situation).

 

[Eetu]

I say 1) forget about it, start from scratch and go all DIY. The flimsy front baffle and cheap cabinet is the 30X series' weak point. And once you have 4(+) channels of DSP and amplification you can do anything you want.

Or 2) undo the mod and raise the speaker a few cm's higher so you avoid the lobing at your listening position.

The folks at JBL have spent a lot of time perfecting the waveguide and as it is with the mod you have made the performance significantly worse.

Coaxials like the better KEFs and Genelecs are end-game stuff. The waveguide-modulation is not an issue if the speaker's a 3-way. Probably not an issue either if high-passed and used with a sub..

 

[Sancus]

Hmm, the tweeter waveguide changes shapes depending on what bass you play...

This doesn't really matter in a 3-way. Yes, it is relevant in a 2-way, though Kef has done a good job of mitigating it in the LS50(the tweeter waveguide has a significant static center portion). Personally the only manufacturers I would look at for coaxials are Genelec, ME Geithain(availability outside Germany sucks), Kef, and Kali in that order.

 

[ernestcarl]

I'm a little bit more open-minded when it comes to coaxes...

BTW, don't forget, Amir liked the small, cheapo Fluid FX50:

Caveat emptor: my somewhat detailed FIR correction example above is unit specific so some minor adjustments with non-perfectly matching units is to be expected if one wants to play with EQ above transition zone.

 

[Mario Sanchez]

I am fairly certain that the tweeter filter is designed to work in conjunction with the "built in" 2nd-order rolloff of the tweeter driver in order to sum to a 4th-order LR highpass on that driver, just as the DSP'ed 3rd order highpass and protective cap (~1st order) on the M2 system sum, with the built in response of the compression driver + horn, to a 6th order crossover with the woofer. I have a feeling that if you factor that extra rolloff from the tweeter into the output signal at crossover, the two drivers' output should be of proper alignment...