notch filter depth / W22 breakup

[charliemouse]

Hello,
Just wondering if anyone can suggest how deep to make a parallel notch filter (i.e. L,C&R in parallel with each other, in series between crossover and driver driver).

As I understand it, the L&C of an LCR effects the frequencies targeted, and the 'R' of an LCR effects how much the trargeted frequency is allowed to pass, lets say how 'strong' the notch filter is.

This is an LCR with no resistor making a superb job of taming the breakup of two W22s in parallel


This is using a 5.6R resistor


I notice this thread suggests 5.6ohm resisitor, but I don't understand why that value is chosen.

Surely unless the filter is adversley effecting the response of the driver in the range in which it is being used, a very low (zero) resistor value should be used.

Note, I'm using this in the following design which I'm working on... hence the bandpass explaining the lack of low frequency response from the W22s



I'll write a separate thread about these speakers which I'm aiming to use 100% passive filters (two amps needed) just as soon as I'm happy with the crossover. The odd person might recognise them being a slightly more developed version of those mentioned a thread I posted last year here, and incorporating advice offered here

[Dennis H]

L,C&R in parallel with each other, in series between crossover and driver

I think Taco's filter is LCR in series with each other, parallel with the driver, so it will be different. Also, as Capslock said in that thread, the other components of the XO can affect the notch. If you have real measurements of the SPL and impedance of the driver, trust what the sims tell you.

[jimangie1973]

The notch depth is chosen from a variety of factors. One is the best R value, or notch shape, to give flat response, or in other words best notch shape to approximate desired acoustic response. Another is the best R value to give desired phase response, so the drivers are in phase (or whatever relative phase is desired) at the crossover region.

In some cases the best R is no R at all. So in the case of the series LCR, the notch depth is limited only by the inductor series resistance.

[Coconutout]

wow they look sexy. sorry for being impatient but what would be the general budget for them?

[Dave Bullet]

Surely unless the filter is adversley effecting the response of the driver in the range in which it is being used, a very low (zero) resistor value should be used.

Hi Charlie,

I see what you are saying - notch it out as much as possible, but there are 2 reasons not to (and to include a resistor):
1. If the notch is required in the stopband, a resistor of low or no value might cause too much attenuation (ie. > desired rolloff) which may upset phase tracking with the tweeter. Ideally, you want phase tracking down to -40dB relative to average sensitivity on either side of the crossover point

2. The LC circuit will form a very low impedance at resonance. This might cause too much of an overall impedance dip. I you are aiming for say a 3+ ohm load, an LC alone, when a tweeter with low nominal resistance is used, and possibly woofer a low inductive rise could result in a resistance load too low.

PS: low pass slope looks great for the W22 pair.

PPS: I can understand why you are covering the TV with the speakers. If Only I could build speaker like that. They look awesome! ;x(

Cheers,
David.

[TacoD]

My suggestion was to use a series notch (components in series) parallel to the woofer. And also tuned for only one driver. Using more drivers and as parallel notch changes the behavior.

[charliemouse]

Lot of advice there... thanks all...
Dennis...yes I had a feeling Taco's might be different as it happens this filter uses 0.15mH and 5.7uF - very similar to his, but I just used a notch filter calculator then the good old 'passive crossover designer' excel sheet to select the exact value... and then played around to get the best values. Thanks for encouragement: I did indeed went with real measurements in the end.

TacoD: Yes, I'd imagine my bandpass on the W22s and the fact there are two would throw your values completely, but it was a good starting point - thanks. Oddly enough I couldn't get a series filter to do anything useful at all... I'll try again, and use the parallel if necessary.

Dave/jimangie1973 - both really helpful. I wasn't paying much attention to phase, I will do that. Throughout this project I've worked as much by experimentation as by theory: building lots of baffles and working both theoretically/toward perfect flat responses, but equally to what I like the sound of - which is usually anything but a perfectly-flat-response-component-stuffed-crossover. I've probably got a little fighting with phase at crossover - in particular to XXLS peerless in the M frame to the W22, where the 4th order XXLS crossover goes to the second order W22.

Generally, All crossovers excepet the W22 are 2nd order. My ears are just not happy with anything more abrupt. I've read similar elsewhere, so I'm not being too strange on this. If I had a Behringer I'd love to compare it to 4th order digital crossovers, but I'm being cheap - and to be honest I'm more than happy with the passive crossovers i've tried to date.

Dave, again, I'm quite sure you could build something like that! I've nothing but a makita circular saw, a makita jig saw, a cheap 30 pound Bosch router, and a few clamps...and my Grandad's old portable work bench which wobbles a lot. I just used B&Q brazillian ply (probably softwood with hard wood face, despite it being labled 'hardwood' (generic DIY store quality: gaps in intermediary layers to fill, and average surface, sawmarks on surface veneer etc). Just sanded with 120 grit (higher stopped stain taking) and used Ronseal 'rosewood - refined woodstain' (6 coats!) and briwax (5 coats).

Dave, yet again, I'm not sure about the W22 response: they're a little enthusiastic between 300-500Hz - presumably related to baffle dimensions, but I'm working on whether I'll tame that. And yes... you can tell how interested I am in TV I remember buying that and being excited it was widescreen - glad drivers don't evolve at that rate...enthusiasts like us would all be bankrupt.

