Couple quick questions re my project

Been away for awhile working on my spouse-acceptable 4-horn speaker and it’s done, more or less - I’ll have to post a show-and-tell sometime soon. (I’m certainly no expert, but it’s been an interesting project, and maybe someone can learn from my mistakes.)

Anyway, _with equalization_, using cheapo measurement tools (Dayton audio imm-6 mic) I can get a decent looking response curve composite of the 4 drivers, using pink noise and the FFT part of the AudioTools app on an iPad:

...but in listening tests, the vocals sound pretty well defined but “distant“ and a little thin (like the singer was literally some ways behind the rest of the band.)

Since I’m after good _sound_, not a pretty eq graph, I’ll probably try playing with the equalization and see what sounds best, but I’m curious. Without spending hundreds on good test gear I’ll use once, am I doing it right? ie _Pink_ noise, with an FFT graph? Or should I try to equalize to white noise, or with the RTA module?
 

Attachments

  • 92DC8E6C-FC24-44F1-BA76-9432C978D3C2.png
    92DC8E6C-FC24-44F1-BA76-9432C978D3C2.png
    406.4 KB · Views: 189
Last edited by a moderator:
I'm not sure I can see what you are showing with this image. The plots overlap rather than cross so maybe you could explain this.

Secondly, horns are not that simple to cross. Without either polar plots, or at least a photo or description of them it would be difficult even to begin to answer your question.
 
Pdq2,

pink noise has got a specific downward tilt response that has to be compensated for in the graph. Audiomatica Clio 4.51 has got one-third octave RTA analysis and pink noise as stimulus to correctly present a measurement. This may or may not be an issue in your case. Read the manual.

Secondly, your individual driver responses do not cross as they should, which is what AllenB has already brought to attention.
 
Pdq2,

pink noise has got a specific downward tilt response that has to be compensated for in the graph. Audiomatica Clio 4.51 has got one-third octave RTA analysis and pink noise as stimulus to correctly present a measurement. This may or may not be an issue in your case. Read the manual.

Secondly, your individual driver responses do not cross as they should, which is what AllenB has already brought to attention.

With regard to your second comment, I think what you’re saying is that they should cross at -3 db, but that’s not what this is. This is a superimposed graph of each driver, tested individually with equalization for each to make the flattest response, but without a crossover. This allowed me to determine (or guess) where the best crossover points would be (I’m using a (cheap) adjustable 4 way electronic (not digital) crossover). So, based on this combined graph, I picked 5K and 1K for the tweet/mid and mid/mid-woofer crossover.

But I guess my question is, if you use pink noise as a tool to set equalization, should it show the downward slope? Or be flat? And should I be using this FFT graphing or RTA? I don’t have Audiomatica Clio, so I can’t read the manual, unless it’s online somewhere.
 
FWIW, this is the response I get when I put it all together, 4 drivers+ equalization + crossover (crossovers at 180, 1K and 5K):

attachment.php


(There is still the problematic dip around 117 Hz - can’t seem to get rid of this, I think it’s from a bad design decision in my subwoofer - and adding the crossover seems to reduce the low end response from the subwoofer, as though there is a low end cutoff. The _crossover_ does have one, but it’s switched off.)

But again, before I get too deep in the details, my question is should I be shooting for a flat frequency response or a downsloping one, when using pink noise?
 

Attachments

  • DCA48BF2-429A-471A-AC30-E09B08311754.png
    DCA48BF2-429A-471A-AC30-E09B08311754.png
    337.4 KB · Views: 206
Is this all being measured / setup in room at seating ? If yes then try moving the speakers outdoors as far away from walls / boundaries as possible and redo it all to measure flat. Then put them back in the room and have a listen.

The last graph posted shows no subwoofer at all. It hasn't been high passed by mistake or left switched off or something ?

Rob.
 
Is this all being measured / setup in room at seating ? If yes then try moving the speakers outdoors as far away from walls / boundaries as possible and redo it all to measure flat. Then put them back in the room and have a listen.

The last graph posted shows no subwoofer at all. It hasn't been high passed by mistake or left switched off or something ?

Rob.

Yup, in the room where it will live. And I appreciate your suggestion, but part of it is built-in, so I can’t really move it at this point.

And I don’t know what the deal is with the low response on the last graph, but I agree - I need to check my wiring again.
 
In that case you need to set up the system to have a room curve. Google the JBL synthesis curve for a start and see if that improves things. Flat in room sounds thin / lacking body / no slam etc in my experience.

Have a read of Floyd Tooles book Sound Reproduction (there might be some pdf papers available on line) It will explain better than I can why a speaker should not measure flat in room.

Read this


Rob.
 
Last edited:
Rob,

Will do - thanks. (Unfortunately, your link doesn’t work for me - “not found”)

And I appreciate the earlier comment about checking the phase of each of the drivers - I will check that too. And finally, I will look again at the discrepancy in the sub woofer response.

Got out-of-town visitors right now, but when they leave, I’ll have some work to do and come back and report here.
 
Just a quick update - turns out I _did_have one driver out of phase, and things sounded better after correction, but I’m still not quite happy with it and I’m thinking there may be significant time alignment issues with the larger horns. (The low-midrange horn is around 4.5 feet long, and the subwoofer more like 12 feet - my home theater receiver may be compensating for the latter (it has Audessy) but not for the low-midrange horn).

The low end FFT response curve was a problem with the iPad I was using, I think - Apple sets up a strong low-bass rolloff/cutoff in iPhones and iPads, and once I made sure that the app was using the Dayton imm-6 mic and its cal file, the low end reappeared in my FFT tracings.

I’m now considering getting a MiniDSP and uMic and using the REW software. It’s about $200 total, but it would replace the equalizer, external crossover, and the imm-6 I’m using now and I could probably get half of that back reselling the older stuff on eBay.

Oh well, it’s a hobby, right?
 
Umik, minidsp, REW. Great idea, exactly what I use and they are all very easy to use. Also keep in mind that with horns, you will want a room curve at listening position that leans down even more in the treble than a standard room curve. I take 6 measurements in the listening position moving the mic several inches each time. Then click the "all spl" tab and average them. You'll want something like a flat response up to 1k, then it will slant downwards towards 20k, ending up down maybe 15db at 20k.