Active multiway speaker design?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
IME, there isn't very much posted on this forum about active loudspeaker design. That's probably because the people who do it already know the theory quite cold.

That doesn't apply to me. So, what are the major differences between active and passive loudspeaker design? One question I have always had, is acoustic rolloff essentially ignored? In other words, since higher order crossovers are easily achieved with more inexpensive op amp circuitry, is the bandpass that the designer wants to see simply imposed upon the driver? I think that must be the case for a 4-way speaker. TIA.
 
leadbelly said:
IME, there isn't very much posted on this forum about active loudspeaker design. That's probably because the people who do it already know the theory quite cold.

That doesn't apply to me. So, what are the major differences between active and passive loudspeaker design? One question I have always had, is acoustic rolloff essentially ignored? In other words, since higher order crossovers are easily achieved with more inexpensive op amp circuitry, is the bandpass that the designer wants to see simply imposed upon the driver? I think that must be the case for a 4-way speaker. TIA.

Active crossovers are just as 'difficult' to design as passive ones. Sure you have more flexibility with an active, but the basic requirement for measuring is just the same.

Most active crossovers will follow the same well known acoustic targets as passive too, 1st-4th order with whatever Q your target maybe.

The only difference between active and passive is the way the filters are applied, the rest is the same.

Now where it could be said, that the effect of a drivers impedance on a passive xover, makes it less predictable then an active version, you're still only half way there if you cant measure.

Some people view active crossovers as a sort of a panacea, this is about as far from the truth as you can get. Having said that, you may be able to get closer to a desired target with active crossovers then with passive if you can't measure, as actives remove the drivers impedance but also the filter stages do not interact with one another.

For example, a simple shelving network with an active crossover, that compensates for bafflestep loss, is very easy to implement and also allows good flexibility in altering the frequency and amount of gain. This network works entirely independently from any other stage, say a low pass at 2khz. Whereas in the passive loudspeaker the series inductor maybe a large value that sorts out baffle step but it also contributes to the low pass at the same time, making it far harder to eventually get right.

I suppose I've sort of contradicted my opening statement. Yes with an active crossover you can probably get closer to a desired goal then with a passive loudspeaker, but you 'will' need to measure to get that final step in performance. Of course, if you are happy to sit there and twiddle resistor values all day long from now until Christmas you may get as good as a result.

Edit - But RE 4 ways. The more complex the system, the longer and harder it will be to tune it by ear, as there is FAR more interacting then with a simple loudspeaker. Measuring in a complex multi-way is a must imo.
 
Re: Re: Active multiway speaker design?

Thanks for the reply.

5th element said:
Most active crossovers will follow the same well known acoustic targets as passive too, 1st-4th order with whatever Q your target maybe.

Really? I was under the impression that a tradional weak point with passive multiway is that at its upper end, a cone mid or woofer becomes beamy. Wouldn't imposing a higher order filter and ignoring the acoustic rolloff be a sort of panacea?
 
Re: Re: Re: Active multiway speaker design?

leadbelly said:
Thanks for the reply.



Really? I was under the impression that a tradional weak point with passive multiway is that at its upper end, a cone mid or woofer becomes beamy. Wouldn't imposing a higher order filter and ignoring the acoustic rolloff be a sort of panacea?

Maybe I misunderstood the question.

If you take a figure of say 24dB as being required such that its low enough down in level as to be inaudible, relative to the passband. Then it would make sense you need 1 octave of 'flat' behavior before the xover point, off-axis, using a 4th order network, to effectively be free from beaming. (figures made up to illustrate a point)

If you were to now use a steeper xover, say 48dB or 8th order then you would only require half an octave before the xover point.

Or looking at it another way, if you have a driver thats flat from 100hz to 3000hz, if you place an 8th order xover at 2250hz, you can be sure that you are getting true 8th order roll off, as the acoustic roll off of the driver will have very little impact on the desired acoustic target.

From this stand point, yes it would be a sort of panacea.

Most analogue active xovers do not tend to exceed 4th order electrical, the same cannot be said for digital however, where really high order roll offs are just as easy as a 1st order. Ringing and pre-ringing cannot be ignored however although I am not particularly well versed on the topic, the higher the order the worse this becomes and afaik is not something to be desired.

Another benefit of lower order networks, I suppose could be called driver 'blending'. If the tonality of your midbass and tweeter are slightly different, a steep, brick-wall type filter may expose this difference quite brutally. With a steep filter its easy to see that as a pianist increases in pitch along the scale, one note could come from the mid/bass, then quite abruptly the next note comes from the tweeter, the two notes could sound quite different. Whereas if there is a region of overlap between the two it would help to mask this via the drivers transition period, making for a more coherent sound. There has to be some reason why lots of people prefer the sound of 2nd order acoustic over 4th order, this could be one of them.
 
I wrote up a simplified guide to designing active crossovers and speaker systems that tried to dumb down Linkwitz to a "no math needed" level. http://mysite.verizon.net/vze3xvxs/audio/id20.html There is also a filter calculation spreadsheet there. Sorry, the boards are sold out. The website will be dying in a few weeks, and I haven't figured out how my new ISP's site builder works, so you may want to download the information. Linkwitz does a really good job of explaining the process and the math behind it, if you are up to that.

