Inductor Voltage Calculations

[MrAl]

Hi,

I can help a little here.

First, inductance books vary quite a bit as you found out.  They dont all use the same units either, and because of that nagging 4*pi*10^=7 factor sometimes they dont even include any of that and work in some other units and other times they just use u0, and still other times they use u/u0 for when they want to include relative permeability, and that's just the start of things to go wrong :-)

The best bet to combat the above and similar is to home in on a formula you like and try to find another one even if in different units, so you can compare results.  Ideally, the results should come out identical within several decimal places, and even better as the ratio of two integers when possible.  If they come out close but not exact, then you might be ok but you're better off trying to find out why there is a variation even if small, and possibly by simply trying some other parameter values.  This allows you to be more sure of the result you need.

The worst part about this is that when they hand you numbers and formulas they dont always show the derivation and sometimes that is because it is very long and complicated and takes some mathematical tricks to get an answer.  Without that derivation though you have no idea what is going on, so it makes you feel kind of lost.
What you could do is either find a better book or start to do some derivations of your own.  To that end, you can start with the Biot Savart Law.  That's the basic law behind a lot of magnetic calculations that can come out exact.  There is always the practicality of it all though, which really makes everything we do outside of pure theory just an approximation, but at least then you can figure out why.

To compute inductance behavior in a converter you have to know the voltage across it for all times, at least the maximum, and the expected maximum 'on' time.  If you know those two you can get pretty far because an important aspect of that is in terms of volt seconds.  In short, if you have more volt seconds than your winding can handle, it saturates.
The max sometimes comes from normal operating conditions but if you dont have a slow start mechanism in place then you end up with more volt seconds than you expected, so you really need a slow start mechanism for any reasonably sized converter.
If you vary the input voltage you need to consider the maximum input voltage and it's relation to the output voltage when the inductor topology puts it in series with the input and output.  But again, this will bring in some form of slow start so the volt seconds can be limited during start up.

You can estimate the fringe flux if you want too, and there are several articles online.  Here is one i found:
http://www.encyclopedia-magnetica.com/doku.php/flux_fringing
There may be more information on that Magnetics Inc site too.
BTW you can always measure the characteristics of your inductor, and change the gap as needed during that test.

The 'risk' of saturation...
If you design it right you should not encounter this problem, right?  Or did you find something that became a problem?
If your inductor volt seconds are limited, it will not saturate.  It's not really about current except when there is a DC current present as well as any AC.  When there is DC also present, then yes, the DC could cause saturation, but that's something you include in the calculations for the inductance.  Overloading must be avoided by current monitoring, if an overload of some type is expected, which is almost always a consideration.

Not sure what you are asking about the MOSFETs here, but if you are doing a full bridge then the voltage ratings of each transistor must be higher than the expected max voltage across each transistor.  It's that simple.  Of course if you have significant spikes then you need to snub them, and possibly go up on the transistor rating.
If you intend to output say 100v peak to the primary, then you should expect 100v peak, then add some head room for spikes and use snubbers.  When a lower transistor turns on it's about 0v, but when the diode conducts it could be -2v, and when an upper turns on it's going to put 100v peak on one transformer lead, so you can figure maybe 102v if you like to be more accurate, but really we'd have to model this to get batter estimates.  So you see using 120v MOSFET's would probably do the job unless we had significant spikes, and then maybe we would go up a little and improve the snubber.
After all is said and done however, the unit must be tested.  This comes from input a DC voltage (or whatever you are going to input) and turning it up slowly and watching for known problems such as saturation and excessive spike amplitude.  There's almost no way around this phase of the project.

I have to ask you a question now too, and that is, did you ever design and build a buck circuit?
A buck circuit is a great way to test an inductor that is to be used for a switcher.

I hope this helps at least somewhat, and if you really want to get into the deep end of magnetics, take a look at the Biot Savart Law which appears at many web sites.  It's at least one of the laws that leads to everything else in magnetics aside from relativistic effects which we usually dont need :-)

[jbb]

The 'risk' of saturation...
If you design it right you should not encounter this problem, right?  Or did you find something that became a problem?
If your inductor volt seconds are limited, it will not saturate.  It's not really about current except when there is a DC current present as well as any AC.  When there is DC also present, then yes, the DC could cause saturation, but that's something you include in the calculations for the inductance.  Overloading must be avoided by current monitoring, if an overload of some type is expected, which is almost always a consideration.

Depending on the mode of the converter, there could be little DC current (discontinuous conduction mode) or a lot of DC current (continuous conduction mode).  It can be a bit hard to tel, because your converter may go between the two as instantaneous line voltage changes.

The mechanism for saturation is that too much magnetic field strength H (Amps per metre) is applied to the (gapped) core set.
\Sigma NI = \Sigma H l

Where:
  N = # of turns
  I = current (A)
  H = magnetic field strength (A/m)
  l = path length (m)
Note: the \Sigma symbols are there to handle multiple windings (e.g. transformers and coupled inductors), and can be ignored for a simple inductor.


Then magnetic flux density B can be worked out:

B = \mu H
\mu = \mu_r * \mu_0

Where:
  B = magnetic flux density (T)
  \mu_r = relative permeability (see core material datasheet or 1 for air/plastic)
  \mu_0 = permeability of free space (4 *pi * 10^-7 H/m)

I forget how to handle the air gap, though.  I think you set the B values in the core and air to be equal, but I'm not certain. I could look that up tonight...

Anyhow, we see that the easiest way to look at saturation of an inductor is simply to check the peak current.  (Transformers are quite different!). As it happens, a lot of current mode control schemes are designed to immediately switch off the power transistor if the current reaches a peak value.  This combination is great, because it means that with a little care on the controls side you can be very confident about not saturating your inductor :-).

[hanzdolo30@gmail.com]

Hi,

I can help a little here.

First, inductance books vary quite a bit as you found out.  They dont all use the same units either, and because of that nagging 4*pi*10^=7 factor sometimes they dont even include any of that and work in some other units and other times they just use u0, and still other times they use u/u0 for when they want to include relative permeability, and that's just the start of things to go wrong :-)

The best bet to combat the above and similar is to home in on a formula you like and try to find another one even if in different units, so you can compare results.  Ideally, the results should come out identical within several decimal places, and even better as the ratio of two integers when possible.  If they come out close but not exact, then you might be ok but you're better off trying to find out why there is a variation even if small, and possibly by simply trying some other parameter values.  This allows you to be more sure of the result you need.

The worst part about this is that when they hand you numbers and formulas they dont always show the derivation and sometimes that is because it is very long and complicated and takes some mathematical tricks to get an answer.  Without that derivation though you have no idea what is going on, so it makes you feel kind of lost.
What you could do is either find a better book or start to do some derivations of your own.  To that end, you can start with the Biot Savart Law.  That's the basic law behind a lot of magnetic calculations that can come out exact.  There is always the practicality of it all though, which really makes everything we do outside of pure theory just an approximation, but at least then you can figure out why.

I spent all day homogenizing the equations from all of the books to use u₀ properly.

Biot Savart Law is EXACTLY what I've been looking for. I just didn't know how to find it.

To compute inductance behavior in a converter you have to know the voltage across it for all times, at least the maximum, and the expected maximum 'on' time.  If you know those two you can get pretty far because an important aspect of that is in terms of volt seconds.  In short, if you have more volt seconds than your winding can handle, it saturates.
The max sometimes comes from normal operating conditions but if you dont have a slow start mechanism in place then you end up with more volt seconds than you expected, so you really need a slow start mechanism for any reasonably sized converter.
If you vary the input voltage you need to consider the maximum input voltage and it's relation to the output voltage when the inductor topology puts it in series with the input and output.  But again, this will bring in some form of slow start so the volt seconds can be limited during start up.

