Measurements of the Natalie P., and simulations... is something wrong?

[fjhuerta]

Hi all!

I've posted in the past about how I found the Natalie P's a bit shrill and cold. So I decided to take some measurements. At first, I thought everything was OK, but then I found something was probably very wrong... (graph 1, measured.gif).

There's an almost 3 dB hump between 1KHz and 2KHz. This measurement was taken after I added 1 Ohm of padding to the tweeter, so as to try and tame what I perceived as too loud a tweeter.

So, I tried to find out if I had built the crossover wrong. The impedance plot is identical to the reference one, so I was really confused.

At this point, I went to PE and downloaded the impedance and frequency response plots for the RS180 and RS28. I then used LSPCad's KHF Tool to simulate diffraction from my baffle (which is identical to the one posted in the Natalie P's thread, except that it's deeper - 38 cms, actually), and simulated the crossover (with driver layout included) on LSPCad.

These are the results, with my measurements overlaid as a reference (Comparison.gif). They measured almost the same! (my measurements are quite poor, but in this case, they matched the simulation pretty well, I think).

I'm a bit confused. Could this be because I didn't round the corners of the speaker? I assume a 3 dB boost in this region would be perceived as a glaring midrange, so maybe this is what I've been listening to all along.

I played a bit with the crossover values, and got this as a result (schematic.gif).

It slows down woofer roll-off a bit, but it still manages to get rid of HF garbage, and it also eliminates the 3 dB hump I was getting, by changing 3 elements. (graph2.gif)

I wonder whether this is a step on the right direction (for my case), and how could I have controlled baffle step diffraction without modifying the crossover. I don't feel good about this, especially since I know this crossover has been built by countless persons and everyone has enjoyed it.... what do you think?

[kirknelson]

If you didn't make the baffle the same way as the original XO designer had his, then yes your BSC is going to be off.

I'm no expert but the 1-2k range seems to be about right (maybe a little high) for baffle anomolies to show.

[beto]

Hola Javier,

Being that you have gone this far, measured, layout the crossover in lspcad, it makes sense IMO, for you to adjust the crossover components to your liking, after all, you are the one stuck listening to your speaker. I think you are doing the right thing, just tweak it and enjoy it..

I have done the same thing with some other designs from other experts and I have suited the sound to my taste...

[fjhuerta]

Kirk, the baffle is the same size (width); the drivers are in the same relative position, but I built a tower speaker. It's a bit more deep, too. I don't know if altering those dimensions change the amount of BSC needed - I thought it didn't. Anyway, you are right - it seems I have baffle anomalies in that region.

Hola Beto! I think I'll do that. I'm too confused about the sound quality of the speakers, and I always thought this was the region that was giving me problems. I'll definitely try my "fix" and measure again.

[cjd]

Actually, I think that the issue you may be hearing is not so much the ~2dB gain from 1-2k, but that gain matched against the drop in response from 2-5k.

This drop is likely diffraction induced, from changes you made vs. the baffle that Jon used when he measured. The extra series impedance from that resistor you added will also affect it in ways less predictable in this crossover implementation (a hybrid series, if you want to call it that), so eliminate that as well and see where that leaves you.

Before you go re-designing the crossover, try an application of heavy wool felt on either side of the tweeter - 1/4" to 3/8" thick.

This could also be an issue with what seems to be a small variation in driver response that, due to the crossover, has a larger impact than it might seem would be expected. In that case, some crossover changes would be in order.

C

[kirknelson]

I came across a website once that had a study of the effects on different baffle shapes on measured FR. Unfortunately I can't seem to find it again. Maybe I have it bookmarked at home, I'll take a look.

Unless of course someone else here knows what I'm talking about and can link to it for me. :T

Basically it showed that it is not just the dimensions and were the driver is placed that effect BS. The edges being chamfered or rounded over, having sidewalls or leaving it open baffled, angleing the side walls or having them at 90 degrees, all cause a change in the FR.

The best (smallest FR change over free air) according to their measurements was asymetrical driver placement, smallest possible baffle wideth and 90 degree side walls with a round over. At least if I remember correctly.

[JonP]

If you want to explore baffle diffraction, go to the FRD Consortium site, and amongst the many good tools there check out the Baffle Diffraction Simulator. (BDS)

It's part of the larger suite of programs that work together, you use it to create a baffle response file, which you then add to your driver's responses, then use that with the crossover design program...

You'll be able to create a baffle and drivers, move them around to see what the effect will be, add chamfering or roundovers of various sizes, etc... Just playing around and watching the resulting graphs will show you what effects various things have and give you a feel for diffraction.

There's a program called The Edge which also does this, though I've yet to play with it..

[fjhuerta]

Update... I built the new crossover, and I listened to it for around 1 our yesterday. As far as I can tell, the midrange glare is gone....

One thing that maybe I wasn't very clear about is that I didn't change any important baffle dimensions. Baffle width is the same, driver position is the same, the distance from the top edge is the same. I only changed the design to a tower speaker (but I still placed the drivers at the same distance from the top edge), and the speaker box is deeper (when compared to PE's box). I don't think this would make any significant change at all to frequency response.

C, I had already tried felt around the tweeter. It helped a bit, but it didn't make the problem go away. I left the felt - I like the results. I also took out the extra padding resistor - it wasn't helping at all with my problem.

I'm beginning to wonder whether the drivers just need some break-in, or if, indeed, every driver batch requires slight crossover adjustments.

[cjd]

Good to hear you got things tweaked to your satisfaction. The power of having the tools to measure.

Some of these high-slope crossovers are indeed quite picky - I have pondered in the past whether they were *too* picky. Not really any good way to tell without measuring the results.

C