About every three weeks, we get a new thread with the exact same misguided questions.
I do not blame the people who join and ask the questions. I blame the hucksters that market audio equipment. They have succeeded with totally bamboozling the public on topics like power, watts, frequency response, etc. Never mind what engineers have to say on the topic!
Should be stickied!
I do not blame the people who join and ask the questions. I blame the hucksters that market audio equipment. They have succeeded with totally bamboozling the public on topics like power, watts, frequency response, etc. Never mind what engineers have to say on the topic!
Wattage (power) is usually measured with a sine wave test signal, at one single frequency. The amount of power delivered to the speakers is determined by the amplifier and speaker. An ideal crossover has no losses, and will have no effect on power delivered to the speaker - as long as the sine wave is within the frequency range that the crossover lets through itself.
Real crossovers have some small amount of power loss, because of the resistance of the inductor. So they will slightly lower the amount of power delivered to the speaker.
Things are different if you are talking about music, rather than a sine test tone. If music is played through the speaker, then the crossover divides up some frequencies to send to the woofer, and some to send to the tweeter. If you measure average power, usually most of the power ends up going to the woofer, and very little to the tweeter. But there may be very short bursts of higher power levels to the tweeter.
Since most of the power goes to the woofer, you can think of the system as still having only 40W of power handling capability.
I should add that the "40 W" number may be optimistic - some manufacturers inflate their power ratings to make the advertisement look more tempting. I have seen "25 W" speakers that would burn up or fall apart if you fed them even 10 watts.
15W and 40W are not the actual power delivered to the speaker - they are only a manufacturers specification for the maximum amount of power that speaker can handle without damage.
The actual amount of power delivered depends on the amplifier (how many volts is it putting out?), and speaker impedance. That may be only 0.1 watt, or it may be 1 watt, or it may be 10 watts, or anything in between.
An analogy: if your car has a 15 gallon gas tank, that doesn't mean you have 15 gallons of gas in it. You may only have 5!
Forget all about that 55W, it is confusing the heck out of you.
The power you need is whatever is enough to play music at the loudness you want.
The speaker power handling needs to be able to cope with this. Hopefully, you are satisfied with the volume level before you get to 40 watts.
A small amount of power is wasted in the resistance of the coils in the crossover network. This power heats up the coils, and also creates a magnetic field in the iron core (you can see the core in the Amazon.com photos you linked to.)
The 150W rating of the crossover means that the coil will survive the amount of heat, and magnetic field, that will be created if you pump 150 watts of music through that crossover, from amplifier to speaker.
First, I'm going to give you the super-simplified version: the crossover has no effect on impedance. If you have two 8 ohm speakers, your combined (crossover + woofer + tweeter) is still an 8 ohm system.
In reality, the question you asked is a big can of worms to open. The crossover network you bought from Amazon will work perfectly if loudspeakers were resistors, but they are not. Real speakers have big peaks and dips in their own impedance at different frequencies, and this "confuses" the crossover network, which ends up presenting even bigger peaks and dips in impedance to the amplifier. Real tweeters and woofers also have very different characteristics from each other (dispersion at the crossover frequency, sensitivity, time response), and a good crossover network has to take all that into account.
Real crossover networks in good loudspeaker systems are far more complex than the ones you can buy from Amazon or Parts Express. Designing them is part engineering, and part art. It is a very complicated thing to do correctly, and needs a lot of knowledge, a lot of expensive equipment, and an even more expensive anechoic chamber in which to do the measurements. Because of this, the vast majority of DIY loudspeakers don't get crossovers right.
But the wonderful thing about loudspeakers is that, however bad the crossover, enclosure, or drivers, some sort of sound will come out of them. This is enough to make many DIY speaker builders happy!
-Gnobuddy
Should be stickied!
Actually, oddball speaker impedances are the norm in consumer oriented integrated devices. This includes mobile applications.
My Subaru had a powered "subwoofer" under the seat. The "subwoofer" driver was rated 1.6 ohms, stamped right on the frame. The amplifier was a cheap and dirty digital amplifier.
Another Subaru had a 6.5" "subwoofer" mounted in the rear deck. It disintegrated (big surprise) and I was charged with fixing it. As far as I could tell it was a 1 ohm subwoofer. I ordered a cheap dual voice coil "2+2 ohm" subwoofer and wired the voice coils in parallel. It blew the original subwoofer away and only cost around $20.
