Thanks Eva,
My 60 Hz "class D" amp needs only 20 uH @ 20 Amps... I am using the 42/21/15 core with 0.62 mm gap at 19 kHz. ...Hoping my eddy current losses will be acceptable. My sims with MPP are telling me that I can lower core losses but at a high cost in copper loss. Parts are on the way... we'll see.
Any bodu have any bright ideas as to how to measure core and eddy losses in an indutor like this? The calibrated finger is a good start of course.
My 60 Hz "class D" amp needs only 20 uH @ 20 Amps... I am using the 42/21/15 core with 0.62 mm gap at 19 kHz. ...Hoping my eddy current losses will be acceptable. My sims with MPP are telling me that I can lower core losses but at a high cost in copper loss. Parts are on the way... we'll see.
Any bodu have any bright ideas as to how to measure core and eddy losses in an indutor like this? The calibrated finger is a good start of course.
-_nando-_ said:I'm almost sure that my amp have problems, as if I don't connect the output inductors (if I leave the output from the amp floating) it still get that heatsink hot !
This is normal. If you don't connect output inductor, then parasitic capacitances must be charged and discharged by FETs inside IC, while with inductor this charge comes from inductor.
Depends on how much trouble you want to go to. Run it in a bomb calorimeter? Or you could try estimating the heat evolved over a short period from the temperature rise, the weight of copper, the weight of the core, and the specific heats of the materials. Keep the period short to minimise the effects of convection cooling in air. In fact you could surround it with cotton wool. Or an aerogel. Course the heating may be uneven. Photograph it with an IR camera? How you going to measure the energy going in?
Energy going in is no problem... I can measure the voltage and current... scope will multiply and integrate.
And... I am thinking EXACTLY about throwing this thing in some water and tracking the rise!
Either that... or just measure output power and deduce the losses on a comparitive basis alone.
And... I am thinking EXACTLY about throwing this thing in some water and tracking the rise!
Either that... or just measure output power and deduce the losses on a comparitive basis alone.
Pafi said:
You are correct.
However, it is not eddy current which is causing the problem per se. It is current constriction in the high field areas of the inductor. Eddies are not exactly the correct term, but rather, inducted voltage potentials within the conductors, which redistribute the current density.
As frequency goes up, skin effect raises the resistance per unit area of the wire. Most consider this on a single wire basis. For wound inductors, each wire will react to it's own magnetic field as well as that of it's neighbors. Proximity effect.
Litz will help solve the problem also. But the best way is to reduce the highest field level, this is what the larger inductors are doing.
If anybody wishes to read more on this, I recommend Dr. Sullivan, he is a Professor at Dartmouth. When he contacted me a few years ago, I found some really good papers of his on this exact topic. (hey, I wuz interested in what he wuz gonna do with da stuff he wanted me to make..)
I'll see if I can find any. But kinda busy at the moment..
ps...ahhh, it was easier than I thought.
http://thayer.dartmouth.edu/inductor/papers/aspectratio.pdf
http://thayer.dartmouth.edu/inductor/papers/optshape.pdf
http://thayer.dartmouth.edu/inductor/papers/anagen.pdf
http://thayer.dartmouth.edu/inductor/papers/cooscost.pdf
Note: do not confuse core losses due to fringing at the gap with losses within the wire as a result of the flux distribution..
Cheers, John
pps...Hey poobah, Happy new year..
Oh, and BTW
Why in the name of sam hill, are you pointing the solenoid flux lines at the circuit????????
Rotate the small inductor 90 degrees to the circuit, see if that reduction in coupling to the active circuit removes the hf oscillation at the zero crossings.
As an added bonus, you may find the heatsink gets cooler.
If you go with a ferrite based inductor, that'll constrain the coupling much more.
Cheers, John
Why in the name of sam hill, are you pointing the solenoid flux lines at the circuit????????
Rotate the small inductor 90 degrees to the circuit, see if that reduction in coupling to the active circuit removes the hf oscillation at the zero crossings.
As an added bonus, you may find the heatsink gets cooler.
If you go with a ferrite based inductor, that'll constrain the coupling much more.
Cheers, John
poobah!
Why do you think this? Why do you think that magnetic coupling is the only way to oscillation on output? Do you think inductor doesn't have capacity? Do you think output capacitor doesn't have ESL? Of course they have. These ast as a resonant circuit, wich reacts to the switching. There is no need to assume over this an inductive coupling. Especially since inductive coupling would leads to square wave error signal! And inductive coupling would not be weaker with bigger choke!
jneutron!
Thank you for emphasizing difference between skin effect and eddy current! There is difference indeed.
From your first link, first sentence of introduction:
"THE fringing field from the air gap in an inductor can
cause severe eddy-current losses in the winding"
And this is correct. Skin effect is caused by current flows in the wire, so it can eliminate current density inside the wire, but can't reverse it, since then a cause of effect would reverse too. Contrary to this, eddy current can reverse the current in one side of wire, and increase it on other side (because it is generated by all of the currents in all turns, not only the current inside the observed wire). Eddy current can be many times stronger than the original current. This is why you may have to use very fine litze (eg 144*0,07), although there is no significant skin effect under 0,3 mm diameter at 100 kHz.
Why do you think this? Why do you think that magnetic coupling is the only way to oscillation on output? Do you think inductor doesn't have capacity? Do you think output capacitor doesn't have ESL? Of course they have. These ast as a resonant circuit, wich reacts to the switching. There is no need to assume over this an inductive coupling. Especially since inductive coupling would leads to square wave error signal! And inductive coupling would not be weaker with bigger choke!
jneutron!
Thank you for emphasizing difference between skin effect and eddy current! There is difference indeed.
From your first link, first sentence of introduction:
"THE fringing field from the air gap in an inductor can
cause severe eddy-current losses in the winding"
And this is correct. Skin effect is caused by current flows in the wire, so it can eliminate current density inside the wire, but can't reverse it, since then a cause of effect would reverse too. Contrary to this, eddy current can reverse the current in one side of wire, and increase it on other side (because it is generated by all of the currents in all turns, not only the current inside the observed wire). Eddy current can be many times stronger than the original current. This is why you may have to use very fine litze (eg 144*0,07), although there is no significant skin effect under 0,3 mm diameter at 100 kHz.
Pafi,
The inductive coupling would be worse with a smaller choke becuase the field is more concentrated.
My point is... the fact that the oscillations are not SYMMETRIC... implies that a NON linear... NON passive element is at work here... what else could it be?
I am NOT offering EXCLUSIVE SINGULAR REASONS here... just suspicions/suggestions that could easily be proven correct or otherwise. Air core and gapped inductors have to be respected for the flux they throw... simple. It's about loops and changing flux... that's all.
The inductive coupling would be worse with a smaller choke becuase the field is more concentrated.
My point is... the fact that the oscillations are not SYMMETRIC... implies that a NON linear... NON passive element is at work here... what else could it be?
I am NOT offering EXCLUSIVE SINGULAR REASONS here... just suspicions/suggestions that could easily be proven correct or otherwise. Air core and gapped inductors have to be respected for the flux they throw... simple. It's about loops and changing flux... that's all.
jneutron said:
Thanks for those links John, the other stuff at that site is very interesting. I found a link there to a Unitrode seminar that explains the skin effect and proximity effect in a pretty easy-to-understand way:
http://focus.ti.com/lit/ml/slup197/slup197.pdf
Regards, Joakim
Sorry, I was trying to simplify this issue, not oversimplify it, but before proceeding to the finer points of inductor design it's better to have an appreciation of their DC characteristics.
Most of this eddy current stuff has to do with fringing in gapped inductors.
The original discussion was about air-cored coils and there are mininal out of alignment lines of flux in a multi-turn air-core inductor when the leads are kept short.
Proximity effect is a frequency dependent phenomenon as is skin effect, and it's magnitude is dependent on form factor, wire dimensions and packing arrangement.
I didn't really want to get into this, when I made my earlier post somebody beat me to it with the stuff about the flux density and the eddy currents, I didn't see that post until after I posted.
I just thought there were enough unstated dimensions to account for the differences in heating considering that the frequencies involved were the same (pretty low, although the harmonics go up higher the faster the rise and fall times), and anyway as I understand it proximity effect is more related to the fields from immediately adjacent conductors than the overall field strength which produces eddy currents.
In this case since we know neither the wire gauge, nor the length, nor the winding regime then the contribution of proximity effect seems to me a needless complication to an already ill-understood situation, given the changes in form factor, turns and wire gauge necessary to produce inductors of different air-core diameters but identical inductance and DC resistance (which tells you your max Q).