The good book I found(in SI units), actually has a number of equations precisely for boost inductors, covering;  I_L, V_L, D (which I knew already but forgot due to the long break  ), ripple current, I_L(RMS), V_L(RMS), L, Apparent Frequency and I already know how to calculate N from L then augment the value by the square root of u_m/u_e, based on: u_m/1+u_m(l_g/lm)

You can estimate the fringe flux if you want too, and there are several articles online.  Here is one i found:
http://www.encyclopedia-magnetica.com/doku.php/flux_fringing
There may be more information on that Magnetics Inc site too.
BTW you can always measure the characteristics of your inductor, and change the gap as needed during that test.

I've actually seen that, that article it was very helpful. Is fringe flux factor a big deal at all? Because you and another person that I asked about it don't seem to give it much importance, but the books make it seem pretty nasty.

The 'risk' of saturation...
If you design it right you should not encounter this problem, right?  Or did you find something that became a problem?
If your inductor volt seconds are limited, it will not saturate.  It's not really about current except when there is a DC current present as well as any AC.  When there is DC also present, then yes, the DC could cause saturation, but that's something you include in the calculations for the inductance.  Overloading must be avoided by current monitoring, if an overload of some type is expected, which is almost always a consideration.

I haven't built the inductors yet, I want to get the equations all correct first. I don't want to short out my new monitor, lol.

Not sure what you are asking about the MOSFETs here, but if you are doing a full bridge then the voltage ratings of each transistor must be higher than the expected max voltage across each transistor.  It's that simple.  Of course if you have significant spikes then you need to snub them, and possibly go up on the transistor rating.
If you intend to output say 100v peak to the primary, then you should expect 100v peak, then add some head room for spikes and use snubbers.  When a lower transistor turns on it's about 0v, but when the diode conducts it could be -2v, and when an upper turns on it's going to put 100v peak on one transformer lead, so you can figure maybe 102v if you like to be more accurate, but really we'd have to model this to get batter estimates.  So you see using 120v MOSFET's would probably do the job unless we had significant spikes, and then maybe we would go up a little and improve the snubber.
After all is said and done however, the unit must be tested.  This comes from input a DC voltage (or whatever you are going to input) and turning it up slowly and watching for known problems such as saturation and excessive spike amplitude.  There's almost no way around this phase of the project.

My guess was the MOSFET voltage should be more or less double the V_DS(in)

The reason for the MOSFET/Output voltage question is because I know it's probably better to boost to about 400V, but I know there's less current stress on the transformer if the input voltage is higher.

Let me explain the project to you in detail:
I'm building a DC-AC-DC converter with 2 PFC inductors ,using 2 mosfets(FCH060N80), boosting to 400VDC (or 800V with 1200V 45A SiC MOSFETS to reduce current stress on the transformer), operating in opposite phase of each other. The total output power of the PFC stage will be 1kW(500W output from each inductor), powering 4 FCH023N65S3_F155 MOSFETs, providing an alternating square wave with a deadtime to avoid shorting. I've chosen this topology to maximize on core useage. The secondary winding will provide a rectified output voltage of 100VDC(I'll probably add a tertiary 12V for cooling fans), buck-boost output regulator to enable a variable output voltage. Size or part count isn't of primary concern, efficiency of operation is (I have 100s of kickass MOSFETS, among other components). I may program an MCU to control the FETS to monitor voltage and current, but I believe this can be achieved with linear TTL logic(more fun that way).

I know I can get a lot more output power from the converter I've described, however I only have 20A available in total (unless I go in the kitchen where there's 30A) and I don't have a sufficient load to test the device above 1.4kW however I will add the appropriate components and windings to increase output power as needed.

I have to ask you a question now too, and that is, did you ever design and build a buck circuit?
A buck circuit is a great way to test an inductor that is to be used for a switcher.

No, but what actually got me into electrical engineering were boost converters. I bought a few cheap $5 converters, tied them to the 12V rail of an ATX PSU and after a few days of being hooked they kept blowing the MOSFET when I applied a 30v 100w load (COB LED) to them, so I took one apart, re-wrapped the toroid with some homemade litz wire and center tapped it to take the stress off of the MOSFET(which worked very well,lol), what I called a flyback autotransformer at the time, and made a freaky boost converter that worked amazingly well, being I totally did it spatially, no inductor calculation at all. All I calculated was the winding ratio between the primary(lower) and secondary(upper=2*lower)

Look at this crazy ass thing :

I had that thing running as lamp in here for months(MOSFET heatsink stayed slightly warm the whole time) until it fell off of the table and the heatsink fan on the LED shattered (never bothered to reassemble it).
I've learned so much since then. The crazy thing is that it actually did what I expected. At a 50% on time it was 36V(40V no load) because the upper portion of the inductor had twice as many windings as the part that I was exciting. I was able to get that thing up to 45V before its little 10A breaker button would pop. It killed a couple of red COBs but the white one could handle excessive voltage with no problem. After that thing I was hooked on power electronics.

[T3sl4co1l]

I spent all day homogenizing the equations from all of the books to use u₀ properly.

Biot Savart Law is EXACTLY what I've been looking for. I just didn't know how to find it.

I would suggest not worrying too much about the nitty-gritty.  Biot-Savart is great for physics problems, but you shouldn't need it in electronics problems.

The biggest drawbacks are:
1. There's no good way to integrate over the surface of a winding.
2. There is better symmetry to shortcut these problems with.

#1 is particularly complicated by the fact that the current doesn't flow uniformly through the wire, and the wire is not transparent to magnetic fields.  It would be easy enough to integrate along, say, a helical path, but wires are not infinitesimally thin; if you add a cross-section to it, now you're assuming the current is distributed somehow over that cross-section (which it isn't).

It's another way of saying: that one particular equation is magnetostatic.  But conductors at AC have resistance, so the EMF (from self-induction, causing skin effect; and the total induction, that gives us circuit inductance) is nonzero.

It's important to understand how this stuff works, but it's devilishly hard to work a real problem with.

As for symmetry, Maxwell's got you covered: for an infinitesimal cube, with field lines parallel and perpendicular to the sides, all the equations are simple proportions.  Flux is flux density times area; EMF is loop area times flux change; field intensity is current over path length; and flux density is field intensity times permeability.  (Clearly, if EMF and current have a relationship -- as in a resistive material -- you need to add extra terms to represent that.)

Nice thing about magnetic cores: since the flux is almost all (99%+) confined to a core, we don't much care that the core is itself going around some kind of path, we just care that it has length and area -- we can consider it as one rectangular chunk, where the field lines are parallel and perpendicular to the sides!

Just as we don't need to integrate over SMPS waveforms if they are straight-sided, we don't need to integrate over the core if it behaves straight-sided.

So these "square wave" or "square block" symmetries help out a lot!

So for analyzing a core, it's enough to use Maxwell's laws directly, where each law is simply a ratio.  We don't need to do vector calculus, because we've applied geometry that gets rid of all the nasty dimensionality and leaves us with scalar proportions (width and height, magnitude and sign).  We don't need any constants, no random pi's and such, only bare fundamental constants; namely, B = mu * H.

I've actually seen that, that article it was very helpful. Is fringe flux factor a big deal at all? Because you and another person that I asked about it don't seem to give it much importance, but the books make it seem pretty nasty.