My buddy's Navigator had a 1.6 ohm subwoofer that was blown. We put a box in the back with two 4 ohm subs wired in parallel. The cheapo OEM digital amplifier does a nice job of driving them.
Integrated home systems (cheapo tabletop stereos) often have 3 ohm speakers.
This stuff is not designed to be mixed and matched any more. You buy it, use it until something goes wrong, then you buy another cheap and crappy system. That's the market now.
Unfortunately, I think you are dead on the money.
And I thought 6-ohm speakers ("Our CrappMaxPlus speakers will work with all 4 and 8 ohm systems!") were bad enough!
-Gnobuddy
I have a rant about "4-8 ohm" speakers somewhere in the forum. "4-8- ohm" = 4 ohm. "6 ohm" = 3-4 ohm. I've done the measurements.
Some popular speakers make great demands from amplifiers to only produce modest performance. These popular speakers sound thin and weak unless driven by an amplifier with generous current reserve. That's what makes them "high end"
.I was thinking about Old'n'Cranky's comments, and I wonder if this is partly a disconnect between old imperial units, and more contemporary metric units.
In American and British English, it seems to be common to use "acre" as a unit of land area, and "acreage" to describe a piece of land. In the same way, distance is measured in miles, and people speak of the "gas mileage" their car gets.
We also have feet and footage ("...a property with 150 feet of waterfront footage..."). "Yardage" is quite commonly used too, and is listed in several online dictionaries I checked.
A number of online dictionaries also list "poundage" as a description of weight, and of course, "tonnage" has been used for centuries to describe the freight carrying ability of ships, and more recently, trains and trucks.
You can count on Imperial units being inconsistent, so it's no surprise that nobody uses "inchage" or "fathomage" or "yearage".
In the world of metric units, on the other hand, you never hear of "millimetreage" or "meterage" or "grammage" or "kilogramage".
But all my life, I have heard "voltage" and "wattage" being used by educated people - in countries still using imperial units. Maybe that's a hangover from their being used to adding "age" to various imperial units?
Just a thought.
And I close with an example of how to use the best word in this entire thread: "How much milkerage do you get from your cow?"
-Gnobuddy
...and they use "condensor" to mean different things: it's used to mean the capacitor in an old-fashioned ignition system, and also used to describe the little radiator that cools off the high-pressure refrigerant coming from the compressor in the air-conditioning subsystem, turning it back into liquid.
I'm technically an amateur at most things I do. Quite possibly a disconnected one as well.The same disconnect sometimes applies to amateurs
-Gnobuddy
Grammage is actually commonly used in food industry or retail in continental Europe, the same for meterage ;-)
Last edited:
The speaker power rating figure is the maximum continuous power rating (i.e. for a 1 kHz sine-wave input) of the amplifier they can handle music and speech signals from at maximum level before clipping. E.g. 100 W rated speakers can't handle 100 W continuous power, they can handle music and speech signals at maximum level before clipping from an amplifier rated for 100 W continuous power. Music and speech have much lower average power than a continuous sine-wave of the same peak level, since the peaks are far less frequent. The continuous power the speakers can handle will be far less than 100 W.
Still, the maximum sound level the speakers are capable of producing, the less they will distort the sound at lower levels. The sound quality will still be improved even if they are not used near the maximum level they can handle without burning out.
What kind of complexity is required, aside from compensation for speaker response dips and peaks and the speaker impedance variations at different frequencies, speaker efficiency differences (which basic, ready-made crossovers don't have)?
Wow, nearly two years later, the thread lives again!
In practice, the simple textbook-formula crossover network ignores everything that actually matters, starting with two very important things you mentioned (impedance curves, efficiency differences.) But there are several other important factors that are also left out of the ready-made crossover design:
1) As you hinted, "Textbook" crossovers don't work with the real impedance curve of real drivers. The acoustic response of the tweeter, in particular, can be utterly horrific if nothing is done to flattend the tweeter impedance curve. Impedance compensation is a must for the tweeter, unless you're willing to settle for really poor speaker frequency response.
