ErinsAudioCorner

[napilopez]

Seriously, though, thanks for letting me know. I was staring at the screen, clicking everywhere, checking settings... couldn't figure it out for the life of me!

@jhaider first pointed me to vituixcad and how to get it started properly, saved me a lot of trouble. I was looking for an easy way to calculate sound power and it solved all my troubles.

Quick follow up question. When you have a mirrored horizontal response (such as +10 being the same as -10, 20, 30, etc), do you need to do anything additional ... like import the file twice or explicitly name the files ± 10, 20, etc? I guess it doesn't matter if it's an overall average but I just want to make sure.

By default, the software will automatically mirror missing angles, so you needn't do it prior to import. But if the mirrored angles are already there, you can either lable them + and -, respectively. Or if you don't VituixCAD will just tell you that it didn't import the suspected duplicated and just mirror the angles itself.

P.S. you need to load all measurements at once, you can't import twice without deleting the prior selection as far as I know
P.P.S. Make sure you hit the 'CTA-2034-A' button in 'Options.' By default the settings for listening window and DI are a bit different than the spinorama standard.

 

[hardisj]

Got it. Yea, this is nice. I may just load up the files in to VCAD and let it do the math but still use my own Matlab script to plot the spin curves so I can specify how they look.

For example, I prefer my own spectrogram (top) to the one vcad produces (bottom):

Still, it's nice to know I can use VCAD to help do some number crunching for the CEA 2034 curves.

 

[napilopez]

Got it. Yea, this is nice. I may just load up the files in to VCAD and let it do the math but still use my own Matlab script to plot the spin curves so I can specify how they look.

For example, I prefer my own spectrogram (top) to the one vcad produces (bottom):

View attachment 66850


View attachment 66851




Still, it's nice to know I can use VCAD to help do some number crunching for the CEA 2034 curves.

Yeah I feel you - I mean, VCAD can present a full spin, but I still import the curves back to REW because I have more control over the data presentation (e.g. I like dark mode )

 

[HammerSandwich]

For example, I prefer my own spectrogram (top) to the one vcad produces (bottom)

Try "contour lines" in the graphic's right-click menu.

 

[hardisj]

Try "contour lines" in the graphic's right-click menu.

Good to know. Still, I prefer my own. I just think it looks more "crisp".

 

[HammerSandwich]

Fair enough. Note that the SPL scaling buttons (under reference angle at top) also affect the polar map, though they still can't give the Klippel's hard edges.

Have you tried the surface chart? (And with mouse wheel!) If I had to choose only one...

 

[hardisj]

Just to be totally clear, the early reflections curve does technically require all the rear measurements from +/-90 to 180. However using just 90 degrees and 180 provides a good enough approximation for most speakers.

The CTA-2034A is written ambiguously on this but Todd Welti from Harman clarified the matter a few months back. The good news is since it is was also clarified the curve is a weighted averag(average of averages), the rear component doesn't affect the final calculation much for monopoles even if it's a bit off.

Did you say previously that you contacted the makers of VCAD and talked to them about the miscalculation regarding this matter? I thought you did but maybe I am just missing it. If so, I think it has yet to be resolved.

I have completed full 360 spins on the Buchardt S400. It took 3 hours and a total of nearly 2 miles of walking. Seriously. But I'll get back to that later... right now I've more pressing concerns.

... back to the Early Reflections...
I have asked this of Mr. Toole as he has been kind enough to offer advice thus far. I am sharing my question and reasons for asking here as well.

I created a Matlab script to generate the CEA-2034 curves with my data and I also used software called VituixCAD to check my code. I noticed that my Early Reflection and my Sound Power curves looked different. After a couple of hours of digging in my script I finally realized the error was that VituixCAD is apparently using only the ±90° & 180° curves to calculate rear wall reflections, rather than using the entire rear hemisphere like I had coded my script to do.