Fringing flux is a problem for two reasons:
1. The intense field strength, and its nonuniformity (that is, it doesn't behave as a simple parallel rectangular region), induces eddy currents in nearby wires, causing heating (proximity effect).
2. The fringe location acts like a winding of the same size, which transmits fields out into space.  Thus, it's a source of emissions.  You often see transformers with a copper tape wrapped around the outside (notice it goes around the whole core, not just the winding where it would be a shorted turn!), which acts to short out (shield) this mode.  In other words, it keeps the flux inside the core, where it belongs.

Related:
http://power.thayer.dartmouth.edu/shapeopt_intro.html
This calculates where to optimally place wire, to avoid those losses.  Cool!

Tim

[MrAl]

I spent all day homogenizing the equations from all of the books to use u₀ properly.

Biot Savart Law is EXACTLY what I've been looking for. I just didn't know how to find it.

I would suggest not worrying too much about the nitty-gritty.  Biot-Savart is great for physics problems, but you shouldn't need it in electronics problems.

The biggest drawbacks are:
1. There's no good way to integrate over the surface of a winding.
2. There is better symmetry to shortcut these problems with.

#1 is particularly complicated by the fact that the current doesn't flow uniformly through the wire, and the wire is not transparent to magnetic fields.  It would be easy enough to integrate along, say, a helical path, but wires are not infinitesimally thin; if you add a cross-section to it, now you're assuming the current is distributed somehow over that cross-section (which it isn't).

It's another way of saying: that one particular equation is magnetostatic.  But conductors at AC have resistance, so the EMF (from self-induction, causing skin effect; and the total induction, that gives us circuit inductance) is nonzero.

It's important to understand how this stuff works, but it's devilishly hard to work a real problem with.

As for symmetry, Maxwell's got you covered: for an infinitesimal cube, with field lines parallel and perpendicular to the sides, all the equations are simple proportions.  Flux is flux density times area; EMF is loop area times flux change; field intensity is current over path length; and flux density is field intensity times permeability.  (Clearly, if EMF and current have a relationship -- as in a resistive material -- you need to add extra terms to represent that.)

Nice thing about magnetic cores: since the flux is almost all (99%+) confined to a core, we don't much care that the core is itself going around some kind of path, we just care that it has length and area -- we can consider it as one rectangular chunk, where the field lines are parallel and perpendicular to the sides!

Just as we don't need to integrate over SMPS waveforms if they are straight-sided, we don't need to integrate over the core if it behaves straight-sided.

So these "square wave" or "square block" symmetries help out a lot!

So for analyzing a core, it's enough to use Maxwell's laws directly, where each law is simply a ratio.  We don't need to do vector calculus, because we've applied geometry that gets rid of all the nasty dimensionality and leaves us with scalar proportions (width and height, magnitude and sign).  We don't need any constants, no random pi's and such, only bare fundamental constants; namely, B = mu * H.

I've actually seen that, that article it was very helpful. Is fringe flux factor a big deal at all? Because you and another person that I asked about it don't seem to give it much importance, but the books make it seem pretty nasty.

Fringing flux is a problem for two reasons:
1. The intense field strength, and its nonuniformity (that is, it doesn't behave as a simple parallel rectangular region), induces eddy currents in nearby wires, causing heating (proximity effect).
2. The fringe location acts like a winding of the same size, which transmits fields out into space.  Thus, it's a source of emissions.  You often see transformers with a copper tape wrapped around the outside (notice it goes around the whole core, not just the winding where it would be a shorted turn!), which acts to short out (shield) this mode.  In other words, it keeps the flux inside the core, where it belongs.

Related:
http://power.thayer.dartmouth.edu/shapeopt_intro.html
This calculates where to optimally place wire, to avoid those losses.  Cool!

Tim

Hello there,

I think that post was well written for sure, and i dont want to take much away from that by disagreeing with one point only, but i must disagree a little when it comes to the Biot Savart Law, and theories that lie in that same natural world base class.

We have basic theories and first principles for a reason.  That reason is not to mock the absurdly abstract, but to use them as a basis of further understanding.  Even if we cant calculate the six dimensional integral required for a certain inductance calculation, we at least know where the inductance originates.  We can also sometimes turn to other simplifications that allows us to go back to the original statements and use them after substituting the simplifications into the original problem.  For example, the geometric mean distance and it's use in magnetic inductance calculations, and other simplifications like that.  So there's never going to be a loss because we studied Biot Savart a little because everything that comes after that builds on that basic theme.  It's good to be able to see the forest, but sometimes it helps to see the leaves too (a reference to environmental analysis based on leaf count).

Also, some things that were not subject to direct calculation yesterday are able to be calculated today.  Elliptic integrals being one example, where many years ago the goal was to come up with formulas that did not have to use them, while today they are much more easily evaluated using symbolic math programs (Wolfram for one example).

If you disagree, you might ask yourself if we dont need it then why did you learn it :-)
Because you had learned it you know about the difficulties and how it may be used, but if you did not learn it you would be asking right here about what significance it has in magnetics.

We have a lot of theories on hand, we dont always use every one of them every day, but when we want to know something about the nature of a problem we have to turn back to them to know for sure why something is happening the way it is.

Granted many people will not be interested in the basic theories of nature, but when questions like the ones i've heard in this thread so far come up, i feel that the askee has the kind of mind that wants to know the whole story, not just part of it.  We sometimes dont want formulas thrown at us.  Yes, Biot Savart is a sort of formula too, but it's a more basic one that digs right to the root of nature.  If you want to find a derivation for that though, you could do that too, and after all, you are willing to accept Maxwell as a basic concept of nature even though that one has a few open ends yet due to Einstein.

In the end, we all have our level of perfection that we want to strive for.  Everything is an approximation to some extent, but sometimes we want to get as close to nature as possible.  The Biot Savart Law is a beautiful way of looking at the nature of magnetics, it would be a shame to miss out on it for someone who really wants to understand.

[MrAl]

The 'risk' of saturation...
If you design it right you should not encounter this problem, right?  Or did you find something that became a problem?
If your inductor volt seconds are limited, it will not saturate.  It's not really about current except when there is a DC current present as well as any AC.  When there is DC also present, then yes, the DC could cause saturation, but that's something you include in the calculations for the inductance.  Overloading must be avoided by current monitoring, if an overload of some type is expected, which is almost always a consideration.

Depending on the mode of the converter, there could be little DC current (discontinuous conduction mode) or a lot of DC current (continuous conduction mode).  It can be a bit hard to tel, because your converter may go between the two as instantaneous line voltage changes.

The mechanism for saturation is that too much magnetic field strength H (Amps per metre) is applied to the (gapped) core set.
\Sigma NI = \Sigma H l

Where:
  N = # of turns
  I = current (A)
  H = magnetic field strength (A/m)
  l = path length (m)
Note: the \Sigma symbols are there to handle multiple windings (e.g. transformers and coupled inductors), and can be ignored for a simple inductor.


Then magnetic flux density B can be worked out:

B = \mu H
\mu = \mu_r * \mu_0

Where:
  B = magnetic flux density (T)
  \mu_r = relative permeability (see core material datasheet or 1 for air/plastic)
  \mu_0 = permeability of free space (4 *pi * 10^-7 H/m)

I forget how to handle the air gap, though.  I think you set the B values in the core and air to be equal, but I'm not certain. I could look that up tonight...

Anyhow, we see that the easiest way to look at saturation of an inductor is simply to check the peak current.  (Transformers are quite different!). As it happens, a lot of current mode control schemes are designed to immediately switch off the power transistor if the current reaches a peak value.  This combination is great, because it means that with a little care on the controls side you can be very confident about not saturating your inductor :-).

Hi,

What i was saying basically was that if you are designing the converter and you dont know if it will saturate, then you dont know if your design will work in the real world, and that sounds a little flakie to me.

[MrAl]

Hi,

I can help a little here.

First, inductance books vary quite a bit as you found out.  They dont all use the same units either, and because of that nagging 4*pi*10^=7 factor sometimes they dont even include any of that and work in some other units and other times they just use u0, and still other times they use u/u0 for when they want to include relative permeability, and that's just the start of things to go wrong :-)

The best bet to combat the above and similar is to home in on a formula you like and try to find another one even if in different units, so you can compare results.  Ideally, the results should come out identical within several decimal places, and even better as the ratio of two integers when possible.  If they come out close but not exact, then you might be ok but you're better off trying to find out why there is a variation even if small, and possibly by simply trying some other parameter values.  This allows you to be more sure of the result you need.

The worst part about this is that when they hand you numbers and formulas they dont always show the derivation and sometimes that is because it is very long and complicated and takes some mathematical tricks to get an answer.  Without that derivation though you have no idea what is going on, so it makes you feel kind of lost.
What you could do is either find a better book or start to do some derivations of your own.  To that end, you can start with the Biot Savart Law.  That's the basic law behind a lot of magnetic calculations that can come out exact.  There is always the practicality of it all though, which really makes everything we do outside of pure theory just an approximation, but at least then you can figure out why.

I spent all day homogenizing the equations from all of the books to use u₀ properly.

Biot Savart Law is EXACTLY what I've been looking for. I just didn't know how to find it.

To compute inductance behavior in a converter you have to know the voltage across it for all times, at least the maximum, and the expected maximum 'on' time.  If you know those two you can get pretty far because an important aspect of that is in terms of volt seconds.  In short, if you have more volt seconds than your winding can handle, it saturates.
The max sometimes comes from normal operating conditions but if you dont have a slow start mechanism in place then you end up with more volt seconds than you expected, so you really need a slow start mechanism for any reasonably sized converter.
If you vary the input voltage you need to consider the maximum input voltage and it's relation to the output voltage when the inductor topology puts it in series with the input and output.  But again, this will bring in some form of slow start so the volt seconds can be limited during start up.

The good book I found(in SI units), actually has a number of equations precisely for boost inductors, covering;  I_L, V_L, D (which I knew already but forgot due to the long break  ), ripple current, I_L(RMS), V_L(RMS), L, Apparent Frequency and I already know how to calculate N from L then augment the value by the square root of u_m/u_e, based on: u_m/1+u_m(l_g/lm)

You can estimate the fringe flux if you want too, and there are several articles online.  Here is one i found:
http://www.encyclopedia-magnetica.com/doku.php/flux_fringing
There may be more information on that Magnetics Inc site too.
BTW you can always measure the characteristics of your inductor, and change the gap as needed during that test.

I've actually seen that, that article it was very helpful. Is fringe flux factor a big deal at all? Because you and another person that I asked about it don't seem to give it much importance, but the books make it seem pretty nasty.

The 'risk' of saturation...
If you design it right you should not encounter this problem, right?  Or did you find something that became a problem?
If your inductor volt seconds are limited, it will not saturate.  It's not really about current except when there is a DC current present as well as any AC.  When there is DC also present, then yes, the DC could cause saturation, but that's something you include in the calculations for the inductance.  Overloading must be avoided by current monitoring, if an overload of some type is expected, which is almost always a consideration.

I haven't built the inductors yet, I want to get the equations all correct first. I don't want to short out my new monitor, lol.

Not sure what you are asking about the MOSFETs here, but if you are doing a full bridge then the voltage ratings of each transistor must be higher than the expected max voltage across each transistor.  It's that simple.  Of course if you have significant spikes then you need to snub them, and possibly go up on the transistor rating.
If you intend to output say 100v peak to the primary, then you should expect 100v peak, then add some head room for spikes and use snubbers.  When a lower transistor turns on it's about 0v, but when the diode conducts it could be -2v, and when an upper turns on it's going to put 100v peak on one transformer lead, so you can figure maybe 102v if you like to be more accurate, but really we'd have to model this to get batter estimates.  So you see using 120v MOSFET's would probably do the job unless we had significant spikes, and then maybe we would go up a little and improve the snubber.
After all is said and done however, the unit must be tested.  This comes from input a DC voltage (or whatever you are going to input) and turning it up slowly and watching for known problems such as saturation and excessive spike amplitude.  There's almost no way around this phase of the project.

My guess was the MOSFET voltage should be more or less double the V_DS(in)

The reason for the MOSFET/Output voltage question is because I know it's probably better to boost to about 400V, but I know there's less current stress on the transformer if the input voltage is higher.

Let me explain the project to you in detail:
I'm building a DC-AC-DC converter with 2 PFC inductors ,using 2 mosfets(FCH060N80), boosting to 400VDC (or 800V with 1200V 45A SiC MOSFETS to reduce current stress on the transformer), operating in opposite phase of each other. The total output power of the PFC stage will be 1kW(500W output from each inductor), powering 4 FCH023N65S3_F155 MOSFETs, providing an alternating square wave with a deadtime to avoid shorting. I've chosen this topology to maximize on core useage. The secondary winding will provide a rectified output voltage of 100VDC(I'll probably add a tertiary 12V for cooling fans), buck-boost output regulator to enable a variable output voltage. Size or part count isn't of primary concern, efficiency of operation is (I have 100s of kickass MOSFETS, among other components). I may program an MCU to control the FETS to monitor voltage and current, but I believe this can be achieved with linear TTL logic(more fun that way).

I know I can get a lot more output power from the converter I've described, however I only have 20A available in total (unless I go in the kitchen where there's 30A) and I don't have a sufficient load to test the device above 1.4kW however I will add the appropriate components and windings to increase output power as needed.

I have to ask you a question now too, and that is, did you ever design and build a buck circuit?
A buck circuit is a great way to test an inductor that is to be used for a switcher.

No, but what actually got me into electrical engineering were boost converters. I bought a few cheap $5 converters, tied them to the 12V rail of an ATX PSU and after a few days of being hooked they kept blowing the MOSFET when I applied a 30v 100w load (COB LED) to them, so I took one apart, re-wrapped the toroid with some homemade litz wire and center tapped it to take the stress off of the MOSFET(which worked very well,lol), what I called a flyback autotransformer at the time, and made a freaky boost converter that worked amazingly well, being I totally did it spatially, no inductor calculation at all. All I calculated was the winding ratio between the primary(lower) and secondary(upper=2*lower)

Look at this crazy ass thing :

I had that thing running as lamp in here for months(MOSFET heatsink stayed slightly warm the whole time) until it fell off of the table and the heatsink fan on the LED shattered (never bothered to reassemble it).
I've learned so much since then. The crazy thing is that it actually did what I expected. At a 50% on time it was 36V(40V no load) because the upper portion of the inductor had twice as many windings as the part that I was exciting. I was able to get that thing up to 45V before its little 10A breaker button would pop. It killed a couple of red COBs but the white one could handle excessive voltage with no problem. After that thing I was hooked on power electronics.

Hello again,

Yes i forgot to say sorry to hear about your monitor problem.  Maybe you should find a way to isolate that from whatever else you are doing.  I hope it was not too expensive.

The fringe flux is a contributor to lower efficiency.  That's the main concern in modern converters i think because the goal today is to get as high as possible.  You do have to realize though that what you are doing is partly academic in nature, as often we dont have the luxury to pick and choose everything we can use.  Sometimes we have to start with whatever is already on hand or accepted as standard and go with that.
For example, a company may have purchased a large number of EI laminations of a certain size, so rather than order a bunch more of a certain more optimum size we have to make use of what we already have.  It  sounds like you are doing this too i think  to some extent.
If we had our choice and if we could find a suitable core of course, we would go with a lower mu core.  A larger gap produces more fringe, so going with a smaller gap would be more efficient, but then we might not be able to get the right inductance and DC current handling we need, because really we need a bigger gap with a certain fixed mu core.  But what about using 2 gaps.  We can use two smaller gaps in the same core and thus get smaller gaps and less fringing, but why stop there.  What if we go to N smaller 'g' gaps, where g=Gap/N.  That's what a distributed gap core looks like, and so that would help efficiency.
But if we are stuck with a core we have on hand, then we are very limited to what we can do.  If the inductance is too high and/or DC current causes saturation, then all we can do is add a gap and add more turns.  If we can create more gaps, then we can go with the g=G/N idea, but that's not usually easy to do either.  So we usually end up with one or two gaps and that's about the end of it.

So your optimum gap is the smallest gap you can get away with and still get the other requirements you need.

The gap isnt really ignored, it's just hard to deal with when you dont have the free choice of core metal to use.

Before ASIC's and uC's came about, TTL was one way to make commercial converter control circuits.  The gov purchased many of these things, maybe because the reliability is so decent.

As a final note, gaps should be ground to high precision.  What we dont want is a partial core saturation which effectively lowers the core permeability which in turn could cause a higher current in the winding.

[hanzdolo30@gmail.com]

Hi, It's good to hear from you again,

Hello again,

Yes i forgot to say sorry to hear about your monitor problem.  Maybe you should find a way to isolate that from whatever else you are doing.  I hope it was not too expensive.

The monitor was a rather inexpensive SEIKI, but there was no dynamic image processing, so it served my purpose well (nice clear text). It was an irritating loss, because I'm a developer/IT professional by trade (I've been coding proficiently since I was 11, in 1988 when you could write ASM code into RAM with DEBUG and save the memory segments to disk as a .COM file) and I'd purchased it to have a higher resolution, in order to enable me work on multiple modules simultaneously. The replacement I'm using now is a Toshiba with some sort of dynamic image processor with poorly written firmware, better suited to be used exclusively as a television(slightly blurred and discolored text). However, when life serves lemons, one must make lemonade!   

At the time I was unaware that a short in my little startup circuit, (which was completely my fault because I allowed the legs of a voltage divider to cross each other, because I didn't cut the leads) could cause a feedback current that would affect other devices on the AC end of the rectifier (which was completely undamaged after the incident).

Now I use a separate power strip with surge suppressor on a mains line that doesn't have anything important attached to it.

The fringe flux is a contributor to lower efficiency.  That's the main concern in modern converters i think because the goal today is to get as high as possible.  You do have to realize though that what you are doing is partly academic in nature, as often we dont have the luxury to pick and choose everything we can use.  Sometimes we have to start with whatever is already on hand or accepted as standard and go with that.
For example, a company may have purchased a large number of EI laminations of a certain size, so rather than order a bunch more of a certain more optimum size we have to make use of what we already have.  It  sounds like you are doing this too i think  to some extent.
If we had our choice and if we could find a suitable core of course, we would go with a lower mu core.  A larger gap produces more fringe, so going with a smaller gap would be more efficient, but then we might not be able to get the right inductance and DC current handling we need, because really we need a bigger gap with a certain fixed mu core.  But what about using 2 gaps.  We can use two smaller gaps in the same core and thus get smaller gaps and less fringing, but why stop there.  What if we go to N smaller 'g' gaps, where g=Gap/N.  That's what a distributed gap core looks like, and so that would help efficiency.
But if we are stuck with a core we have on hand, then we are very limited to what we can do.  If the inductance is too high and/or DC current causes saturation, then all we can do is add a gap and add more turns.  If we can create more gaps, then we can go with the g=G/N idea, but that's not usually easy to do either.  So we usually end up with one or two gaps and that's about the end of it.

So your optimum gap is the smallest gap you can get away with and still get the other requirements you need.

The gap isnt really ignored, it's just hard to deal with when you dont have the free choice of core metal to use.

When I initially looked at gapped cores it seemed a bit counterintuitive, being the fringe flux is actually feeding MMF back into the coil creating what would appear to be an unwanted "mutual inductance" at the gap area. Which is why I initially asked you if it was a good idea to create a bridge between the gap and the at the residing region of the gap, on the bobbin, proportional to the distance of the flux radius, to avoid the undesirable effect.

I may take the i(t)  (current as a function of time) approach, where tao=L/R, where 5tao is the full charge time, and use Euler's Method to time the charge of an inductor (easily calculated and implemented using RC timing and a number of other factors, I'm sure you're aware of so I won't bore you with the details). If the u_e is calculated correctly the slope of the B-H curve can be made perfect. It's all about the gap and how the nasty fringe flux is handled, from what I can gather.

I'm quite aware of the academics involved. Freelance IT work is dead as fried chicken (which is where the $ used to be),lol, and I'm not enjoying development as much as I used to, due to the constant change in "climate", due to the constant emergence of new API's (we used to call them libraries ) and the need to use them because; "It's the new cool thing everyone wants!", which is why I'm turning in the direction of engineering. My long term is to pursue a PhD in applied physics (planning to start this fall). 

Before ASIC's and uC's came about, TTL was one way to make commercial converter control circuits.  The gov purchased many of these things, maybe because the reliability is so decent.

Oh I know, I'm 41, I grew up in the 80's and 90's and RadioShack was my favorite place to be (I eventually ended up working there, being I knew the manager so well). TTL circuits are actually quite reliable, as they are not susceptible to ESD damage, due to the lack of CMOS transistors on the die of the IC's. Which is why I'm leaning in that direction when it comes to power converter control circuitry. I know I can write a kickass advanced control scheme into one of those ARM Cortex M7 chips, but what if there's a nasty power surge? DEAD PSU that's what. Which is probably why they're usually placed on daughter boards, for easy replacement (I've seen this in some of the better PSU's that I've dissected).

As a final note, gaps should be ground to high precision.  What we dont want is a partial core saturation which effectively lowers the core permeability which in turn could cause a higher current in the winding.

Have a couple of shattered E cores that I've practiced grinding down with a diamond file set. I've noticed that the best approach is to measure it with the caliper, marking the core with a white pencil, just below the point where I would want to stop, filing it in a circular motion using decrementing grits, to get it within +/- .01mm of accuracy. It takes a while but, I carve wood and wax sculptures, so I have the patience for it.

I would Like to add, This has got to be the BEST forum I've used on electrical engineering. Because of the members . Everyone here has been friendly and knowledgeable, though I've noticed everyone seems to have their own style of doing things, which is cool, being a non polarized thinker, I can merge the knowledge. The only complaint I have (because I am a developer and I could rewrite this HTML editor within a week and that's an overestimation), is the inability to use some sort of script to lay out the mathematics properly(like MathJax). I mean this is an engineering forum and we use complex equations all the time.
When I lay out a formula it's like talking to Excel(which were not in the 80's anymore and we've been in the age of GUI for over 2 decades, so Microsoft should have improved on it by now ). I just had an epiphany, I'm going to start writing my own spreadsheet program properly suited for scientific mathematics. If one exists I would love to know. I can't seem to find one.

[hanzdolo30@gmail.com]

Hi Tim, Great to hear from you again,
 

I would suggest not worrying too much about the nitty-gritty.  Biot-Savart is great for physics problems, but you shouldn't need it in electronics problems.

