Ok, so the "visual wrap" is visible when phase is bigger than 180deg or smaller then -180deg, because the graph is 180/ -180deg. Thanks for making it clearer.
This is program is awesome. The idea of measuring drivers independently on their axis is great. It makes it easier to design RLC circuits and makes it possible to see how a driver would behave if you move it in Z or Y axis with a filter on. As I see it now, the box should be tilted not 3 but 4,5 degrees, because woofer would be not 19mm but 29mm in front of the midwoofer. Those 10mm would make it much easier to integrate phase correctly.
Kimmo you are genius. Thank you.
Kimmo you are genius. Thank you.
Yes exactly. I just changed some values of RLC circuit in the bass driver around it's resonant frequency.
Last edited:
Here is good example about tiny variation in phase response causing dramatic phase wrapping. It's Epe-3W demo with minimum phase FIR crossover calculated with 1/3 oct. smoothing and plotted without smoothing, all drivers inverted and not inverted. Nothing really wrong so it's just visual harm due to continuous phase response trace wrapped within 360 deg.
Attachments
Just to clarify, there is no real need to extend the scale range up to positive 1800 degrees.
To illustrate, the first attachment shows a standard wrapped phase response. The second attachment shows the same response with the phase unwrapped.
Attachments
is it possible to remove the function of a protective cap when import measurements?
say i have a ribbon tweeter that is measured with a 47uf protective cap and i want to use this response curve to simulate a crossover, but i want the function of the cap removed, how is this done? is there a way to get an inverse function from a component in vituixcad?
say i have a ribbon tweeter that is measured with a 47uf protective cap and i want to use this response curve to simulate a crossover, but i want the function of the cap removed, how is this done? is there a way to get an inverse function from a component in vituixcad?
^Option 1 is to measure with ARTA, dual channel connection and mode. Reference channel connected between cap and driver, or between line level high pass filter and input of power amp. See measurement with ARTA pdf.
Option 2 is to measure with SoundEasy using MLS and pre-emphasis. Protection cap might not be needed, depending on how power amp reacts with close to 0 Ohms impedance at LF.
Option 3 is to compensate error in response measurement. Needs electrical response measurement from driver terminals, and flat response / measured response with Calculator.
Option 2 is to measure with SoundEasy using MLS and pre-emphasis. Protection cap might not be needed, depending on how power amp reacts with close to 0 Ohms impedance at LF.
Option 3 is to compensate error in response measurement. Needs electrical response measurement from driver terminals, and flat response / measured response with Calculator.
Member
Joined 2003
Option 2 pre-emphasis in SoundEasy is a great option to remove low frequency information from the measurement signal without affecting the measured result, however if you have a true ribbon a capacitor will still be necessary, as the ribbon is transformer coupled and will present a near dead short to your amplifier without a capacitor in place.
For SoundEasy and use with a true ribbon, the Option 1 procedure will work the same.
For SoundEasy and use with a true ribbon, the Option 1 procedure will work the same.
Hello colleagues! I ask for help! I'll tell you step by step.
I take driver measurements on-axis and off-axis + -90 degrees using a two-channel analog microphone measurement system.
Next, I extract the minimum phase from the axial response.
Question! What should I do with the off-axis response data? Is it the same to highlight the minimum phase? Thank you!
Oleg
I take driver measurements on-axis and off-axis + -90 degrees using a two-channel analog microphone measurement system.
Next, I extract the minimum phase from the axial response.
Question! What should I do with the off-axis response data? Is it the same to highlight the minimum phase? Thank you!
Oleg
^You should not extract minimum phase! That destroys timing differences between drivers and between off-axis angles in measurement data by removing excess delay from all measurements. Removing excess delay is based on assumptions about response extensions and radiator geometry which are not facts by default.
Some amplifiers can handle true ribbons with pre-emphasized (1st order HP) signal, but signal level is better to be much lower than standard 2.83 Vrms. For example 0.5 V with 4 kHz high pass. This is difficult to know without testing.
One possible problem with capacitor without shunt resistor is impedance peak at resonance.
Another passive option is 1-2 Ohms series resistor, but also that requires low level, pre-emphasis and MLS or white noise to limit power and excursion at LF.
I would probably use active high-pass filter or some other line level HP and semi-dual connection (reference connected between filter and power amp) and amplifier which can handle low impedance if current stays low enough. That creates ideal slope no matter Z peak at resonance, and allows also any greater than 1st order high pass. Both ARTA and SoundEasy supports this.
Noise at LF is one problem with high-passed signal, but fortunately that does not matter.
