Passive crossover into active crossover

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I designed a bookshelf loudspeaker with Visaton AL130 and Scan Speak D2904/9800. The preliminary crossover was designed with
Boxsim, a freeware which can be tricked into accepting other speakers than Visaton brand, such as the ScanSpeak tweeter.
The final design was made with LAUD measurements and listening tests.
As I am going to make 4 or even more of these I tried to replace the passive crossover with an active. The reason for this is simple:
a second power amp costs me virtually nothing as all parts are on stock and insofar are written off, but the high quality air conductors and special capacitors of the passive crossover are
really expensive. I took the frequency and phase response of the low pass and the high pass passive crossover with LAUD and designed an active filter which has very close the same characteristics. Although the acoustical response of the active and the passive speakers match, the listening tests clearly favored the passive version especially in complex chorus peaces of music with female voices dominating. Somehow the acoustical resolution of the passive is better.
I don't have a plausible explanation.

Dieter F.
 
hahfran said:
I designed a bookshelf loudspeaker with Visaton AL130 and Scan Speak D2904/9800. The preliminary crossover was designed with
Boxsim, a freeware which can be tricked into accepting other speakers than Visaton brand, such as the ScanSpeak tweeter.
The final design was made with LAUD measurements and listening tests.
As I am going to make 4 or even more of these I tried to replace the passive crossover with an active. The reason for this is simple:
a second power amp costs me virtually nothing as all parts are on stock and insofar are written off, but the high quality air conductors and special capacitors of the passive crossover are
really expensive. I took the frequency and phase response of the low pass and the high pass passive crossover with LAUD and designed an active filter which has very close the same characteristics. Although the acoustical response of the active and the passive speakers match, the listening tests clearly favored the passive version especially in complex chorus peaces of music with female voices dominating. Somehow the acoustical resolution of the passive is better.
I don't have a plausible explanation.

Dieter F.


As Magnetar said, you have to derive the electrical transfer function of your passive crossover while taking the reactive nature (impedance) of the drivers in account. If you simulate the transfer function of your passive crossover into a fixed impedance it will not have much in common with the transfer function of it into a varying impedance of a real world loudspeaker.
So, you have to export the real transfer function of your passive from your sim, then construct and optimize an active circuit to that transfer function- down to at least 0.5dB. It might take shelving or parametric filters in addition to the standard LP and HP filters to arrive at a close match.
I use lspCAD which makes this fairly easy.
 
hahfran said:
I designed a bookshelf loudspeaker with Visaton AL130 and Scan Speak D2904/9800. The preliminary crossover was designed with
Boxsim, a freeware which can be tricked into accepting other speakers than Visaton brand, such as the ScanSpeak tweeter.
The final design was made with LAUD measurements and listening tests.
As I am going to make 4 or even more of these I tried to replace the passive crossover with an active. The reason for this is simple:
a second power amp costs me virtually nothing as all parts are on stock and insofar are written off, but the high quality air conductors and special capacitors of the passive crossover are
really expensive. I took the frequency and phase response of the low pass and the high pass passive crossover with LAUD and designed an active filter which has very close the same characteristics. Although the acoustical response of the active and the passive speakers match, the listening tests clearly favored the passive version especially in complex chorus peaces of music with female voices dominating. Somehow the acoustical resolution of the passive is better.
I don't have a plausible explanation.

Dieter F.


Excellent thread, I wrote a similar thread in another forum, the response was very poor indeed.

I have limited experience with passive x-over design however I am now using active x-over, digital DCX2496, which one are you using, analog or digital?

There is even a difference between pure analog and digital x-over. I used an analog Behringer before and now I am using a digital x-over, ie DCX2496.

The adv. of DCX is that signal from the source is digitised all the way till the analog stage prior reaching the amplifier. DCX also accepts analog input from the source , digitises it and converts it back to analog. There is Real difference between the 2 options, digital is much better.