2) Adding tweeter impedance compensation components will now change the acoustic response of the tweeter. You need an expensive accurate calibrated measurement microphone and an even more expensive anechoic chamber (or a good substitute, such as hanging your speaker and measurement microphone high up in the air by a cable run between two tall trees) to measure this change, so you can correct for it - by redesigning the crossover network as necessary.
3) Also as you hinted, textbook crossovers don't allow for efficiency differences between drivers. Adding a series resistor to lower tweeter efficiency now changes the tweeter's acoustic frequency response curve once again.
So it's back to the anechoic chamber with the calibrated mic to measure this change and compensate for it by re-designing the tweeter crossover network.
Now that you've changed the tweeter crossover, the combined impedance plot (woofer + tweeter) may no longer be well behaved. You may have to add more impedance compensation components. Which may change the acoustic response again, so you may have to do the whole thing a few times, iterating gradually until you have a good enough result.
4) Textbook crossovers don't allow for the fact that all woofers are themselves acoustic low-pass filters in the region of the crossover frequency, and all tweeters are themselves acoustic high-pass filters.
A good crossover network often ends up with asymmetric filter responses for woofer and tweeter, because what matters is the combined acoustic frequency response, not the electrical frequency response measured at the driver terminals.
And once again, once the crossover provides a smooth acoustic frequency response, you may find the combined impedance plot of (woofer + tweeter) is once again unacceptable, with big peaks and dips in the crossover region. Again, you may have to add new compensation components, which again may change the frequency response a little, which again will need the crossover network values tweaked a little. Round and round and round, until everything converges to good-enough.
5) Textbook crossovers don't allow for the fact that acoustic coupling between the closely-spaced woofer and tweeter can change the frequency response in the crossover region. This can be ignored if you're making crude speakers with big peaks and dips in the frequency response, but if you're trying to make a good system flat to +/- 1 or 2 dB in the midrange, you can't ignore it.
6) Textbook crossovers don't allow for the fact that the moving parts in woofers are physically heavier than the moving parts in tweeters, so the woofer is "slow" and has extra phase lag around the crossover frequency. This can cause dips and peaks in the frequency response near the crossover, and tweaking the crossover (by trial and error) can reduce these.
In a nutshell, it's hardly exaggeration to say that virtually nothing works the way the textbook formula predicts, and virtually everything interacts with everything else.
Because of this, there is no purely theoretical way to design a really good crossover network, even with good software tools. It always comes down to cut and try, repeatedly, guided by accurate measurements of acoustic frequency response, electrical frequency response, and electrical impedance curve, until you have a good-enough result.
Most speaker manufacturers don't care enough to bother to do all this. They settle on "good enough" after a couple of rough measurements and tweaks.
Using active crossover networks helps a lot by reducing the need for impedance compensation and by reducing the electrical interaction between the high-pass and low-pass sections of the crossover network. But you still have all the other stuff to deal with (driver frequency responses are not flat, drivers interact acoustically, woofer is slow and laggy compared to tweeter at the crossover frequency.)
For a couple of years I worked side-by-side as part of a team with a speaker engineer with a good pair of ears, a quest for perfection, and access to some very expensive and high-quality measurement software and hardware. For me, it was a revelation to find out just how much work and time it took to create a really good passive crossover network. The finished network might not have a whole lot of components in it - it might not look super complex - but getting to that final circuit and those specific component values might have taken weeks or even months of effort by several people.
-Gnobuddy
It would be nice if that were true. Sadly, both speakers and amplifiers frequently have dishonest ratings. Ebay and Amazon are filled with ads for "100 W" amplifiers that would struggle to put out 20 W RMS. Speaker outlets are filled with adds for "100 W" speakers that will burn up if you feed them the full unclipped output from a 100 W amplifier playing representative music.
If this was true, 1000W P.A. speakers designed to produce up to 130 dB SPL would sound better than 50 W studio reference monitors when playing at a sane 80 dB SPL. But that isn't the case.
Basically the simple ready-made crossover would work perfectly if all drivers had perfectly flat impedance and frequency response curves - in which case, we wouldn't need crossover networks in the first place!
