Soft Sabotage

NAD manufactured some really good sounding amplifiers. Until the 80's when they stopped making the 3020. While later NAD amps were encumbered with marketing devices such as their "Power Envelope" circuit, the 3020 was afflicted only by a somewhat fragile output stage (despite the fact that low impedance drive capability was claimed; I bet they held their breaths during the demo) and its "Soft Clipping" circuit, activated by a switch on the rear. Other than Soft Clipping, the 3020's limitations should be viewed in context of its low cost and as such are totally forgivable. Almost 40 years on it is still not only a good sounding amp, but an iconic design that NAD never equaled, let alone superseded. While Soft Clipping, in theory, may be useful, (then again Communism works, in theory) the Power Envelope circuit is a rabid basket of amplifier destroying bad ideas. I'm not going to let the PE debacle deflect me too far from the target of this polemic (Soft Clipping), because unlike Soft Clipping, one can't just turn it off (although it's probably already done so itself in a yet undetected Seppuku) so let's just despatch PE quickly and move on. The idea behind PE is that it allows the amplifier to output relatively high power for a short time, reducing to lower continuous power, hopefully before the output transistors are destroyed. The circuit is implemented in a technically cumbersome fashion via a second power supply, another set of power transistors (which got bigger over time as the failures racked up), and some less expensive components. It exists, not so much to provide high peak power and low continuous power, but as a marketing implement. If those extra components, cost and PCB space were allocated to a bigger power supply and a better output stage, the amp could have higher continuous power all the time, with a simpler, more reliable, better sounding circuit to boot. I'm glad we got that out of the way. 

Soft Clipping, on the other hand, sounds rational. I'll try to keep this technically simple, so apologies in advance to any Comrades expecting an in depth treatise, which would bore most folks.

An amplifier can only swing an output signal from its positive power supply rail, through zero to its negative rail and back again, as the signal demands. As the voltage rises and falls, your speaker moves in and out like so:

The chuffing sound you can hear is an air leak, which is why air leaks are bad. Bass ports can make chuffing noises too, which is why passive radiators are good. The speaker is being moved within the limits of the amplifier's supply rails. The + rail pushes it out, via the NPN output transistors, the - rail sucks it back in, via the PNP output transistors. But the output voltage can't rise any higher, or fall any lower, than the limit set by the amplifier's power supply rail voltages, and if we try to force it to, it is stopped in its tracks, unable to rise or fall, a flat line; this is called clipping, and this is what it sounds like:

Ive used this unrealistically low frequency so you can actually see the cone movements, and also so I wasn't deafened; the speaker in the clipping video is dissipating over 120 watts. Its voice coil was starting to smoke after only ten seconds of clipped, square-like waves. It could easily handle much more unclipped power. Why this is so will soon become apparent. The oscilloscope screenshot below shows an amplifier approaching clipping, but still relatively undistorted. The horizontal scale is time, the vertical, amplitude. That yellow line at the top is the scope's second channel displaying the amplifier's + power rail; the signal can't swing higher than that. Note that the rail is a flat line; that's because its DC. The AC signal will be swinging the speaker back and forth, as in the first video.  

This is relatively clean power, although the amp is approaching its limits. In practice, the signal can't swing all the way to the positive and negative rails, due to losses and inefficiencies in the circuit. Unless excessive (which they sometimes are), these losses are not an issue. Lets turn up the volume:

The signal can't swing any higher or lower. It flattens out as it hits the rails; hard clipping. Look at the top and bottom of the wave. Its a flat line; DC, and at a high amplitude. Unlike the previous screenshot, the amplitude at the top and bottom of the wave is not changing along the time scale. During the flat sections of the wave, the speaker is held firmly in place, moving neither forward nor backward, robbed of the cooling air its movement provided, while DC proceeds to melt its voice coil. Sometimes the molten coil will short out, also destroying the amplifier's output stage. An unclipped wave of the same amplitude, or even considerably higher for that matter, will not have the same destructive effect, which is why it is better to err on the side of high, rather than low, power. Clipping is not only speaker destroying, it sounds dreadful. So, soft clipping, which rounds off the top and bottom of the wave somewhat, rather than just shearing them off, must be a good idea then, right? Wrong. Here's why:


On the scope is a NAD 742 receiver, soft clipping engaged, beginning to clip. The wave is indeed rounded off somewhat. Now let's turn off Soft Clipping and take it to full power again: 

Again the amplifier begins to clip, same load, same channel, same day. But not the same power output! Look at the voltage measurement in the top right of the scope. With Soft Clipping engaged, clipping begins at 25.82 volts. Without it, clipping begins at 33.55 volts; that's about 83 watts RMS with Soft Clipping versus about 140 watts RMS without it! At 80W, Soft Clipping is attempting to mitigate the awful sound and potential damage of clipping, when, without it, clipping wouldn't even begin for another 60 watts (more if you factor in peak power). Soft Clipping is attempting to solve a problem that wouldn't exist without the solution; like invading Iraq! If you choose to engage Soft clipping, you will lose up to half of the amplifier's useful power, depending on the model. One could make the argument that, in purely Decibel terms, the last 60W is not that much. But why pay for a 140W amp when you only get 80? The question we need to ask is; does Soft Clipping provide enough, if any, sonic benefit, to justify such a loss of power? Let's listen to find out. Remember, Soft Clipping will reduce maximum power, hence volume, so the hard clipped signal will be louder:

If you were wondering, the quivering at the top and bottom of the wave sans Soft Clipping is caused by power supply ripple; more energy storage eliminates this. Too many capacitors are barely enough. Transformers are never too big.

Hard clipping sounds bad. Soft clipping sounds bad. Clipping, always, sounds bad. Maybe NAD's Soft Clipping circuit makes clipping less excruciating but it's still an angle grinder exfoliation for the ears. And it happens a lot more often with Soft Clipping engaged. There is also the deleterious effect of the Soft Clipping circuitry itself, which is not all bypassed by switching it off. I haven't covered issues like signal compression, or how failure of the circuit can kill your amp, or how the lower powered NAD's can't afford to lose half their already meagre power, because the case against Soft Clipping is already clear cut. At least you can switch it off. Doing so is the cheapest improvement you can make to any NAD amplifier. If you really love your 3020, and you should, I can excise the Soft Clipping circuit entirely for you, and perhaps upgrade those fragile output transistors while we're at it.

You may wonder, if you buy the odd Hi-Fi magazine, why this entirely empirical dissection of Soft Clipping is the first time you've heard that it may be a problem? We should all keep in mind that magazines (on and offline) do not exist to deliver good Hi-Fi to us. They exist to deliver us to advertisers.   

William Crampton