Speaker Workshop for passive line-level xo design

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Sorry about this basic question, but I couldn't find the answer anywhere.

Can Speaker Workshop be used to design and optimise passive line-level xo's?

Basically, for the FR curves of the drivers, I'll use an amp and a mic as usual. For the impedance curves, I'll make it a flat impedance of value equal to whatever is my power amp's input impedance. (Let's assume it is directly coupled and resistive.) With these two sets of curves for each driver, can I then design an optimised passive line-level xo?

Of course, one of the reasons to go line-level is to get out of inductors. Therefore, I'll probably want to design purely RC networks. How far can I go with such networks? Is it practical to do 4th order slopes with pure RC?

Tarun
 
I don't think this would work except for the most basic, 6dB per octave xover. Just having resistors and caps to work with would not give you the ability to contour the response of the drivers. And the insertion losses of using multiple stages of caps and resistors would be high. That is why people use active line-level crossovers.
 
jbateman said:
I don't think this would work except for the most basic, 6dB per octave xover. Just having resistors and caps to work with would not give you the ability to contour the response of the drivers. And the insertion losses of using multiple stages of caps and resistors would be high. That is why people use active line-level crossovers.
Thanks for taking the trouble, Jbateman.

Couldn't I do at least a 2nd order with RC? And would insertion losses really matter if all this is happening just before the power amp in the same chassis?

And could I do better line-level xo if I used LCR? Maybe SW would help me do that too?
 
it is certainly possible to daisy chain RCs to get a fourth order slope. The catch is it is VERY lossy (You'll need lots more drive voltage) and the response is dependent on the input impedance of the following stage. So, if you design for one amp and then upgrade to another with different input impedance, your filter needs to be reworked.

LCR would work as well, with the same losses and huge inductors.

You're better off going active if you need fourth order response. It is not that hard and more flexible. See Rod Elliot's site, the Linkwitz site or get hold of Don Lancaster's book "The Active Filter Cookbook"
 
Ex-Moderator
Joined 2002
Take a look at Planet 10's website here which has some articles on PLLXO's. You'll be limited to 2nd order filters and success will depend on the amps input impedance and the power output of your pre-amp.

One word of caution, be sure to insert a cap (hi-pass filter) between your amp and hi-frequency driver(s). This will of course affect the final filter response and will need to be taken into account. However, you do not want to leave your tweeter unprotected from the low frequency signalsthat can be generated in the amp.
 
Ex-Moderator
Joined 2002
Tarun

PLLXOs are only really any good for low order crossovers. For a 4th order, you will be losing so much signal that you will need a buffer stage after your crossover, so you might as well go properly active.

Unfortunately, SW doesn't seem to support these either.
 
Thanks for the responses, guys. :)

Let me first clarify one point about my idea. I was thinking of this with the intention of building active loudspeakers, with the PLLXO, power amps and the PSU for the amps inside the speaker enclosure. Therefore, (i) I'd have full control over the power amp's input impedance, and (ii) that power amp would never change to a Krell or a Canasya tomorrow.

BobEllis said:
it is certainly possible to daisy chain RCs to get a fourth order slope. The catch is it is VERY lossy
I was wondering whether this higher drive voltage would still remain well within the limits of modern opamp-based preamps. Most preamps probably reach less than 10% of their line-level voltage limit before the listening levels in most homes reaches high volumes. I was getting the feeling that this extra headroom which is normally unused would happily compensate for the loss in the PLLXO. Instead of listening at 9 o'clock volume settting as I now do most of the time, I'll listen at the 12 o'clock or even 2 o'clock setting. That should not be a problem, right?

LCR would work as well, with the same losses and huge inductors.
Ouch. I haven't done the math, but I was hoping that line level would mean smaller inductors rather than larger. I was somehow under the impression that moving from speaker-level xo (SLXO) to LLXO would mean higher resistor values, but smaller C and L values. Now that you've got me thinking.... I think I usually find that when C values shrink and R values rise, then L values change in opposition to C. This means that L values will probably be larger. Sigh... if this is the case, I'm not getting into PLLXO.

You're better off going active if you need fourth order response. It is not that hard and more flexible. See Rod Elliot's site, the Linkwitz site or get hold of Don Lancaster's book "The Active Filter Cookbook"
I know all that. I have drooled over Don Lancaster's book and over "The Art of Electronics" for more than a year now. I've also studied the Linkwitz site and I have the LR4 PCBs from Rod Elliott, which I'm using in some other project. But the thing that's missing from my quiver is the network optimisation for active xo. For that, I'll have to abandon SW and move to LSPcad (at least). That's the reason I began this entire line of exploration of seeing whether, if I stick with passive, SW will do. Without network optimisation, how do I tune the xo slopes to give me the final acoustic slopes I want?

In fact, I did quite a bit of reading on LSPcad's capabilities too, and I discovered to my dismay that for notch filters, you need to use L even in active filters in LSPcad. The typical opamp-based tuned circuits which are used for active notch filters are not supported in LSPcad, and the developer of LSPcad himself shows how he's used an active LCR circuit for notch filtering in his demo project. This means for a better support for active xo, I'll probably have to go higher, to SoundEasy or somethng similar. I don't want to spend money on a commercial speaker design software just to get support for active xo, and then have to wind my coils whenever I want a notch filter in my xo.

All this made me wonder whether PLLXO and SW may work.

roddyama said:
One word of caution, be sure to insert a cap (hi-pass filter) between your amp and hi-frequency driver(s).
Will remember this. :)

pinkmouse said:
PLLXOs are only really any good for low order crossovers. For a 4th order, you will be losing so much signal that you will need a buffer stage after your crossover, so you might as well go properly active.
Actually, if the PLLXO is housed in the same box as the power amps, I guess I can get away without the buffer? Specially since I'll know what the Zin of my power amps will be?

And guys, how does the Pass Labs Rushmore do it? I know it's a proprietary design, but if somehow one can do good designs using passive rather than active, it has its advantages, doesn't it? (Eg. no separate opamp power supply needed, no distortion/noise addition through the supply lines and the opamps, etc.) The Rushmore was one of the important inspirations for me to think along these lines, in fact.
 
The buffer need not be an op amp - you can use a single jfet as Grey Rollins showed in a recent active filter thread.

Also see Jens Rasmussen's site, http://www.delta-audio.com/Active cross overs.htm for some more ideas on how to set up active filters. He shows a family of curves at various Qs so you can visualize what you'll need to correct a droopy or peaked response in the crossover region.

As Linkwitz shows on his site, it doesn't matter whether you use a real inductor or an op amp equivalent. So just model it as a real inductor and then make it as an op amp circuit.

If you build something flexible like the MOX for prototyping (see Jen's site for good documentation) you can just plug away, tweaking frequencies and Q of each section until the system measures flat in SW. You can mute on driver or the other to measure its response and phase, then put them together. Not quite as easy as modeling first and building once, but you'll probable tweak the modeled design anyway.

If you are planning to do your phase compensation the Linkwitz way with an allpass section, you won't be able to model it anyway. Adjust the delay to get the greatest FR dip when the phase of one of the drivers is reveresed.

BTW, The Art of Electronics is Horowitz. The Active Filter Cookbook only set me back aboout $30 at Borders
 
I have designed line level crossovers (mostly active) using Speaker Workshop by filtering the measured response in SW. I create a goal, overlay that with a copy of the measured response, and then filter the measured response. This is easy with simple active filters because all the stages are buffered - a first order at this frequency, a 2nd order at this other frequency, etc. Lots of undoing things until you find something that looks good... A passive design might be more difficult if you are getting any kind of interaction between the stages, but it should still be possible. If I wanted to add notches or shelving filters, I'd design something in Excel, model in Circuit Maker, and then build and measure the results. This isn't bad if you're doing it on a bread board - if it doesn't work, tweak the values and try again.