The biggest drawbacks are:
1. There's no good way to integrate over the surface of a winding.
2. There is better symmetry to shortcut these problems with.

#1 is particularly complicated by the fact that the current doesn't flow uniformly through the wire, and the wire is not transparent to magnetic fields.  It would be easy enough to integrate along, say, a helical path, but wires are not infinitesimally thin; if you add a cross-section to it, now you're assuming the current is distributed somehow over that cross-section (which it isn't).

It's another way of saying: that one particular equation is magnetostatic.  But conductors at AC have resistance, so the EMF (from self-induction, causing skin effect; and the total induction, that gives us circuit inductance) is nonzero.

It's important to understand how this stuff works, but it's devilishly hard to work a real problem with.

As for symmetry, Maxwell's got you covered: for an infinitesimal cube, with field lines parallel and perpendicular to the sides, all the equations are simple proportions.  Flux is flux density times area; EMF is loop area times flux change; field intensity is current over path length; and flux density is field intensity times permeability.  (Clearly, if EMF and current have a relationship -- as in a resistive material -- you need to add extra terms to represent that.)

Nice thing about magnetic cores: since the flux is almost all (99%+) confined to a core, we don't much care that the core is itself going around some kind of path, we just care that it has length and area -- we can consider it as one rectangular chunk, where the field lines are parallel and perpendicular to the sides!

Just as we don't need to integrate over SMPS waveforms if they are straight-sided, we don't need to integrate over the core if it behaves straight-sided.

So these "square wave" or "square block" symmetries help out a lot!

So for analyzing a core, it's enough to use Maxwell's laws directly, where each law is simply a ratio.  We don't need to do vector calculus, because we've applied geometry that gets rid of all the nasty dimensionality and leaves us with scalar proportions (width and height, magnitude and sign).  We don't need any constants, no random pi's and such, only bare fundamental constants; namely, B = mu * H.

For the sake of this and other DC-DC/DC-AC conversion projects, in order to expedite the process, I totally agree that Maxwell's Equations are sufficient, but I do want to learn all about the Biot Savart Law for the sake of knowledge. Which is why I was so happy to find out about them, as it provides a lot of the details. 

Fringing flux is a problem for two reasons:
1. The intense field strength, and its nonuniformity (that is, it doesn't behave as a simple parallel rectangular region), induces eddy currents in nearby wires, causing heating (proximity effect).
2. The fringe location acts like a winding of the same size, which transmits fields out into space.  Thus, it's a source of emissions.  You often see transformers with a copper tape wrapped around the outside (notice it goes around the whole core, not just the winding where it would be a shorted turn!), which acts to short out (shield) this mode.  In other words, it keeps the flux inside the core, where it belongs.

Related:
http://power.thayer.dartmouth.edu/shapeopt_intro.html
This calculates where to optimally place wire, to avoid those losses.  Cool!

Tim

I knew that nasty fringe flux was an issue , because just looking at it in an illustration of how it exits and re-enters the core, it's obvious that it will cause a secondary(mutual) feedback inductance in the primary coil(its source), which would obviously cause inductance issues, as it is pumping current back into itself . I know the gap is essentially an MMF resistor, so like a resistor the gap must dissipate the energy in some form(like resistors do heat).  I was trying to reason how to shield against it, I thought about maybe wrapping some copper foil around the region of the gap on the bobbin, prior to applying the litz wire, to catch it, but quickly realized that it would act as a shorted secondary winding like you mentioned above, so I never bothered asking if that one would work. So my second Idea was to simply calculate the distance(blast radius) of the MMF radiation produced by the gap and put a few layers of teflon over it of the same thickness. Would that work? I know it must change A_e at the gapped region so it would change inductance values proportionally but wouldn't there be a way to calculate that proportionality?

I've only had a chance to briefly go over that link you sent. I just printed the Optimal Core Dimensional Ratios for Minimizing Winding Loss in High-Frequency Gapped-Inductor Windings, PDF.
I'm gonna go read it right after I'm done with this post.

Thank you, you always drop some really good Gems on me! 
BTW, did you see the schematic of my first boost converter ?
It's the BJT Astable one I was telling you about.

[T3sl4co1l]

I was trying to reason how to shield against it, I thought about maybe wrapping some copper foil around the region of the gap on the bobbin, prior to applying the litz wire, to catch it, but quickly realized that it would act as a shorted secondary winding like you mentioned above, so I never bothered asking if that one would work. So my second Idea was to simply calculate the distance(blast radius) of the MMF radiation produced by the gap and put a few layers of teflon over it of the same thickness. Would that work? I know it must change A_e at the gapped region so it would change inductance values proportionally but wouldn't there be a way to calculate that proportionality?

It could be shielded by wrapping the core with solid copper foil, that is insulated so it doesn't form a shorted turn where it overlaps.  This way, the field fringes outward at the gap, smacks into the foil, induces current*, and pushes the field lines back towards the gap.  What you would see in a plot is, a couple field lines push out and penetrate into the foil (a few skin depths; if the foil is thinner than the skin depths, a noticeable amount will leak through, still going on to touch the wires; if thicker, it will bend back and follow along through the foil, until coming out at the other core face).  The rest of the field lines are confined within the core gap area, but are particularly dense around the corners (because there's a "pressure" for those lines to spread out -- hence fringing in the first place -- and the copper strip packs them back in around the edges).

*Way more total current than would be induced in loose wires, but the same resistivity (copper is copper), so you can guess what this will do to the losses...

So as is often the case -- better to let it happen, than to try to fight it.  Moving the wire away from the gap by adding a large spacer around the middle of the bobbin, or using Litz (that simply has a smaller eddy current cross section), is the way to go.

The fringe is about as wide (distance from side face of core to nearest wire) as twice the gap, and about as tall.  That is, keep wire away from about one gap-length above and below the gap, and two gap-lengths beside.  Or maybe three.  It's inverse-cube in any case, so it drops off quickly.

Tim

[MrAl]

Hi, It's good to hear from you again,

Hello again,

Yes i forgot to say sorry to hear about your monitor problem.  Maybe you should find a way to isolate that from whatever else you are doing.  I hope it was not too expensive.

The monitor was a rather inexpensive SEIKI, but there was no dynamic image processing, so it served my purpose well (nice clear text). It was an irritating loss, because I'm a developer/IT professional by trade (I've been coding proficiently since I was 11, in 1988 when you could write ASM code into RAM with DEBUG and save the memory segments to disk as a .COM file) and I'd purchased it to have a higher resolution, in order to enable me work on multiple modules simultaneously. The replacement I'm using now is a Toshiba with some sort of dynamic image processor with poorly written firmware, better suited to be used exclusively as a television(slightly blurred and discolored text). However, when life serves lemons, one must make lemonade!   

At the time I was unaware that a short in my little startup circuit, (which was completely my fault because I allowed the legs of a voltage divider to cross each other, because I didn't cut the leads) could cause a feedback current that would affect other devices on the AC end of the rectifier (which was completely undamaged after the incident).

Now I use a separate power strip with surge suppressor on a mains line that doesn't have anything important attached to it.