2.0.58.3 (2020-12-16)
Enclosure
* End correction list boxes replaced with text boxes to allow any value within 0.6 and 2.4.
Note! Open, modify and save old/existing enclosure projects to save end correction factor also to project files (vxe).
Enclosure
* End correction list boxes replaced with text boxes to allow any value within 0.6 and 2.4.
Note! Open, modify and save old/existing enclosure projects to save end correction factor also to project files (vxe).
***Thanks for the answer! I didn't quite understand you! Not to extract minimum phase only on off-axis responses? Or at all?
^See instructions: Measurement with ARTA or Measurement with REW or Measurement with SoundEasy.
Recommended measurement and design procedures do not contain a word about minimum phase extraction. It's valuable if phase response is unknown or user has screwed up measurements so badly that "some" phase should be generated mathematically.
Recommended measurement and design procedures do not contain a word about minimum phase extraction. It's valuable if phase response is unknown or user has screwed up measurements so badly that "some" phase should be generated mathematically.
Last edited:
***Thank you! I'll see! However, I am completely confused! Why don't we need a minimum phase? I am building a two-way loudspeaker! I get frequency responses from two microphone positions in the finished box and I want to keep real off-axis responses under control, so I measure them, not simulate them! Don't understand what I am doing wrong?
^VituixCAD is simulation program, and everything I have published for measurements is for producing measurement data for simulation.
Phase does not matter i.e. can be anything or totally missing if you use SPL (phase hidden), Power & DI and Directivity charts of the main program for QC measurements of complete speaker with crossover.
BUT if you like to see actual phase and actual group delay in GD & Phase chart, you must use actual measured phase. Not minimum phase extraction because it destroys phase response of multi-way with normal IIR (not linear-phase FIR) crossover.
Phase does not matter i.e. can be anything or totally missing if you use SPL (phase hidden), Power & DI and Directivity charts of the main program for QC measurements of complete speaker with crossover.
BUT if you like to see actual phase and actual group delay in GD & Phase chart, you must use actual measured phase. Not minimum phase extraction because it destroys phase response of multi-way with normal IIR (not linear-phase FIR) crossover.
^You can also play without active TF block by processing acoustic measurements with Calculator tool. Divide measured acoustical responses (A) by electrical transfer function at driver terminals (B). Response B should be scaled so that gain=0 dB at HF. Result is response without series capacitor (and possible shunt resistor).
Impedance response should be measured without extra components of course.
Impedance response should be measured without extra components of course.
This is the approach I took (protection cap, about 3uF from memory) with ARTA in full dual channel mode with the 2nd reference channel connected to the driver after the cap.
In theory this will cancel out the roll off of the capacitor. However there is a problem with this method - ARTA does not seem to be able to cope with too much attenuation of the "reference" signal at low frequencies relative to high frequencies. It is expecting the reference signal to be approximately flat.
Once it is more than about 20dB down at low frequencies vs high frequencies the SNR of the measurement at low frequencies will be poor and this causes random DC offset in the measured impulse and therefore phase errors in the frequency response derived that make the measurements unusable for absolute phase.
I actually raised this issue in the ARTA thread with Ivo (you might remember reading the discussion) but the end result was Ivo's advice to "don't do that" and that it's expected to happen if the low frequencies in the reference channel are attenuated too much.
I actually worked around it by just taking many measurements one after another until I got one where there was no DC offset in the impulse (each measurement would have a different random DC offset) and this seemed to give correct phase results. (Which is what makes me think this should be fixable in software and that it is a deficiency in the processing in ARTA not a fundamental limitation of the technique)
If I was doing the measurements again now I think I would not use the cap method but instead use noise as the excitation signal because you can set a roll off frequency for the noise in the generator window in ARTA - just set this to about 3000Hz and it will roll off the low frequency content of the noise over a few octaves below this so that the ribbon doesn't experience excursion, but without causing the impulse DC offset problems I experienced with the cap method. I did a quick test like this and it seemed to give good results.
Some care needs to be taken though:
1) Don't switch on the amplifier with the tweeter directly connected... connect it just before the measurement.
2) Make sure your computer can't produce any other sounds like windows alert sounds that could send low frequencies to the tweeter - best to enable exclusive mode and/or use a separate USB sound device for ARTA which is not the default windows sound device.
3) Start with the volume turned down and slowly turn it up during a test run to find a suitable level that is loud enough but doesn't cause the tweeter any distress.
4) Disconnect the tweeter after the measurement and certainly before turning off the amplifier or doing anything else that might generate a noise or a pop.
Last edited: