SMPS for Hi-Fi?

[mr_push_pull]

I know it has been done, even with audiofile grade amplifiers, but I want to know what are the special characteristics of the SMPS's for audio amps. I'm pretty much a novice with SMPS's, I only know the basics, but I have this Motorola handbook on voltage regulators (including switch mode) at home, and I thought about using a switch mode power supply for my future Quad 707, I think it would be a right place to use it, since the amp has an output rating of 240 Watts. Can anyone help with some info?

[Mr Evil]

The first thing that springs to mind is noise. SMPSs will have switching noise on the rails, and will radiate it too. For audio, this needs to be as low in amplitude as possible within the audio band. This means very good filtering and shielding, and as high a switching frequency as possible. Things like efficiency and regulation are of lesser importance for most audio applications.

[mr_push_pull]

Originally posted by Mr Evil
Things like efficiency and regulation are of lesser importance for most audio applications.

Yes, but small is cool I saw once this class-D amp with a SMPS, it had an output rating of 700W if I recall correctly and it had only a few centimeters of height.

The noise issue can be overcome I think, currently the switching frequencies are high enough. But still intermodulation can become a problem.

One problem with SMPS's that someone mentioned once is that the common ones are not fast enough to cope with the dynamics (and corresponding drawn current varriations) found in music. Is that correct?

[Eva]

'Noise' is a classic misconception about SMPS

Speaking properly, there is no noise in SMPS, since noise by definition is a signal of random nature and what SMPS produce on their output is periodic 'ripple' as any 50/60Hz supply. We could call this 'conducted ripple'

This ripple is usually a sawtooth waveform and it may have some RF ringing on transients [on the peak and valley of the sawtooth cycle]. There's not much difference from 50/60Hz supplies, except the ripple frequency is about 1000 times higher

As in 50/60Hz supplies, ripple is easily filtered with pi filters and it may be attenuated as desired [even below noise floor], but with the advantage of size since 100Khz pi filters are much smaller and cheaper than 100Hz pi filters

There is also 'radiated ripple' like in any 50/60Hz supply. This radiation happens both due to quick changes in electric fields and quick changes in magnetic fields in the SMPS circuit [as in any non-linear circuit with diodes or switches]

Electric fields are shielded by just mounting the SMPS inside a grounded metal case and it's production may be reduced by means of optimized PCB layout [reducing the lenght and shielding high dV/dt tracks]

Magnetic fields get only about 20dB attenuation with metal cases, but it's production may be also reduced by means of proper layout [essentially reducing the area of high dI/dt PCB loops]

There are also field-cancellation layout techniques based on dipoles or multi-poles

Ringing at switching transients is produced by parasitistic resonance of capacitive/inductive elements in the circuit [unexpected RLC resonators excited during switching transients], like parasitistic capacitances of MOSFETs, parasitistic inductances of PCB traces and parasitistic self-inductances/capacitances of windings. [Increasing or decreasing parasitistic R may damp the oscillation but it's not allways practical]

Using bipolar transistors or IGBTs causes a dramatic reduction of ringing since they show capacitances an order of magnitude smaller than MOSFETs and have inherently limited switching rise and fall speeds as opposed to MOSFETs, reducing dI/dt and dV/dt

Schottky diodes and SiC diodes also reduce ringing dramatically since Schottkys show no reverse-recovery period and SiC diodes show very small reverse recovery. Reverse recovery is allways a pain in non-transformer-coupled buck and boost converters and the source of most radiated ripple in these converters

Radiated ripple atenuates 12dB every time the distance is doubled, so placing the SMPS away from small signal equipment is also useful. Anyway, ambient noise usually dominates over SMPS far-field radiated ripple

A final note : Paralelling different kinds of capacitors, like electrolytics with ceramics, ceramics with films, electrolytics with films, etc... Usually produces high Q resonant RLC networks and lots of ringing. Each case must be studied carefully before paralelling different type of capacitors in an SMPS, since it usually increases ringing instead of reducing it!!!

This is also true for audio circuits

[mr_push_pull]

This is what I call an answer! Thanks a lot!