Understood, and agree. Does your software let you create a target curve to work against? I'd expect you to spend considerable time behind the scenes developing such a curve representing what needs to be done to each driver. This curve, as you say, doesn't relate specifically to one axis. However if your particular method requires it to be translated to one axis you could do some curve arithmetic, but I wouldn't expect normally that you'd need to do that.
Also, you can cross correctly and still have a bad sounding speaker. Consider keeping your global equalisation separate from the crossover side as it so easily gets out of hand and mixed up.
It's not strange at all, for some reason every person who creates a digital parametric EQ decides to use a different definition of what Q does, some are constant, some are proportional, some are gain bandwidth dependent.
It is a total PITA matching them sometimes so I can understand using the Hypex designer software to avoid this.
I used the Hypex software to help someone out remotely and matched it in REW, I'll see if I can dig up the files to see what I did.
As I said above you can run some electrical transfer function measurements with different Q's and gain from Vituix and Hypex to see what needs to change.
One option is to create FIR filters in rephase of the different variants and then use those, as FIR filters are self describing and don't rely on variable definitions. Vituix can use them directly in the simulation process.
Actually the process does not take very long, although I do not have a lot of experience... I have done it just three times. The latest was with the Satori textreme drivers in a foam-board prototype box:
https://www.diyaudio.com/forums/multi-way/343831-sb-acoustics-textreme-62.html#post6431289
I was able to get a good working DSP 3 way filter set up in about 3 hours. I do not use the HFD software built in measurement system, I am more comfortable with Omnimic. I have Omnimic software and HFD running at the same time, and as I make filter adjustments, I get rapid feedback. This helps the process move along.
I expect to be going through this process on Saturday. I plan to take screen shots and photos, and to document my process and results to share. I do not consider myself an expert at this, but I know that some people have struggled with this aspect of active speaker setup, so if someone can learn from what I am doing, I will be pleased.
Well in one respect it is nice to know I was not going crazy... but it is a shame that such things are not more standardized.
I just got my idea for a cabinet and layout for drivers and filters, from Heissmann... already did the work - I just confirmed it by measuring almost exactly the same here at home with the same filter and some EQ.
But this VituixCAD... thought I'd give it a try. But even after building speaker and measuring for years... I cant seem to understand how to use the program. Any guides anywhere... I would really like to understand the bafflestep/edgecalculation/diffraction better.
But this VituixCAD... thought I'd give it a try. But even after building speaker and measuring for years... I cant seem to understand how to use the program. Any guides anywhere... I would really like to understand the bafflestep/edgecalculation/diffraction better.
Have you read through these
https://kimmosaunisto.net/Software/VituixCAD/VituixCAD_Measurement_Preparations_REW.pdf
VituixCAD help 2.0
and watched these ?
kimmosto - YouTube
https://kimmosaunisto.net/Software/VituixCAD/VituixCAD_Measurement_Preparations_REW.pdf
VituixCAD help 2.0
and watched these ?
kimmosto - YouTube
Not a true thread hijack, I have printed Kimmo's document so I can truly digested. I need to upgrade my measuring gear. I have looked at the Arta page and been searching ARTA threads, but all the recommended audio interfaces recommended are discontinued or there are complaints about them. So, what should I get? Maybe something in the $150.00 or less range. What are you guys using? Again, sorry for the hijack.
Last edited:
For cheap but good enough I would look at this
https://www.behringer.com/product.html?modelCode=P0BJZ
https://www.behringer.com/product.html?modelCode=P0BJZ
When hifijim says "at the same time" I believe he means he sets the filters and then makes a measurement to confirm the filter's effect. I've done the same with a mini-dsp and REW using a laptop. I can set up the mini-dsp and then make a measurement without having to move away from the laptop. Very conveinient!
Thanks ernperkins, I assumed as much. I suspect it may be showing a deficiency in the process somewhere.
As we've covered, you take anechoic polars and process them behind the scenes producing a result which is not directly related to any one axis. Does the software not display various curves, such as where you are and where you're going?
Remeasuring would undo all the good processing work and remove the target response, and potentially introduce new measurement errors and room aberrations.
As we've covered, you take anechoic polars and process them behind the scenes producing a result which is not directly related to any one axis. Does the software not display various curves, such as where you are and where you're going?
Remeasuring would undo all the good processing work and remove the target response, and potentially introduce new measurement errors and room aberrations.
Yes, as ernperkins says... I have both Omnimic software and Hypex HFD filter design software running on the laptop. The laptop is connected to the Hypex amp via USB cable, and the Omnimic is connected to the laptop with another USB cable.
I make a sweep, capture the FR curve, and assess it. Then I make an adjustment to the DSP filter, load it into the amp, and make another sweep. rinse... repeat... it is an iterative process, but with care it can be quite accurate.
This is basically the process recommended by Hypex, except that they want me to use their built-in measurement capability. I prefer to use Omnimic for several reasons. However, I might give the Hypex built-in process a try. I now own a Dayton EMM-6 microphone and a Behringer UMC202HD audio interface.
My expectation is that for each driver, the on-axis response and the listening window response will be nearly the same through each driver’s pass band. If it is not, then I have done something wrong in the design. Once I have confirmed this, there is no longer a need to measure the listening window response until final system testing.
Using the boost/cut filters and shelf filters (first and second order), I adjust the eq of each driver such that it is smooth and flat when viewed with 1/6 octave filtering. I need to be careful to distinguish between driver responses and diffraction responses. I do not correct for diffraction responses. The flat response needs to extend for at least one octave beyond the crossover frequencies.