In practice, the simple textbook-formula crossover network ignores everything that actually matters, starting with two very important things you mentioned (impedance curves, efficiency differences.) But there are several other important factors that are also left out of the ready-made crossover design:
1) As you hinted, "Textbook" crossovers don't work with the real impedance curve of real drivers. The acoustic response of the tweeter, in particular, can be utterly horrific if nothing is done to flattend the tweeter impedance curve. Impedance compensation is a must for the tweeter, unless you're willing to settle for really poor speaker frequency response.
2) Adding tweeter impedance compensation components will now change the acoustic response of the tweeter. You need an expensive accurate calibrated measurement microphone and an even more expensive anechoic chamber (or a good substitute, such as hanging your speaker and measurement microphone high up in the air by a cable run between two tall trees) to measure this change, so you can correct for it - by redesigning the crossover network as necessary.
3) Also as you hinted, textbook crossovers don't allow for efficiency differences between drivers. Adding a series resistor to lower tweeter efficiency now changes the tweeter's acoustic frequency response curve once again.
So it's back to the anechoic chamber with the calibrated mic to measure this change and compensate for it by re-designing the tweeter crossover network.
Now that you've changed the tweeter crossover, the combined impedance plot (woofer + tweeter) may no longer be well behaved. You may have to add more impedance compensation components. Which may change the acoustic response again, so you may have to do the whole thing a few times, iterating gradually until you have a good enough result.
4) Textbook crossovers don't allow for the fact that all woofers are themselves acoustic low-pass filters in the region of the crossover frequency, and all tweeters are themselves acoustic high-pass filters.
A good crossover network often ends up with asymmetric filter responses for woofer and tweeter, because what matters is the combined acoustic frequency response, not the electrical frequency response measured at the driver terminals.
And once again, once the crossover provides a smooth acoustic frequency response, you may find the combined impedance plot of (woofer + tweeter) is once again unacceptable, with big peaks and dips in the crossover region. Again, you may have to add new compensation components, which again may change the frequency response a little, which again will need the crossover network values tweaked a little. Round and round and round, until everything converges to good-enough.
5) Textbook crossovers don't allow for the fact that acoustic coupling between the closely-spaced woofer and tweeter can change the frequency response in the crossover region. This can be ignored if you're making crude speakers with big peaks and dips in the frequency response, but if you're trying to make a good system flat to +/- 1 or 2 dB in the midrange, you can't ignore it.
6) Textbook crossovers don't allow for the fact that the moving parts in woofers are physically heavier than the moving parts in tweeters, so the woofer is "slow" and has extra phase lag around the crossover frequency. This can cause dips and peaks in the frequency response near the crossover, and tweaking the crossover (by trial and error) can reduce these.
In a nutshell, it's hardly exaggeration to say that virtually nothing works the way the textbook formula predicts, and virtually everything interacts with everything else.
Because of this, there is no purely theoretical way to design a really good crossover network, even with good software tools. It always comes down to cut and try, repeatedly, guided by accurate measurements of acoustic frequency response, electrical frequency response, and electrical impedance curve, until you have a good-enough result.
Most speaker manufacturers don't care enough to bother to do all this. They settle on "good enough" after a couple of rough measurements and tweaks.
Using active crossover networks helps a lot by reducing the need for impedance compensation and by reducing the electrical interaction between the high-pass and low-pass sections of the crossover network. But you still have all the other stuff to deal with (driver frequency responses are not flat, drivers interact acoustically, woofer is slow and laggy compared to tweeter at the crossover frequency.)
For a couple of years I worked side-by-side as part of a team with a speaker engineer with a good pair of ears, a quest for perfection, and access to some very expensive and high-quality measurement software and hardware. For me, it was a revelation to find out just how much work and time it took to create a really good passive crossover network. The finished network might not have a whole lot of components in it - it might not look super complex - but getting to that final circuit and those specific component values might have taken weeks or even months of effort by several people.
-Gnobuddy
I think what Allan meant was that for a given speaker, distortion will be less if operated at a lower level - your example 1000W/130dB speaker will sound cleaner at 10w/110dB than at 100W/120dB for example.
You're perfectly right that that breaks down when comparing two different speakers however, as they will have been designed with different goals, and thus different max SPL capabilities, distortion thresholds etc etc.
Cheers,
David.
Wow, nearly two years later, the thread lives again!