Now, I realize what VituixCAD is doing isn't of real concern to you. I was merely providing that as anecdotal evidence I am indeed calculating things right (depending on which method for rear wall bounces is used). More importantly, I have attached the two graphs (ignore sound power for now; I believe it may also be subject to the same concern of rear wall reflections; though, I did take care to use the weighting values per angle via the CEA-2034 specification Appendix C). I was surprised to see such a difference made here by using all the rear hemisphere angles. And this begs the question: which is right? According to page 13 of the ANSI/CEA-2034-A rear is:
"Rear: ± 90°, ± 100°, ± 110°, ± 120°, ± 130°, ± 140°, ± 150°, ± 160°, ± 170°, 180° horizontal, (i.e.: the horizontal part of the rear hemisphere)."

And the follow-up question would be: do I need to treat the sound power curve in the same manner? Either use the entire rear-hemisphere or only the (3) angles (±90° & 180°).

The first picture below uses the full rear hemisphere as prescribed in Appendix C:

The image below is created using only ±90° & 180° for rear wall reflections:

This is what VituixCAD shows. You can see the Early Reflection matches the one where I do not use the entire rear-hemisphere and instead used only ±90° & 180°.

And just for reference, I am attaching my horizontal and vertical polars I created from the spin data:

 

[napilopez]

Did you say previously that you contacted the makers of VCAD and talked to them about the miscalculation regarding this matter? I thought you did but maybe I am just missing it. If so, I think it has yet to be resolved.

I have completed full 360 spins on the Buchardt S400. It took 3 hours and a total of nearly 2 miles of walking. Seriously. But I'll get back to that later... right now I've more pressing concerns.

... back to the Early Reflections...
I have asked this of Mr. Toole as he has been kind enough to offer advice thus far. I am sharing my question and reasons for asking here as well.

I created a Matlab script to generate the CEA-2034 curves with my data and I also used software called VituixCAD to check my code. I noticed that my Early Reflection and my Sound Power curves looked different. After a couple of hours of digging in my script I finally realized the error was that VituixCAD is apparently using only the ±90° & 180° curves to calculate rear wall reflections, rather than using the entire rear hemisphere like I had coded my script to do.

Now, I realize what VituixCAD is doing isn't of real concern to you. I was merely providing that as anecdotal evidence I am indeed calculating things right (depending on which method for rear wall bounces is used). More importantly, I have attached the two graphs (ignore sound power for now; I believe it may also be subject to the same concern of rear wall reflections; though, I did take care to use the weighting values per angle via the CEA-2034 specification Appendix C). I was surprised to see such a difference made here by using all the rear hemisphere angles. And this begs the question: which is right? According to page 13 of the ANSI/CEA-2034-A rear is:
"Rear: ± 90°, ± 100°, ± 110°, ± 120°, ± 130°, ± 140°, ± 150°, ± 160°, ± 170°, 180° horizontal, (i.e.: the horizontal part of the rear hemisphere)."

And the follow-up question would be: do I need to treat the sound power curve in the same manner? Either use the entire rear-hemisphere or only the (3) angles (±90° & 180°).

The first picture below uses the full rear hemisphere as prescribed in Appendix C:

View attachment 67248



The image below is created using only ±90° & 180° for rear wall reflections:
View attachment 67249



This is what VituixCAD shows. You can see the Early Reflection matches the one where I do not use the entire rear-hemisphere and instead used only ±90° & 180°.

View attachment 67252




And just for reference, I am attaching my horizontal and vertical polars I created from the spin data:

View attachment 67257


View attachment 67258

Weird! It was definitely fixed @hardisj , as I triple checked this with several speakers.

I thought that maybe VCAD's creator had accidentally gone back to the old calculation by accident in a recent update, but I just repeated the manual ER curve calculation for my measurements of the D&D 8c and get identical results in REW and VCAD.

Looking at your measurements, I think your first image, the full-rear ER curve, tilts down significantly more than expected, and your second one doesn't match VCAD's as perfectly as it should.

