keithe@tekgvs.UUCP (Keith Ericson) (10/18/85)
Gentlefolk - The discussion of passive crossover design and the problems involved in modelling the driver(s) is getting lots of net.audio space lately. But, friends, it's *old technology* and not really current in todays audio market-place, especially for those capable - as I believe lots of net.audio reader are - of some relatively simple electronic construction. (Right now you're saying 'what IS he talking about! - bear with me...) OK - the problem in a conventional (i.e., passive-crossover) system is that you've got this monstrous power amplifier, with a fraction of an ohm output impedance, trying to control an electro-mechanical motor. The motor - the speaker - has a "mind of it's own" as to what it wants to do with the signal thrown at it and it will, if left alone, proceed to do whatever it thinks best with the signal. One way you can start to "leave it alone" is to interpose a crossover network between the amplifier and the driver. This prevents the amplifier from damping out the extraneous motion that the driver wants to make. It's like this: Put a long-duration pulse to the speaker (long duration being several times longer than 1 over the resonant frequency). The cone moves out, then moves back. But when it moved out, it probably over-shot the "correct" location and oscillated into place, finally getting to where it was supposed to be. Thes oscillations generate a "back emf" (remember that from your power engineering classes?) that tries to drive the amplifier. But if the amplifier (1) has a low output impedance and (2) is connected pretty close (i.e., through a very low impedance connection) then the back-emf will have to generate a lot of current to exist. Instead, the back emf "gives up" and the speaker doesn't oscillate around the resting location (yeah, I know that's really simplified, but it gets the point across). With the passive crossover there's lots of circuitry that can allow the current from the back-emf to flow and not be damped out by the low output impedance of the power amplifier. I's like trying to drive your car with a spring between your hands and the steering wheel. You can steer, but bumps will throw the car all over the place 'cause you can't correct for them with the springs in the way. The best way around this is to use a power amplifier for each speaker, and put active crossovers between the pre-amp and the power amplifiers. The advantages include the following: 1) The crossovers are in a well controlled impedance. (You could even use a passive implementation - if you really wanted to for some reason.) 2) The crossovers work with low-level signals instead of having to be capable of handling tens (hundreds?) of watts of power. 3) The speakers are driven directly by the (low output- impedance) power amplifier. 4) Active crossovers completely do away with inductors, the hardest part of any passive, high-level crossover to fabricate. In the "old days" when good power amplifiers were prohibitively expensive it was important to reduce their number. That just isn't the case now days. Today almost anyone can afford a multiple power amplifier system. Except that to *buy* the active crossovers is prohibitively expensive. But they are quite simple to build, and there is really no "magic" involved. Just check into National Semiconductor's "Audio/Radio Handbook" publication (circa 1980). (But use the Signetics 5534 op-amp.) Well this has gone on long enuff. If you haven't hit the 'n' key by now you must !really! be interested in this stuff... keith Keith Ericson at TekLabs (resident factious factotum) Tektronix, PO 500, MS 58-383 Beaverton OR 97077 (503)627-6042 uucp: [ucbvax|decvax|ihnp4|(and_many_others)]!tektronix!tekgvs!keithe CSnet: keithe@tek ARPAnet: keithe.tek@rand-relay -- Keith Ericson at TekLabs (resident factious factotum) Tektronix, PO 500, MS 58-383 Beaverton OR 97077 (503)627-6042 uucp: [ucbvax|decvax|ihnp4|(and_many_others)]!tektronix!tekgvs!keithe CSnet: keithe@tek ARPAnet: keithe.tek@rand-relay
rdp@teddy.UUCP (10/21/85)
In article <1259@tekgvs.UUCP> keithe@tekgvs.UUCP (Keith Ericson) writes: > The discussion of passive crossover design and the problems >involved in modelling the driver(s) is getting lots of net.audio space >lately. But, friends, it's *old technology* and not really current in >todays audio market-place, especially for those capable - as I believe >lots of net.audio reader are - of some relatively simple electronic >construction. (Right now you're saying 'what IS he talking about! - >bear with me...) > OK - the problem in a conventional (i.e., passive-crossover) >system is that you've got this monstrous power amplifier, with a >fraction of an ohm output impedance, trying to control an >electro-mechanical motor. The motor - the speaker - has a "mind of it's >own" as to what it wants to do with the signal thrown at it and it >will, if left alone, proceed to do whatever it thinks best with the >signal. One way you can start to "leave it alone" is to interpose a >crossover network between the amplifier and the driver. This prevents >the amplifier from damping out the extraneous motion that the driver >wants to make. It's like this: > Put a long-duration pulse to the speaker (long duration being >several times longer than 1 over the resonant frequency). The cone >moves out, then moves back. But when it moved out, it probably >over-shot the "correct" location and oscillated into place, finally >getting to where it was supposed to be. It will only overshoot the position if it is both electrically (as your are alluding to) and mechanically underdamped. It is interesting to note that total system Q does not change dramatically with source impedance (note Small, et