No load protection for Class D amp

[Gaz]

Hey All,

Got my hands on an Icepower Class D Module. There are some built protection features, but not against no load conditions. Are there any simple precautions I can take like one would on a tube amp? Flyback diodes, several hundred Ohm dummy load, etc? Thanks for any advice. I am new to Class D.

[Reader]

Your class DM is solid state so it will not be damaged under no load conditions. Actually, for solid state, you need to make sure that nothing shorts the output. Solid state amps are quite happy when there is no load on the output.

[R.G.]

Reader is right. Tube amps are unusual in that an open load is a problem. I'd have to read the tech data sheets on the Icepower amps, but I suspect that they are short circuit protected, and don't need open circuit protection.

Tube amps have problems with open circuits because they can, under certain circumstances, oscillate themselves into creating flashover voltages for tubes and the innards of their transformers when their loads are too high. Tube amps without feedback would seem to be less susceptible to this. I have personally ... um... forgotten to connect loads to an amp I was developing. No damage, although I slapped myself a bit for doing that. I have some personal theories about why the no-load-death would happen, but that's another story. Tube amps with feedback, like most modern ones, should never be left open-loaded.

Solid state amps die a different way. Unlike tube amps, they generally don't go into death-oscillation with open loads. What kills solid state amps is trying to let through enough current to make +/40V or so appear across 50 milliohms of wire. The output devices actually support full supply voltage at huge currents, develop hot spots, vaporize bonding wires and convert carefully doped silicon into puddles of molten goo. Then it gets worse...

Different beasts. Class D amps are yet different beasts, but they are much more like linear AB amps than tube amps. Don't sweat running the Icepower without a load unless the datasheet says not to.

[J M Fahey]

Maybe.

Personally I´d love to study this a little, I just can´t consider Class D amplifiers same as "plain" SS amps (think class AB).

To begin with, they usually *do* have significant "iron" at the output, that alone makes them different.

"Tube" vs "Solid State" is not the main point, but mode of operation.

OTL tube amps have no problem with open loads.

Transformer output SS amps, think old style 12V car battery powered ones (we call them "watermelon seller amps" ) or even autoformer out Guitar amps (some Peavy Steel amps, a certain 100W Marshall, some high powered Mc Intosh) are in BIG trouble if driven without a load.

[Reader]

- - - autoformer out Guitar amps (some Peavy Steel amps, a certain 100W Marshall, some high powered Mc Intosh) are in BIG trouble if driven without a load.

I never considered that situation. It makes sense that it's the magnetics good causes the issue not the solid state vs. Tube driving device. I look forward to hearing more details if you're willing.

[J M Fahey]

Oh, no mystery.
Big inductors create high power inductive spikes which *normally* get absorbed, "snubbered" if you wish, by load.
And since speaker loads are typically inductive, and don´t absorb very well higher frequency sharp edges, snubbers are used even in conventional Class AB amplifiers, go figure: the Zobel network.
AND often clamping protective diodes ; generally in high power amps but I have seen them suggested (by official datasheet) even in humble TDA2030.

They often are there even if they seem not to: typical Darlington (TIP142/147 , BDVxx, etc.) and all switching type power MosFets have built in backwards/clamping *power* diodes in parallel.

Now if even conventional Class AB amps have use for snubbers and clamping diodes, I can easily imagine Class D amps may be even more critical with weird loads or lack of it.

But this is just an educated guess, in any case I bet somewhere there is some .PDF full of Math analysing that.

If not in the amplifier realm, in the power supply none.

**For me** a Class D "amplifier" is nothing more than a specialized SMPS, which supplies a "slowly" varying (for their standards) modulated DC output, which "we" , inhabitants the "slow World" call "Audio".

A lightning fast gut wrenching race

[R.G.]