I found this thread with a Google search of the topic.
That is also true but I pointed out the actual meaning of honest ratings because misinterpretation of those might also be a factor in disappointing results, if the buyers are testing the speakers with continuous, sinewave signals rather than just playing music through them.
In addition to music having a lower average power level than a continuous sinewave of same peak level, the amplifier can deliver music signals around 15% higher in level than a continuous sinewave because there is a lower voltage drop at the amplifier's supply rails for a music signal compared to a continuous signal and the filter capacitors can sustain the less frequent transient, maximum peaks and prevent clipping of them.
My reasoning is that if a speaker driver is built to handle more power, the main reason for this is so that it can reproduce sound at higher levels and hence not only is the voice coil bigger in size to be able to dissipate more heat, but the cone and suspension must be capable of making greater excursions without nonlinearities.
There are other reasons for higher power capacity, such as lower efficiency speakers or use in lower-efficiency system designs. But the usual reason is for the sake of producing higher sound levels.
PA speakers are designed primarily for sound dispersion for filling a large area with sound rather than sound quality and any given PA speaker unit is not designed to reproduce full-range sound. They are also built for durability in live settings and for high output levels so big compromises have to be made for how they operate at home listening levels, which they would never seriously be used for.
Thank you for this detailed response (not a pun). I was actually wondering how complex the circuit would need to be but I can see that the complexity is in finding the right circuit, due to the complexities of the speaker system it is designed for.
Out of curiosity, just how bad is the sound quality of a speaker system with basic crossover design? How many dB, for instance, will the dips and peaks be in size?
Let's say a basic three-way system, standard closed-box enclosure, with respectable electronics store drivers, with driver centers vertically aligned and drivers closely spaced, and a well-designed but basic first-order crossover (low-pass, band-pass, high-pass), with crossover points placed well within the "flat" regions of the responses.
Last edited:
My point was that a driver designed to produce higher maximum sound levels should have better linearity (i.e. in it's excursions) at lower levels than a driver designed to produce lower maximum sound levels. In other words, build your system like a Rolls Royce: The ride will be better even if you never drive it at top speed.
I agree with you. But you might be surprised, there was a thread on this very forum in which several people expressed the opinion that P.A. speakers were great choices to use for home Hi-Fi.
Exactly!
From memory, probably the worst problems occur because of the huge peak in the tweeters impedance at its fundamental resonance.
By interacting with the high-pass filter in the crossover network, this impedance peak can really mess up the tweeter's actual frequency response curve. So this the first thing to correct.
Unfortunately, it's not the easiest thing to correct! You need a way to measure the impedance curve, for starters, then you have to calculate values of the L-C-R components in the impedance compensation network, then buy them, build the circuit, and re-measure to see if it worked. It doesn't always work exactly as predicted, so you may need to tweak the values of L or C or R. More money, more parts-buying, more time spent, more building and measurements. (It helps when you have a few parts bins full of an array of inductors, capacitors, and resistors right there at your fingertips!)
I wanted to give you a better answer than just hand-waving as I've done above, so I did a little simulation work in LTSpice. Imagine we have a 4 ohm tweeter with its fundamental mechanical resonance (Fs) at 1 kHz. We decide for various reasons to use a textbook second-order high-pass Butterworth filter with a nominal crossover frequency of 2 kHz (which is one octave above Fs.)
I looked up some sample tweeter impedance curves online, then used LTSpice to design and simulate an electrical circuit that more or less duplicates a typical tweeter impedance curve (image attached.)
Because I'm driving the network with a 1-amp constant current source, one ohm of impedance translates to 1 volt across the network. Therefore the Y-axis values can be read as impedance in ohms (though labelled as volts by LTSpice.)
This particular (hypothetical) tweeter has Rs = 3 ohms, voice coil inductance is 30uH, and the impedance peaks up to about 9 ohms at 1kHz, the mechanical resonance frequency of the speaker suspension. These are reasonably typical values.
In my next post, we'll look at what happens if we hook up this sort of an impedance to a textbook 2nd order Butterworth LC high pass filter.
