irb
pfm Member
There’s been some discussion about this on another thread. To avoid dragging it further off topic with a mammoth post, I thought I’d start a new thread. Apologies for the length of what follows.
The idea that extra transducers should be removed from the room has a longish history: in the 1980s, Ivor Tiefenbrun of Linn insisted that his dealers use ‘single-speaker dem rooms’ because, he claimed, extra undriven speakers in the room would wreak all kinds of havoc with the sound.
I was never convinced. It seemed too convenient that Linn made their dealers literally remove the competition from the room before conducting a demo. (Some of the ground is covered in this Stereophile article from 1985.)
But maybe the point was still correct. Maybe undriven speakers in the same room ‘sing along’. It’s well known that undriven speaker cones will vibrate, in response to environmental sounds. But that’s not necessarily a problem. They might just be moving with the air, absorbing sound from the environment. For it to be a problem, they’d have to continue vibrating after the initial stimulus, and they’d have to do it at a high enough level that it could audibly interfere with the sound of the driven speakers. And it’s far from clear that that happens.
There’s a fairly simple way to find out how much a cone is vibrating in response to environmental sound – measure the voltage generated by the magnetic circuit, at the speaker terminals. So that’s what I did today, with the speakers I had to hand: a pair of Scanspeak 10 inch woofers in DIY sealed enclosures, along with a single Apogee Stage full range ribbon.
I tried it both ways round – playing tones through the Scanspeak woofer pair and measuring the voltage generated across the undriven Apogee; and also playing sounds through the unnatural pairing of one Scanspeak and one Apogee, while measuring the voltage generated across the undriven Scanspeak.
Driving the Scanspeak pair, 3.62 volts was enough to make several things in the room rattle loudly, including, to my consternation, one of the Scanspeak DIY enclosures. I played three different frequencies in turn: 44 Hz, which I reckoned would stimulate the Apogee’s fundamental resonance, as well as 74 and 121 Hz, tones that would be some distance away from it.
The Apogee Stage was placed 60cm in front of one of the Scanspeak woofers. Sure enough, as soon as I played a tone through the Scanspeaks, my meter registered a voltage across the terminals of the undriven Apogee. With a 121 Hz tone, I measured 3.2mV. With a 74 Hz tone, I got 4.8mV. With a 44 Hz tone, I got 12mV.
Is that more or less than expected? Well, an output of 12mV at 44Hz is 49.6 dB down from the input voltage of 3.62 volts driving the Scanspeak Woofer. An output of 3.2mV at 121 Hz is 61dB down. That’s a long way down in level.
Next I switched the connections and played tones through the Scanspeak/Apogee combo, while measuring the voltage produced by the remaining undriven Scanspeak. I set the volume a bit lower this time, just 1.26 volts at the driven speaker terminals. At the terminals of the unconnected Scanspeak, I measured the following output voltages: with a 121 Hz tone, I got 4.8mV; the 74 Hz tone generated 3.3mV; and the 44Hz tone generated 41.7mV.
That’s clearly a much bigger output, and, on the face of it, bad news for me. The 74 Hz and 121 Hz output is still around 50 dB down on the input signal, but the 44 Hz output is only 29.6dB down.
I wasn’t ready to admit defeat, though. I tried a few different things: moving the driven and undriven speakers further apart helped a little – the voltage generated by the 44 Hz tone dropped from 41.7 mV to 38.4mV.
I thought about current flow: if there’s a low impedance path for current to flow through the undriven woofer, that current will resist the movement of the cone, damping the resonance (I think! – no doubt someone will tell me if I’m wrong). Most speakers, including the Apogees, have passive crossovers which provide some kind of current path (although the impedance at low frequencies may still be pretty high, as there’s likely to be a capacitor in the way). But my Scanspeaks are run actively so there’s no passive crossover, and no current path at all. I put a 100 uF capacitor across the Scanspeak’s terminals and measured again. Now the output was down a little more - to 34mV. Not a dramatic drop, admittedly.
The obvious solution would be to short the woofer terminals of the undriven speaker, providing a really low impedance current path. Of course, with a short across the terminals there’s no point trying to measure voltage there. What I did was rather unscientific, I admit, but it’s the best I could come up with. I fixed a wire to one of the undriven speaker’s terminals, so that it could easily be touched to the other terminal, shorting the woofer. Then I started the 44 Hz tone. I rested the fingers of one hand very gently on the undriven woofer’s rubber surround. Its vibrations could be felt quite easily. With the other hand I touched the wire to the second speaker terminal, to short its voice-coil. The vibration of the woofer stopped completely – as far as my fingers could tell. Opening and closing the shorting path confirmed this – with the path open, the vibrations could be felt easily. With the path closed, the vibrations appeared to stop instantly.
So, what have I learned? Well, I admit I’m surprised that an undriven woofer in a sealed box moves enough to generate a voltage that’s only 30 dB down from the input voltage. That’s more than I expected. However, I’m still not convinced this is a problem. For several reasons.
First, the voltage I measured shows that the undriven speaker was moving in response to the vibrations of the air around it. It shows that the undriven speaker was absorbing sound, in other words. It doesn’t show that the undriven speaker also emits sound, by resonating. If it is resonating, we need to know at what level, and for how long. Unfortunately my crude experiment can’t answer those questions. Here’s one opinion, though, from J. Gordon Holt, in that Stereophile article:
“[The] undriven cone has mass, and it takes energy to set any mass in motion, so practically none of the sound-wave pressure reaching that cone will ever leave it. Virtually all of the sound-pressure energy will have been used to move the mass of the cone. The only re-radiation that would occur from that "undriven" cone would be due to the release of stored-energy resonances from it after the cessation of air-pressure-induced motion. And in any loudspeaker we would give house room to, that stored energy should be more than an additional 50dB lower in amplitude below that of the cone vibrations which stored it. It is exceedingly unlikely that they would be audible.”
I’m sure this is right in most circumstances, but does it really apply in the case of an unshorted, undriven speaker? Well, think of what happens if you gently push a woofer cone in then release it. (I bet everyone’s done it, even though it’s potentially hazardous for the woofer.) Does it show any tendency to oscillate back and forwards as it returns to rest? No. Not even when the speaker is unconnected. Not even when the woofer is in free air. I tried it with the unconnected Scanspeak, (and also some unboxed sub drivers). In each case, the cone moved back quickly to its resting position, with absolutely no detectable sign of oscillation. OK, this isn’t exactly rigorous science, but I reckon J. Gordon Holt is probably right: any resonance will be well below the level of the initial movement.
Second, the fundamental resonance of most speakers is low – often below 50 Hz. Our ears are not very sensitive at those frequencies – perhaps some 20dB less sensitive than with midrange frequencies. So the contribution of a resonating undriven woofer will effectively be at least 50dB below the perceived midrange level (and if J. Gordon Holt is right, it will be at least 100dB below it). Also, we will not be able to locate the direction of sounds at those frequencies.
Third, there will be other things in pretty much every room that will produce more troublesome vibrations than the ones produced by undriven speakers.
Fourth, any undriven speaker in a room will interfere with the sound of a driven speaker, through the mechanisms of reflection and diffraction. It will do this because it is a large solid object, not because it is a transducer. I had an interesting confirmation of this as I was taking my measurements, in fact: as I measured the induced voltages, the meter was in one place, but I was free to move around the room. And as I moved, the meter reading changed, quite dramatically at times, doubling or halving even. Why? Because I was reflecting and diffracting sound, altering the modes in the room. (When I realised this was happening, I had to redo my measurements, making sure I stood in the same place for each one.)
So, my conclusion. Even if an undriven woofer does act like a passive radiator, it does so very weakly indeed. An ‘actual’ passive radiator will produce close to the level of the main driver. A passive radiator which produced 30dB less than the main driver would probably be inaudible. A passive radiator 50 or 60 dB down (even 80 dB down, if J. Gordon Holt is right) couldn’t possibly spoil the sound.
I was genuinely surprised at the voltage I measured from the Scanspeaks with a 44 Hz tone playing, but I’m actually reassured that undriven speakers in the room are not a problem in the real world. I won’t be lugging my extra speakers out the room any time soon.
And if you’re not convinced, there’s a simple solution: just short the undriven speaker terminals. If you’re still not convinced, go ahead and remove extra transducers from the room. But if you do, and the sound improves, it’ll almost certainly be because you removed a problem reflection. Or because, like me, you had a dodgy cabinet singing along.
