19-07-2016, 09:09 PM
For higher power amps, a lot are bridged. Once you get above ~300 watts into 8 ohms, the supply rails needed get a bit "hairy", and semiconductors become somewhat specialist. A pair of 300W amps bridged should give 1200W, but in practice the bridged output is always less because the doubled current means some voltage loss. But 1kW should be attainable with a suitably massive PSU.
Naturally, each half of a bridge "sees" half the load impedance, so you end up having to shovel in more output devices to cope. What doesn't help is irresponsible loudspeaker designers not paying attention to the impedance curve, so a nominal 8 ohm loudspeaker might well be substantially less than that at certain frequencies. The phase angles make a mess of things, too... And lowering the impedances helps to "massage" the sensitivity figures, as this is measured at 2.83V RMS (which is 1 watt into 8 ohms). They have to specify a voltage rather than a power because of this varying impedance malarkey (power amps being voltage sources).
Today, finding good, chunky output devices is not a problem - there's loads of decent ones (at a price!). The small-signal transistors are starting to become tricky, however. The phasing out of the TO92 plastic package began a while back, meaning that favourites like the MPSA06/56 are now obsolete. Others exist, but experience and confidence counts for a lot when you're designing amplifiers for production. As a result, it's tempting to use different topologies that have gain in the output stage, meaning you can use more "jellybean" transistors, or indeed op-amps for the small-signal stuff. Loop stability gets "interesting" when you go down that path, especially when you consider the multitude of "difficult" load impedances encountered.
Most PA amps are class G or H, using multiple rails. These can suffer from switching distortion, but that rarely matters for PA work. Domestically, bridged amps seem more common than G/H - now that output transistors with reduced "beta-droop" are available, driving low-Z loads is less of an issue. And, bridging gets the output current away from the ground system, which is nice...
I ignore class D. Too many compromises for high quality stuff...
Still, I enjoyed the WW reprint. As Rod says, some important points in there. Most obviously is PMR - peak to mean ratio. This is typically 20-30 dB on reasonable programme material that hasn't fallen in the name of the "Loudness Wars". Now, we judge loudness in a complex way, but essentially, it's the mean level that we pay most attention to. The peaks might require 100 to 1000 times the power of the mean level - this means that if you run a 100 watt amplifier on the verge of clipping, the average power delivered to the load might only be 1 watt! Or 0.1 watt. Scary, but true. And thank goodness, as if that wasn't the case, how would you make loudspeakers that didn't catch fire? But if you accept that you wish your music to be undistorted, it's surprising how much power you really need, even in a small domestic setup.
The large dynamic range of decent audio is what makes audio difficult and interesting. Throughout a recording chain, or throughout the broadcast chain, the dynamic range is gradually reduced as operational staff do their job, but still, you need to maintain plenty of it until it hits an Optimod.
Naturally, each half of a bridge "sees" half the load impedance, so you end up having to shovel in more output devices to cope. What doesn't help is irresponsible loudspeaker designers not paying attention to the impedance curve, so a nominal 8 ohm loudspeaker might well be substantially less than that at certain frequencies. The phase angles make a mess of things, too... And lowering the impedances helps to "massage" the sensitivity figures, as this is measured at 2.83V RMS (which is 1 watt into 8 ohms). They have to specify a voltage rather than a power because of this varying impedance malarkey (power amps being voltage sources).
Today, finding good, chunky output devices is not a problem - there's loads of decent ones (at a price!). The small-signal transistors are starting to become tricky, however. The phasing out of the TO92 plastic package began a while back, meaning that favourites like the MPSA06/56 are now obsolete. Others exist, but experience and confidence counts for a lot when you're designing amplifiers for production. As a result, it's tempting to use different topologies that have gain in the output stage, meaning you can use more "jellybean" transistors, or indeed op-amps for the small-signal stuff. Loop stability gets "interesting" when you go down that path, especially when you consider the multitude of "difficult" load impedances encountered.
Most PA amps are class G or H, using multiple rails. These can suffer from switching distortion, but that rarely matters for PA work. Domestically, bridged amps seem more common than G/H - now that output transistors with reduced "beta-droop" are available, driving low-Z loads is less of an issue. And, bridging gets the output current away from the ground system, which is nice...
I ignore class D. Too many compromises for high quality stuff...
Still, I enjoyed the WW reprint. As Rod says, some important points in there. Most obviously is PMR - peak to mean ratio. This is typically 20-30 dB on reasonable programme material that hasn't fallen in the name of the "Loudness Wars". Now, we judge loudness in a complex way, but essentially, it's the mean level that we pay most attention to. The peaks might require 100 to 1000 times the power of the mean level - this means that if you run a 100 watt amplifier on the verge of clipping, the average power delivered to the load might only be 1 watt! Or 0.1 watt. Scary, but true. And thank goodness, as if that wasn't the case, how would you make loudspeakers that didn't catch fire? But if you accept that you wish your music to be undistorted, it's surprising how much power you really need, even in a small domestic setup.
The large dynamic range of decent audio is what makes audio difficult and interesting. Throughout a recording chain, or throughout the broadcast chain, the dynamic range is gradually reduced as operational staff do their job, but still, you need to maintain plenty of it until it hits an Optimod.