So correct me if I'm wrong, but this makes me think you are not performing an average of averages but rather a single average of all the ER angles? If you were, I don't think your first image would be so different from the rest.

The manual process described clarified by Welti involves creating five averages out of the angles listed in the ER curve before you compute the final ER average. So first you compute the individual front, side, rear, ceiling, and floor averages. Then you average those five averages such that each of those portions contributes equally to the final ER curve - essentially making the ER a weighted average (RMS averages throughout, btw).

For example, with the D&D 8C, here are how the ER curve differs depending on how it's calculated:

Red: Manual calculation (Average of averages, full rear)
Blue: VCAD calculation
Green: Single average with three rear hemisphere angles (±90, 180)
Yellow: Single average with all 19 rear curves

As you can see, my manual calc and VCAD are identical, perfectly overlapping each other. The Green method is pretty similar, but as a single average, it weighs all angles equally, so it misses out on the dip caused by vertical lobing visisble with the proper methodology. Meanwhile the yellow is completely off on its SPL level due to having so many more rear curves factored into a single average.

(Note that since the 8C is a constant directivity cardioid down to 100Hz, the entire curve is lowered by several dB, but with a normal speaker this curve would be more like a steeper downward tilt).

 

[napilopez]

By the way, other than the on-axis curve my spin for the S400 looks very similar to your vituixCAD one( above 1k, you have much more resolution below that), even though at the time I was only using front hemisphere data and used 15 degree intervals, so I had to 'fake' the missing data.

So as far as I'm concerned, your vituixcad-generated curves are on point.

 

[hardisj]

I am completing an average of averages. In fact, I purposely broke the individual sets out so I could plot those on a separate graph.

I’ll have to look at this again and see if there’s something going on.

 

[hardisj]

So correct me if I'm wrong, but this makes me think you are not performing an average of averages but rather a single average of all the ER angles? If you were, I don't think your first image would be so different from the rest.

I am completing an average of averages. In fact, I purposely broke the individual sets out so I could plot those on a separate graph. I wish VCAD broke these out individually so I could check to see if one of these curves is off. As a double check, here's the result of me not using an average of averages. You can see the tilt is indeed more severe and another 5dB down compared to using the average of averages.

I’ll have to look at this again and see if there’s something going on.

 

[napilopez]

I am completing an average of averages. In fact, I purposely broke the individual sets out so I could plot those on a separate graph. I wish VCAD broke these out individually so I could check to see if one of these curves is off. As a double check, here's the result of me not using an average of averages. You can see the tilt is indeed more severe and another 5dB down compared to using the average of averages.

View attachment 67313


I’ll have to look at this again and see if there’s something going on.

Ah, sorry for assuming you were trying otherwise. VituixCAD does break down the curves individually though! Just right click on the spin graph and enable 'early reflections 5'. You can also export them from the export menu -- they are included under the CTA-20434A export option. You can export as a CSV to get all the curves in one sheet or as txt to get them as multiple separate text files.

 

[hardisj]

Ah, sorry for assuming you were trying otherwise. VituixCAD does break down the curves individually though! Just right click on the spin graph and enable 'early reflections 5'. You can also export them from the export menu -- they are included under the CTA-20434A export option. You can export as a CSV to get all the curves in one sheet or as txt to get them as multiple separate text files.

No worries. If I were confident in my calculations I wouldn't have asked. So I appreciate you keeping me honest.

Good to know I can export those graphs as well. Honestly, the driver for me building my own data set was for those because I didn't think VCAD did. At this point, I can just import them all in to Matlab instead of calculating them myself, I reckon'. But I still wanna try to figure out just what it is I am doing wrong.

 

[napilopez]

No worries. If I were confident in my calculations I wouldn't have asked. So I appreciate you keeping me honest.

Good to know I can export those graphs as well. Honestly, the driver for me building my own data set was for those because I didn't think VCAD did. At this point, I can just import them all in to Matlab instead of calculating them myself, I reckon'. But I still wanna try to figure out just what it is I am doing wrong.