I like my DCX because it is very flexible and I can design many different types of speakers and don't have to spend lots of money on passive components. If I like the sound speakers I may change the X-over to passive. That is the option I have.

I am sure there is a difference between active and passive x-over. I hope there will be more contributors here than I experienced before. I am sure there are many here who have experience in both technologies.

cheers.
 
I too am very interested in some of the background to this subject, as I am contemplating going digital active with a dcx2496 sometime soon.

hahfran: What active crossover are you using? Does it simply split the signal or is it capable of using slopes/filters such as Butterworth, Bessel and Linkwitz-Riley?
Maybe this is the problem, and you are not achieving the ideal integration of your drivers?

I have been led to believe that an active crossover should be able to achieve better results as it does not have passive components in the signal path. In theory the drivers are performing unhindered and free, which must have similar advantages as fullrange drivers.
 
Re: Re: Passive crossover into active crossover

ttan98 said:


If I like the sound speakers I may change the X-over to passive. That is the option I have.

Well, that might not be a trivial task to do, if you want to achieve an exact match.

There are pros and cons for both active and passive, what really matters is the end result. Passive crossovers are tightly coupled to the drivers, while actives are separated: this can be an advantage or a drawback, it depends on the application.

I compared once an analog active, OPAMP-based crossover with a passive on the same drivers. While I wasnt been able to achieve the exact same response, I ended up with the passive version, as being both simpler and better sounding.

Depending on the driver combination, sometimes you can achieve a simple and elegant solution using passive components.

There is however one area where I definitely think active is much better: very low (100-200 Hz) crossover points. I wouldn't bother using can-sized coils there, where active can do it a lot better.


There are also hybrid approaches, as JohnK describes on his site: use of passive for the actual filter and active equalization to flatten the response curve.
 
Re: Re: Re: Passive crossover into active crossover

bzfcocon said:


There is however one area where I definitely think active is much better: very low (100-200 Hz) crossover points. I wouldn't bother using can-sized coils there, where active can do it a lot better.


There are also hybrid approaches, as JohnK describes on his site: use of passive for the actual filter and active equalization to flatten the response curve.

When I say if I like the sound I will change to passive I mean x-over between the mid and tweeter. I will still maintain active x-over between the woofer and mid. That we are in agreement.
 
Thanks for the responses. So when I dig out my sparse knowledge of physics , the acoustical side of a loudspeaker is the
radiation resistance which is frequency dependent and complex.
It depends also on the acoustical properties of the listening room which is never an echo-free half-sphere.
This radiation resistance must be transformed into an electrical circuit consisting of L,C,R which is the electrical load of the crossover.
The current which a voltage source ( the ideal amp ) drives depends then on the paralleled crossover and the electrical equivalent of the speaker's radiation resistance.
So far so good or bad.
Insofar the active crossover has to simulate the passive plus a correction which is due to the interaction of the passive loaded with the electrical equivalent of the radiation resistance.
That should be solvable at least for the simulated loudspeaker.
 
The best way to obtain a transfer function (electrical response) of an existing passive crossover is to actually measure it. Perform a differential measurement from the crossover input (reference point) to each drivers terminals (measurement point.) This will yield the electrical response of what you would need to duplicate with an active crossover to obtain the same acoustic result. Most of the soundcard-based speaker/analyzer design programs will do this. ARTA, Spectraplus, SW, etc, etc.

A really good simulation would also get you there, but since you already have the speakers it's easier and better to take an actual measurement. The result can be saved as a text file and used for other purposes as well. Imported into Jan's excellent program....."target" file for some other speaker design program......whatever.

However, I think if you're going to use an active crossover setup you can most likely improve upon your existing design quite a bit. :)

Cheers,

Dave.
 
Davey said:
The best way to obtain a transfer function (electrical response) of an existing passive crossover is to actually measure it. Perform a differential measurement from the crossover input (reference point) to each drivers terminals (measurement point.) This will yield the electrical response of what you would need to duplicate with an active crossover to obtain the same acoustic result. Most of the soundcard-based speaker/analyzer design programs will do this. ARTA, Spectraplus, SW, etc, etc. (snip)
Dave.

