I've long been curious about systems that use a 8" (Lowther or similar) in a front horn + a separate bass box. I'm trying to adapt this basic idea to use a smaller, no-whizzer driver, because they seem to have better response. I realise this won't be as efficient a system, but I think this is the compromise I prefer. It will also cost much less :)

For a trial, I built a rough approximation of a Le Cleac'h horn from light plywood, using the excellent export function of Hornresp: 40*80cm mouth, loaded with a basic pro 4" driver.

BLUE: 4" driver on horn, with just a HP filter (amp --> 15ohm/12uF in parallel to each other --> driver)
RED: same horn and filter with a less efficient 4" driver, graphs offset for clarity (this is a generic replacement speaker from an electronics shop) + bass section*
OVERLAID: the first, blue plot, translucent & resized to overlay the plot of a Lowther PM2A in a professional Le Cleac'h horn (from the excellent Azurahorn website)

Questions:

What would the notch around ~7.5kHz be from? It is there on every on axis measurement I take. Off axis it shifts / goes away.

Off axis, I get a -5dB notch 450-500Hz. Is that something that may go away as I build a better horn?

Off axis, the 8-16kHz range looks pretty spiky, the pic I've posted is the nicest one for that range. Should I just mentally apply smoothing to these graphs and forget about it, or is this something I can tweak and improve?

*Bass section for tests is 1 Peerless and 1 Vifa 6.5" in BR, via 5.6mH coil. These are running in parallel, so sensitivity here = 94dB based on specs.

 RE: rough front horn (semi fullrange) - SpeakerScott 12:38:35 11/13/14 (4)
 RE: rough front horn (semi fullrange) - hollowboy 16:42:48 11/13/14 (3)
 RE: rough front horn (semi fullrange) - SpeakerScott 16:57:51 11/13/14 (2)

...

Show full thread outline!

 RE: rough front horn (semi fullrange) - SpeakerScott 12:38:35 11/13/14 (4)
 RE: rough front horn (semi fullrange) - hollowboy 16:42:48 11/13/14 (3)
 RE: rough front horn (semi fullrange) - SpeakerScott 16:57:51 11/13/14 (2)
 RE: rough front horn (semi fullrange) - hollowboy 03:08:31 11/19/14 (1)
 RE: rough front horn (semi fullrange) - Joseph Crowe 04:35:56 11/19/14 (0)
 RE: rough front horn (semi fullrange) - Mr_Steady 17:03:47 11/12/14 (1)
 RE: rough front horn (semi fullrange) - Brian Levy 07:35:56 11/13/14 (0)
 RE: rough front horn (semi fullrange) - SpeakerScott 19:14:47 11/11/14 (3)
 RE: rough front horn (semi fullrange) - hollowboy 03:34:06 11/12/14 (2)
 RE: rough front horn (semi fullrange) - Mr_Steady 17:15:15 11/12/14 (1)
 RE: rough front horn (semi fullrange) - hollowboy 01:51:33 11/19/14 (0)

Not a drawing, but this shows the shape / my high tech construction.

It's a 74.5mm*74.5mm throat, going to a 400mm*800mm mouth
--> that's about 3" and 16*32" for all y'all in the USA, Myanmar and Liberia

The pictured sides stay more or less flat, the other sides curve. I just took the numbers from Hornresp (a high T Le Cleac'h with a cutoff of 200Hz) and tweaked them slightly:

- oversized the length of the flat sides by 150mm, so it looks a little like a Smith Horn (hoping to extend low frequency support and vertical pattern control)
- oversized the cross sectional area slightly, so I could later fill in the corners to get an octagonal rather than square throat ...which I've now done, but not yet measured.

No impedance plot, I'm not equipped. I'm keen to get a better measuring system sometime. Happy to take recommendations.

• http://www.commonsenseaudio.com/nirvana.html

Thanks for the info.

I measure with white or pink noise. I capture with Audiotool + Dayton Audio iMM-6 + phone, load the file to my PC and graph it in a spreadsheet.

In the graph I originally posted, it is a standard log scale, but the numbering is set by where I decide to chop the axis off. So, each minor gridline from 2kHz to 20kHz was a 2kHz increment. The notch is deepest at 7.2kHz.

1. This got me thinking in good directions. 7.2kHz = wavelength 47mm, which is a match for the driver radius. It may have been from a leak: adding a gasket has improved it. The throat interface, by the way, is rough and basic :)

2. Nope, just noise, so the window is totally open. The low end will be lumpy from room effects, the high end will have reflection artefacts from the mic apparatus, phone, and my hands.

I could improve the HF part by mounting the mic in a foam block and/or adding a lead to separate the mic.

3. I've heard good things about REW. I'm also tempted to get the Dayton Audio OmniMic V2 system: it seems like a more costly but easier way to improve my measurement capability.

The attached image (with the axis numbered more conventionally) shows where I'm at: added a gasket, filled in horn corners (now it has an octagonal rather than square throat), dragged an old pair of 12" boxes out, am running them in parallel for extra sensitivity. The horn is about as flat as I could hope for at this stage, but is still about 3dB hotter than the bass. This is much more efficiency than I was expecting: looks like it wants a pair of 15" direct radiators (or equivalent), separate amplification, or padding of some sort.

I purchased four of these 4" drivers (and four 3") so that I can sacrifice one or two with non reversible treatments. It looks like I can aim to add resonance-eating cone treatments which cost a dB or three, and this would make it easier to integrate with a reasonably sized bass system.

Thanks for the info. A quick look at some REW tutorials make it seem achievable, not too crazy a learning curve.

I tried the ground measurements (but not yet outdoors) and some other tricks to get more intelligible plots with my current gear. Some physical (placement) changes plus some smoothing make it easier to follow what's going on. Louder test levels help too, reducing the hash a little, and also gets above any LF rumble (nearby cars etc).

Sadly, my last few tweaks haven't done much good: I cut a new throat, tried a different gasket and crossover, a few other things. Apart from removing that original (leak?) notch just above 7kHz, nothing really improves what I initially measured. I actually made the lows worse - the new gasket material I tried is too leaky to LF, I think.

Hopefully, a more solid and well executed horn will be an improvement, as will dialing in the filter, but I think the response I've already got from 1-10kHZ is already pretty decent, and extra tweaking will have diminishing returns there.