The fringe flux is a contributor to lower efficiency.  That's the main concern in modern converters i think because the goal today is to get as high as possible.  You do have to realize though that what you are doing is partly academic in nature, as often we dont have the luxury to pick and choose everything we can use.  Sometimes we have to start with whatever is already on hand or accepted as standard and go with that.
For example, a company may have purchased a large number of EI laminations of a certain size, so rather than order a bunch more of a certain more optimum size we have to make use of what we already have.  It  sounds like you are doing this too i think  to some extent.
If we had our choice and if we could find a suitable core of course, we would go with a lower mu core.  A larger gap produces more fringe, so going with a smaller gap would be more efficient, but then we might not be able to get the right inductance and DC current handling we need, because really we need a bigger gap with a certain fixed mu core.  But what about using 2 gaps.  We can use two smaller gaps in the same core and thus get smaller gaps and less fringing, but why stop there.  What if we go to N smaller 'g' gaps, where g=Gap/N.  That's what a distributed gap core looks like, and so that would help efficiency.
But if we are stuck with a core we have on hand, then we are very limited to what we can do.  If the inductance is too high and/or DC current causes saturation, then all we can do is add a gap and add more turns.  If we can create more gaps, then we can go with the g=G/N idea, but that's not usually easy to do either.  So we usually end up with one or two gaps and that's about the end of it.

So your optimum gap is the smallest gap you can get away with and still get the other requirements you need.

The gap isnt really ignored, it's just hard to deal with when you dont have the free choice of core metal to use.

When I initially looked at gapped cores it seemed a bit counterintuitive, being the fringe flux is actually feeding MMF back into the coil creating what would appear to be an unwanted "mutual inductance" at the gap area. Which is why I initially asked you if it was a good idea to create a bridge between the gap and the at the residing region of the gap, on the bobbin, proportional to the distance of the flux radius, to avoid the undesirable effect.

I may take the i(t)  (current as a function of time) approach, where tao=L/R, where 5tao is the full charge time, and use Euler's Method to time the charge of an inductor (easily calculated and implemented using RC timing and a number of other factors, I'm sure you're aware of so I won't bore you with the details). If the u_e is calculated correctly the slope of the B-H curve can be made perfect. It's all about the gap and how the nasty fringe flux is handled, from what I can gather.

I'm quite aware of the academics involved. Freelance IT work is dead as fried chicken (which is where the $ used to be),lol, and I'm not enjoying development as much as I used to, due to the constant change in "climate", due to the constant emergence of new API's (we used to call them libraries ) and the need to use them because; "It's the new cool thing everyone wants!", which is why I'm turning in the direction of engineering. My long term is to pursue a PhD in applied physics (planning to start this fall). 

Before ASIC's and uC's came about, TTL was one way to make commercial converter control circuits.  The gov purchased many of these things, maybe because the reliability is so decent.

Oh I know, I'm 41, I grew up in the 80's and 90's and RadioShack was my favorite place to be (I eventually ended up working there, being I knew the manager so well). TTL circuits are actually quite reliable, as they are not susceptible to ESD damage, due to the lack of CMOS transistors on the die of the IC's. Which is why I'm leaning in that direction when it comes to power converter control circuitry. I know I can write a kickass advanced control scheme into one of those ARM Cortex M7 chips, but what if there's a nasty power surge? DEAD PSU that's what. Which is probably why they're usually placed on daughter boards, for easy replacement (I've seen this in some of the better PSU's that I've dissected).

As a final note, gaps should be ground to high precision.  What we dont want is a partial core saturation which effectively lowers the core permeability which in turn could cause a higher current in the winding.

Have a couple of shattered E cores that I've practiced grinding down with a diamond file set. I've noticed that the best approach is to measure it with the caliper, marking the core with a white pencil, just below the point where I would want to stop, filing it in a circular motion using decrementing grits, to get it within +/- .01mm of accuracy. It takes a while but, I carve wood and wax sculptures, so I have the patience for it.

I would Like to add, This has got to be the BEST forum I've used on electrical engineering. Because of the members . Everyone here has been friendly and knowledgeable, though I've noticed everyone seems to have their own style of doing things, which is cool, being a non polarized thinker, I can merge the knowledge. The only complaint I have (because I am a developer and I could rewrite this HTML editor within a week and that's an overestimation), is the inability to use some sort of script to lay out the mathematics properly(like MathJax). I mean this is an engineering forum and we use complex equations all the time.
When I lay out a formula it's like talking to Excel(which were not in the 80's anymore and we've been in the age of GUI for over 2 decades, so Microsoft should have improved on it by now ). I just had an epiphany, I'm going to start writing my own spreadsheet program properly suited for scientific mathematics. If one exists I would love to know. I can't seem to find one.

Hi,

The way i handle math format differences is i write programs to convert between styles.  That's the only way i have found so far to use whatever program i want to use when i want to use it, or else i just cant use some programs.

Since i grew up on text however, i have no problem reading text equations, and i usually prefer that format because then i can copy and paste it into a program and i am up and calculating within minutes.

For example, i use one interpreted language for my home made formulas (as well as others really) that does not know how to handle the exponent symbol commonly used for exponents like in 2^9 which represents 2 to the ninth power.  Since i got tired of changing all of those occurrences in a formula by hand, i started writing small easy to modify programs, one of which converts a^3 for example into a*a*a and if there is more than one occurrence it makes a constant a3=a*a*a and then inserts a3 into any place that formerly had a^3 in it.
There are so many variations though that i ended up creating several small programs, another one that removes "abs" from the equations, because sometimes we dont need that when dealing with all positive numbers, yet a solution may come out with several of those in it.
The simplest way i have found so far to facilitate this kind of conversion is to have the program interface with the clipboard.  Then it is just a copy, run, paste, and the new equations are pasted into the required app.
You may have some other good ideas here too since you do a lot of programming.

[hanzdolo30@gmail.com]

It could be shielded by wrapping the core with solid copper foil, that is insulated so it doesn't form a shorted turn where it overlaps.  This way, the field fringes outward at the gap, smacks into the foil, induces current*, and pushes the field lines back towards the gap.  What you would see in a plot is, a couple field lines push out and penetrate into the foil (a few skin depths; if the foil is thinner than the skin depths, a noticeable amount will leak through, still going on to touch the wires; if thicker, it will bend back and follow along through the foil, until coming out at the other core face).  The rest of the field lines are confined within the core gap area, but are particularly dense around the corners (because there's a "pressure" for those lines to spread out -- hence fringing in the first place -- and the copper strip packs them back in around the edges).

*Way more total current than would be induced in loose wires, but the same resistivity (copper is copper), so you can guess what this will do to the losses...

So as is often the case -- better to let it happen, than to try to fight it.  Moving the wire away from the gap by adding a large spacer around the middle of the bobbin, or using Litz (that simply has a smaller eddy current cross section), is the way to go.

The fringe is about as wide (distance from side face of core to nearest wire) as twice the gap, and about as tall.  That is, keep wire away from about one gap-length above and below the gap, and two gap-lengths beside.  Or maybe three.  It's inverse-cube in any case, so it drops off quickly.

Tim

Hi Tim, I hope your day went well,

I'm actually amazed that I was remotely correct about how to shield the fringe flux radiation. Though I am aware of the existence of a Faraday's cage, and came the foil assumption based on that. However, I did think distancing the air gap from the litz wire would actually work.

There's one thing I'm not clear on. Being there will be a slight increase in the coil area where the spacer exists, wouldn't that increase the A_E at the spacer, affecting the inductance value?

I just want to be sure I'm understanding this correctly. Are you telling me that If I use Litz I don't have to use a spacer ?
I was always planning on using Litz Wire for this project. Being I'll be operating the inductor at 100kHz, I figured it was a prerequisite, due to skin effect.

Thanks .

Gilbert

[hanzdolo30@gmail.com]

Hi,

The way i handle math format differences is i write programs to convert between styles.  That's the only way i have found so far to use whatever program i want to use when i want to use it, or else i just cant use some programs.

Since i grew up on text however, i have no problem reading text equations, and i usually prefer that format because then i can copy and paste it into a program and i am up and calculating within minutes.