Dave, is there much information on doing this measurement?
 
Kevin,

Not much. :)

The basic concept is pretty simple. You perform a two-channel (differential) measurement and have the program you're using plot the difference between the two. The reference "probe" would at the input of the device-under-test and the measurement "probe" at the output. You can use the same basic setup on line-level gear, a speaker-level setup, (like we're talking about here) or just about anything else. You just have to make sure the voltage levels do not blast the soundcard inputs.

I like to use SpectraPlus with a white noise excitation and average over a minute or so. You'll end up with a totally smooth curve that shows the exact electrical response of the device-under-test.

ARTA works well too. You can use the STEPS application to sine sweep slowly over a period of time, or the ARTA application which uses an MLS stimulus.

Hope that helps.

Cheers,

Dave.
 
Somehow I fail to get this. With a differential measure you get the impedance of the respective crossover - hipass or lowpass -
under its real load, the driver ( built in the enclosure, with dampening etc. ) only if the driver itself were a real, frequency independent load.
But in your case, you measure the voltage drop across the crossover filter and that depends on the driver's impedance which is around 20 ohms at resonance of the total system and tends to rise with frequency.
Wouldn't it make more sense to measure the voltage across the driver terminals? As the amp is considered a voltage source, the impedance variations of the driver do not apply.

I just made such a measurement and I see a lowpass slope that differs considerably from the filter loaded with 8 ohms.
The slope of the 12 dB design is actually 8 to 9 db per oct but this isn't even constant.
 
Yep, that is essentially what you're doing. You're taking a reading across the driver terminals and simultaneously across the input to the crossover. The comparison (difference) of those two measurements yields the exact response of the crossover itself.

If the amplifier is indeed functioning as a true voltage source and you're sure the input signal is flat then the reference measurement should be flat and...in theory...a single measurement would suffice. But, what if the amplifier is not a true voltage source. Many are not. Or what if your signal generator is not flat across it's response? In those cases a two-channel measurement will exclude all the other variations (elsewhere) in response that may sneak into the measurement.

If your measurement differs considerably from what you expected then there may be another setup factor that you don't have set correctly yet. Or maybe it is correct and this technique is highlighting (very well) the response variations of the complex load of driver and crossover. The electrical response of crossover networks seldom looks "generic" and usually reflects quite a bit of tweaking by the system designer to get the ultimate acoustic slopes correct. That is the essence of crossover design.

Anyways, this differential technique is a really powerful measurement scenario in most conditions.

Cheers,

Dave.
 
Thanks yes I get it now. The complex or vector sum across the filter and the driver must be equal to the amp output voltage.
Since I have just one pair of ground-free inputs I have to derive the diference manually from two separate measurements.

The design of the active crossover is a perfect task for a sim.
 
I'm not totally familiar with LAUD, but I thought a dual-channel measurement like this should be straightforward. You have both a "Cal" probe and "Measurement" probe, yes?

I'm not sure why you would need "ground-free" inputs. The ground circuit within the crossover network is most likely common to both input and output so that would be the reference for both probes.

The basic process should be the same as a normal measurement in LAUD except your measurement probe is attached directly to the driver terminal vice being used with a microphone/preamp for an acoustic measurement.

Cheers,

Dave.
 
I have finally managed to derive the passive filter characteristics.
The high pass is relatively easy as it exhibits a constant slope of 16 dB / oct beyond the -3dB.
That can be realized with an active filter which depends largely on the open loop gain , such as a Rauch filter. One has to design
a discrete op amp with a well defined open loop gain to achieve the 16 db characteristics.
The lo pass however looks weird especially in the frequency range that matters most for music, between 800 and 2000 Hz.
Only beyond the crossover frequency it exhibits an almost constant slope of 10 dB.
An approximation are 3 directly coupled RC with a small amount of positive feedback.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.