I want to have a discussion regarding the design of output filter inductors for class D power amplifiers. I already know the basics - how to design inductors with regard to cross sectional area, Bmax and Imax, but I want to dig deeper - what materials are the best to use, and why?
I want to leave the powder iron core (amidon toroidal -2 material) I am using now and instead use the more space efficient ferrite cores available (RM-cores). I have understood that the B-H curve can be vastly linearized by introducing a large air gap, but I do now know what ferrite parameters that are important to gain an output inductor that introduces as little distortion as possible (is as low permability as possible after gapping desired? Does 1/mu correspond to linearity?). Is the area of the hysteresis loop important with regard to distortion, or does it only indicate power loss? Is the permeability versus frequency relationship important? I guess it is, and therefore I suspect 3F3 material to be good, correct?
Finally, can I hope to outperform an amidon T106-2 inductor using say an RM-10 core with ferrite material?
Edit: what is a reasonable limit on Bmax for hifi-use? Is there any formula to calculate the distortion introduced from a certain inductor?
I want to leave the powder iron core (amidon toroidal -2 material) I am using now and instead use the more space efficient ferrite cores available (RM-cores). I have understood that the B-H curve can be vastly linearized by introducing a large air gap, but I do now know what ferrite parameters that are important to gain an output inductor that introduces as little distortion as possible (is as low permability as possible after gapping desired? Does 1/mu correspond to linearity?). Is the area of the hysteresis loop important with regard to distortion, or does it only indicate power loss? Is the permeability versus frequency relationship important? I guess it is, and therefore I suspect 3F3 material to be good, correct?
Finally, can I hope to outperform an amidon T106-2 inductor using say an RM-10 core with ferrite material?
Edit: what is a reasonable limit on Bmax for hifi-use? Is there any formula to calculate the distortion introduced from a certain inductor?
zilog said:
This is a great topic that I know some members here are capable of contributing to. There is also some good info out there that just takes some time to uncover.
Here's a good read to start out with: http://focus.ti.com/lit/ml/slup124/slup124.pdf#search="inductor core saturation"
As far as inductors with as little distortion as possible, you need to find a good compromise between the inductor parameters that will be largely dictated by the switching frequency and power output of the amplifier the inductor will be used in.
Never allow a core to saturate and you will probably ok in terms of linearity. As you know, an air core inductor will not saturate, but the field is not nearly as well contained as it is with other core materials.
This can be a very interesting thread! I can think of at least EVA who is an expert on this, and hopefully will contribute.
The choke is always a crucial contributor of performance in a ClassD amplifier, and there are so many parameters in the design of the choke, that it can almost be considered a science of it's own right. Therefore, i think many commercial producers will protect their choke design as part of their IP.
The choke is always a crucial contributor of performance in a ClassD amplifier, and there are so many parameters in the design of the choke, that it can almost be considered a science of it's own right. Therefore, i think many commercial producers will protect their choke design as part of their IP.
I remember another thread talking about choke design for Tripath based amps.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=81719&perpage=10&pagenumber=18
Started here
http://www.diyaudio.com/forums/showthread.php?s=&threadid=81719&perpage=10&pagenumber=4
http://www.diyaudio.com/forums/showthread.php?s=&threadid=81719&perpage=10&pagenumber=18
Started here
http://www.diyaudio.com/forums/showthread.php?s=&threadid=81719&perpage=10&pagenumber=4
I'll repeat - this could be a very interesting thread! 
I have worked with about 10 different inductors on the little T-Amps. Small bobbin types, shielded bobbins, aircore both home made and factory, and toroids of various types.
They do sound different, sometimes a lot, sometimes a little. Mostly I find that if the core is big enough and doesn't saturate, the inductors sound much the same.
But there are other concerns.
In the RF domain they are certainly different, different core materials attenuate the RF more or less. I have some measurements comparing the RF output of two different inductors. Will try to dig them up and graph them - then post here. Choosing a core material that is designed to work at and above the amp's switching frequency is key to reducing the RF harmonics. High permeability allows few turns, lower DCR, but will saturate earlier. Aircore inductors radiate a lot of RF. The radiation needs to be addressed or it will find its way back into the audio circuits. Sounds bad.
I have been trying for months to get some good IM distortion measurements done on various inductor types. That would tell us a lot.
A good start to a good thread. I hope that this discussion can remain both technical and subjective. We are most interested in the subjective change to the music, but want to know the technical how and why the changes are happening.
Thanks to Zilog for starting this.
I have worked with about 10 different inductors on the little T-Amps. Small bobbin types, shielded bobbins, aircore both home made and factory, and toroids of various types.
They do sound different, sometimes a lot, sometimes a little. Mostly I find that if the core is big enough and doesn't saturate, the inductors sound much the same.
But there are other concerns.
In the RF domain they are certainly different, different core materials attenuate the RF more or less. I have some measurements comparing the RF output of two different inductors. Will try to dig them up and graph them - then post here. Choosing a core material that is designed to work at and above the amp's switching frequency is key to reducing the RF harmonics. High permeability allows few turns, lower DCR, but will saturate earlier. Aircore inductors radiate a lot of RF. The radiation needs to be addressed or it will find its way back into the audio circuits. Sounds bad.
I have been trying for months to get some good IM distortion measurements done on various inductor types. That would tell us a lot.
A good start to a good thread. I hope that this discussion can remain both technical and subjective. We are most interested in the subjective change to the music, but want to know the technical how and why the changes are happening.
Thanks to Zilog for starting this.
N87 by EPCOS seems to be quite similar to 3F3:
http://www.epcos.com/web/generator/...ERRIT,property=Data__en.pdf;/PDF_SIFERRIT.pdf
It can also be had as gapped halves of EFD cores:
http://www.epcos.com/inf/80/db/fer_01/05140515.pdf
If you use a core that is already gapped from the beginning you will end up with less stray-field outside the coil compared to the insertion of insulation material.
You will still have to take care of the strayfield inside however. Take some measures to avoid this area with your windings.
Regards
Chasrles
In the Epcos range the ferrites that are of interest (in terms of increasing performance are:
N27
N67 (similar to Ferroxcube 3F3)
N87
You can get EFD and ETD cores in all these materials. These cores have the gap in the centre leg, so you don't get much of a stray external field.
N27 and N67 are easier to obtain than N87, which is a more recent material.
Thinking of output inductors: has anyone cut a slot in a ferrite toroid with a diamond faced tile-cutting tool? I think that a tile-cutting disc is about the only thing that would produce a neat cut in ferrite. The problem is that the slot would be a bit wider than required.
N27
N67 (similar to Ferroxcube 3F3)
N87
You can get EFD and ETD cores in all these materials. These cores have the gap in the centre leg, so you don't get much of a stray external field.
N27 and N67 are easier to obtain than N87, which is a more recent material.
Thinking of output inductors: has anyone cut a slot in a ferrite toroid with a diamond faced tile-cutting tool? I think that a tile-cutting disc is about the only thing that would produce a neat cut in ferrite. The problem is that the slot would be a bit wider than required.
I would like RM or some upright standing core both from PCB area point of view and because I want the magnetic stray field to lie in a plane perpendicular to my upright standing modulator PCB. How do I get hold of a couple of gapped RM-10 or RM-12 cores in 3F3, N87 or similar material? Gapping probably needs to be ~1mm for these cores to get a permeability low enough to avoid too high Bmax at my Imax.
zilog said:
For practical purposes, you may assume that the linearity gain when you gap a magnetic circuit equals the ratio of ungapped to gapped permeabilities/inductances.
Hysteresis is one of the non-linearity mechanisms in magnetic materials, therefore it will not only cause losses but will also increase distorsion.
The first-order effect of permeability dependence on frequency is linear. It is a bit similar to the dielectric absorption in capacitors. Now, many audiophile dislike DA, but in itself, it doesn't introduce non-linearities. It may just affect the group delay.
I doubt you'll be able to surpass your powder core: powder, even the cheap iron one can take a substantially higher B than ferrites, and you also intend to switch from a near-canonical shape (toroid) to a non-canonical one (RM, etc). In such complex shapes, the flux density is heterogeneous; in general this is of no concern to the manufacturers, because in the parameters such as Aeff, they take into account the small regions that become saturated and unusable (if you examine the datasheet, you'll notice differences between the physical and magnetic areas).
This is OK for pure power applications, but if you have linearity in mind as well, it is a problem, as these small high-B regions will contribute more to the distorsion than the bulk of the core.
Be also aware that the stray field generated by a well wound toroid is always very low, even at low permeabilities.
This is not the case for other shapes.
The maximum reasonable induction is the one where you stay comfortably away from saturation at all temperatures. Do not be tempted to overdesign because hysteresis f.e. is prevalent at low inductions.
Calculating the distorsion caused by an inductor is no simple matter; it has to be done graphically, or simulated if models are available. In addition, the "HF-bias" effect caused by the switching also complicates the matters as it tends to reduce strongly the non -linearities (but at least, it's also good news).
LV
http://www.hypex.nl/docs/emi.pdf#search=%22ucd%20inductor%22
Talks against toroid inductors because of their bad EMI performance at high frequencies, this is why I want to avoid using them. My PCB area restraints also require me to use an upright standing toroid.
Talks against toroid inductors because of their bad EMI performance at high frequencies, this is why I want to avoid using them. My PCB area restraints also require me to use an upright standing toroid.
zilog said:
Just drop me a mail with your address, and i will send you.
My mail is: info@newclassd.com
All the best from
Lars Clausen
At high frequencies, single layer toroidal inductors (start and finish leads are at one end of the inductor across from each other at the same point along outer circumference) look like a single turn wire loop of the core's average diameter. This functions as a single turn loop antenna to radiate EMI (magnetically AND electrically). With a pot type (RM, etc.) core, the winding is enclosed (i.e., shielded at high frequencies) magnetically and often electrically (if large, electrically grounded core clips or clamps are used).zilog said:
A toroid can be wound in such a way as to largely eliminate the single turn effect. The technique is to wind half way around, then run the wire straight across the center hole (to back by the start), wind up the other side of the core half way around again, and then run the wire straight across the center hole one more time to make the finish end back again by the start. (Make sure the two half windings always pass through the center hole in the same direction.)
This is sort of like splitting the output toroid into to smaller toroids that are then oriented in such a way that the effective high-frequency single turn for each oppose each other and more or less cancel. Of course, for both methods just described, the electrical field problem still remains.
Regards -- analogspiceman
zilog said:
zilog said:
It would be a neat comparison between you RM cores and this one.
http://www.diyaudio.com/forums/showthread.php?postid=1024461#post1024461
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.