Crossover Frequecy Specifications - Confusion

[radiogram]

In Klsipch Hertiage speakers what does the crossover frequency really stand for? Is it the frequency at which the electrical outputs of the different sections actually cross OVER (i.e. the point at which the plots of the two sections meet, as is normally defined in classic theory) or is it based on -3db or -6dB point of just one section taking over from another section ? The reason I have this confusion is because, first of all, for the older heritage speakers, the mid sections are run full out in the highs (no bandpass) and its electrical response will never cross the high section and secondly due to the following:

Take the following examples:

1. Circuit 1 (Type AA). - ----II (13uF)-------Tap5---T2A---Tap4-----K55 - The crossover frequency is specified as 400Hz. Now assuming an average impedence of 14Ohms for k55 (effective impedence 2* 14 = 28) , the -3dB point is 437Hz. So the 400Hz roughly corresponds to -3dB point.

2. Circuit 3. (BEC's cornscala) ------4 uF--------CT125 - Specified crossover frequency is 4500 Hz. Taking 8 Ohms for CT125, -3dB point is 4973Hz. Here the specified freq seems to about the -3dB Point.

3. Circuit 2 (Heresy). ------|| (2 uF)------K77 - The crossover frequecy is specified to be 6000Hz. Taking 8 Ohms for K77, the -3dB point is 9947Hz and -6dB point is about 5800Hz. Here the 6000Hz specified corresponds to about the -6dB point not -3dB.

Now classical theory dictates that for first order crossovers the ideal crossover point is the summation point of -6dB of both the sections for flat electrical output at and around the crossover frequency. Of course, this is purely based on eclecrical output only without any regards to acoustic output and such theory will work only if all the drivers have flat acoustic output at that crossover frequency and ideally even an octave beyond.In reality of course this is not the case.

In any case, can someone throw some light on the what the specified crossover frequencies is supposed to really mean?

Thanks

[Al Klappenberger]

Radiogram,

You are thinking like someone who understands filters to some degree! The fact is, the Klipsch networks are so goofy that none of the classic thinking you are using is applicable. This is the reason for your confusion. In a true contiguous "multiplexer" the lossless 3 dB points should be the crossover for constant impedance but if the network depends on the squawker driver pooping out, and there is no actual filter in the network doing the crossing over, the only way to define it is by the acoustic outputs of the drivers! Another factor is that the electrical cross should be the 6 dB points for best summation. I think this is part of what's called Linkiwitz-Riley alignment. The fact is that component loss add to the 3 dB point of a lossless network which causes it to approach 6 dB, especially in a woofer / squawker crossovers. If the crossover loss in the network is exactly 3 dB you will get a slight rise in output at the crossover assuming the two are in phase. They probably will not be though. This is true on a properly designed network, but not the Klipsch networks. I suggest that you stop trying to figure out the Klipsch networks. I have been doing computer analysis on them for years and they are all over the place! If you are interested in the actual response of a specific network, I can post the computer analysis on it for you. I have most of them on my computer as design files.

Al K.

[mikebse2a3]

Hey radiogram

Advertising specifications often contain to little information to completely define what is going on in a design and this is pretty much the norm in the majority of all audio componets.

It's important to know that PWK correctly called what some might call the crossover networks in the Heritage Line "Balancing Networks". They were designed to properly integrate/balance (ie:EQ) the individual horns/drivers to reach an accurate acoustical response for the system which means they took into account the electrical/acoustical response of the drivers/horns and their physical positioning.

To design optimized networks for combining the different elements of a loudspeaker requires that acoustical mesurements be made of the completed loudspeaker design otherwise it's just guess work at best.

mike tn

[radiogram]

You are thinking like someone who understands filters to some degree!

Al - You are quite right. While I do have some background in these, but as you have rightly preceived only to some extent. While I did formal courses in Filter Networks (I believe we used book by Van Valkenberg!, among others), Signals & Systems etc, as part of my EE, I never really practised that discipline, having swithced to computer Networking & Telecom Software as my Career. Looking back I regret having strayed off from my passion - Analog, RF, Phased arrays, etc.

Anyway, thanks for your explanation. It confirms my guess that klipsch had their own definition in mind (and that too not seemigly a consisent one), when they specified the crossover frequencies.

I also did not think about the component losses you mentioned. Appreciate your time and feedback.

BTW: I did some measurements today and I found that I have a 10% loss in signal voltage in the range 12000Hz and above (very much in full passband) with a 4uF Sonicap in series compared to without.

Amp Out ------8 Ohm Resistor - Voltage @ 12000Hz = 1 Volt

Amp Out ------|| 4uF------- 8 Ohm Resistor - Voltage @ 12000 Hz = 0.9 volts

Is this acceptable?

Thanks

[Al Klappenberger]

radiogram,

BTW Mike is correct in what he said.

I think .9V represents .92 dB loss. That is actually better than what I expected. You didn't say what network you had. The AA will show loss around 2 dB because it's an old constant K design working in a singly terminated environment plus a few other problems with it. The later networks have a singly terminated filter having a single finite transmission zero. It's a better filter. That is probably what you have. It has less loss. Older networks usually have dried out caps that increase loss also.

Al K.

[WMcD]

I think the crossover specs may be very close. However, it is not because of the performance of the networks.

Generally, people look at network performance based on magnitude and phase and how they add. A really nice explanation is here:

http://www.rane.com/note160.html

This sort of thing gets forgotten by me soon after reading. So I have to read it again and again.

Maybe we have to start with some things you see in crossover filter design. 1) It would be nice to have brick wall filters at the crossover point. 2) Since we can't have that (at least in the old days), we should have filters with predictable flat summing -- assuming there is no part of the "system" thereafter which screws this up.

The real issue in addition to 2) is that the filter explanation misses out on a very big part of the system. It assumes that we are using perfect drivers with flat acoustic response to some distance away from the selected x-over freq. That big part is not true. There is part of the system thereafter which messes up things. The driver/horn acoustic response is often falling off rapidly AND has some bumps and glitches not too far away.

You can see this in some freq response curves of the K-Horn bass bin. It is struggling at 400 Hz and drops off quickly with some output bumps up higher. Also the K-400 and K-55 combination has very sharp rolloffs at 400 and 6000 Hz. but then there is a glitch higher up. IIRC about 48 dB per octave or more. That is before you start with any electrical filtering. The K77 will go down to 3500 but then it drops suddenly.

The bottom line is that slavish attention to filter design can work out to perfection in summing of phase and magnitude elecrically . . . but then the drivers/horn mess up perfection because they have additional roll off.

As you can see from the above, the crossing over is achieve to a large extent by the driver/horn. So PWK's 3rd order designs mostly just keep the unwanted power out of the voice coils. At least at one time he believed that rapid phase change due to any steep slopes in acoustic output can be heard. So maybe he figured there was no use to increasing the acoustic roll off by adding steeper electical slopes.

There is a modern take on all this. It has to do with the residual bumps and glitches in the acoustic outputs somewhat distant from the crossover points -- and a bit of the summing issue. The modern take is that with very steep electrical slopes, we narrow down the range in which the drivers share any acoustic output. Therefore the summing issue acoustically is minimized to the extent it is reduced to a narrow range. We're gettng closer to a brick wall situation.

Also, the bumps and glitches of driver output are killed off.

Al K. has championed the high slope passive crossover. Others have used higher slope active crossovers starting years ago. In recent times, digital signal processing has allowed even more control over the situation.

Wm McD