After the drivers are flat for one octave beyond the crossover points, I set the levels of the three amps so that the drivers are at equal level. The next step is to measure the acoustic delay between the drivers and make that adjustment in the filter software. At this point the drivers have a flat response, have the same level (sensitivity), and due to digital delay, they all originate from the same plane. Since drivers are minimum phase devices, once they are eq'd flat, they have phase alignment. The final step is to apply crossover filtering. Because of the adjustments and manipulations, idealized crossover slopes can be simply applied.
Of course, it is possible to load all the raw driver measurements into a simulation program, such as the very powerful Vituixcad, and develop the DSP filtering algorithm by simulation. However, given the ease of developing the filters based on iteration and rapid feedback, this seems unnecessarily complicated. It is analogous to tuning a guitar. One way is to measure the actual frequency of the string and compare it to the desired frequency, then using physics and math, determine how many degrees the tuning peg must be rotated. The more simple approach is to just start turning the tuning peg until the digital tuning meter says we are in tune.
I use Vituixcad to assess my filters after the fact to make sure I have not done anything stupid. I expect close alignment between simulation and measurement. But the actual development is conducted based on iterative measurements and adjustments.
I make a sweep, capture the FR curve, and assess it. Then I make an adjustment to the DSP filter, load it into the amp, and make another sweep. rinse... repeat... it is an iterative process, but with care it can be quite accurate.
This is basically the process recommended by Hypex, except that they want me to use their built-in measurement capability. I prefer to use Omnimic for several reasons. However, I might give the Hypex built-in process a try. I now own a Dayton EMM-6 microphone and a Behringer UMC202HD audio interface.
My expectation is that for each driver, the on-axis response and the listening window response will be nearly the same through each driver’s pass band. If it is not, then I have done something wrong in the design. Once I have confirmed this, there is no longer a need to measure the listening window response until final system testing.
Using the boost/cut filters and shelf filters (first and second order), I adjust the eq of each driver such that it is smooth and flat when viewed with 1/6 octave filtering. I need to be careful to distinguish between driver responses and diffraction responses. I do not correct for diffraction responses. The flat response needs to extend for at least one octave beyond the crossover frequencies.
After the drivers are flat for one octave beyond the crossover points, I set the levels of the three amps so that the drivers are at equal level. The next step is to measure the acoustic delay between the drivers and make that adjustment in the filter software. At this point the drivers have a flat response, have the same level (sensitivity), and due to digital delay, they all originate from the same plane. Since drivers are minimum phase devices, once they are eq'd flat, they have phase alignment. The final step is to apply crossover filtering. Because of the adjustments and manipulations, idealized crossover slopes can be simply applied.
Of course, it is possible to load all the raw driver measurements into a simulation program, such as the very powerful Vituixcad, and develop the DSP filtering algorithm by simulation. However, given the ease of developing the filters based on iteration and rapid feedback, this seems unnecessarily complicated. It is analogous to tuning a guitar. One way is to measure the actual frequency of the string and compare it to the desired frequency, then using physics and math, determine how many degrees the tuning peg must be rotated. The more simple approach is to just start turning the tuning peg until the digital tuning meter says we are in tune.
I use Vituixcad to assess my filters after the fact to make sure I have not done anything stupid. I expect close alignment between simulation and measurement. But the actual development is conducted based on iterative measurements and adjustments.
If you can export the impulse from OmniMic as a text file in the correct format then you can import that in HFD and do your settings there and immediately see the effect. That is probably a simpler way but I don't really see anything wrong with what you are doing now if you are familiar with it.
Code:
[I]Load measurements for speaker
On the Graphic filter design screen, select the channel you want to import and click
“import”. On the advanced screen, right click a most left channel-box and click “Load
impulse file”. The filter designer expects the impulse response measurement as a text
file with one sample per line. There is no restriction on the absolute gain of the
impulse response data. The only thing that matters is that the absolute gain be the
same for all three measurements. The filter designer computes a gain offset based on
all loaded responses to centre them collectively on the vertical scale.[/I]By the way, thanks for this tip... It became quite useful for one spot.
I do wish I could just work in solid wood... more satisfying on many levels...
I think I remember trying to get omnimic software to export an impulse response, and I could not figure out how to do it. I suspect it does not have that capability, but perhaps I missed it somewhere.
Last edited:
What if you had a target on your development screen, and tweaked filters until you had that response, maybe like you do now only you didn't need to press measure after each change.
<rant> A number of unexpected practices have cropped up since Hypex came about. Seemingly as the result of their instructions, which appear to be workarounds for their software limitations, rather than necessarily through their expertise in designing speakers.. which they seem to want to make appear to be as easy as possible. </rant>
<rant> A number of unexpected practices have cropped up since Hypex came about. Seemingly as the result of their instructions, which appear to be workarounds for their software limitations, rather than necessarily through their expertise in designing speakers.. which they seem to want to make appear to be as easy as possible. </rant>
You're welcome glad it helped.
Omnimic can export IR's as a wav file, according to the manual a right click gets you to the options.
I tried with REW and I could export an IR as text and get it into HFD but it did not give the right response. Just testing this reconfirmed my previous description of their software...
As I brought it up I thought I should see if I could do it.... I think using REW and it's EQ and target function would be a better way. The generic EQ should be pretty close to the boost/cut in Hypex or stick to what you know.