Again, it is not strictly necessary, only convenient (or when you need the actual group delay values, without any offset - seldom needed).
Last edited:
This the phase data contained in the ABEC output.
it looks like that is the minimal phase already, and unwrapped.
I checked with the Hilbert transform and it seems correct.
it looks like that is the minimal phase already, and unwrapped.
I checked with the Hilbert transform and it seems correct.
This leaves us with relative acoustic offset, then? Which would be solved when a woofer is co-located in the same box. However, if everything that is needed is the delay, extracting this as a number could be helpful to use it with the ability of VCad to create driver data from the SPL chart of actual woofers.
Typically we don't want minimum phase data but a real phase including all the propagation delays relative to some fixed (but otherwise more or less arbitrary) point in time. Such data you can directly feed into a crossover simulator and be sure it's correct, without thinking about anything else. For example in vertical polars of a TM setup, as you go up the tweeter becomes closer to the microphone - that's what you want to be captured in the phase data automatically. Minimum phase won't provide you that, so you then need to fiddle with the relative offsets, etc. - something I always try to avoid. It's much easier to use the real phases all along.
Last edited:
Kartetian cms50 vs BMS 4550 (twice the price...)
30 dB less 2nd, equal 3rd between 4-6k else better, no data for higher orders on the cms50...
Viva la france 🙂
Exit angle?
//
30 dB less 2nd, equal 3rd between 4-6k else better, no data for higher orders on the cms50...
Viva la france 🙂
Exit angle?
//
The experience I know of with the Kartesian drivers was very bad, they returned the drivers as unusable (a QC issue?). Perhaps they got better since, I wouldn't risk it. They looked great on paper but the reality was a disaster.
Okey - very good to know - thanks! Sorry for littering the thread with driver stuff... but this payed off 🙂 - will stop now...
I suppose we should have heard more about ot if it really was as good as advertised...
//
I suppose we should have heard more about ot if it really was as good as advertised...
//
Last edited:
No, three times the excursion, 0.1mm vs 0.3mm.
At lower frequencies, the limitation to any high frequency compression driver is excursion, rather than power.
Any HF compression driver output basically becomes unusable at the excursion at which the diaphragm contacts the phase plug- "clickity-clack", no distortion measurements are needed to detect that sound 😉
If the diaphragm suspension somehow limits excursion without contact, a hard limit in SPL would occur at Xmax.
The magnetic gap height is 3.5mm for the DFM-2544R00-08, and 3mm for the DFM-2535R00-08, so we can assume the excursion difference at Xmax simply is the diaphragm to phase plug distance.
Assuming the horn gain is the same for each driver, the difference in output between the two drivers at 800 Hz would be around 13dB.
Not sure what driver model Mabat used in this simulation, but output was 110 dB at 800Hz at 0.3mm excursion:
If the driver used in the above simulation was a 1.75" diaphragm, you could expect the 1.4" 0.1mm driver to "crap out" at around 97dB at 800 Hz on the ST 260 horn, while a 12" loudspeaker could do 119 dB at 0.1mm.
Art
Adding 6 dB at crossover frequency 800 Hz, when the crossover is an LR4-filter. That would make it a hypothetical maximum of 103 dB/1 m. I don’t know if my microphone calibration is really right (set the level via the voltage method), but with peaks, I could register >100 dB at 2.7 m/LP. Admittedly levels where higher than usual, but still possible. So this sounds a but constricted (if the SPL read out was correct).
Thank you!
Someone could try to make a lumped element model of the Peerless(es). Then it's possible to generate the above plots and see how capable they really are with various horns.
This was the LE model used in the above example (I think it's still the one bmc0 made):
This was the LE model used in the above example (I think it's still the one bmc0 made):
Code:
Def_Driver 'Drv1'
dD=44mm
Mms=0.8g
Cms=30e-6m/N
Rms=3.0Ns/m
Bl=7.5Tm
Re=6.3ohm
fre=35kHz ExpoRe=1
Le=0.1mH ExpoLe=0.618
System 'S1'
Driver 'D1' Def='Drv1' Node=1=0=10=20
// Rear volume
Enclosure 'Eb' Node=20
Vb=50cm3 Qb/fo=0.1
// Front volume
Duct 'D1' Node=10=200
dD=44mm Len=0.5mm
// Phase plug (simplification)
Waveguide 'W1' Node=200=300
STh=1.52cm2 dMo=20mm Len=22mm Conical
// Conical section between phase plug and exit
Waveguide 'W2' Node=300=400
dTh=20mm dMo=25.4mm Len=22mm Conical
RadImp 'Throat' Node=400 DrvGroup=1001
Last edited:
The low driver will continue getting louder at 800 Hz, while the HF driver won't.
Since Mabat just confirmed the dD=44mm (1.75") was used in the ST260 sim in post 9828, the 110dB low frequency output potential of the DFM-2544R00-08 should be similar, though ending at 0.3mm rather than 0.5mm. The 35mm (1.4") DFM-2535R00-08 should be around 13dB less before hitting it's 0.1mm excursion limit.
Still curious about whether these drivers suspension limits excursion before the diaphragm hits the phase plug.
Using REW's CEA-210 Burst tone generator, and the LZ peak logger, you could easily see and hear where "the end of the road" is in the bottom end of your horn/driver without burning it up.
You may be surprised how the burst sounds if you have only used the sine and pink noise functions.
Well, as I wrote, the question is a bit if the SPL calibration of the microphone is correct, but I will give it a try in a weak, when I can measure again and share my results.
In ABEC, solve but ONLY for meshing.
Do that, and tell us what the results are.
You only have 16GB and if the mesh is too fine, it will never finish.
It is not a bug, per se, but I found that ath 4.3 beta will not finish if the gap between the waveguide and enclosure walls is too small.
For instance, when I set the gap to 5mm it wouldn't finish.
I believe this is because the resolution is too coarse to accommodate such a small gap.
I increased it to 20mm and everything worked fine.
For instance, when I set the gap to 5mm it wouldn't finish.
I believe this is because the resolution is too coarse to accommodate such a small gap.
I increased it to 20mm and everything worked fine.
Removing the mic to speaker propagation delay to remove unnecessary phase wrapping but maintaining the relative offsets is ideal.
This is true, I wonder though if the simulated woofer will give an accurate phase or if it will just be some other version of not quite right but closer to reality, much like minimum phase with an estimated z offset.
The problem for me and anyone else using VACS text export is that it seems to like to wrap the phase and export it that way, even though the data underneath is complex and can be represented as continuous phase, I have struggled to get it out even when generating a continuous phase plot and exporting it with the preserve continuous phase option. But this is most likely just me not knowing the right buttons to press.
If you or maiky76 use your scripts to extract the data directly from the spectrum file instead of going through VACS, that problem is avoided.
In the mean time getting minimum phase in Vituix was actually very easy.
Select all the frequency response files for the driver and tick the Minimum phase box, click OK and the wrapping is gone.
vs
I don't share your concerns, or maybe don't see why you are concerned. Phase estimation errors typically appear in situations where you don't have a complete information (like data for a limited frequency band only). I don't think this is the case.
There's nothing wrong with phase wrapping. All the complex numbers used in the calculations are completely blind to this - it makes absolutely no difference if the phase is 0° or 3600° in the input data - both give the same complex number.
If the second plot was a correct measured response, i.e. including the actual delay relative to the selected reference, it would be absolutely correct data then. There's nothing wrong about that. A lot of delay causes of lot of wrapping (=phase rotations) but that's how it is. It seems to be a mess only because the programs connect the data points with lines - we know the lines are not correct (because there is not enough data points to capture that), but the points are (or should be).
There's nothing wrong with phase wrapping. All the complex numbers used in the calculations are completely blind to this - it makes absolutely no difference if the phase is 0° or 3600° in the input data - both give the same complex number.
If the second plot was a correct measured response, i.e. including the actual delay relative to the selected reference, it would be absolutely correct data then. There's nothing wrong about that. A lot of delay causes of lot of wrapping (=phase rotations) but that's how it is. It seems to be a mess only because the programs connect the data points with lines - we know the lines are not correct (because there is not enough data points to capture that), but the points are (or should be).
Last edited:
No, that's the trouble. Complex numbers per se can't represent an unwrapped phase - it's always inherently wrapped (0° - 360°, that's all you have). So whenever you need it unwrapped (like e.g. for numerical group delay calculations) you have to unwrap it somehow - not always easy. But the wrapped form is still the absolutely correct one for all the other calculations that work per each single frequency point, like a summed frequency response.
Last edited:
Is there, as for now, a way to set up parameters both in ABEC and in VCad, so to have correct phase information out of the, seemingly already available, minimum phase data?
For example, setting distance parameter in ABEC/ath script to 1m, which is also the standard distance for driver measurements, and the microphone position for the woofer in VCad diffraction tool at woofer level, while the position in VCAS is at waveguide level. Would we then, by dialing in the respective offsets in crossover tool, that were used for waveguide in ath and for woofer in diffraction tool, have legitimate information?
For example, setting distance parameter in ABEC/ath script to 1m, which is also the standard distance for driver measurements, and the microphone position for the woofer in VCad diffraction tool at woofer level, while the position in VCAS is at waveguide level. Would we then, by dialing in the respective offsets in crossover tool, that were used for waveguide in ath and for woofer in diffraction tool, have legitimate information?
It may be correct but I don't find it easy to work with wrapped all over the place. It is recommended in Vituix's preparation instructions to remove the measurement delay and have the tweeter phase be more similar to minimum phase at the top end, this makes sense to me.
Continuous Phase Technique
In VACS complex data are either stored as real and imaginary or as amplitude and continuous phase, where the latter is the polar notation of a complex-valued function with the special feature that the phase is continuous, i.e. the phase does not jump by 360 degree. This means it is un-wrapped. In this way it is possible to perform complex interpolation, smoothing, re-sampling and forming mean-values without the usual interference problems.
All that's necessary is that we know what we are doing and why. It just needs to be correct in the end and there are many ways how to do that.
OK but that doesn't work by itself and within the field of complex numbers. For example, how do you add those numbers and retain the continuous phase? You just have to keep a side track of the phases, that's of course possible. But not within complex numbers alone.
Last edited: