Class D Amp. Design

It may be possible to remove the LPF - the speaker coil has a certain amount of inductance.

The speaker coil also has a certain amount of parallel capacitance. For single layer windings this can be quite low - of the order of a pF - but for multilayer windings it can be much higher - of the order of hundreds of picofarads. The can make the speaker leads and the speaker itself into a sort of antenna, radiating the high frequency components of the switched output, messing up radio reception over an appreciable area.

If the speakers uses a metal magnet - alnico, neodynium doped iron, or samariam cobalt - the speaker coil can induce high frequency current around the magnet, warming it up and reducing its magnetic strength.

Depends.

----------- Bill Sloman, Nijmegen

The 'raw unfiltered' output of any Class-D amplifier is indeed PWM i.e rectangular 'square waves'.

What did you expect ?

See TI's excellent application note about this. It's sometimes possible when amplifier and speaker are located close nearby.

The best result without a filter is obtained with a 'phase-shift' version of PWM. Requires a differential output.

Why is an LPF filter normally needed you mean ?

Several reasons including reduced dissipation in the load ( esp in the no-signal condition ) and minimisation of EMI.

That's a tricky one to answer. Some ppl have more acute hearing than others.

For 'low-fi' I'm sure the filter is not needed. For 'hi-fi' I'm sure it is needed.

Graham

I tend to agree.

A speaker is a non-linear device, much like the amplifier itself. Feeding it with inaudible HF signals will still cause additional LF intermodulation distortion. How much? Ahmmm...

Kevin Aylward snipped-for-privacy@anasoft.co.uk

Ken Smith: Unless you spend extra on the speaker, the not much of the high frequency power in the PWM signal will convert into motion in the cone, but I don't think you want to inflict it on any dog you like.

How about for that neighbor's dog I DON'T like? GG

There are good reasons to use an LC low pass filter.

The PWM signal contains lots of RF. That RF will radiate from the speaker. It is very hard to make a EMI shielded speaker that you can hear.

The voice coil of a speaker is a lossy inductor at high frequenies. You don't want to send any power to the speaker that won't end up as sound you can hear. The power that you do is just losses to the system and results in a shorter battery life. The reason to use Class-D is mostly to get good battery life.

Unless you spend extra on the speaker, the not much of the high frequency power in the PWM signal will convert into motion in the cone, but I don't think you want to inflict it on any dog you like.

The purpose of the LPF is to remove harmonic element of the switching waveform and also average the switching waveform (at higher freq outside audio bandwidth) to represent lower freq waveform (within audio bandwidth)

However since inductor is not lossless, some power will be lost depending on magnitude of the current. To minimise this, apply low losses core (ferrite) and maximum cross section area copper, fit as much as you can within winding space. Also keep inductance optimal value to minimise losses.

PWM is perferred becuase the frequency is generally fixed outside audio range, but sometime they become pulse skipping when power drive level become low, this goes inside the audio frequnecy. That may be the reason why LPF could be useful.

I not sure how the speaker inductance become LPF since the capacitance in order of pf seem relatively low to be effective LPF.

It not expensive hardware to add in LPF, unless you dealing with 10Amp output drive currents.

In article , Glenn Gundlach wrote: [...]

In that case you may want to, but you know it would be wrong to do so.

Actually it shouldn't be too problematic with 200-500 kHz switching f.

The motor assembly in most speakers is a pretty effective shield.

TI have an interesting app note about this. It looks at those power calculations and EMI - worth reading.

Graham

Actually, iron powder cores are preferred for this application. The distributed gap makes them more practical to use.

Toroids from the likes of Micrometals are v popular.

Graham

That's not quite right. The filter removes not just the harmonics of the switching waveform. It removes most of the high frequency components including the switching frequency and the sidebands near it.

Any good Class-D amplifier will have its switching frequency above the audio range. Some really cheap stuff doesn't.

There is no reason to have pulse skipping when there is any sound being reproduced. The basic circuit is a switch that switches between two voltages. The duty cycle of the switching controls the average output voltage.

In article , Pooh Bear wrote: [...]

The folks that did the last batch of EMI testing for us, said that they have never seen a Class-D amplifier pass the FCC specs without a filter.

It only works for a co-located amp speaker combination.

E.g. a radio set with internal speaker.

Read the app note.

Graham

True, but when using phase shift ( differential ) PWM it's rather less of a problem.

Graham

that makes sense. regardless of how good the speaker is as a shield, the dangly wires betwix amp & spkr will spew EMI everywhere.

Cheers Terry

No net current with zero signal.

Graham

Yes...what you described is actually harmonics(!), the high frequency component of the switching waveform prioir to the LPF is defined as harmonics.

The pulse skipping occurs if there is too much power transfer by single pulse when light loaded. The PWM cannot reduce pulse width too narrow due to circuitry so they went to pulse skipping. Try this on DC/DC with PWM and vary the load. The problem with the pulse skipping that the harmonics and fundamental shifted to lower frequnecy into audio frequnecy.

BTW, with fixed load by the speaker, pulse skipping may not likely to occurs, but when you reduce the volume to very low level, PWM generates narrow pulse width and hence more likely to pulse skips.

Generally higher freq *may* give compant circuit and perhaps reduce losses, but as the current start to increase to say 20A, the switching frequnecy may need to be reduced due to MOSFET capacitance (major losses element). It very depends on input voltage, output voltage and circuit topology.

Thank for information on core... I worked on ferrite core for specilist application involving high temperature (150 and 180degC), which is more suitable.

Terry, perhaps installing amp inside the speaker would minimise the EMI or use LPF. The LPF must be situated next to the switching elements as this is more effective.

I don't know for sure but I assume that the lab in question had even tested some products where the speaker was in the same box.

In our case, the load was external so it didn't count as the TI suggested case.

why would that be so?

Cheers Terry