FAQ regarding sensitivity and the choice of crossovers.

If we want deep bass from small loudspeakers, the price to be paid is efficiency.
Read below:
(2021: this is an old file and most links are no longer working)

Q:
I was just having a question or comment on your xovers.
I notice that most of your xovers use higher order and most of them have parallel components. I know that one has to use Zobel network and notch filters for the obvious purpose, but the more parallel components a xover has, the less efficient the speaker will be.  I do believe that efficiency is important and it sure helps improve sound transparency. I was reading about an article with regard to the upgrade that was made to the previous Living Voice speakers xovers in which the Zobel network was removed to improve the bass and transparency of the speaker.    Anyway, that is just my opinions on xover in general.
Thanks for your time.
Andy

A:
Thanks for your response to my website. I would like to use your questions and comments as an add to my website for a "discussion" on pros and cons on simple vs. complex crossovers, notch-filters, efficiency, etc.
I have no fixed preference for any type of crossover, although I very seldom come across 1st order filters that I can live with for an extended period of time. Very few drivers manage the immense overlap properly - and the mix of "sounds" - to my ears - colour most instruments and vocals. Very few drivers have a reasonably flat response - if combined with large magnets/high sensitivity - unless mounted on infinite baffles. And getting 95 dB sensitivity is usually at the expense of anything below 50 Hz - unless we are talking 100 litre cabinets or more. A few e.g. Vifa drivers have an intrinsic flat response, but due to small magnets and/or highly damped suspensions, i.e. poor transient response, etc. And the result is low sensitivity, e.g. Vifa. P13WH.
The statement that high efficiency should improve transparency is new to me. I can find a lot of very high efficiency speakers that sound terrible, e.g. Cerwin Vega.
Notch filters: 3-5 dB bumps in any (audible) frequency domain is likely to throw a lot of details never heard before, but was this what the performer and recording engineer had in mind? Is this high-fidelity in the strict sense of the word? Any deviation from a "flat" response profile must be a "distortion" of the original signal put into the grooves - or dots. A lot of manufactures "voice" their speakers to a specific sound they feel - or know - will sell. Because they need to sell, otherwise they're out of business.
With regard to the Living Voice upgrade, the improvement in sound may be significant, but not necessarily because the zobel network has been removed. What we don't know from this tweak is how the frequency response, phase tracking, etc. now looks compared to what it was before the change. The zobel network will produce a linear impedance profile of the driver but leaving out the zobel network can be actively used in the construction of the crossover. With drivers featuring symmetric drive, like Scan-Speak, it's hardly necessary to ad zobel networks due to an intrinsic linear impedance profile.
I don't think I have any constructions with zobel networks. Not because I don't like them, but because they were not necessary. If I did 1st order networks, I'd probably use them from time to time.
Regards, Troels

Q:
The thought of sending you an email came up when I was reading Kevin opinions of his TJL speaker.  The things that caught my attention was his comments that the TJL seems to be inefficient (he had to turn up the volume more than usual) and lacking in dynamics, so I thought it might have to do with the design of the xover.  The xover of the TJL is 3rd order.  I also remember reading that the Thor from Madison Sound which uses the same driver
http://www.madisound.com/index.html as the TJL albeit in a D'Appolito configuration.  The xover of the Thor uses 1st order with only a series inductor and a notch filter at 4.5khz.  I don't mean to compare the two speakers but I was wondering if you had considered using 1st or 2nd order filter for the TJL?
Andy  

A:
OK, so let's get this right before we proceed: The impact on efficiency has nothing to do with the choice of order of crossover. Check graph below: This is a 18W/8535 driver applied a 1st order filter (top) and (below) a 4th order filter to get "flat" response in the 200-2000 Hz range. As can be seen: same SPL level for both crossover topologies. Actually the 4th order is 1 dB better than the 1st order - because I made it that way.

The TJL has a sensitivity similar to the 2.5 clone and is in line with most 6"+1" speakers derived from this kind of philosophy, where we want decent bass from a small bass driver, thus using a driver with the appropriate TS-data to get the desired F3 of the construction. One problem here is that most commercial constructions exaggerate the sensitivity of their constructions - sometimes grossly.

Above is the THOR crossover applied to the W18001 driver and the magnesium cone break-ups are only 20 dB down in the 5-8 kHz range, and this will clearly have a negative impact on tweeter performance.
There's a commercial element in any construction released from a driver manufacturer, and this may be just about good enough for most people not to be aware of the intrinsic cone break-up, but we will never render the full potential of these drivers with this simple approach.
Read a good discussion on how "Phat" is "Flat":http://www.speakerbuilder.net/web_files/Articles/phat/hpif.htm
Regards, Troels  

Q:
About TJL Kevin comment:With loudspeaker with high order filter, complex network. You should write, more if the filter is complex, it's need an amplifier with highest current capacity. The passive components ( capacity and inductor) need current to work fine. It's very important the link between the amplifier and the loudspeaker. It's depending of the number of drivers too. With modern loudspeaker, you need current to have a good sound: I think about Thiel : http://www.thielaudio.com/ see for example :
http://thielaudio.com/THIEL_Site/Pages/models/Current_Models/cs6/cs6.html and ATC :
http://www.atc.gb.net/ . If you can, listen the passive close box SCM20, very heavy with 82dB sensitivity, the dealers don't like it ;-).
Their filter are complex and they need great amplifier with high current output and good musicality. I think it's a philosophy like we can find in high sensivity system with horn and other ...SEAS with the magnesium driver is a very good compromise.
Regards, Rodolphe

A:
I have a webpage planned regarding crossovers as it is a common misunderstanding that complex crossover are responsible for high current demands and low sensitivity. This is no way the case. The efficiency of a construction is dependant on the sensitivity of the drivers and the current demand mostly dependant of this and the bass-reflex tuning where the relation between voltage and current may cause large amounts of power to be drawn from the power amplifier. We can make a 2-way speaker with 6 dB filters that draws high current and a 4th order filter with high sensitivity that is mild on the amplifier. All depends on the drivers (impedance) and the cabinet tuning. If you have a low impedance like 2 ohm (like the Watt/Puppy) and at the same time a phase that is largely capacitive, then you will draw high current.
Another problem here is that commercial speaker manufacturers usually state 3 dB more sensitivity than is real. The Response 2.5 is a good example, where ProAc claim 86 dB and John Atkinson measures 83 dB - the same I measure from the 2.5 clone.
Regards, Troels

Comments from Sven, DK:
High efficiency drivers have less distortion due to reduced current flow through the voice coil at a given acoustic output, i.e. reduced magnetic un-linearity and reduced voice coil temperature and rise in impedance.

A:
Good point!
I may modify the statement by saying that a high efficiency driver may have reduced distortion; that is if the driver is properly constructed. We can find high-efficiency drivers (PA) that are quite terrible. But let's leave out the poor examples and assume we're dealing with well designed drivers.
And again, I'd love to only make speakers with a sensitivity above 90 dB/W, but as stated before, if we wan't deep bass from small cabinets, we have to sacrifice efficiency.

More comments are highly welcome and will be added to the website.