diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Analog Line Level (http://www.diyaudio.com/forums/analog-line-level/)
-   -   Adjustable Linkwitz-Riley filter: Help needed with SPICE (http://www.diyaudio.com/forums/analog-line-level/210740-adjustable-linkwitz-riley-filter-help-needed-spice.html)

NebuK 14th April 2012 06:27 AM

Adjustable Linkwitz-Riley filter: Help needed with SPICE
 
Hello,

after having searched through various resources for a nice LR filter design online i've started to make my own schematic and PCB - as i'm still a beginner i've tried to keep close to what i've found. My goal is a LR-Filter that can nicely be adjusted by plugging four precision resistors mounted on a IC socket into another IC socket. After having finished the PCB design i had the idea to simulate my schematic to save the hassle of building a prototype that most probably wouldn't work yet ;P. So i read a little into ltspice (which seemed quite useable from what i heared) and tried:

http://pics.kanojo.de/acrossover-spice.png

then i clicked simulate, picked my two outputs and BAAAM:

http://pics.kanojo.de/acrossover-plot.png

Nope ... thats not what a good crossover should look like. Maybe for some magic strange usecase, but certainly not for audio.

As a friend i asked presented me with a working LR SPICE model i'm not totally out of options here, however ... i'd really like to understand what the hell is going on here and why this isn't working? Also, i'd be nice if i wouldn't have to re-do the PCB :P. By the way: the PCB is designed to be home-etchable and... well, if someones interested to have a look, it's here:
ActiveCrossOver - Adjustable active 24db LR crossover


Thanks in advance - and best regards!
- NebuK

godfrey 14th April 2012 08:05 AM

Are you trying to make second order (12dB/oct) or fourth order (24dB/oct) filters?
That circuit doesn't resemble any kind of filter I've seen before. Where did you find the design? Links to anything similar?

NebuK 14th April 2012 08:14 AM

Heyyas,

i've seen the circuit in
http://www.rane.com/pdf/linriley.pdf
as a abstract concept. Later, i've found another reference in the manual of some product not available in europe, its:
http://www.marchandelec.com/ftp/xm1man.pdf

I was under them impression that what i schematic'ed was a 24db/oct L-R filter... :/

Thanks so far! :P

godfrey 14th April 2012 09:14 AM

2 Attachment(s)
OK, I simulated the circuit and got reasonable results. Maybe LTSpice got it's panties in a twist because you didn't put a ground connection at the midpoint of your power supply between V1 and V3. I've been caught like that before, but wouldn't have thought it would make a difference with opamps.

In my simulation, the phase response of the high-pass and low pass are identical (good), but the amplitudes cross at -9dB (bad, should be -6dB).

It looks like you copied the circuit from the second link. I notice the resistor ratios are different to what's recommended in the first link. Maybe if you change to those ratios, you'll get the correct response, crossing over at -6dB.

p.s. Nice filter design! Thanks for sharing.:)

NebuK 14th April 2012 12:16 PM

Mh, i was trying to get my model working in ltspice first ... i've tried both, connecting the ground between my voltage sources as well as using a (ideal) opamp model without supply voltage connections ... both resulted in strange and even stranger plots. I'm still totally at loss at whats going on here. Haven't tried the resistor ratios yet since i wanted to get the rest sorted out first.

As for the filter design itself, i was planning to use this for a flexible, modular crossover and subwoof equalisation system. So i could easy and cost-effectively activate any speaker designs as well as use it as a standalone subwoofer eq (in conjunction with a appropriate amp)... Once this works (and maybe i'll have a bunch of questions more, be warned ;P) i wanted to make a nice pcb design so everyone could use it (if it doesn't suck badly ... i'm still a beginner, so ... ;P).

Thanks a bunch! Regards
- NebuK

AndrewT 14th April 2012 12:57 PM

Which resistors are changed to change the filter frequency?
The seven 24k9, or the one 5k23, or the four 105k?
Or all 12?

Why does Godfrey's show eight 24k9 resistors?

NebuK 14th April 2012 01:09 PM

Hey,

uh, unless i'm blind Godfreys design shows 5 (not eight) 24.9K resistors. Its 5 and not 6 because i believe be left out the input buffer(?) opamp - thats why he has only 5 opamps as opposed to my 6(??), right?

Anyways, i was under the impression that the four 105K resistors would need changing for adjusting the crossover frequency.

Thanks, Regards
- NebuK

PS: Just tinkered with leaving out the input buffer opamp just as godfrey did, now my curve looks a little better, with the crossover point at -8.8db / 459hZ... but the low-pass starts rising at ~18k again. It decreases until ~18k / -122db, then rises at a not quite so fast slope... i'm even more confused now?!

AndrewT 14th April 2012 01:12 PM

49k9 should be two 24k9 connected in series giving 49k8
There's 100r error there.

A four gang resistive adjuster to change the frequency. Can you get sufficiently accurate tracking of the four resistive elements to maintain your 4pole LR characteristic?

NebuK 15th April 2012 07:26 AM

Hey,

i've corrected the 49K9 to 49K8, sadly the rise at ~18k in the lowpass curve didn't vanish :/. I'm now toying around empirically since my understanding of how this circuit works stops there ...

As for the resistive adjustment, i planned to get a set of resistors and use ic sockets to plug the resistors. So i only get a stepped adjustable system, but better than nothing! If this works and proves nice, i thought about either a digital pot (maxim has a few nice ones i think) or resistor banks and switches just as rane proposes...

Thanks once again... :P Regards
-NebuK

NebuK 15th April 2012 08:08 AM

Also, very strange: I've just tried to copy godfreys circuit 1:1 in ltspice using the ideal opamp (opamp.sub with .lib opamp.sub)... all i get at *any* point in the circuit is a linear frequency response Oo

*EDIT: Okay, that was ******** just now ... and i've found my mistake. Some of my plots hat autoranged mangnitude axis with so many ticks that i wouldn't see the bottom is -500 or something dB - which left me under the impression that my filter was not steep enough ... now i'm going to niceify everything. SOLVED :P Yay, thanks so far! :) (but the strange hump in my original schematics lowpass remains :()


All times are GMT. The time now is 10:22 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2