notch filter help/suggestions- thx

tajanes

I've been working on developing a notch filter to reduce the mid's rise at 5,000 Hz, from chart the blue is pre-notch, the green is w notch R 12.5 ohm, cap 32uF, and L .03, and the purple notch is R 12.5ohm, cap 10.uF2, and L .08- but I cannot seem to get it down and narrow it out further (I've been adjusting values but can't seem to pull out this rise any further- ideas?  While the purple level is close its too wide- any suggestions- thanks 

its an 8ohm mid (actually a coax but this is just a chart of the mid), 

rjj45

You are on the right track - the ratio of cap <-> coil controls the frequency, and the resistor controls the Q - whether narrow or broad. You might look at a Zobel - I've found that when a crossover just won't behave on the "top end" of a woofer, usually a Zobel will help out.

Wolf

Let's start from the beginning...

Which type of notch are you using? Parallel (with each other placed in series with driver)? or Series (with each other placed across the driver)?

Where is the tentative xover point to be? Active or passive?

If this is a parallel notch, it's not unheard of to have a 40 or 50 ohm resistor there. In your simulation, set it really high to 1000 or something in that range. Then adjust the L/C to get it centered on the peak, and the dip in the sim will be where you want it. Then just reduce the resistor value until it's dialed in. This is mainly if it just needs to be flattened out as far as this process goes.

If this is above the xover point, and it is a nasty breakup, then you really should suppress it to -40dB from reference with the xover applied. That will minimize it's audibility. -25dB minimum, -40dB desired, and -50dB makes it no longer a concern at all.

If this is a series notch, and you are running active xovers, the circuit will not have the upstream voltage divider necessary for it to work appropriately. I have however found in some cases where a brute-force LC shunt or 'bottomless trap' (with only coil's DCR for R value) is most effective at reducing higher Q peaks and/or breakups.

tajanes

a) parallel, b) tentative xover depending on net rolloff of mid net the notch filter, and passive btwn mid and tweeter c) this is the program http://www.mh-audio.nl/parallelnotchfilter.asp , d) this notch filter will be just below the xover point (using the modified fall off to set the tweeter HP level and filter slope)

I appreciate you help with this,
Andy

tajanes

fyi, bummped resistor to 46.5, cap at 20 and L .04, getting closer... 
going in right direction- thanks

jhollander

Can you post the FRD file somewhere?

tajanes

I can attach a REW file to an email (but not here?) if you have REW send me an email at taj.janes@gmail.com and I'll attach and reply

here is the last freq chart per 46.5 R, cap 20 and L .04

duanebro

Is this breakup at 5K? I would try to cross below this if I could.

jhollander

I've never messed with REW. Ed P could probably tell us if that's an FRD file that needs to be exported from REW.

Wolf

I'm with Duane here, as I feel that 5.5k is likely way to high, even for a coaxial of the type you have there.

I'm betting this is the same coaxial that Keith Etheredge used in his 'Keeping up with the Jones' winning entry build.

He did a good job making it smooth out, though did hammer it into place. It might provide some ideas for what worked well.

tajanes

thanks, I'll take a look
what I'm concerned with and trying different component values to try to work around, is the notch filter is causing wave disruption / ripple- take a look at first posted pic, the blue is pre-notch, and the green and purple are with notch filters- thinking I have a miss-match of the cap and coil values causing ripple 2000 to 4000 ?

dynamo

What kind of driver is this? That looks like the response of a silver flute driver, which I cave found to sound poor if pushed that high up anyway. I’d hammer it flat at a minimum, or even make it a dip, with a parallel notch, then still cross it steep before that point. At least with the sf drivers with a nearly identical breakup, whatever is going on with the cone breakup up there is ugly sounding imo.

rjj45

tajanes said:

thanks, I'll take a look
what I'm concerned with and trying different component values to try to work around, is the notch filter is causing wave disruption / ripple- take a look at first posted pic, the blue is pre-notch, and the green and purple are with notch filters- thinking I have a miss-match of the cap and coil values causing ripple 2000 to 4000 ?

Welcome to "Q". Since you want the notch to cover 3.5K to 8K, of course it will affect frequencies above and below that range. XSim should be able to model this pretty easily, you can use the sliders to change any of the values and see results in near real time.

tajanes

I really appreciate all suggestions and info !

driver is a celestion coax FTX0617

tried a zobel, not much dif, but putting toys away for the night and starting up in am with new componet values I'm now seeing light at the end of the tunnel - still have work to do, but its getting there -this chart is w inclusion of tw 2nd L-R above notch level on mid (using notch to let mid roll off)

jhollander

Just a caution to make sure you test in place.  With box resonance an rear wave addition the results may not match your sim.

tajanes

this last graph (previous post) was measured at listening area, don't see box reasonance issue as its fairly open design and padded, below is chart from rear (top speaker only) at about 18 inches

jhollander

That rear response looks pretty good imo.  What's the front response at 30 inches or so look like?  Also what's the vertical scale divisions?

tajanes

Vertical scales were 10, so I've got some more work to do.  I have a few more parts ordered...

tajanes

notch filter ended up with: 60ohm R, cap 8.2uf, and L 0.1, also put in an Lpad (from an old project) for the coax tw, and ran the mid with just the notch, put in a 3rd order for tw.  Measurements shown at 15deg approx 18in.  Coax dips around 2150 so eq'ed a bit w dsp.

tajanes

and with the dipole Ws...