If after you took the second set (post crossover, shown in orange), you then measured only the tweeter part and focussed on that, then you removed the crossover without touching the mic or the measurement setup and straight away measured the tweeter again, in theory it should show the blue response... but if it doesn't...
This suggests either,,,
1. Your pressure measurement came in different the second time (it can be challenging to get them identical between sessions), or...
2. You measured the impedance wrong in which case the uncrossed measurements would be near the same, only the crossed measurements would be different.
This suggests either,,,
1. Your pressure measurement came in different the second time (it can be challenging to get them identical between sessions), or...
2. You measured the impedance wrong in which case the uncrossed measurements would be near the same, only the crossed measurements would be different.
Your ND25 zma file in post #1 still contains FRD data. I don't think it is a post cache issue at my end. Can you please check?
Bogus ZMA data when I import it is creating a 4KHz peak, not there when removed (e.g. and VCad is using a flat 8 ohm impedance default)
Bogus ZMA data when I import it is creating a 4KHz peak, not there when removed (e.g. and VCad is using a flat 8 ohm impedance default)
Today I repeated all measurements. This time I started by comparing the amp I used before, a cheap TDA7498E board I got fro aliexpress, with my Denon DRA700AE amp. This is the result of the finished speaker measurement:
(Because the measurement is gated, with a 4.5ms window, the response below 222Hz is removed.)
Quite a lot of difference. I didn't use the Denon amp before because it's much more bulky and harder to move around. I expect the TDA7498E amp is sensitive to varying impedance loads, which could explain the (rather puzzling) difference between modelled and the real-world crossover I saw before.
Next, I measured the response of the individual drivers, and both drivers in parallel (without crossover). Using the latter as an overlay in VituixCAD, I noticed I indeed did have to modify the Z offset (22mm) to get the modelled response of both drivers (w/o XO) to overlap with the measured response. I didn't expect this, as I use a loopback for timing reference in REW, from which REW does calculate speaker-to-mic distance for all individual measurements.
Next I imported all the new Z and FR measurements in VituixCAD for the original crossover, and loaded the measured response of the finished speaker (incl. crossover) as an overlay. To my satisfaction, they overlap pretty nicely:
The orange line represents the measured response, the black one is the modelled response. Up high the measured response is about 1dB off, but that's not much of a concern to me. In conclusion I learned two things. The first is that a decent amp is needed if you want to make trustworthy measurements. The second is that Z-offset correction is still needed, even if a loopback reference is used in REW.
Thanks to everyone that offered tips and insights getting to the bottom of this!
(Because the measurement is gated, with a 4.5ms window, the response below 222Hz is removed.)
Quite a lot of difference. I didn't use the Denon amp before because it's much more bulky and harder to move around. I expect the TDA7498E amp is sensitive to varying impedance loads, which could explain the (rather puzzling) difference between modelled and the real-world crossover I saw before.
Next, I measured the response of the individual drivers, and both drivers in parallel (without crossover). Using the latter as an overlay in VituixCAD, I noticed I indeed did have to modify the Z offset (22mm) to get the modelled response of both drivers (w/o XO) to overlap with the measured response. I didn't expect this, as I use a loopback for timing reference in REW, from which REW does calculate speaker-to-mic distance for all individual measurements.
Next I imported all the new Z and FR measurements in VituixCAD for the original crossover, and loaded the measured response of the finished speaker (incl. crossover) as an overlay. To my satisfaction, they overlap pretty nicely:
The orange line represents the measured response, the black one is the modelled response. Up high the measured response is about 1dB off, but that's not much of a concern to me. In conclusion I learned two things. The first is that a decent amp is needed if you want to make trustworthy measurements. The second is that Z-offset correction is still needed, even if a loopback reference is used in REW.
Thanks to everyone that offered tips and insights getting to the bottom of this!
If our measurement technique is perfect, in theory, the measurements will fully capture both the time-of-flight and the phase of the drivers. No further z-offset will be needed... However, in practice there is almost always some errors that are introduced. I almost always need to empirically fine-tune the delay (or z-offset) to get simulation to match measurements.
Nicely done... you had several measurement issues to resolve, and now it all looks good !