Can you tell us what is the optimum shape?Now I have researched toroid optimization and found that it's not rocket science - it's more advanced than that
The maths turns out to be the same as for nuclear fusion reactor toroids, wonderful, and there's some nice results.
Not the same as your formula but you are correct that it's not very sharp peaked optimum.
Bear in mind that any practical output coil has very few turns.
I read that too, but I'm very tempted to create a Wikipedia entry with my explanation
(probably too much Randall Munroe and XKCD)The optimum solenoid "window" is close to a D section.
So the toroid looks a bit like a sphere with a cylindrical hole but rounded over at the "poles".
The hole is proportionally quite small, down to about 20% of the diameter.
"Very few turns" is not optimal, better to use more turns but make the diameter smaller. This lowers the resistance for a specified inductance.
Obviously there are practical limits, I plan to wind a few experiments.
But mostly because I think nuclear fusion research derived toroids are so cool to have in an amp
An externally hosted image should be here but it was not working when we last tested it.
View of the plasma inside the tokamak MAST in the United Kingdom. (Source: UKAEA-Culham)
Best wishes
David
Note that the shape in MiiiB's Dartmouth link is constrained for depth, and then further compromised to make it easy to fabricate with their particular methods.
Not that this has a major performance cost, just explains the difference.
Last edited:
I wonder if you will enjoy the Scottish leg of your tour?Angela and I are leaving for a much-needed 2-week vacation to Europe tonight.
I'll be in very limited contact by email and for posting, so don't feel like I went dark
Cheers,
Bob
optimums do depend on what are the weightings for the application
2nd ref in 1st post shows a ~1 uH air core foil inductor proto, measurements
http://www.diyaudio.com/forums/clas...-inductor-class-d-amp-output.html#post3082043
single digit milliohm AC R @20 kHz - are we there yet?
2nd ref in 1st post shows a ~1 uH air core foil inductor proto, measurements
http://www.diyaudio.com/forums/clas...-inductor-class-d-amp-output.html#post3082043
single digit milliohm AC R @20 kHz - are we there yet?
https://engineering.dartmouth.edu/inductor/papers/pesc2007b.pdf & Dave Zan's UKAEA version.
I was assuming we would use Bob's "wind a 'long' coil & twist into a toroid" method. Both the above methods require a former ... but then you can talk about formers & cores hand carved from solid Unobtainium by Virgins
________________
Formers would also be needed to fix the turns so they don't move. You can't just dip in glue as you might with Self's short closed wound coils.
________________
Anyone got any 'real life' contributions on the importance of glueing the turns?
I was assuming we would use Bob's "wind a 'long' coil & twist into a toroid" method. Both the above methods require a former ... but then you can talk about formers & cores hand carved from solid Unobtainium by Virgins
________________
Formers would also be needed to fix the turns so they don't move. You can't just dip in glue as you might with Self's short closed wound coils.
________________
Anyone got any 'real life' contributions on the importance of glueing the turns?
So no one is interested in trying to keep the inductor inside the feedback loop?
I thought perhaps the likes of Dave Zan and Edmond Stuart might be up for the challenge?
I thought perhaps the likes of Dave Zan and Edmond Stuart might be up for the challenge?
have to solve the problem of getting it to look like an inductor around the gain intercept frequencies so it does its intended job there
could probably do something looking like TMC - the improvement from feedback is going to be limited by 20 kHz
leakage in practical toroids come from the lumpiness of the winding - the math for a uniform sheet of current on a toroid gives no leakage - practical single layer winding do have a net single loop around the hole as it were and the gaps between the wire turns limits the geometric cancellation
in the thread I linked earlier it was pointed out that lower mu iron powder cores, u_r ~ 10 are probably a fine compromise for flat with excitation current added L
it also has been pointed out in transformer design that leakage is a winding geometry issue - it is always there and adding a perm core only adds more flux through the core doesn't change the winding geometry leakage
Last edited:
Halcro output coil. I vaguely recalling when they were getting all the reviews that somewhere it was mentioned that the pictured 'output inductor' is actually a piece of hardline coax. Never made any sense to me why you'd want that. Can someone verify? If you own a Halcro, please disassemble and take pictures so we can document. ;-)
This is a good question. It’s not clear to me how well-contained is the field of a air-core toroid. I think there would still be crosstalk concerns for a stereo amp.
I assume we want to minimize the resistance for specified inductance.
Pretty much the same as "use the shortest (and cheapest) piece of wire".
My experiments so far have been "wind and twist" because it's so easy.
But a toroid former would be hardly more difficult, it's a simple shape.
And then less need to dip in epoxy or whatever, so more or less a wash I think.
Maybe not "Unobtainium" but I remember that inductors for the National Bureau of Standards used to be wound on cores of marble. Nice!
Or could wind onto a "D" shaped rod of slick plastic as "K.man" proposed, then slide off and twist.
Toroids do have a lower Q, for any specific piece of wire and core, air or other.
So the trade off is more wire to meet a resistance/inductance requirement, in return for essentially zero external field.
I'm happy with that, as JCX noted -it's not hard to hit lower resistance than any realistic need.
The loop around the hole is easy to cancel.
The gaps between the wires is a separate problem but pretty much academic I think, foil looks nice however.
Best wishes
David
Hi Harry
If the poloidal component is cancelled (the loop around the hole) then the containment could be just about perfect I think.
So I really do think this is a solved problem, no need to move the inductor inside any feedback.
But thanks for your confidence that I could do it
Best wishes
David
Last edited:
Hi HarryDymond,
An interesting thought. Compensate the feedback loop for the output inductor - like EQ if you will. You pre-distort the signal and allow the output coil to correct it. The coil then remains outside the feedback loop and it's effects largely eliminated.
-Chris
To the best of my knowledge, that coil is supposed to shield the feedback loop from capacitance to common. Putting it within the loop only defeats the purpose of why it's there in the first place.
An interesting thought. Compensate the feedback loop for the output inductor - like EQ if you will. You pre-distort the signal and allow the output coil to correct it. The coil then remains outside the feedback loop and it's effects largely eliminated.
-Chris
Yes, feedback amplifiers typically have a certain range of capacitance (for example, 10nF to 1uF)
for which there are stability problems. In feedback voltage regulators this is dealt with by using
a large enough load capacitance to avoid the problem. This cannot be done with an audio amplifier,
so moderate capacitive loads often can cause problems, hence the output inductor used for load isolation.
Last edited:
I think the output L for a class D has another more important function: energy storage and integration of the output pulse train.
You could arguably have two output L's - the main filter coil, and then a separate one to deal with cap loads. I would assume the cap load L could be compensated also but, and it's a big but, the whole comp scheme would be pretty complex IMV. Did'nt someone mention BP Ncore SMPS were 4 or 5 order comp loops?
You could arguably have two output L's - the main filter coil, and then a separate one to deal with cap loads. I would assume the cap load L could be compensated also but, and it's a big but, the whole comp scheme would be pretty complex IMV. Did'nt someone mention BP Ncore SMPS were 4 or 5 order comp loops?
Yes, this is true. However, what is relevant in this discussion, is that in class-D amplifiers where the feedback is taken after the LC output filter, said LC filter is used to provide two poles of the feedback compensation.
The point I’m trying to make is that as there is already significant “C” in the LC filter, a second L that is outside the feedback loop is not required, as any additional C added by the load cannot make the feedback loop go unstable. The challenge is that the feedback loop will now probably have to be fourth or fifth order.