-Gnobuddy
Attachments
Next, I used LTSpice to simulate the frequency response of a 2nd-order LC high-pass network with a 2 kHz crossover frequency, operating into a perfectly constant textbook 4-ohm load. L & C values were calculated using an online crossover network calculator. I added 0.2 ohms to represent the resistance of the inductor itself as well as the loudspeaker cables running from the amplifier to the loudspeaker.
The result is shown as the green curve (Vtweeter1). There are no surprises - it has the characteristic smooth shape we expect from a textbook Butterworth high-pass response.
However, as you can see, I also simulated the frequency response of the same LC Butterworth filter, but loaded with our realistic tweeter impedance, which varies with frequency, unlike the constant 4-ohm load resistor.
The result is the purple curve (Vtweeter2). And just one glance is enough to show that it is UGLY! There are departures of up to +/-3 dB from the "textbook" green curve. The curve isn't smooth through the critical crossover region, with an ugly kink at the tweeter resonance frequency (1 kHz).
I don't have the time to do a similar simulation for the woofer and/or combined woofer/tweeter response. But it's easy to see that if the tweeter curve is bumpy and distorted by up to +/- 3dB, the composite woofer/tweeter curve certainly isn't going to be any better. (It will be worse, because the woofer response won't be perfect, either.)
Keep in mind, this is only half the story: we've only looked at the electrical signals at the driver terminals. What we really care about is the acoustic response of the driver, and we know that will be even worse than the bumpy purple line - because the tweeter has its own acoustical imperfections, as does the woofer.
-Gnobuddy
The result is shown as the green curve (Vtweeter1). There are no surprises - it has the characteristic smooth shape we expect from a textbook Butterworth high-pass response.
However, as you can see, I also simulated the frequency response of the same LC Butterworth filter, but loaded with our realistic tweeter impedance, which varies with frequency, unlike the constant 4-ohm load resistor.
The result is the purple curve (Vtweeter2). And just one glance is enough to show that it is UGLY! There are departures of up to +/-3 dB from the "textbook" green curve. The curve isn't smooth through the critical crossover region, with an ugly kink at the tweeter resonance frequency (1 kHz).
I don't have the time to do a similar simulation for the woofer and/or combined woofer/tweeter response. But it's easy to see that if the tweeter curve is bumpy and distorted by up to +/- 3dB, the composite woofer/tweeter curve certainly isn't going to be any better. (It will be worse, because the woofer response won't be perfect, either.)
Keep in mind, this is only half the story: we've only looked at the electrical signals at the driver terminals. What we really care about is the acoustic response of the driver, and we know that will be even worse than the bumpy purple line - because the tweeter has its own acoustical imperfections, as does the woofer.
-Gnobuddy
Attachments
Oh-oh. Venturing into subjective territory? Almost always a bad move on an audio forum. Particularly an audio forum where some very vocal people claim to able to hear 0.002% THD or sonic improvements created by NOS papyrus-in-mermaid-oil capacitors.
That said, my opinion, my experience, is that most P.A. speakers sound bad, to varying degrees, at all audible SPL levels. They sound bad playing quietly, and they sound bad playing loudly. Compared to, say, a good studio reference monitor.
I'm sure there is an exception somewhere, a combination of an exceptionally good (and probably exceptionally expensive) P.A. speaker, and a truly rotten white-van-special "studio monitor". But as a rule, P.A. systems compromise quality to achieve SPL, and the compromises tend not to be subtle.
My pet hate is the absolutely horrid horn-loaded compression driver midrange/tweeters used in so many relatively affordable P.A. systems. The way those things mangle vocals, particularly female vocals, ought to be a crime.
I have no professional involvement with P.A. systems, and never have. But like most people interested in music, I've been exposed to a fair number of them over the years.
-Gnobuddy
I have complete faith in the Taylor Series ( Calculus II - Taylor Series ), if that's what you were wondering.
The Taylor Series is the mathematical basis for the fact that most real-world nonlinear systems are very nearly linear for small enough disturbances. This applies to everything from suspension bridge cables and simple pendulums to loudspeaker drivers.
As an aside, IMO P.A. speakers aren't usually bad because of their nonlinearites - the ear will tolerate a surprisingly large amount of nonlinear distortion.
P.A. systems are bad because they have awful frequency response curves and awful polar response curves. And human ears are much more sensitive to those two things than they are to small amounts of THD.
-Gnobuddy
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.