I wholeheartedly agree with 5th, accurate measurements are a must. Speaker design is all about interactions between the drivers' responses, the filters, baffle and room. You reduce the complex math needed when you go active and eliminate filter section interaction, as 5th pointed out. Going by ear you're likely to end up with something that sounds a lot like what you are used to which is not necessarily accurate.

IMHO, it is easier to go from manufacturer's published data and predicted baffle response to a passable speaker I've also found myself much more willing to go through more measure, tweak, measure iterations when the parts I am throwing away only cost a few cents.

Of late, I am using Cauer-Elliptic filters that approximate a 8-10th order L-R response for the first 60 dB and keep the bounce back below 50 dB. It really lets me push the tweeter crossover down.

Edit to specifically answer your question: No, the acoustic roll off is NOT ignored. Sure, with high order filters far enough away from the roll off you minimize its effects. This is true whether active or passive.

At one point Linkwitz used a 2nd order shelving filter along with a 2nd order filter to get a 4th order roll off that compensated for the tweeter's natural 2nd order roll off below its resonance that was about an octave below the crossover. He doesn't seem to do this any more, so perhaps he found it inaudible.
 
leadbelly said:
question I have always had, is acoustic rolloff essentially ignored?....

Hi leadbelly,

I have been looking at just this issue recently, as I am designing some two-way active speakers.

Using a 4th order crossover, I found that the crossover frequency needed to be roughly >3 times higher than the tweeter resonance, if the overall response is to be OK.

Originally, I had a crossover freq of 2kHz, and the tweeter resonance is 650Hz. This gave almost 1dB dip in the response. Increasing the crossover freq to 2.5kHz, and the dip reduced to just 0.3dB.

CS
 
At one point Linkwitz used a 2nd order shelving filter along with a 2nd order filter to get a 4th order roll off that compensated for the tweeter's natural 2nd order roll off below its resonance that was about an octave below the crossover. He doesn't seem to do this any more, so perhaps he found it inaudible.

I don't think that a 2nd order rolloff would cause an inaudible error. Maybe SL just used another method for including the tweeter's natural response.

Regards

Charles
 
There are those who argue that the topology shown for passives is required for phase coherence. Start with a buffer, add baffle step, EQ and A bit of delay in the appropriate legs. Linkwitz mostly does it that way...

Others make active XOs as you show in the second topology. When properly phase adjusted, I would think that would work just fine. It would probably require more phase adjusting stages than the previous one.
 
I think Linkwitz did the Pheonix mostly in the second topology, however, the all pass filters used for the tweeter did show that he cared about the phase / delays. The mid and bass each has its own shelving pass filter, notch, etc., suggesting it is closer to the second topology.

In John K's NaO, it is similar. The dipole EQ, notch filter, tone control are at the root, suggesting it is using the first topology. However, the woofer EQ is at a branch, which would change the phase, similar to what is in the second topology.

Leonard Audio obviously uses the second topology.

From my memory, Rod Elliot suggests using the first topology.

So I am confused. I have not sufficient knowledge to be 100% sure about it. Of course, measurements will show if the results are correct or not. But I guess I should understand it first, design it then measure it.

My thought is that drivers are minimal phase devices, which means their phase is determined by their frequency response. If that is the case, if not considering the driver offset, provided that the active or passive XO network shapes the driver response to the target frequency response, such as LR4, LR2, BR3, etc, the network should sum up as predicted. In that case we don't need to worry about the phase but only the frequency response curves that determine the phase. If this assumption is correct, then I don't need to use the first topology, the second topology should do fine. This is very important because with the second topology, there are far less components (opamps, caps) in the signal path for each leg, beneficial especially for the higher frequencies.

I am currently designing my new WWMTMWW + Sub 4 way hybrid active / passive, hybrid dipole / cardiod / monopole speaker. For the 2.5k XO, I will use passive. I am planning to use the second topology. If I had to use the first topology, I would abandon the project because I don't believe the speaker can sound good with that many components in the signal path for a 4 way. I understand that driver offsets will introduce errors. But given the XO points are 300Hz, 65Hz the kind of phase errors would be at a minimum, and perhaps moving one leg of the XO higher or lower will adjust it perfectly, I hope.

Please correct me if I am wrong and save me from my troubles down the line.

Thanks in advance.
Bill
 
Seems reasonable to me, as I think all that matters is that the phase is correct in the end.

However, I don't have any problem with my current 3 way using the first topology. My tweeters end up being driven through 7 op-amps, still extremely clean using OPA2134s. NE5532s were not quite as good. I cannot put my finger on exactly how they weren't as good, so it may just be that I spent a fair amount of money on my OPAs and want them to sound better.
 
HiFiNutNut said:
The normal 4 way passive XO would be something like this:

An externally hosted image should be here but it was not working when we last tested it.



Bill,

I think what you've got there is the normal active crossover layout. Parallel passive lay-out (this for a 3-way) is more like:
An externally hosted image should be here but it was not working when we last tested it.

If you ran a passive (4-way) the way indicated in your diagram, then you'd have the tweeter signal passing through 3 different high-pass filters before it got to the tweeter. You'd be hearing nothing at all, or at least, nothing good.
A parallel passive filter is completely independant of the filters for each of the other drivers in the system (in theory), which is why you can drive each of them with an amplifier.
I don't know what John Murphy was thinking when he came up with that block diagram that you posted...
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.