The idea that extra transducers should be removed from the room has a longish history: in the 1980s, Ivor Tiefenbrun of Linn insisted that his dealers use ‘single-speaker dem rooms’ because, he claimed, extra undriven speakers in the room would wreak all kinds of havoc with the sound.
I was never convinced. It seemed too convenient that Linn made their dealers literally remove the competition from the room before conducting a demo. (Some of the ground is covered in this Stereophile article from 1985.)
But maybe the point was still correct. Maybe undriven speakers in the same room ‘sing along’. It’s well known that undriven speaker cones will vibrate, in response to environmental sounds. But that’s not necessarily a problem. They might just be moving with the air, absorbing sound from the environment. For it to be a problem, they’d have to continue vibrating after the initial stimulus, and they’d have to do it at a high enough level that it could audibly interfere with the sound of the driven speakers. And it’s far from clear that that happens.
There’s a fairly simple way to find out how much a cone is vibrating in response to environmental sound – measure the voltage generated by the magnetic circuit, at the speaker terminals. So that’s what I did today, with the speakers I had to hand: a pair of Scanspeak 10 inch woofers in DIY sealed enclosures, along with a single Apogee Stage full range ribbon.
I tried it both ways round – playing tones through the Scanspeak woofer pair and measuring the voltage generated across the undriven Apogee; and also playing sounds through the unnatural pairing of one Scanspeak and one Apogee, while measuring the voltage generated across the undriven Scanspeak.
Driving the Scanspeak pair, 3.62 volts was enough to make several things in the room rattle loudly, including, to my consternation, one of the Scanspeak DIY enclosures. I played three different frequencies in turn: 44 Hz, which I reckoned would stimulate the Apogee’s fundamental resonance, as well as 74 and 121 Hz, tones that would be some distance away from it.
The Apogee Stage was placed 60cm in front of one of the Scanspeak woofers. Sure enough, as soon as I played a tone through the Scanspeaks, my meter registered a voltage across the terminals of the undriven Apogee. With a 121 Hz tone, I measured 3.2mV. With a 74 Hz tone, I got 4.8mV. With a 44 Hz tone, I got 12mV.
Is that more or less than expected? Well, an output of 12mV at 44Hz is 49.6 dB down from the input voltage of 3.62 volts driving the Scanspeak Woofer. An output of 3.2mV at 121 Hz is 61dB down. That’s a long way down in level.
Next I switched the connections and played tones through the Scanspeak/Apogee combo, while measuring the voltage produced by the remaining undriven Scanspeak. I set the volume a bit lower this time, just 1.26 volts at the driven speaker terminals. At the terminals of the unconnected Scanspeak, I measured the following output voltages: with a 121 Hz tone, I got 4.8mV; the 74 Hz tone generated 3.3mV; and the 44Hz tone generated 41.7mV.
That’s clearly a much bigger output, and, on the face of it, bad news for me. The 74 Hz and 121 Hz output is still around 50 dB down on the input signal, but the 44 Hz output is only 29.6dB down.
I wasn’t ready to admit defeat, though. I tried a few different things: moving the driven and undriven speakers further apart helped a little – the voltage generated by the 44 Hz tone dropped from 41.7 mV to 38.4mV.
I thought about current flow: if there’s a low impedance path for current to flow through the undriven woofer, that current will resist the movement of the cone, damping the resonance (I think! – no doubt someone will tell me if I’m wrong). Most speakers, including the Apogees, have passive crossovers which provide some kind of current path (although the impedance at low frequencies may still be pretty high, as there’s likely to be a capacitor in the way). But my Scanspeaks are run actively so there’s no passive crossover, and no current path at all. I put a 100 uF capacitor across the Scanspeak’s terminals and measured again. Now the output was down a little more - to 34mV. Not a dramatic drop, admittedly.
