Help with Class D Amplfier Design (feedback)

[Pierre]

In fact type-2 material was also used by me with no problems.
If you can, give also a try to Wilco drum core inductors, they offer free samples. The only issue is that they don't have a closed magnetic field so may produce interference with the rest of the circuit or other equipment, but definitely worth a try.

 

[JohnW]

I never had any results with the Amidon red cores, I never tried it. I just said JohnW recommended it as being highly linear and can handle the higher frequencies nicely. He gave a measurement of something like .008 THD in open loop.

In fact with the larger Amidon Type 2 red cores I achieve about 0.0015% @ 120W 8 Ohms, Open Loop (384 KHz). The primary distortion components being 3rd harmonic & odd order components.

The primary cause of the odd order distortions is the rounding effect of the B-H characteristics. Ideal magnetic materials have a Square-Loop characteristic with very high permeability and insignificant stored energy until finally driven into saturation. This is called a “Sharp saturation” characteristic.

A rounded or “Soft saturation” characteristic exhibits a gradual reduction of incremental permeability until finally the core is completely saturated. Magnetic “Hard spots” and inside corners will also cause rounding of the B-H characteristics. In some inductor filter applications, this rounding effect is akin to a “Swinging Choke”, and is actually desirable – but not for Class D output filters!

Rounding effects in metal-alloy cores are normally quite low. However, in composite metal powered cores, non-magnetic “Gaps” exist between the discrete magnetic particles. It's these distributed non-magnetic regions that cause the significant rounding of the B-H characteristics – these regions also are responsible for storing energy within the core. Similar non-magnetic areas occur among the sintered particles in ferrite cores.

1. The distributed reluctance of these tiny “Gaps” causes the flux and the flue change to be discrete across the entire core, rather then as a discrete flux change boundary moving from inside to outside for a ungapped idealized metal alloy core.

2. At lower flux densities, the flux tends to concentrate via the easiest paths (i.e. paths with the lowest reluctance), where the magnetic particles are in close proximity too each other. As the flux density increases, these “easier routes” are the first to saturate. These portions of magnetic particles saturate first becoming non-magnetic, resulting in their flux paths becoming “less easy”. Incremental flux increase shifts to adjacent paths where the magnetic material has yet to saturate, but where the gap is somewhat wider. This process continues, effectively widening the incremental distributed gap as the flux density increases. Thus, the incremental permeability (and therefore inductance) is progressively reduced, and is observed by the rounding of the B-H characteristics.

The solutions to reducing the effect are: -

1. Manufacture the cores with a material that exhibits Square-loop characteristics

2. Reduce the irregular non-linear gap area by manufacturing the cores with highly polish magnetic material thereby reducing their uneven surface area and - higher density material (i.e. reduced particle size).

3. Adding a discrete air gap, the rounding effect of the B-H characteristics disappears –dominated by the linear high reluctance of the Air gap, thus linearizing the inductance characteristics. If an Air gap is used, then the copper windings must be kept well away from this gap to reduce the “Fringe” effects.

Cheers,

John

 

[subwo1]

Thanks, JohnW, excellent!

 

[classd4sure]

Yup that's an excellent post.

On one hand it sounds like a distributed gap just isn't the thing to use but the Amidon cores performance as measured by you ..... seems good enough for any amp I'd be making.

I trust their core is composed of a finer material than others use in order to get performance like that?

What kind of power level would such a thing be good for though?

Thanks
Chris

 

[Randy Knutson]

JohnW,
WOW!!! Thanks for taking your time to write such an amazing and detailed post!!! I'm going to order some type 2 torroidal cores from amidon tonight.

I am shooting for 500watts RMS into 4 ohms... I am going to order a few of each.. Does these sound good? all type 2 material

T106-( ) ID-1.06" OD-0.56" Ht-0.44"
T130-( ) ID-1.30" OD-0.75" Ht-0.44"
T157-( ) ID-1.57" OD-0.95" Ht-0.57"

Thanks!
-=Randy

 

[phase_accurate]

If he states up to what induction they can be used with low distortion then it is not a big problem to calculate the max current/turns for each core size.

regards

Charles

 

[JohnW]

For the record, I would like to thank Jaka Racman for his extremely useful leads while I researched the cause of the odd order distortion introduced by magnetic cores.

I’ve not built a high power Class D amplifier above 150W 8 Ohms or so, so I cannot comment on the performance of the Amidon cores above this power level.

I use the T106 core, and can just about wind 18uH on a single layer using 1.5mm Dia. Wire (for 18uH I counted about 36 turns 1.5mm, on the T106 type 2 core).

It’s not just a simple matter of material type, particle size & shape is critical, finer particles (denser material) result in lower distortion – which is a real concern for QC control. To achieve finer / smoother particles the magnetic powered is “rumbled” in a large drum for many days. The longer the rumbling process, the finer / smoother the particles.

I understand that Micro-Metals also have a similar core to the Amidon type 2 – however I have no experience of there variant.

Bigger cores and thicker wire can only be better - so I don't see higher powers being a problem if you scale up the core size.

A interesting effect I’ve observed with the Type 2 core material is that the distortion level remains the same for any power from about 1W to 120W, below 1W there’s a sharp step drop in distortion to the level of the HBridge output.

John

 

[sx881663]

A interesting effect I’ve observed with the Type 2 core material is that the distortion level remains the same for any power from about 1W to 120W, below 1W there’s a sharp step drop in distortion to the level of the HBridge output.

John

John,
Thank you for sharing all this with us. This is what makes this forum special.
I am curious about the low level distortion disappearing. Have you observed a corresponding decrease in low level inductance? This would make some sense if the energy dropped below a critical threshold level and could no longer bridge the gaps.
If this is the case and the inductance measures a lot less what will be the effect on very low level filtering? Would this mean a lot of low level HF stuff would get through?
If all this is the case I don’t know if it is relevant as the cores are biased with so much switching energy anyway. Could it show up as a tiny bit of switching residual around zero crossing?
Roger

 

[JohnW]

I have not measured a step change in inductance at lower power levels with the Amidon cores, so I’m at a loss to explain the distortion “Step Effect”, and can only speculate that it’s some “weird” hysteresis effect – but I would be grateful for a better explanation!

Air-cores not surprisingly, do not exhibit this effect.

Attached is some info I grabbed from an ST app note AN1013, which mentions a “Step” in the THD vs. Pout plots. However the app note suggests that this mechanism is akin to Class AB bias in the output stage, which is a valid explanation only when THD of the HBridge output is “High” - however my inductor Test Jig output stage has a typical THD at the HBridge node Pre inductor of 0.000025% 100W (highest harmonic below -120dB).

I’ve found that even with high order correction Post inductor, the absolute quality of the inductor is still very audible. The Amidon type 2 have the effect of “sharpening” detail very slightly, however with most normal Class D circuits this effect will be buried under many other problems.

In listening test, the Amidon cores performed the best out of Magnetic cores I’ve tested.

Sorry to have digressed this thread,

John

 

[sx881663]

Low level effects

John,
It never ceases to amaze me how good ears are at recovering low level information. This lack of distortion/filtering of the low level information must be of incredibly tiny values. Leads me to think there is research to be done on inductors. What about a core made from an extremely thin ribbon of hi nickel allow magnetic material like used in some of the better audio transformers? I know some of these type cores can go into MHZ with no problem. They would have to have good hysteresis characteristics as well to do good audio. What I don’t know is how they handle large field strengths. Any thoughts?
Roger

 

[skaara]

I would really like to have final schematic for this to have some fun here too

 

[Enlight]

I would really like to have final schematic for this to have some fun here too

I agree with you there, maybe you could post some pictures two? i'm very interested in the cooling you need for this design.