Output Stage Protection on Amplifiers
Hi Michel (and others interested!),
Thank you for your input on this vexing issue.
Protection is an important consideration for audio amps, particularly SOAR protection. Almost none of them are transparent sonically, however, particularly where series relays are concerned. It's like autos; we all want performance, but we must meet emissions standards, and we must have fuel economy. To some extent, these requirements are mutually exclusive. And so it is with SOAR protection...........
Why do we need it? It is required in the event the output goes DC to the rail and burns up the voice coils on direct coupled drivers, usually the woofer, which fortunately has a large voice coil and is in fact not too vulnerable in a large system. The midrange and treble, however, with passive crossovers, do not suffer in this scenario since they are DC blocked with a cap.
An output stage generally fails once in its life - if only every 20 years. The idea of taking out an expensive NHT1259 woofer is unthinkable. So how do we protect the woofer?
The output devices, in my view, are expendable. They are much cheaper than the driver. Someone has to lay down their life, after all, to protect the queen bee. So let's design the output stage so it is absolutely stable, never oscillates, and not worry too much about SOAR which almost invariably has major compressive effect on real music.
OK, so we can dispense with SOAR. Now, the failure mode. Let's select transistors which fail by gradation. I have designed the AKSA so the NPN output will go before the PNP on most loads. When this happens, the positive rail takes out the fuse, and the NPN device is now a short wire leading nowhere. However, the PNP is still alive, so the amplifier's global offset control remains effective. You might have 1 or 2 volts DC on the output, but the voice coil on the woofer is intact. None of my AKSAs has ever, to my knowledge, taken out a voice coil.
Overcurrent protection is important on speakers, and I recommend the polyphase switch, which is peanuts at RatShack and works with no audible signature I can detect. Let's use it.
Series relays on loudspeakers are anathaema. The only ones I have found which are inaudible are mercury relays, using the liquid metal and physical orientation to connect the signal. But they are slow, so not very practical. Perhaps the best solution is the shunt relay to ground. Here we set up a DC detection circuit on the output - very simple with a comparator or even a simple transistor/diode/cap circuit - and use it to trip a relay which effectively shorts the output to ground. Of course, a direct short will kill the output devices, so maybe a 1.5R resistor will offer the remaining intact device some small chance of survival. This short will, within about 200mS, blow the rail fuse of the intact device, removing all power from the output. The PNP device may even survive this trauma. The beauty of it is that the relay is open during normal operation, and thus has no sonic effect whatever.
This protection regime I would consider on some of my future designs. However, it is bulky, and quite expensive, and if the amp does not take out voice coils as it fails (and one can only rely on large numbers of amps being sold and used over a period of months to verify this is indeed the case), then even this approach is probably unnecessary. Testimony to this philosophy is the large number of high end amplifiers which use no output protection aside from fuses.
Thank you for bringing up this complex topic, Michel. I hope my take on this serves to explain the anatomy of output stage failure.
Cheers,
Hugh
Hugh R. Dean
www.printedelectronics.com
Melbourne AUSTRALIA