The obvious solution would be to short the woofer terminals of the undriven speaker, providing a really low impedance current path. Of course, with a short across the terminals there’s no point trying to measure voltage there. What I did was rather unscientific, I admit, but it’s the best I could come up with. I fixed a wire to one of the undriven speaker’s terminals, so that it could easily be touched to the other terminal, shorting the woofer. Then I started the 44 Hz tone. I rested the fingers of one hand very gently on the undriven woofer’s rubber surround. Its vibrations could be felt quite easily. With the other hand I touched the wire to the second speaker terminal, to short its voice-coil. The vibration of the woofer stopped completely – as far as my fingers could tell. Opening and closing the shorting path confirmed this – with the path open, the vibrations could be felt easily. With the path closed, the vibrations appeared to stop instantly.
So, what have I learned? Well, I admit I’m surprised that an undriven woofer in a sealed box moves enough to generate a voltage that’s only 30 dB down from the input voltage. That’s more than I expected. However, I’m still not convinced this is a problem. For several reasons.
First, the voltage I measured shows that the undriven speaker was moving in response to the vibrations of the air around it. It shows that the undriven speaker was absorbing sound, in other words. It doesn’t show that the undriven speaker also emits sound, by resonating. If it is resonating, we need to know at what level, and for how long. Unfortunately my crude experiment can’t answer those questions. Here’s one opinion, though, from J. Gordon Holt, in that Stereophile article:
“[The] undriven cone has mass, and it takes energy to set any mass in motion, so practically none of the sound-wave pressure reaching that cone will ever leave it. Virtually all of the sound-pressure energy will have been used to move the mass of the cone. The only re-radiation that would occur from that "undriven" cone would be due to the release of stored-energy resonances from it after the cessation of air-pressure-induced motion. And in any loudspeaker we would give house room to, that stored energy should be more than an additional 50dB lower in amplitude below that of the cone vibrations which stored it. It is exceedingly unlikely that they would be audible.”
I’m sure this is right in most circumstances, but does it really apply in the case of an unshorted, undriven speaker? Well, think of what happens if you gently push a woofer cone in then release it. (I bet everyone’s done it, even though it’s potentially hazardous for the woofer.) Does it show any tendency to oscillate back and forwards as it returns to rest? No. Not even when the speaker is unconnected. Not even when the woofer is in free air. I tried it with the unconnected Scanspeak, (and also some unboxed sub drivers). In each case, the cone moved back quickly to its resting position, with absolutely no detectable sign of oscillation. OK, this isn’t exactly rigorous science, but I reckon J. Gordon Holt is probably right: any resonance will be well below the level of the initial movement.
Second, the fundamental resonance of most speakers is low – often below 50 Hz. Our ears are not very sensitive at those frequencies – perhaps some 20dB less sensitive than with midrange frequencies. So the contribution of a resonating undriven woofer will effectively be at least 50dB below the perceived midrange level (and if J. Gordon Holt is right, it will be at least 100dB below it). Also, we will not be able to locate the direction of sounds at those frequencies.
Third, there will be other things in pretty much every room that will produce more troublesome vibrations than the ones produced by undriven speakers.
Fourth, any undriven speaker in a room will interfere with the sound of a driven speaker, through the mechanisms of reflection and diffraction. It will do this because it is a large solid object, not because it is a transducer. I had an interesting confirmation of this as I was taking my measurements, in fact: as I measured the induced voltages, the meter was in one place, but I was free to move around the room. And as I moved, the meter reading changed, quite dramatically at times, doubling or halving even. Why? Because I was reflecting and diffracting sound, altering the modes in the room. (When I realised this was happening, I had to redo my measurements, making sure I stood in the same place for each one.)
So, my conclusion. Even if an undriven woofer does act like a passive radiator, it does so very weakly indeed. An ‘actual’ passive radiator will produce close to the level of the main driver. A passive radiator which produced 30dB less than the main driver would probably be inaudible. A passive radiator 50 or 60 dB down (even 80 dB down, if J. Gordon Holt is right) couldn’t possibly spoil the sound.
I was genuinely surprised at the voltage I measured from the Scanspeaks with a 44 Hz tone playing, but I’m actually reassured that undriven speakers in the room are not a problem in the real world. I won’t be lugging my extra speakers out the room any time soon.
And if you’re not convinced, there’s a simple solution: just short the undriven speaker terminals. If you’re still not convinced, go ahead and remove extra transducers from the room. But if you do, and the sound improves, it’ll almost certainly be because you removed a problem reflection. Or because, like me, you had a dodgy cabinet singing along.