John
 
line level crossovers

It is possible to make crossovers at line level using only c-r sections, but these are restricted to q=.7 for first order and q=.5 for second order, higher orders have even lower q's.
I have built succesful multi amped sytems using only passive components but if you want the best results you are restricted in your choice of drivers to those that have very accurate roll off characteristics that occur at the right frequencies.
The advantage of fourth order L-R type filters is that as Bullock showed in a AES paper, they give the closest to ideal performance for an arbitrary woofer and tweeter and in practise if the tweeter resonance is at least one octave below the crossover frequency the frequency abherations left are mainly due to acoustic center offset and can be largely removed by anti phase connection.
 
John Sheerin said:
I have designed line level crossovers (mostly active) using Speaker Workshop by filtering the measured response in SW. I create a goal, overlay that with a copy of the measured response, and then filter the measured response. This is easy with simple active filters because all the stages are buffered - a first order at this frequency, a 2nd order at this other frequency, etc. Lots of undoing things until you find something that looks good... A passive design might be more difficult if you are getting any kind of interaction between the stages, but it should still be possible. If I wanted to add notches or shelving filters, I'd design something in Excel, model in Circuit Maker, and then build and measure the results. This isn't bad if you're doing it on a bread board - if it doesn't work, tweak the values and try again.
This is a very interesting approach. I am not sure I'll actually try to do one this way myself, but it teaches me a lot about what's possible. Maybe I'll just buy LspCAD once I begin actually working with active xo.
 
BobEllis said:
Also see Jens Rasmussen's site, http://www.delta-audio.com/Active cross overs.htm for some more ideas on how to set up active filters.
Actually, this site was a great help for a different reason: I got the ready-made design for building a 4-way 4th order active xo whenever I want one. Really cool. This is probably the only such active xo design where the Gerber plots are available for free download, isn't it?
As Linkwitz shows on his site, it doesn't matter whether you use a real inductor or an op amp equivalent. So just model it as a real inductor and then make it as an op amp circuit.
Yes, good idea.

If you build something flexible like the MOX for prototyping (see Jen's site for good documentation) you can just plug away, tweaking frequencies and Q of each section until the system measures flat in SW. You can mute on driver or the other to measure its response and phase, then put them together. Not quite as easy as modeling first and building once, but you'll probable tweak the modeled design anyway.
Absolutely right. With a tweakable active xo board, one can do a lot just by rapid prototyping and a measuring setup. This is a whole new angle.... takes a lot of the pain out of active xo. In fact, till now I thought this kind of rapid prototyping will only be possible when I shell out for a Behringer programmable xo. And I don't like that approach because I will then need a Behringer for each pair of speakers I design and build, thus raising the cost of each project. (Can one reverse-engineer a Behringer's settings into an analog line-level active xo schematic easily and accurately?)

If you are planning to do your phase compensation the Linkwitz way with an allpass section, you won't be able to model it anyway. Adjust the delay to get the greatest FR dip when the phase of one of the drivers is reveresed.
Got it.

BTW, The Art of Electronics is Horowitz. The Active Filter Cookbook only set me back aboout $30 at Borders
:D I'd never said the Art of Electronics was written by Lancaster. I'd said that I've read both Lancaster and the Art.

Your responses were a great help. Thanks a lot. :)
 
I have had success using line level crossovers, I think they sound better than active crossovers because they are simpler.

To implement them, I line up some headers on prototype boards, wire together the headers and plug in different combinations of L's and C's and R's. I approximate what the L's and C's will be by simulating the filter in Tina. Then, when the circuit is together, I sweep the circuit with a signal generator and measure the response with a voltmeter. Usually, something has to be adjusted until the response is just right in real time....

Not having all those buffers in the circuit removes a veil from the sound...

I originally built a Marchand active crossover, and while it was good, passive is better.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.