Beyond the Linkwitz ASP

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I'm putting the finishing touches on a crossover for my current speakers, and looking to move it out of my computer - I currently use Thuneau's Allocator for all XO, EQ and time allignment, and am very happy with it (I just wish it weren't so buggy). The sonic results are great, but having to use a computer every time I want music is a pain, doesn't work well for turntables and other analog sources, and the multichannel volume control is a pain. So anyway, I want to build an analog version of my current digital XO...

I'm looking at various options, and have some experience with Linkwitz's designs, which I hold in high esteem. Rather than copying his work, which he makes readily available, I'd like to combine multiple filters around as few opamps as possible. I know its possible, I just don't know how, and I haven't figured out how to simulate them yet.

Any pointers?
 
Allocator allows has a graphical interface which allows you to see the transfer function combined with your driver's measured response, which seems to work quite well for developing crossovers.

My current system is a two way:

The lows are crossed at 4th order at 125Hz, Q=.71 and highs crossed at 130Hz, same order and Q. There is no gain, both connected in phase, but the highs have a 1.1mSec delay. Last time I checked, the acoustic slopes were pretty close to 4th order.

In the highs, I use a shelving filter at 200Hz (+2dB, Q=1.0), and a notch filter at 600Hz (-4dB, Q=1.0). On the low end, I use the following notch filters - 80Hz, +3dB, Q=3.47 - 100Hz, -2dB, Q=13.13 - 110Hz, +4dB, Q=12.38.

These aren't final, but they have been stable for a couple of months, and are based both on measurement and listening. I will be changing the baffle, and have to remeasure then. If anyone is wondering, I'm using an Eminence Alpha 15 in an H frame on the bottom, and a Visaton B200 open baffle on top, modified with a phase plug and Enabled according to Planet10's recipe.

Next I'll probably post a couple opamp based filters, like Linkwitz uses - then I'm hoping someone can show me how to start combining them around one opamp. :D
 
Hmm, I hope someone is interested...

I got the sim working this weekend, and did some interesting stuff. Here are three 'equivalent' 4th order high pass circuits, the first is passive, the second is how Linkwitz does his active LR4 crossover, and the last is an simulated equivalent that I pieced together. It uses one opamp instead of two, the way SL does it.
 

Attachments

  • cir copy.jpg
    cir copy.jpg
    56.6 KB · Views: 756
Here is the simulation. There is something interesting going on - each filter has a slightly different 'knee', the passive circuit having the softest knee, and the single opamp circuit having the hardest knee. Y axis is in decibels..

(the green line is the passive circuit, the yellow is the double opamp, and the blue is the single opamp)
 

Attachments

  • 4th order comp run copy.jpg
    4th order comp run copy.jpg
    75.9 KB · Views: 697
I've been successful at combining notch filters around a single opamp, and plan on doing that with the analog equivalent of the digital XO I'm currently using. Linkwitz uses 22 opamps per channel in his 3 way design, I'm hoping to use 3 or 4, depending on if I keep it a 2 way or make it a three way, which I'll probably do. Anyone have any objections?

In another thread ('correcting valleys', I think), someone asked about a peaking filter. I simmed one, and here are the results.
 

Attachments

  • peaking filter cir copy.jpg
    peaking filter cir copy.jpg
    22.8 KB · Views: 672
Hmmm, subtractive crossovers look ideal, but in the real world, I'm not so sure. Aren't they only 'transient perfect' on axis, at a specific distance (if this isn't true for subtractive crossovers, please correct me)? Subtractive XOs seem to have relatively shallow slopes, which isn't ideal: when you throw in the amplitude/phase variation of real drivers, and any additional equalization, it seems like there would be no chance of perfect summation, even on axis at a specific distance. Perhaps thats why I haven't been able to find any actual measurements of 'transient perfect' designs. Off smooth off-axis response seem like it might be more important than reducing phase distortion, and higher order XOs are better suited for that.

I'm interested in subtractive XOs, I'm just not sure they actually work, and even IF they are able to pass a square wave over a wide listening angle, I still haven't seen much evidence that lower phase distortion is more important than off axis power response irregularities.
 
cuibono said:
Subtractive XOs seem to have relatively shallow slopes, which isn't ideal: when you throw in the amplitude/phase variation of real drivers, and any additional equalization, it seems like there would be no chance of perfect summation, even on axis at a specific distance.

hi
They (subtractive) work as xover's in situations, that would allow for 1st order designs. An example of their use could be from subs to a 2 way.

Here is a protype PCB that might intrerest you, it's not the most eff. use of opamps tho. see links
Schematic http://mysite.verizon.net/vze3xvxs/sitebuildercontent/sitebuilderfiles/filter_schematic.pdf
Manual
http://mysite.verizon.net/vze3xvxs/audio/id20.html
Group Buy thread
http://www.diyaudio.com/forums/showthread.php?s=&threadid=74420
 
You would do very well if you could bring down the number of opamps from 22 to 3 or 4.

I guess 4 or 5 might be OK for the tweeter and the midwoofer but 3 would be extremely challenging. For the lower bass, who cares how many opamps are used?

I have not built the Orion but built its "counter-part" John K's NaO and like it a lot.

The NaO has 10 opamps instead of 22 for its active EQ/XO. I have reduced it to 7 opamps while maintaining 100% of its transfer functions. For the tweeter and the midwoofers, they go through only 4 opamps.

I remembered that I also spent about an hour or so trying a completely different topology with 2 opamps for the tweeter and midwoofers, while trying to maintain the same transfer functions. The NaO uses a modular approach with dipole EQ, notch filter and LP/HP filters. I tried to use a high Q (Q>1) Chebyshev filter which could create the "notch" and dipole EQ at the same time. I got fairly close (up to 90%?) but was not happy with it so I ditched the plan. But even with 4 opamps in the signal path the NaO still sounds very clean without any harshness.

With the Orion, I am not sure if it is possible to use 3 or 4. The delay network (all pass filter) would require that many. Unless you make a passive crossover between the tweeter and the midwoofer.
 
Although I'm only part way through it, it looks like each driver will only go through 2 opamps each - the input goes to an opamp that does all the eq'ing, then the signal is split by three opamps, one for each driver, high, mid, and low. So the signal will only go through two opamps before it hits the amp and then driver (I hope..). I might have to add more, but we'll see. To delay the tweeter I'm just going to do physical offset after measuring the drivers phase.

The reason I'm going to this trouble is due to my experience with SL's Pluto and Orion speakers - IME, both speakers had a similar sonic signature, I felt like there was some softness in the treble, kind of a lack of clarity. While there are many possible explanations, including my own poor judgment, I feel it will be worth while to go to the trouble and see what happens..

I'm hoping someone will speak up if there is some problem with the single opamp 4th order circuit posted above, it'll save me a lot of trouble :D
 
cuibono said:
Although I'm only part way through it, it looks like each driver will only go through 2 opamps each - the input goes to an opamp that does all the eq'ing, then the signal is split by three opamps, one for each driver, high, mid, and low. So the signal will only go through two opamps before it hits the amp and then driver (I hope..). I might have to add more, but we'll see. To delay the tweeter I'm just going to do physical offset after measuring the drivers phase.

The reason I'm going to this trouble is due to my experience with SL's Pluto and Orion speakers - IME, both speakers had a similar sonic signature, I felt like there was some softness in the treble, kind of a lack of clarity. While there are many possible explanations, including my own poor judgment, I feel it will be worth while to go to the trouble and see what happens..
Distort the tweeter and you'll get all the clarity you want, hehe.
 
I have just had a look at your simulation. I don't think it can do it without matching at least 95% of the original transfer function. When I modelled the NaO with 2 opamps, I got result closer than that but still would not dare to use it.

1dB difference at the knee would constitute unacceptable response which is highly audible.

Have you simulated both the Orion and your circuit and got the transfer functions? Which simulation software do you use?

Regards,
Bill
 
In my first post, there is a link to the circuit simulator program I'm using. Its the one SL recommends. I'm not actually doing anything with the Orions, I'm trying to make an analog duplicate of the digital crossover I'm currently using. I have only just started, and haven't constructed the whole transfer function yet. I also haven't yet checked if the complementary high pass and low pass 4th order crossovers (based on a single opamp each) sum flat. I'm going to guess they do, but we'll see..
 
cuibono said:
Here is the simulation. There is something interesting going on - each filter has a slightly different 'knee', the passive circuit having the softest knee, and the single opamp circuit having the hardest knee. Y axis is in decibels..

(the green line is the passive circuit, the yellow is the double opamp, and the blue is the single opamp)

The "soft knee" freq response is due to the circuit Q. So unless they all 3 track well near the HP cutoff freq (-6 dB point for 4th order) they cannot be considered equivalent.
 
What I believe is that text book LR2, LR4, BW3 responses work with the acoustic slope of the drivers, not the electrical one.

Drivers are not completely flat so if we apply text book electrical crossovers the result would not be as intended.

For example, if the tweeter is naturally rolling off towards lower frequencies at 12dB/oct, applying a 2nd order electrical crossover may yield a 4th order acoustic slope.

It is the correct acoustic slopes that allow summing correctly.

You could possibly use Speaker Workshop and design a passive network first and derive the network response out of it. Your active circuit can then mimic the same network response to obtain the correct filters.

Regards,
Bill
 
I'm currently using a program called Allocator which allows you to import a .frd files for each driver, then apply crossovers and EQ. Allocator is then applies the filters to the output of my media player, in my case Winamp. I've gotten my desired acoustic function (mostly), and am going to try and make an electrical analog of the digital function in Allocator, so right now, I'm mostly worried about how the electrical signals will work when combining multiple filters around a single opamp. Later tonight I'm going to check that the XO's do electricaly sum flat.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.