Sallen Key distortion? - simulated vs measured

[HiFiNutNut]

Hi,

I have measured, designed and built an active crossover network for my DIY active loudspeakers.

The measured results deviate from the simulated results. First please have a look at the simulations and the actual measured results.

Regards,
Bill


First, simulated high pass filter for the tweeter circuit.

[HiFiNutNut]

This is the measured result for the tweeter circuit.

[HiFiNutNut]

Simulated midrange bandpass filter.

[HiFiNutNut]

Measured midrange response.

[HiFiNutNut]

Simulated low pass filter.

[HiFiNutNut]

Measured woofer response.

[HiFiNutNut]

The measurements were made using SpeakerWorkshop, i.e. they are acoustic measurements. The microphone was not in the exact location as when the first raw driver measurements were made, so some small differences can be observed.

I understand that those are not high end equipments and the noise may be on the high side, but do not expect the noise to be as high as what you see from the graphs.

Especially worrying, the woofer response is far from what is simulated.

I also understand that ideal opamps were used in the simulations. I added a 75R to the output of the opamp to simulate output impedance and only see deviations from 200kHz, and it is -82dB, rather than from 400Hz -25dB.

I have been using opa627. For the woofer circuit, I replaced the opa627 with LM4562 with absolutely no measurable differences.

The input impedance on the positive side and negative side are not matched. But I do not expect they would cause the deviations as seen in the graphs.

What has gone wrong?

[HiFiNutNut]

The summed response does match the simulated summed response fairly well.

Please help me fix these problems.

[picowallspeaker]

Hi Bill ! It is funny to see how the drivers try to follow the ideal curve ,but they fail !!
Each on its own ,they add peaks and dips ...and those are only the frequency graphs .
My suggestion is to look at the system as a whole : you didn't tell us which amplifiers and which drivers in which enclosures you were using for the test ,which is meaningful by itself and appreciable ...but in a active multi-way system you can't blame only to the crossover .

[HiFiNutNut]

Originally Posted by picowallspeaker

Hi Bill ! It is funny to see how the drivers try to follow the ideal curve ,but they fail !!
Each on its own ,they add peaks and dips ...and those are only the frequency graphs .
My suggestion is to look at the system as a whole : you didn't tell us which amplifiers and which drivers in which enclosures you were using for the test ,which is meaningful by itself and appreciable ...but in a active multi-way system you can't blame only to the crossover .

The peaks and dips are due to comb filters of boundary effect as I was using a 44ms gate in the "in-position" measurements. However, the absolute values are are unimportant, the comparison values are.

All power amps use identical circuit boards and are constructed in the same way with response well into 50k and above with little errors within the audioband.

The devils must be in the line level crossover circuits posted in above.

1) The tweeter response rises at the stop band. This looks like a measurement artifact to me.

2) The midrange also rises at the stop band on the low end. At -30dB it does not look like noise but may also be measurement artifact.

3) The woofer circuit obviously has errors. The problem does not appear to be noise because on the midrange graphs the stop band response at the top end works as simulated right down to -50dB and below. Earlier I found this problem and redesigned and rebuilt the circuit board. But I got the same measurement result so it is not possible to have wrong connection or cold solder joints, etc. I also compared the left channel to the right channel and they showed the same things.

In summary, I guess the high pass circuits may not have a problem. The problems shown with them would probably be measurement artifacts, due to noise or measurement equipment errors. However, the low pass section of the woofer circuit looks very suspicious.

Does Shallen key have problems with low pass filters rather than high pass filters?