@jhollander said:
Hard for me to tell at this scale but the FR is not flat enough from 2.5K to 6K imo
We should create a "MAC standard" for presenting scale. I always struggle trying to anticipate what is most useful for others to see. Here is a more compressed scale for the FR and I think the white background might be better for the minor divisions.
I dropped the Elliptical xo and added a notch to the tweeter, there was a broad but small hump between 1k and 2k that I think is due to baffle step. Increased the woofer inductor to 2.0mH. Pretty much at +/- 1db from 50Hz to 10khz. Looks good on my computer screen, will wire it up and listen tonight.
I think that this is looking very good. From about 600Hz to 2kHz, your target line goes right down the half way point between the on-axis and predicted in-room response curves. The blooming of the in-room curve from 2-5kHz is fairly close to your target curve and I doubt that this would make it sound excessively bright.
@dcibel said:
What’s wrong with IEC263? 25dB/dec keeps common scale for easy comparison and interpretation.
But how many db above and below the on-axis FR do you show to be most useful? I think John's point was that I was so zoomed-out that it wasn't easy to see how far from flat my curve really was. The 25db/dec only scales the horizontal axis, right?
25dB/dec specifies the aspect ratio, ie rise and run. dB = y axis, decade = x axis. Maintaining constant aspect ratio is important for interpretation and direct comparison. How many dB span for y axis will vary relative to the image dimensions.
Easy method is to use image copy/export buttons through right click memory on the graphs. Aspect ratio and graph size is in the options to keep consistency.
I vote for the 25dB/decade aspect ratio as well. I have VituixCAD's aspect ratio set to 25dB/dec, but this only gives me this aspect ratio when exporting images. If I grab a VituixCAD 6 pack image or individual graph, using the Windows snipping tool, then I get an aspect ratio that is based on my SPL/Directivity span setting. In VituixCAD options, you must have the SPL/Directivity span set to 50dB to get a 25dB/dec aspect ratio when grabbing a screen image. (EDIT: This would be for 20-20kHz graphs only. If you change the frequency range to something else, then a SPL/Directivity span of 50dB does not equal a 25dB/dec aspect ratio.)
Fwiw the latest REW 5.20.14 early access release provides buttons in SPL and All SPL controls for strict conformance to IEC263. Image export (camera button) also allows for 25dB/dec aspect ratio conformance, so it’s easy to keep to this standard.
I find with some poor graph scaling selections from online sources that it’s much easier to compare graphs by tracing them into VituixCAD so they can be overlaid in same scaling.
Latest VituixCAD 2.0.103.1 will now show IEC263 25dB/dec aspect ratio when you double click on the SPL or power & DI graph. Main 6-pack graph will still vary aspect ration in order to maximize screen utilization.
@dcibel said:
Here's a quick test run on a small bookshelf. In my basement I have a ledge 8" deep , and the speaker is able to be placed flush to the ledge, so the front of the speaker is 8" from the back wall, but below the speaker the ledge is only about 1" back.
I compared this measurement to placing the speaker on the front edge of a table 2.5ft from the back wall. Measurements were unwindowed at 3ft distance, and smoothed 1/1 oct to show only the overall trend of the response graph.
I then completed magnitude division to show only the difference between these two measurements.
Take that information for what it is, just a quick and dirty comparison.
@dcibel - going back to your post on the first page of the thread and estimating the effect of being near the wall. Your last image gave the estimate of the effect, but can you take it as step further? Could you then take this result and do magnitude multiplication with the SPL curve measured away from the wall and windowed to get a more precise near-wall estimate? For example, would the final measurement (E) below be a relatively realistic estimate of this speaker near the wall? (This speaker is 4-1/2" wide so the BSL starts around 3khz. It has an F3 of 100hz and no attempt to measure nearfield and merge to get results below 250hz is undertaken.)
(Sorry, the colors aren't the easiest to see)
(A) Speaker measured at 1m, near the wall, unwindowed (teal)
(B) Speaker measured at 1m, away from the wall, unwindowed (green)
(This graph is 1:2 smoothing but the magnitude division will be done at 1:1)
(C) = A / B (with 1:1 smoothing)
(D) Speaker measured at 1m, away from the wall, windowed at 4.75ms
I think you've answered your own question . Of course, if C=A/B, then A=B*C. Response C is a differential between in-room and on-wall, as you’ve shown it gives at least a half decent approximation of the change in response on-axis. I’m sure there will be some significant changes off-axis as well, but this process should get you in the ballpark.
You can try completing full near/far/diffraction merge process, then process the multiplication using the calculator tool in VituixCAD, it will allow you to process the full set of measurements in bulk.
@dcibel said:
I think you've answered your own question . Of course, if C=A/B, then A=B*C.
Lol, that's a fair point. But the one difference is A and B are unwindowed and smoothed 1:1 to get C but then I applied that to the windowed result with no smoothing. I wasn't sure whether the result in C would be such a rough estimate that combining it with the windowed result would just be too unreliable. I'm trying to help someone over at diyaudio and don't want to give him bad advice. But, he was buying an off the shelf xo from PE, so I think the estimate of that last teal line is a lot better than what he would have got on his own.