al, JAES). The chances of this "overshoot" occuring are usually only found at the system resonance, see further... > Thes oscillations generate a >"back emf" (remember that from your power engineering classes?) that >tries to drive the amplifier. But if the amplifier (1) has a low >output impedance and (2) is connected pretty close (i.e., through >a very low impedance connection) then the back-emf will have to generate >a lot of current to exist. Instead, the back emf "gives up" and the >speaker doesn't oscillate around the resting location (yeah, I know >that's really simplified, but it gets the point across). With the >passive crossover there's lots of circuitry that can allow the current >from the back-emf to flow and not be damped out by the low output >impedance of the power amplifier. I's like trying to drive your car >with a spring between your hands and the steering wheel. You can steer, >but bumps will throw the car all over the place 'cause you can't correct >for them with the springs in the way. This is not only simplified, it's wrong. Look again at the situation of a woofer at resonance. Sure, it may want to overshoot, and it generates it's back EMF but there is NOT a lot of circuitry in the way. There might be a big mother inductor or two, but at these frequencies, the presented impedance of these inductors is damn near zero, else the woofer is putting out an attenuated signal. At higher frequencies, yes, the inductors now have a significant impedance, but this is unimportant for several reasons. 1), there is usually a shunt capacitor in the network, and it's impednace is now approaching 0, 2) the output of the woofer now is attenuated, so it's anomolies are becoming less and less important (that's the idea of a crossover, yes?) and 3) at these frequencies, the driver is being comtrolled by it's readiation chacteristics, not by it's second-oorder mechanical characteristics. The same holds true for tweeters. The big killer to the "crossovers are evil because they have high impedance and therefore stifle back-emf" argument is th fact that a typical "8 ohm" driver has a pure and simple resistor in series with it on the order of 7 to 7.5 ohms. This is the DC resistance of the voice coil. This resistance is always there. So it don't make diddly-skwat whether there is a .01 or a 1 ohm resistance due to the network (or lack of network). It is an interesting experiment to watch what hppens to the impedance curve at resonance of a woofer when it's hooked up to an amplifier with a source impedance of .01 ohms versus one with a source impedance of, say, 10 ohms. The impedance curve changes are damn near unmeasurable, indicating little, if any change to the second-order electro-mechanical characteristics of the driver. > The best way around this is to use a power amplifier for each >speaker, and put active crossovers between the pre-amp and the power >amplifiers. The advantages include the following: > 1) The crossovers are in a well controlled impedance. (You > could even use a passive implementation - if you really > wanted to for some reason.) ????? > 2) The crossovers work with low-level signals instead of > having to be capable of handling tens (hundreds?) of > watts of power. This is not a problem, given reasonable passive component selection. I have yet to see a crossover inductor saturate. Although I have seen capacitors explode! (Maybe the subject of another "Audio anecdote of the <interval>") > 3) The speakers are driven directly by the (low output- > impedance) power amplifier. As seen above, the same is true of properly designed passive networks. > 4) Active crossovers completely do away with inductors, the > hardest part of any passive, high-level crossover to > fabricate. I have found it a lot easier to make an inductor of audio-use size and accuracy than a capacitor. I mean, I don't even know how to make a capacitor of the sizes I need :-) > > In the "old days" when good power amplifiers were prohibitively >expensive it was important to reduce their number. That just isn't the >case now days. Today almost anyone can afford a multiple power amplifier >system. Except that to *buy* the active crossovers is prohibitively >expensive. But they are quite simple to build, and there is really no >"magic" involved. Just check into National Semiconductor's "Audio/Radio >Handbook" publication (circa 1980). (But use the Signetics 5534 op-amp.) > Well this has gone on long enuff. If you haven't hit the 'n' key >by now you must !really! be interested in this stuff... > >keith > Add to the cost of such active crossovers, another disadvantage is that they can add quit measurable noise and (possibly) distortion to the signal. Dick Pierce
dca@edison.UUCP (David C. Albrecht) (10/21/85)
> Put a long-duration pulse to the speaker (long duration being > several times longer than 1 over the resonant frequency). The cone > moves out, then moves back. But when it moved out, it probably > over-shot the "correct" location and oscillated into place, finally > getting to where it was supposed to be. Thes oscillations generate a > "back emf" (remember that from your power engineering classes?) that > tries to drive the amplifier. I'm kind of interested in this also. I did see a circuit in a TI application guide which used a 5534 as driver for a low powered amplifier < 20 watts directly connected to a speaker. The unusual thing about it was the circuit used current feedback from the speaker by puting a 0.5 ohm resistor in series and feeding back the voltage to the 5534. I was curious if anyone had comments on the use of such feedback in multi-amp speaker designs (i.e. general usefulness, should it be used only for the bass drive or is it good across the spectrum, etc.). David Albrecht