Yeah, so since you are doing the average of averages, what stands out to me is that your two initial Matlab curves are so different. In my experience, using 19 or 3 rear curves when doing the average of averages has a very minimal effect on the ER above the woofer's passband since there's so little rear output regardless of which you choose and it only contributes one fifth to the score regardless. Will give an example when I'm on my PC

 

[hardisj]

Yeah, so since you are doing the average of averages, what stands out to me is that your two initial Matlab curves are so different. In my experience, using 19 or 3 rear curves when doing the average of averages has a very minimal effect on the ER above the woofer's passband since there's so little rear output regardless of which you choose and it only contributes one fifth to the score regardless. Will give an example when I'm on my PC

I understand. I was surprised by that as well. I’m sure it’s something on my end. But I’ll use the individual curves from VCAD to compare against my Matlab curves and see if I can narrow down the culprit.

 

[napilopez]

I understand. I was surprised by that as well. I’m sure it’s something on my end. But I’ll use the individual curves from VCAD to compare against my Matlab curves and see if I can narrow down the culprit.

Maybe it'd help to check it on a different set of data just to make sure it's something in the calculations and not the data itself? For example,if you want to try to use Amir's measurements of the Revel M16, I've already exported the individual angles into separate text files for easy import into VCAD and REW. I'll attach those in a zip at the end of this post.

Here are the different ways the ER can be calculated:

White: Listening Window
Yellow: Simple average with 3 rear curves, AKA what the NFS generates, though Klippel issued a fix that hasn't been implemented in the main software yet
Green: Simple Average with full rear
Blue: Average of Averages, using full Rear
Red dashed: Average of Averages, using +90, -90, 180 for Rear curve

You can see the Blue and Red are almost perfect overlaps, with the red just being a teensy higher at some levels.

Below are the ERDI curves from the klippel and from the proper method.

You can see the blue ERDI curve more matches harman's pretty much perfectly while the klippel one smooths out the crossover error.

Attached below are the 72 Revel M16 angles for you to mess with. Note if you import these into VCAD you'll get a 'missing phase' error, so you'll have to click through a bunch, but it should calculate the spin just fine.

 

[hardisj]

Well, I've found a couple culprits but still unsure why I am getting the error I am getting.

For example, direct comparison between my ER and VCAD's ER. Mine in black. VCAD in red.

I have verified I am pulling in the correct data. I have also verified the calculation I get in matlab matches what I get in excel, averaging ±40, 50, 60, 70, 80. (Well, a single set; averaging doesn't matter since they are a mirror of each other horizontally). Yet, VCAD still gives me a different answer.

I am attaching my ER files if you want to take a look (but I don't expect you to). I'm sure it's something simple I'm just not seeing at the moment.

 

[napilopez]

Well, I've found a couple culprits but still unsure why I am getting the error I am getting.

For example, direct comparison between my ER and VCAD's ER. Mine in black. VCAD in red.

View attachment 67361


I have verified I am pulling in the correct data. I have also verified the calculation I get in matlab matches what I get in excel, averaging ±40, 50, 60, 70, 80. (Well, a single set; averaging doesn't matter since they are a mirror of each other horizontally). Yet, VCAD still gives me a different answer.

I am attaching my ER files if you want to take a look (but I don't expect you to). I'm sure it's something simple I'm just not seeing at the moment.

Weird. Are you including phase in your calculations? Should be amplitude/frequency only

 

[hardisj]

Weird. Are you including phase in your calculations? Should be amplitude/frequency only

I am not. Only F vs SPL.

The plot difference shows up in excel as well. So it’s an actually difference between how I am calculating the reflection vs how VCAD is calculating it.

 

[napilopez]

I am not. Only F vs SPL.

The plot difference shows up in excel as well. So it’s an actually difference between how I am calculating the reflection vs how VCAD is calculating it.

Hmm, what a mystery. Have you tried importing/averaging the curves in REW? Gives me identical results to VCAD