@dcibel said:
I think you've answered your own question . Of course, if C=A/B, then A=B*C.
Lol, that's a fair point. But the one difference is A and B are unwindowed and smoothed 1:1 to get C but then I applied that to the windowed result with no smoothing.
Yes, so the suggestion above would be to run through near/far merge process as you normally would, then apply the compensation curve "C". Based on your result above, you should end up with nearly a IB response but with a boomy bass lump around 100Hz.
Comments
My ears said stop effing with it and put on another track. So I guess I'm running with it.
We should create a "MAC standard" for presenting scale. I always struggle trying to anticipate what is most useful for others to see. Here is a more compressed scale for the FR and I think the white background might be better for the minor divisions.
I dropped the Elliptical xo and added a notch to the tweeter, there was a broad but small hump between 1k and 2k that I think is due to baffle step. Increased the woofer inductor to 2.0mH. Pretty much at +/- 1db from 50Hz to 10khz. Looks good on my computer screen, will wire it up and listen tonight.
I think that this is looking very good. From about 600Hz to 2kHz, your target line goes right down the half way point between the on-axis and predicted in-room response curves. The blooming of the in-room curve from 2-5kHz is fairly close to your target curve and I doubt that this would make it sound excessively bright.
What’s wrong with IEC263? 25dB/dec keeps common scale for easy comparison and interpretation.
But how many db above and below the on-axis FR do you show to be most useful? I think John's point was that I was so zoomed-out that it wasn't easy to see how far from flat my curve really was. The 25db/dec only scales the horizontal axis, right?
25dB/dec specifies the aspect ratio, ie rise and run. dB = y axis, decade = x axis. Maintaining constant aspect ratio is important for interpretation and direct comparison. How many dB span for y axis will vary relative to the image dimensions.
Easy method is to use image copy/export buttons through right click memory on the graphs. Aspect ratio and graph size is in the options to keep consistency.
I vote for the 25dB/decade aspect ratio as well. I have VituixCAD's aspect ratio set to 25dB/dec, but this only gives me this aspect ratio when exporting images. If I grab a VituixCAD 6 pack image or individual graph, using the Windows snipping tool, then I get an aspect ratio that is based on my SPL/Directivity span setting. In VituixCAD options, you must have the SPL/Directivity span set to 50dB to get a 25dB/dec aspect ratio when grabbing a screen image. (EDIT: This would be for 20-20kHz graphs only. If you change the frequency range to something else, then a SPL/Directivity span of 50dB does not equal a 25dB/dec aspect ratio.)
75dB range from 20Hz to 20Khz?
Only if graph is square. usually they are rectangular so vertical span is <75dB
Fwiw the latest REW 5.20.14 early access release provides buttons in SPL and All SPL controls for strict conformance to IEC263. Image export (camera button) also allows for 25dB/dec aspect ratio conformance, so it’s easy to keep to this standard.
I find with some poor graph scaling selections from online sources that it’s much easier to compare graphs by tracing them into VituixCAD so they can be overlaid in same scaling.
Latest VituixCAD 2.0.103.1 will now show IEC263 25dB/dec aspect ratio when you double click on the SPL or power & DI graph. Main 6-pack graph will still vary aspect ration in order to maximize screen utilization.
@dcibel - going back to your post on the first page of the thread and estimating the effect of being near the wall. Your last image gave the estimate of the effect, but can you take it as step further? Could you then take this result and do magnitude multiplication with the SPL curve measured away from the wall and windowed to get a more precise near-wall estimate? For example, would the final measurement (E) below be a relatively realistic estimate of this speaker near the wall? (This speaker is 4-1/2" wide so the BSL starts around 3khz. It has an F3 of 100hz and no attempt to measure nearfield and merge to get results below 250hz is undertaken.)
(Sorry, the colors aren't the easiest to see)
(A) Speaker measured at 1m, near the wall, unwindowed (teal)
(B) Speaker measured at 1m, away from the wall, unwindowed (green)
(This graph is 1:2 smoothing but the magnitude division will be done at 1:1)
(C) = A / B (with 1:1 smoothing)
(D) Speaker measured at 1m, away from the wall, windowed at 4.75ms
(E) C * D
I think you've answered your own question
. Of course, if C=A/B, then A=B*C. Response C is a differential between in-room and on-wall, as you’ve shown it gives at least a half decent approximation of the change in response on-axis. I’m sure there will be some significant changes off-axis as well, but this process should get you in the ballpark.
You can try completing full near/far/diffraction merge process, then process the multiplication using the calculator tool in VituixCAD, it will allow you to process the full set of measurements in bulk.
Lol, that's a fair point. But the one difference is A and B are unwindowed and smoothed 1:1 to get C but then I applied that to the windowed result with no smoothing. I wasn't sure whether the result in C would be such a rough estimate that combining it with the windowed result would just be too unreliable. I'm trying to help someone over at diyaudio and don't want to give him bad advice. But, he was buying an off the shelf xo from PE, so I think the estimate of that last teal line is a lot better than what he would have got on his own.
Yes, so the suggestion above would be to run through near/far merge process as you normally would, then apply the compensation curve "C". Based on your result above, you should end up with nearly a IB response but with a boomy bass lump around 100Hz.