I did what I should have done before my first reply - I went off and read some Icepower data sheets. While they're explicit about their protection circuits preventing any damage from hazardous loads, including output shorts, they don't say anything about running it with no load. The data sheets are workmanlike enough to have mentioned danger from open-load operation. I also pinged my amp tech friend, who I remembered had done some work with Icepower modules. He said he often ran them without loads, no problem.

J.M. is correct - Class D amps are special cases of switching power supplies, with a modulated reference voltage and special crossover provisions. Switching power supplies commonly have limits on the switching speeds of their power switches, and cannot go to infinitely short pulse widths. The shortest pulses are needed for letting the through the least power to the load. So there is usually some maximum load resistance specified, a minimum current load.

However, with modern devices, the shortest pulse widths tend to be so short that the maximum load resistance/minimum load current can be served by a low power resistor, and that is included in the power supply so that the external load resistor can go to open circuit with no damage. Note that this concept is full of asterisks, footnotes and special considerations, but in general, modern switching power supplies have zero minimum load, which means they work into open circuits fine.

And again J.M. is right, it's the inductors that make voltage spikes. Installing catch diodes on the outputs of even linear Class AB amps is a good idea - I do it on all of my amps if they don't already have it, because speakers have regions where they are primarily inductive, as J.M. notes.

The real underlying issue is probably amplifier stability with reactive loads. Having a high power output stage of any technology go into massive oscillation is a Bad Thing. Feedback theory says that you can't get oscillation with less than three time constants in a classical negative feedback setup. There is always at least one time constant involved with each active device. There is another time constant usually present in the feedback loop for amplifiers that are not DC coupled. Modern feedback amp practice is to use fast enough devices that the device time constants/poles are at such high frequencies that a single dominant pole can be set low enough that the amp has unity gain at a frequency below the additional poles, and hence is stable.

A highly reactive load adds another pole/time constant into the mix by adding phase shift to the signal right at the feedback take off point. Amps with linear follower power stages, like most class ab amps, are sensitive to capacitive loading more than inductive. That's why you see specs on linear amps about them being stable into X amount of capacitance, and why damped inductors are in series with the outputs. Amps driving inductive loads - like speakers in the higher part of the audio range, or crossover networks, will have issues with spikes and oscillation unless they are loaded/damped. That's what that Zobel network does. It lets the cap "fade in" the low impedance at high frequencies. These things help keep the output stage stable.

But a transformer is two time constants all by itself. So an amp with feedback around a transformer will have Nyquist stability issues unless the transformer's pass band is way bigger than the signal range, and we already have issues making transformers wide enough for audio. The other fix for keeping transformer amps stable is to lower their open loop gain. This makes compensation easier by lowering the amount of gain to be scrubbed off before the amp goes unstable. So yes, any amplifier, SS or tube, with a transformer in its feedback-covered signal path will have special considerations for stability and loading, as well as a sensitivity to open/short loading on its output if the transformer is at the output.

SMPS will typically have inductor/capacitor filters on their outputs, and Class D amps usually do too. The filters are very closely designed to match the expected load and overall stability. For the icepower modules, I suspect that the output stage was very closely designed for just such issues.

[Helmholtz]

I think that tube amps without global NFB are at risk without a load as well. But as they don't tend to self-oscillate, a signal is required.
With speaker load removed the tubes work into a loosely damped parallel resonant circuit made up of primary (magnetizing) inductance and the distributed capacitance of the primary. Using typical values, the resonant frequency can be expected to lie between 1Khz and 3kHz, i.e. well within the frequency range of a guitar.
The high Q of the resonant circuit strongly increases gain, voltage amplitude and currents around the resonance.

Players tend to turn everything up when they get no sound (speaker being accidentally disconnected) and this may destroy the OT and power tubes.