Although my intuition is that a smaller core diameter results in a larger proximity effect (I'm trying to avoid doing the calculations), if form factor can affect this for a siongle core diameter, then it's inevitable that you can make a small ID inductor of a given value less affected by proximity effect than one with a larger ID and the same inductance.
All this said it is not accurate to characterise the situation in smaller ID coils as 'A bigger part of the flux penetrates into the wire'. The flux density throughout the copper is related to the magnetic field strenth via the permeability. B=u x H.
Most of this eddy current stuff has to do with fringing in gapped inductors.
The original discussion was about air-cored coils and there are mininal out of alignment lines of flux in a multi-turn air-core inductor when the leads are kept short.
Proximity effect is a frequency dependent phenomenon as is skin effect, and it's magnitude is dependent on form factor, wire dimensions and packing arrangement.
I didn't really want to get into this, when I made my earlier post somebody beat me to it with the stuff about the flux density and the eddy currents, I didn't see that post until after I posted.
I just thought there were enough unstated dimensions to account for the differences in heating considering that the frequencies involved were the same (pretty low, although the harmonics go up higher the faster the rise and fall times), and anyway as I understand it proximity effect is more related to the fields from immediately adjacent conductors than the overall field strength which produces eddy currents.
In this case since we know neither the wire gauge, nor the length, nor the winding regime then the contribution of proximity effect seems to me a needless complication to an already ill-understood situation, given the changes in form factor, turns and wire gauge necessary to produce inductors of different air-core diameters but identical inductance and DC resistance (which tells you your max Q).
Although my intuition is that a smaller core diameter results in a larger proximity effect (I'm trying to avoid doing the calculations), if form factor can affect this for a siongle core diameter, then it's inevitable that you can make a small ID inductor of a given value less affected by proximity effect than one with a larger ID and the same inductance.
All this said it is not accurate to characterise the situation in smaller ID coils as 'A bigger part of the flux penetrates into the wire'. The flux density throughout the copper is related to the magnetic field strenth via the permeability. B=u x H.
Maybe due to my poor english, I don't know what do you mean "out of alignment", and why it is dependent of lead length.
Maybe it's not accurate, but it is true and relevant, and there is no need to precisely calculate exact loss values.
Such large diameter at this small current makes impossible heating to be explained by Rdc. ~ 1 A/mm^2
[/QUOTE]The flux density throughout the copper is related to the magnetic field strenth via the permeability. B=u x H.[/QUOTE]
So what? Permeability was constant. Air core!
Yes, max Q, wich is much higher then actual Q, and there is no direct relation between them. This is a useless information in this case.
We know exact wire gauge, and we know the ratio of ID, and we know (at least I'm sure) that L was ~22 uH in all cases, so I see almost the same situation I have already experienced, and based on my measurements (and other theory) I can say for sure that problem was eddy current. This is not needless complication, but the truth. If you can't calculate: don't do it! But this doesn't mean that nobody can tell anything about this situation, does is?
Yes. Something makes rising and falling edges different in speed. This can be in the gate-source, or anywhere, but we don't see inside of the IC. However in an IC there are really tiny loops.
BTW.: on the waveforms you can see 2 different type of "error". With small inductor there is ringing, with big inductor there is a little step. The step can be explained by inductive coupling from inductor to scope measuring cable.
Pafi said:
Cube?
Ahh, I see the difference.. In my work, we consider the field falloff radial to the dipole. It is 1/R, with uniform field inside. For quads, it is 1/R squared, with 1/R in the middle. For sextupoles, it is 1/R cubed, with internal field 1/R squared.
You are speaking of a solenoid..
Pafi said:
Agreed.
Guys..if one considers Faraday's law of induction here, one can see that the most induced voltage being created at the victim circuit occurs when the current within the solenoid is changing the most. I suspect that the control circuit is being upset by the induced voltages, and that this occurs during the highest current slopes, both rising and falling.
I still recommend rotating the small inductor 90 degrees.
While the IC internally has small loops, so will be less susceptible to induction, the external connections are not small.
Bud: Thanks for the link. Note that it is applicable for individual wires, I do not know if it can be scaled accurately for air core inductors where flux enhancement occurs. It may be applicable to single layer toroids.
If one were to plow through Dr. Sullivan's papers, it is seen that much effort has been given to reducing the flux magnitude based loss in the wires, not just the gap issues.
Cheers, John
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.