Coconutout - yea, I quite like appearance of them, but it was more function than form: I had to have WW-WMTMW, had to have tweeter at ear height obviously, wanted two XXLS (simlar to linkwitz designs), really like mtm type format as encouraged by PaulW's wings design etc, I had to have M frame physically separate so I could move the blasted things, etc... I just made a few wallpaper cutouts trying various arcs and shapes for side panels. Inevitably it looks very much like Orion / BOB designs, as they're also presumably designed functionally around a dipole baffle... and given there are certain shapes and angles that just 'work' (usually they're like something in nature I think), there isn't really a lot of choice for the side panels. I showed my various designs for side panels to a few artist friends, who all chose the same angles. Thanks for positive comment on appearance though: appreciated
Cost? Err, 4x Peerless XXLS, 4x Seas W22, 4x visaton TI100, 2xScanspeak 98000. They vary so much by country I couldn't suggest cost. All I can say is I have a pair for quads - the 22L. They're supposedly rather good and like the electrostatics and 'blah blah' in reviews. Cost £900 I think. They sound like a pair of PC speakers, with no bass, no refinement of trebble, and a sound that clearly comes from a box... when compared to these speakers which I'd say cost less than 1.5x the amount the Quads cost. Admitedly, the cost of the drivers in the quads was probably about £50, given they feed trees, plastic and wire in one end of a factory in china, and get speakers out the other. You could save a lot getting the standard seas L22 drivers. But the phase plug looks like an aircraft engine. It costs a lot more if you buy L18s first, cheaper seas tweeters, and Fountek neoCD3 ribbons before buying scanspeak tweeters Aw crap. I should have just started a thread on the speakers I've written far too much here. That serves me right for writing when I get in from an evening out :/

...oh an in relation to the original question, I'll look at effect on phase, and select as low a resistor as possible without any detrimental effect elsewhere. Right, onto taming the TI100s better tomorrow 8O

[Dave Bullet]

Hi Charlie,

Thanks for the detailed build process. I suppose patience is the key here and having a good vision of your finished build.

Dave, yet again, I'm not sure about the W22 response: they're a little enthusiastic between 300-500Hz - presumably related to baffle dimensions, but I'm working on whether I'll tame that

A 300Hz bump = 78cm distance. I'd measure from the centre of the top W22 down to see if it meets with the XXLS enclosure to see if some sort of cavity effect. My limited experience says cavity effects cause a dip not a bump, but maybe a possibility. I can't quite see the XXLS enclosure (my Computer CRT monitor is over 10 years old.. you can tell I put my money into speakers and not computers!) to temporarily check, place a sheet of MDF over any enclosure void at the bottom and re-measure.

Another possibility is a baffle resonance at that frequency, causing the baffle to vibrate in unison with the W22s at that frequency providing a loading effect (ie. gain). Try and temporarily change the mass of the OB - either by bracing to the wall (ie. piece of timber) or blu-tak some heavy mass to the back.

Another test is an impedance sweep. This can highlight any loading effects on the driver due to resonances or "back pressures". Look for an impedance ripple in that area.

I'm not sure of your measurement conditions, but maybe it might also be some measurement artifact.

I haven't looked at others' W22 measurements to see if that is typical of W22s. Seas have had problems with consistency in breakup nodes of some of its metal coned drivers, however down that low (ie. 300Hz) they should be pretty consistent. I don't think Zaph has any tests or published data for the W22.

Cheers,
David.

[charliemouse]

Dave, nice one with that. Lots of ideas to play with. I'll reply and add some waffly meanderings as ever.
I'll certainly have a look at where any peaks might be coming from with respect to the areas you suggest - those measurements were fairly close to drivers, things get much more messy as measurements are taken meters away: but then that is normal. I'm not sure that reflects the baffle any more than the room. It is still nice to have quads as a reference, which suffer many similar distortions as measurement distance increases.
Again, though I won't do this ahead of time, I'll put measurements etc on the board when I'm happy with crossovers. I really don't want to write a thread on it at the moment whilst I'm still experimenting. To me the response I'm getting at a couple of meters are now very flat indeed, and sound just superb, but others may think they're wobbly. I know the main thing is the wmtmw and dipole are just so much more room tollerant than closed boxes i tried before. I know you can't put them up against a wall etc.

the only other thing that might screw the measurements is my mic. I find i get better measurements with an EV vocal mike, but like the difference between a sure SM57 and 58, vocal mikes have odd characteritics themselves at very low and high frequencies. I still find the good quality vocal much more useful than a lower quality instrument mike. Yes, it's better very low/very high, but not flat enough over the rest. I'm afraid like your monitor, my measurements tend to be with various mikes and interpreted with knoledge of the mic's responses. I trust ears in the end. I'm sure experts will cringe.
Still, I'm happy... no, ecstatic with the speakers I have now, and I suppose they only have to make me and visitors happy! I also want to have an option on the crossover to select a slightly inaccurate response: for these as for other things i tried earlier, I find around the 6/7KHz frequency a minutely increased response increases the apparent clarity hugely. Again, some people will cringe, I want a 'faithful resproduction' option too. As was commented on elsewhere, on here, or it might have been diyaudio, the ti100's are very clear - and the 6/7KHz distortion was commented on as possibly being to do with that. I don't want a non-flat response due to their distortion, but I do due to crossover design letting the tweeters do a little more there.