Crossover - physical layout/construction help

[scottvalentin]

Good morning all,
I am near completion of my first self-designed 2-way after running them with a Mini-DSP for awhile testing crossover points.

I completed one crossover board but the speaker sounded really lifeless, so I took a measurement. It is the third attachment. As you can see, the tweeter is not behaving like it should. Note that my sim files are from in-box, in room measurements.

On the physical layout side, I was surprised after doing it that all of the ground points ( -in from amp, shunt connection from tweeter notch, shunt from woofer cap, -woofer, -tweeter) were all at one connection. So 5 points all together that were at one connection on the board. Given the schematic, is this correct? I also think I had some cold solder joints. Would a cold solder joint have affected it to have that kind of response? I have started disassembling the board, but wanted to be sure that I had put it together correctly for when I re-do it. The woofer rolled off correctly, so the problem is in the tweeter section.

10 uf cap, .35mh coil and 2R in series to ground, 6.8uf cap, 16R in series from 10uf cap, then I soldered all 3 leads of the .2mh coil, 10R resistor, and 50uf cap together and put that in series from the 16R resistor, and then tweeter + is off of that.

Thanks for any and all help!

[Chris Roemer]

Not sure what you're notching there (1.6k ?), but your filter looks like about -11 to 14dB of attenuation, rolling off 2nd order around 1.7k?
Your measurement shows a broad (-5dB, is it?) dip from 3-6k. That would sound lifeless to MY ears, I guess.
I think your plan is good. Sim looks good. Something wired wrong?
Can you DATS it? Your HP section SHOULD have rather low Z near 3-4k (as your sim shows), which should NOT be the low pt. in your tweeter's response - right there.

BTW, your "notch" (LCR on tweeter) should NOT go to gnd. - according to your sim, but the shunt coil (w/2n resistor) SHOULD.

[scottvalentin]

Not sure what you're notching there (1.6k ?), but your filter looks like about -11 to 14dB of attenuation, rolling off 2nd order around 1.7k?
Your measurement shows a broad (-5dB, is it?) dip from 3-6k. That would sound lifeless to MY ears, I guess.
I think your plan is good. Sim looks good. Something wired wrong?
Can you DATS it? Your HP section SHOULD have rather low Z near 3-4k (as your sim shows), which should NOT be the low pt. in your tweeter's response - right there.

BTW, your "notch" (LCR on tweeter) should NOT go to gnd. - according to your sim, but the shunt coil (w/2n resistor) SHOULD.

Thanks Chris I am notching the big peak between 1k and 2k on the raw tweeter response which created problems because of the dip between 2k and 3k on the raw response.

I don't have the lcr going to ground. But I did have some poor solder joints so was wondering if it could be that or if I had laid something out incorrectly.

I don't have dats but do have the diy impedence jig - problem is that I get the rising impedence problem due to crosstalk from the uca222 I use.

Does having all the ground points as one electrical point make sense? Just want to rule out layout error before I try and re-solder everything.

[filmslayer]

no problem with the one spot for ground Scott . my CC XO

[Chris Roemer]

Sure, a single (common) gnd. is very common. I'd THINK that (w/the right component values) you wouldn't NEED that notch. You're prob'ly (somewhat) fighting the Fs Z-peak trying to roll off the tweeter where it's putput is the highest.

Try an L-pad to get -12dB (or so) attenuation for your tweeter (which tweeter is it, btw?) instead of just all series resistance. The shunt (parallel) leg in an L-pad helps tame a tweeter's (or mid's) Z-peak at Fs, which makes it easier to roll it off the way you'd like.

[scottvalentin]

Hi Chris Chris Roemer
WOrking with the SB26 STAC from SB Acoustics. Any suggestions? Take out the notch altogether and just add a parallel resistor?
EDIT: Son of a gun I just took it out and it affected the response very little - didn't need those parts I'm thinking!

[scottvalentin]

Kornbread ok thanks, glad to know that it is a regular occurence - I will resolder and leave out the notch components.

[Chris Roemer]

(In your sim) you can effectively turn the LCR (notch) filter off and on by taking the R value WAY down (like 0.01ohms - which bypasses the entire LCR). Easier than adding/deleting all the separate parts. Make the resistor 1000ohms and the notch becomes "bottomless".

[scottvalentin]

I'm at a loss, I can't get this tweeter section to work. I've resoldered everything and tried again. Same response as above.

So, I took it all apart and measured just the tweeter, raw response in teal?. Then I added just the series components first (10uf, 6.8uf, 16R) and got about the expected response in green, then added the coil and resistor to ground. I got the purple response which is NOT what my sim shows. So I played in PCD trying to figure what would give a response like that, and it seemed like maybe the coil was not what it should be, so I swapped coils and got the same response (not shown). As best as I can figure through the PCD fiddling, it's closest to having a 0 instead of the .35mH coil, but my soldering connections are solid and I tried two different coils.


Any suggestions? I don't know what else to try. On a side note, my soldering skills have improved, so I did manage something positive here!

[scottvalentin]

My post didn't show up on front page, so this is a bump.

[jhollander]

I think your troubleshooting is spot on, just start breaking down into smaller pieces.

Try recreating your circuit in xsim and see if you get the same as PCD. I assume you took the notch circuit out of win PCD while you were checking the response and you are measuring the components. You can also measure the impedance and compare that to winPCD. Sometimes it's a bit easier than setting up the microphone.

[scottvalentin]

I have measured the resistors and they are all a bit higher than spec values but within 10%. I'm not sure how to measure the caps or inductors as I don't think my multimeter has that function.

I haven't ever tried xsim . I have an impedence jig so will see if that is still working later!

I will also take a pic of the crossover maybe I'm missing something obvious.

Thanks John for the input

[scottvalentin]

Here are two pics of the crossover, of of the overall board and one showing the inductor connections.

Thanks

[jhollander]

X-o Looks like what you were describing. I assume you scraped off the enamel on the inductor? Only other issue would be to make sure your measurement distance matches the same as what the initial measurements were.

[lasse]

I once encountered something similar. Turned out to be a dry solder on one of the input terminals to the speaker. Have you checked the solders on the driver and the terminals? Just checking..... Regards// lasse