Walnut Smith Selsted

[John W]

At the prodding of another forum member I began looking into the Smith Selsted tweeters from this post:

http://audioheritage.org/vbulletin/showthread.php?t=9871

I took the small sketch in the article, saw that the diaphragm is 1 in and scaled up to see what the size would be. It turns out the height of the horn scales to about 1/2", the same size as the opening on the 2404 ring radiator. So I cleaned up the sketch, made a pattern and duplicated the horn in walnut. Here is the result.
Initial listening tests are very positive. My RTA unit is pretty worthless above 10khz, so I didn't even pull it out.

[John W]

Here is the test setup

[Zilch]

Hee, hee. I see those Minis there.... ;)

[johnaec]

John - 'nice little horns! How low are they crossed over? And I see you finally got the grilles done for the mini's - cool!

John

[John W]

How low are they crossed over?
I was just giving them a trial run on these speakers. They are crossed over pretty low here, about 3k. I figured it was the best way to get an overall impression of there sound without a lot of coloration from other drivers.

[Mr. Widget]

John, you constantly amaze me... those are damned cute!


Widget

[yggdrasil]

Very impressing.

Keep us posted on how they sound, and possibly as a UHF together with your 2397's.

[Mr. Widget]

Interestingly it seems to be roughly the size of Ken's Westlake tweeter horns, but it is narrower and Westlake omitted the vanes. From Ken's comments and measurements, I'd guess that your tweeter horn using the ring radiator will have better performance.

Do you have any 075 diaphragmed ring radiators? An impedance plot along with a FR curve of a 075 on one of these horns would be interesting....


Widget

[John W]

The drivers in the picture are 2402, I think the diaphragms are original. I don't have the equipment to perform the tests, though.

[Mr. Widget]

The drivers in the picture are 2402, I think the diaphragms are original. I don't have the equipment to perform the tests, though.To remedy this, John sent a couple of his Micro Smith horns down to SF for me to run some tests on. One version has vanes and the other does not.

Since I needed a 2404 driver to test John's horns, I asked Zilch if I could borrow one... he graciously offered to lend me one.

To carry out these tests I used my recently upgraded CLIO rig... unfortunately I discovered after the conclusion of my tests and having put everything away that one of the new features, saving the plots directly as jpegs isn't terribly great. The images are slightly fuzzy and the color info has been discarded.:( I think the plots will still be useful. For subsequent plots I'll be saving them the old way which preserves the colors.

This first plot compares a stock 2404 with one of John's Micro Smiths with vanes. The plot shows how the 2404 is significantly more efficient, but also how to two curves are roughly similar.



Widget

[Mr. Widget]

Here are two plots... the first shows the polar response of a stock JBL 2404. (Whether you like the way they sound or not, their polar response is remarkable.) The second plot is of John's Micro with vanes... it's polar response is also pretty good.

[Mr. Widget]

This plot shows the vaned version with only on axis and 15 degrees off axis... I am posting this to more clearly show how well they track.

[Mr. Widget]

John also made a version of these Micro Smiths with no vanes.

Here is a plot comparing the vane version vs. the no vane version. Looking at this on axis comparison there doesn't seem to be much of a difference.

[Mr. Widget]

Here is the polar respose for the Micro Smith with no vanes... it seems to perform a bit better than the one with vanes. I have also posted a simplified on axis and 15 degree off axis plot.

[Mr. Widget]

To sum up... while these graphs look a bit rough, that is the nature of diffraction slots. For the most part the diffraction effect seems to be causing narrowband suck outs... these tend to be difficult to hear. While I did not set these tweeters up in a system and evaluate them subjectively, I have heard quite a number of tweeters under test over the years and these did sound a bit more natural than the baby butt cheek.

Oh yeah, and did I mention that John's tiny walnut tweeters are just about the cutest little critters I have ever seen!



Widget

[John W]

Big thanks to Widget for taking the time chase down the drivers and make these measurements. :applaud: You’re a gentleman and a big asset to the forum. I really appreciate the consistent help, encouragement and good advice you provide.

I am a little surprised at the jagged response compared to 2404, but then again it’s the great engineering that makes JBL great, not just pretty good.

I wonder if there are things that could be improved upon in the design to smooth it out some? A couple of things I will probably try would be rounding over the lips, and smoothing out the throat a little.

Also, I made a couple measurement errors in the original drawing at the top of the post, the height is really 5” rather than 6”, and the radius is 4 7/8, not 5 7/8.
Here is an updated sketch, showing the no-vane version.

[4313B]

Very cool!

Nice thread guys! :yes:

Since I needed a 2404 driver to test John's horns, I asked Zilch if I could borrow one... he graciously offered to lend me one.So that's where the little trade came in...
To carry out these tests I used my recently upgraded CLIO rig... unfortunately I discovered after the conclusion of my tests and having put everything away that one of the new features, saving the plots directly as jpegs isn't terribly great.I would always save each curve as a .mls file for future reference.

To sum up... while these graphs look a bit rough, that is the nature of diffraction slots. For the most part the diffraction effect seems to be causing narrowband suck outs... these tend to be difficult to hear. While I did not set these tweeters up in a system and evaluate them subjectively, I have heard quite a number of tweeters under test over the years and these did sound a bit more natural than the baby butt cheek.Yep, so it's probably best to post at least 1/12 octave smoothing. You can open up the .mls files and apply the smoothing as desired/required.

[northwood]

really really very gorgeous

[Mr. Widget]

I would always save each curve as a .mls file for future reference.
Ideally sure... but I did something like 15-20 separate measurement and saving and naming each one was more of a commitment than I was prepared for. With the plot images 4 or more test plots can be saved at a time along with notes explaining what they are.



Yep, so it's probably best to post at least 1/12 octave smoothing. You can open up the .mls files and apply the smoothing as desired/required.I agree that saving the .mls files would have allowed that, and in this case it would have been nice to have the option. I guess my laziness bit me in the butt this time. :D I usually do run my tests with 12th octave smoothing, but in this case there was little change between 12th octave smoothing and no smoothing so I thought I'd post the full resolution plots.

For additional info on the tests:

They were all done at 1m. The tweeter was 4' off the ground and the test structure supporting the tweeters was wrapped in sonex.
There was also sonex on the floor between the mic and device under test and the time windowing was wide enough to far exceed the frequency response of the device under test.


Widget

[4313B]

Ideally sure... but I did something like 15-20 separate measurement and saving and naming each one was more of a commitment than I was prepared for.I hear ya! I have to keep reminding myself that it is soooo much better than the old days though.