[Reader]

Thank you all for the advanced explanations.
A bit of very relevant information for me is that I have one of the McIntosh SS power amps that uses Autoformers at the output. It is currently in storage and I am trying to make the time to get it out and put it back in service. I know it needs some work and now I know to take the precaution of using a load during any bench testing.
Thanks,
Bill

[Mike Sulzer]

I think the danger to the push pull AB tube power stage with ac coupling and grid bias through large grid resistors is subtle but deadly. When you drive such stage really really hard, you charge the grid coupling capacitors. This increases the negative bias under these conditions and moves the operation from AB through B in the direction of C. This means that one tube turns off before the other turns on. Under normal operating conditions this probably extends tube life. But if you leave off the load, you can have both tubes biased off with significant energy still stored in the magnetic flux, and no load to absorb it. Thus the voltage rises and BANG!

How do I know this? I have been testing an attenuator which is a speaker simulator. The small inductor that simulates the voice coil inductor causes small short voltage peaks in the crossover region under hard overdrive; not dangerous. This is with a load, of course. Imagine what happens with a large inductor and no load.

[emolatur]

J.M. is correct - Class D amps are special cases of switching power supplies, with a modulated reference voltage and special crossover provisions. Switching power supplies commonly have limits on the switching speeds of their power switches, and cannot go to infinitely short pulse widths. The shortest pulses are needed for letting the through the least power to the load. So there is usually some maximum load resistance specified, a minimum current load.

While this is true, removal of that "minimum current load" does not typically lead to destruction of the power supply (/amplifier) -- it simply throws the voltage regulation out of whack. If you have no load, then your load doesn't care that the voltage is wrong.

[Helmholtz]

I think the danger to the push pull AB tube power stage with ac coupling and grid bias through large grid resistors is subtle but deadly. When you drive such stage really really hard, you charge the grid coupling capacitors. This increases the negative bias under these conditions and moves the operation from AB through B in the direction of C. This means that one tube turns off before the other turns on. Under normal operating conditions this probably extends tube life. But if you leave off the load, you can have both tubes biased off with significant energy still stored in the magnetic flux, and no load to absorb it. Thus the voltage rises and BANG!

How do I know this? I have been testing an attenuator which is a speaker simulator. The small inductor that simulates the voice coil inductor causes small short voltage peaks in the crossover region under hard overdrive; not dangerous. This is with a load, of course. Imagine what happens with a large inductor and no load.

Agree, that something like blocking distortion or other current disruption is necessary. Flyback spikes are always produced by a sudden reduction of inductor current. As long as the the current is continous flyback/kick-back is no concern. With load, only leakage inductance+ speaker inductance is involveld. Without load, the whole primary inductance fires back.

[Mike Sulzer]

Agree, that something like blocking distortion or other current disruption is necessary. Flyback spikes are always produced by a sudden reduction of inductor current. As long as the the current is continous flyback/kick-back is no concern. With load, only leakage inductance+ speaker inductance is involveld. Without load, the whole primary inductance fires back.

The leakage inductance+ speaker inductance is in series with resistive part of the load, limiting the rate at which current can change. There is no such protection for the transformer primary when the load is removed.

[R.G.]

While this is true, removal of that "minimum current load" does not typically lead to destruction of the power supply (/amplifier) -- it simply throws the voltage regulation out of whack. If you have no load, then your load doesn't care that the voltage is wrong.

True. I was synopsizing heavily.

I wanted to get in the idea of a minimum load required for proper operation, but not go down the perceived rabbit hole of explaining operating modes and probable failures, and stretching the analogy even more. The point was that while some switching power supplies may have a minimum load, even then it's not tragic, and can be easily handled - even by means other than an internal minimum load resistor, which I didn't go into either.

I was pre-ducking a comment that some switching power supplies have a minimum load, so how does that work with an audio modulation. I neglected to pre-duck other second order comments.

[tedmich]

I too just laid in some IcePower modules



two of their 300AS1 with the custom wires from Ghent Audio, I'll be using them for 2 subwoofers fed with a balanced summed signal <100Hz.

They do provide some protection



and should give a semi clean (THD+N ~.02%) up to 250W into 4 Ohms

its sooo nice not to have the sweat a power supply!