Traced the problem to the fact that my mapping program does not have the exact same impulses as Holm. Not sure where that is coming from, but some sample points got dropped. Also not sure how wide spread this is in my data. All of the sets that I had done previously passed the reasonableness test - I was very familiar with the data and would have caught something like this. But, I had never seen any data for the Orion so I had no idea what to expect.
But maybe other should be aware as well since I read out Holm Impulses into a text file and then read them into MathCAD, but data points ere missing. So one or the other of these programs is acting funnny. Thats computer software for ya!
But maybe other should be aware as well since I read out Holm Impulses into a text file and then read them into MathCAD, but data points ere missing. So one or the other of these programs is acting funnny. Thats computer software for ya!
After a careful point by point tracing, it is clear that Holm dropped writing out a data point. And this occured at precisely the maximum impulse. Thats what made it so easy to find. I only traced one point, but I guess that there could be many. I will try the transfer again and see if this is a fluke or if its consistant.
Minor correction:
I think that SL doesn't sell/ship anything other than plans or circuit boards to users.. rather other suppliers do. Presumably Wood Artistry (for the US).
Corrected the Orion plots.
Thanks for pointing that out John. I would never have caught that myself since I had no base of comparison.
I have no idea what happened in Holm other than the same sample number was dropped from every angle impulse. It just skipped a line when it wrote out the data (each line in the text file is 13 data points at a common sample number). Just goes to show how careful you have to be, its so easy to have an error that goes undetected.
Thanks for pointing that out John. I would never have caught that myself since I had no base of comparison.
I have no idea what happened in Holm other than the same sample number was dropped from every angle impulse. It just skipped a line when it wrote out the data (each line in the text file is 13 data points at a common sample number). Just goes to show how careful you have to be, its so easy to have an error that goes undetected.
The Orion front response looks much better from 1k up but the response below 1k still look strange.
With regard to normalizing the contours I guess we will have to disagree. But consider what the contours would look like for a system with perfectly flat response and perfect CD; horizontal lines. Now put a 6dB dip at 1k Hz in the response and those contours will look pinched in around 1 k while the directivity would be unchanged. I guess it depends on what you really want to show, full SPL map or a representation of the directivity alone. I'm ok with either. The normalized contours just show the variation of directivity with frequency more clearly.
The rise below 1kHz surprised me too, but its clearly in the raw data as well.
I think your example above is perfect. A 6dB dip would drastically change the polar map, pinching it at all angles, but yet the directivity has not changed. A hole on-axis with no affect off-axis is what happens and only un-normalized will this be evident.
I once wrote to John Atkinson when he remarked about an off-axis flare in the response. I pointed out that this apprent "flare" in his "normalized data, was simply a "hole" on axis and what was likly to be hardly audible appeared and was commented on as an audible feature.
An option button is an easy addition I suppose.
I think your example above is perfect. A 6dB dip would drastically change the polar map, pinching it at all angles, but yet the directivity has not changed. A hole on-axis with no affect off-axis is what happens and only un-normalized will this be evident.
I once wrote to John Atkinson when he remarked about an off-axis flare in the response. I pointed out that this apprent "flare" in his "normalized data, was simply a "hole" on axis and what was likly to be hardly audible appeared and was commented on as an audible feature.
An option button is an easy addition I suppose.
Its full space, but based on only horizontal data, i.e. the polar pattern is assumed to be axisymetric - not perfect, but not bad either. The only error would be at the crossover where there is a vertical lobe that is not represented in the horizontal data.
DLR - thats what I suggested above - "An option button is an easy addition I suppose.
"
DLR - thats what I suggested above - "An option button is an easy addition I suppose.
"
I do mine from 0-350 deg., horizontal only too.
On this side of the world we are used to read "power reasponses" instead of DI tho.
But power response is different than DI. DI is the SPL equivalent of the power response normalized to the response on the designated angle. Thats why it changes with angle. The power response would not. Subjectively the DI better indicates the situation where you want the direct sound and the reverberant sound to have the same spectrum. A flat DI indicates this, but a flat power response does not. If the response on the listening axis is not flat, but the power response is then there is a subjective effect.
As stated in the program, when the DI and axis response are both flat then you have the ideal listening angle for the speaker. This is not easy to do. You can see this quite clearly in the Summa at about 20 degrees off axis. The DI should also be high as well as flat. Less early reflections.
As stated in the program, when the DI and axis response are both flat then you have the ideal listening angle for the speaker. This is not easy to do. You can see this quite clearly in the Summa at about 20 degrees off axis. The DI should also be high as well as flat. Less early reflections.
Last edited:
Well, now that you have admitted and found your measurement error, perhaps you should go back and edit your post (#4399) where you made some snide comments about Mr Linwitz's product.
I´m familiar with the difference, pwr vs. DI. Here the "press" has published 0, 20, 60 deg off axis and the power response. Those who could, did their own conclusions from there.
HiFi that is.
I´m more into pro-stuff, but that´s OT.
Why not show both?
Dave
Exactly. The normalized response would show the frequency dependent directivity map. The un-normalized plot would show the SPL map. The option button would be a good addition.
When I build the Orion and compared it to my NaO II I found the NaO II sounded richer and fuller in through the mids. The Orion sounded a little thin to my ear. The NaO II has a boost similar to what your Orion measurements show, though not as great. The Orions I build did not show that rise.
At some point I believe SL did suggest a mode which I believe boosted the response in the lower mids to "warm" the system up. I never made that mod to the Orions I built before I discarded them. Perhaps that is what you are showing?
On another note, when I look at your systems and the Behringer system closely what I see in the un-normalized contour maps is more a represntation of constant beam witdth. When I look at the corresponding polar plots they seem to change significantly with frequency.
Any differences must be in your perception because the data set is exactly the same in all the plots. The data is a matrix - the top plot shows rows and the side plot shows columns - exact same data in all of them. No smoothing done in one of them thats not in the others. The smoothing is 1/20th octave in frequency.
As far as snide comments goes - the error wasn't mine, but HolmImpulse. I applogize for any misunderstandings that resulted from this error.
I just had to check my latest home project.
The DI at the design angle (yes, not on axis) is sloping upwards 1.2 dB from 500 Hz up. It deviates +/- 1 dB from a straight line between 400 and 13 kHz.
Different goals, I guess.
Last edited:
My "on-listening-axis" curve is flat. The power response is falling.
I´d say "dark" compared to Yours... Pity I can´t listen to Summas and others. I´d be VERY interested !
I´ll see what I can do with the data. Was it IR in .txt ? With or without time ??
I showed both at #4396:Why not show both?
http://www.diyaudio.com/forums/multi-way/103872-geddes-waveguides-440.html#post2230071
Ref 0 is normalized. If there's a significant anomaly on-axis, then Ref can be any other angle.
In the third image, I cranked up the sensitivity to clearly show the -6 dB directivity @ Cyan....
Last edited: