Increasing efficiency through output inductor selection

[zerokelvin99]

Hi all,

this should hopefully be a nice basic question for many on here, fingers crossed!

It seems to me that many of the idle losses within ClassD designs are in the output inductors (as confirmed in the TAS5630 datasheet). I intend to modify the Class D TI-600 ( Class D Audio TI-600 Amplifier - Class D Audio Amplifiers - PRODUCTS ) in the output inductor stage to achieve a lower idle current.

What should I be looking for in an inductor to replace their stock inductors? Is it even possible? According to several online sources, 90% of the hear loss is through the wires rather than the core. However, I have the feeling it isnt quite as simple as just changing to a lower resistance inductor (while keeping the audio quality top notch)


If anyone could give me any pointers then that would be greatly appreciated.

I have been using this source http://www.inductors.ru/pdf/doc486_inductorlosses.pdf so far but dont seem to be able to get any further

Many thanks in advance!

[zerokelvin99]

ps. the TAS5630 datasheet http://focus.ti.com/lit/ds/sles220b/sles220b.pdf

specifies output filter inductor values to be 7 min, 10 nominal in BTL

[abraxalito]

The Pacr term in that link is the one to watch - its a function of skin effect. Skin effect means the current flows only at the surface of the conductor as frequency rises - this leads to greater losses at higher frequency. If the coil is only a single layer one, this will be dominant. It can be mitigated by using Litz wire to increase the surface area. Going simply for a lower DC resistance inductor will be very misleading in respect of losses if its only wound with thicker wire, rather than multiple thin wires.

[Eva]

Main sources of losses to consider:

Core losses in output inductor due to magnetic flux excursion.
Straight I^2*R losses in magnet wire.
Skin effect losses in magnet wire just due to the moderately high di/dt of ripple current.
Eddy current losses in magnet wire due to magnetic flux leaking the gap (only in gapped ferrites).
Possoble losses in switching stage due to overlap.
Losses due to hard switching and d-s capacitance of the own output MOSFET.
Losses in snubbers.

Proper dead time, gate resistors, output inductance and snubber value selection can achieve resonant operation and very low idle losses on the output stage, leaving only inductor losses to worry about.

TAS5630 output stage is probably optimized for the recommended inductor and snubber values. Integrated class D chips give little or no chances for tweaking idle losses.

The inductors in the picture look like gapped ferrite toroids. Do they get warm when the amplifier is idle? Winding your own toroids with -2 material (Micrometals) may be an alternative.

[Pafi]

Eddy current losses in magnet wire due to magnetic flux leaking the gap (only in gapped ferrites).

Not only in gapped ferrites! Even in a perfect toroidal choke the flux density is not zero between the wires and the core. On the surface of the core it drops to B/mu_relative and it decreases gradually till the outer surface of wires. While the low core loss type of powder cores have small permeability, this effect is not negligible in ClassD audio.

The most efficient "trick" to achieve low idle loss is BD modulation and combined common+differential mode choke. Ideal core material is ferrite (gapped), ideal wire is litz.

But be careful, because common+diff mode filter choke has magnetic leakage!

[Pafi]

90% of the hear loss is through the wires rather than the core.

The copper loss and core loss should be roughly equal to keep sum the lowest.

In the particular case of this amp, with only 7 turns on a core I guess the core loss wont be insignificant. Somewhat lower AL (bigger/more air gap) could be better, if material can be the same.

[AP2]

Hi,
litz wire is designed only to facilitate the high-current windings. poor performance compared to solid wire in the toroid.
capacity ser-par on the winding should be zero. but this is not possible, then better segmented with a stiff wire wrapped space.
Another problem is the section of the wire. I often see the sections about 1mm. this is wrong and can not be calculated for the current only.

What hinders the development of new ideas,... it's just the theory.

[megajocke]

Originally Posted by Pafi

Not only in gapped ferrites! Even in a perfect toroidal choke the flux density is not zero between the wires and the core. On the surface of the core it drops to B/mu_relative and it decreases gradually till the outer surface of wires.

And what happens for thick wires is basically that the HF AC current flows in a single skin-depth deep layer, but only on the part of the wire facing the core. AC resistance is thus about twice as high as would be predicted if only considering skin effect.

If you have more than one layer of winding losses go south quickly...

[nigelwright7557]

Originally Posted by Eva

Main sources of losses to consider:
The inductors in the picture look like gapped ferrite toroids. Do they get warm when the amplifier is idle? Winding your own toroids with -2 material (Micrometals) may be an alternative.

I started off using power inductors but they got red hot 120 degrees C !

Someone recommended a t106-2 core and I would my own and the coil now doesnt even get warm.

[Pafi]

Originally Posted by megajocke

And what happens for thick wires is basically that the HF AC current flows in a single skin-depth deep layer, but only on the part of the wire facing the core. AC resistance is thus about twice as high as would be predicted if only considering skin effect.

Be careful! Skin depth is only 0.1 mm at 400 kHz!