Troubleshooting level-dependant frequency response in loudspeakers

[a.palfreyman]

1/4 wavelength at 1.5kHz is about 60mm which will be the first node in the standing wave in front of the driver firing at an obstacle.

 

[ToTo Man]

1/4 wavelength at 1.5kHz is about 60mm which will be the first node in the standing wave in front of the driver firing at an obstacle.

The drivers are currently 100mm apart, the grille frame edges are 80mm apart.

 

[a.palfreyman]

OK, if you think about it, that is almost 1/2 wavelength apart which puts them in phase, so you create a standing wave with the peak midway between the drivers which squirts out of the gap. Maybe dangle a hand towel in the gap?

 

[a.palfreyman]

I bet there are a few acoustcians going

 

[davidsrsb]

The level dependence could be an off-centre voice coil rubbing in the gap.

 

[ToTo Man]

Running them in for a week hasn't fixed the issue so I ran a suite of FR measurements on them today to see what's going on.

I think the issue is a faulty tweeter (Celestion HF1300) in the right speaker as there are elevated levels of harmonic distortion between 2kHz-7kHz and the FR is very choppy when measured at lower SPLs. If I measure 20dB louder, the FR gets closer to that of the left speaker but the measured distortion is still much higher.

The left speaker also has elevated levels harmonic distortion in the upper mids but it isn't as bad as the right speaker:

This is how the left and right speakers compare when measured loudly (1 watt) in the middle of my room at 1 metre distance. I find it interesting that the (faulty) HF1300 unit in the right speaker is louder than the left in all measurements apart from the B110 axis measurement:

This is how each driver measures with the mic placed just 5mm away and the volume turned down by -20dB:

 

[ToTo Man]

Does anyone know what orders of crossover filters are used in the IMF TLS 80 / Professional Monitor mk3 / Reference Standard Professional Monitor mk4? The tweeter is supposed to cross to the supertweeter at 13kHz but, as you can see in the above measurements, the HF1300 runs all the way out to 13kHz but the supertweeter comes in much earlier than this so there must be significant overlap between the two?

 

[davidsrsb]

I have never seen a supertweeter with anything more than a very basic low order crossover. Part of the marketing as with 4th order, very little music would come out of it in the vinyl and radio source days.

 

[ToTo Man]

I have never seen a supertweeter with anything more than a very basic low order crossover. Part of the marketing as with 4th order, very little music would come out of it in the vinyl and radio source days.

Can you infer from the above measurements what the order and turnover frequency is? When I selected HF2000 units for my Ditton 66 restoration they all exhibited higher output below 5kHz than above 5kHz so I'm wondering how IMF managed to achieve such a steep roll-off below 5kHz using only a low-order slope? I thought perhaps the steep slope might be result of cancellation from the HF1300 and/or B110 (when the mic is 5mm away from the HF2000 it is 95mm away from the HF1300 and 220mm away from the B110). I therefore repeated the measurements with the HF1300 and B110 covered with pieces of thick foam, but it didn't make much difference:

On a side note, I located measurements I took in 2019 of some spare HF1300 units I have. I measured these without a crossover attached. Many of them exhibit a resonance at 22kHz that's -10dB down in level from their 3kHz-13kHz output. The measurements in post #26 also show a resonance at 22kHz at a similar if not higher level, suggesting there is very little, if any, low-pass filter being applied to the HF1300 by the IMF crossover?

 

[davidsrsb]

Can you infer from the above measurements what the order and turnover frequency is?

Not easily. An impedance curve would reveal more, REW is designed to make that easy to measure.
Photos of the crossover, both sides would allow me to do some reverse engineering

Often supertweeters have nothing more complicated than a series capacitor