For example, i use one interpreted language for my home made formulas (as well as others really) that does not know how to handle the exponent symbol commonly used for exponents like in 2^9 which represents 2 to the ninth power.  Since i got tired of changing all of those occurrences in a formula by hand, i started writing small easy to modify programs, one of which converts a^3 for example into a*a*a and if there is more than one occurrence it makes a constant a3=a*a*a and then inserts a3 into any place that formerly had a^3 in it.
There are so many variations though that i ended up creating several small programs, another one that removes "abs" from the equations, because sometimes we dont need that when dealing with all positive numbers, yet a solution may come out with several of those in it.
The simplest way i have found so far to facilitate this kind of conversion is to have the program interface with the clipboard.  Then it is just a copy, run, paste, and the new equations are pasted into the required app.
You may have some other good ideas here too since you do a lot of programming.

Hi,

I've realized that writing a full blown spreadsheet would probably be overkill for now. So I'm going to take your approach and write a quick desktop application in C# to calculate equations for magnetics, to do things like, plot B-H (Hysteresis) curves, i(t), v(t), U(t), . It'll be a lot easier than dealing with cell numbers and long one line equations in excel. Honestly It'd be easier(on the eyes) to write a script in js and use canvas in an HTML file to plot graphs than to deal with Excel, lol.

 
Thanks

[T3sl4co1l]

I'm actually amazed that I was remotely correct about how to shield the fringe flux radiation. Though I am aware of the existence of a Faraday's cage, and came the foil assumption based on that. However, I did think distancing the air gap from the litz wire would actually work.

Yeah I'm not clear if you meant to wrap the gap, all the way around, with a solid (shorted) turn of foil -- if so, that's a big difference, but just that subtle change (leaving it overlapping but insulated) makes it workable.  (I mean, besides the problem that air gap fringing is just too vicious to shield.)

Placed between windings, this is also how an electrostatic shield is made.

Shields also enforce field uniformity, so they're useful in precision (balanced / bridge) transformers and RF.

There's one thing I'm not clear on. Being there will be a slight increase in the coil area where the spacer exists, wouldn't that increase the A_E at the spacer, affecting the inductance value?

Coil area doesn't matter: very little flux flows in the air beneath the coil.  That's the point of the core.

That said...

The fact that the fringing extends out from the core, means, yes, the effective area is increased at that point!

So you're again not entirely wrong!

The effect is that, as you increase gap, the Ae of the gap increases as well (basically by adding gap length to all sides; if you assume it remains a square, that makes the new area,
Ae' = Ae + 4 * (l_g * H + l_g^2)
Where H is the height of the core.  (Or similarly for a round core, changing H for D (diameter), and 4 for pi, so that the diameter is increased the same way.)

That means you need slightly more gap than calculated (from a linear method), because the gap has lower reluctance (more area) than you were expecting.  And that also means this is only important when the gap is relatively large compared to H, say 5% or more.

Even further into the weeds, we can contemplate the core geometry itself; the fringing flux affects the core as well, so that flux is concentrated on the edges of the core faces (recall how metal filings hang off the edges of a bar magnet!), which makes the corners saturate earlier.  The core face could perhaps be rounded off, which would make the fringing volume even bigger (a downside), but would have the effect of sharpening saturation (i.e., the whole core saturates at once, more nearly).

But that's more of a digression, than anything you'd do; it's much cheaper to simply get the next larger core size, than to try and wring every last mT out of a core.

I just want to be sure I'm understanding this correctly. Are you telling me that If I use Litz I don't have to use a spacer ?
I was always planning on using Litz Wire for this project. Being I'll be operating the inductor at 100kHz, I figured it was a prerequisite, due to skin effect.

Yes, though it depends on the Litz of course.  Regular Litz, for carrying currents in cables, let's say, won't work as well once it's inside a transformer: because of proximity effect.  Proximity effect is exaggerated when you have multiple layers carrying current in the same direction, so it's hardest on multilayer inductors.  Transformers (including coupled inductors like flyback transformers) can be wound with alternating primary and secondary layers, to minimize proximity effect.  (This is one advantage of forward converters: the primary and secondary currents oppose, so that a construction with alternating layers doesn't incur any extra proximity losses.  Flyback converters don't carry primary and secondary currents simultaneously, so the losses aren't reduced by current cancellation: but the reduction of leakage inductance is more critical for this, so the alternate construction is still important.)

Further, because the fringe field is intense, it might be that you need Litz made with several gauges finer strands, to avoid excessive loss near the fringe.

For sure, don't make a marginal design where the Litz is just barely good enough in the rest of the winding -- then the part deeper inside (subject to proximity effect or fringe eddy currents) will overheat.

But in general: absolutely, without a doubt, Litz is better, whether it's placed near the gap or not.  There may still be some advantage to avoiding the fringe area, but likely you'll be fine to put it there: don't worry.

You definitely have to worry if your strands are rather coarse.  Why coarse strands?  You'll find sometimes it's easier/cheaper to build a transformer with several individual strands, laid down in parallel, or twisted together.  Maybe you don't have Litz that size, or don't want to splurge for it.  An inductor designed for low current ripple will even be fine to make out of solid wire.  But if these materials are wound on a cut and gapped core, they may burn up near the gap, due to induced eddy currents (instead of load current).  This would be a situation where you'd add a spacer in the middle.

There are also good reasons to add a different kind of spacer, like this,



Which has more to do with reducing the capacitance between sections.  The insulation voltage also goes up.  These are commonly seen in high voltage transformers (including CCFL drivers).

And, apparently, whatever this transformer was doing.  It was removed from a CRT monitor, so a couple of those windings are mains voltage input (~350V, probably switched at ~140kHz -- flyback), a couple would've been ~100V output (for deflection and CRT driver), and a couple for logic supply (3.3 to 12V, say).

The single split bobbin design is also very common in mains transformers, where the huge isolation voltage is a plus, and the high leakage inductance is not-a-bug-it's-a-feature (crappier regulation, but reduced harmonics, fault current, and in the smaller sizes (< 20VA or so), tolerance of continuous short circuits).

Tim

[MrAl]

Hi,

The way i handle math format differences is i write programs to convert between styles.  That's the only way i have found so far to use whatever program i want to use when i want to use it, or else i just cant use some programs.

Since i grew up on text however, i have no problem reading text equations, and i usually prefer that format because then i can copy and paste it into a program and i am up and calculating within minutes.

For example, i use one interpreted language for my home made formulas (as well as others really) that does not know how to handle the exponent symbol commonly used for exponents like in 2^9 which represents 2 to the ninth power.  Since i got tired of changing all of those occurrences in a formula by hand, i started writing small easy to modify programs, one of which converts a^3 for example into a*a*a and if there is more than one occurrence it makes a constant a3=a*a*a and then inserts a3 into any place that formerly had a^3 in it.
There are so many variations though that i ended up creating several small programs, another one that removes "abs" from the equations, because sometimes we dont need that when dealing with all positive numbers, yet a solution may come out with several of those in it.
The simplest way i have found so far to facilitate this kind of conversion is to have the program interface with the clipboard.  Then it is just a copy, run, paste, and the new equations are pasted into the required app.
You may have some other good ideas here too since you do a lot of programming.

Hi,

I've realized that writing a full blown spreadsheet would probably be overkill for now. So I'm going to take your approach and write a quick desktop application in C# to calculate equations for magnetics, to do things like, plot B-H (Hysteresis) curves, i(t), v(t), U(t), . It'll be a lot easier than dealing with cell numbers and long one line equations in excel. Honestly It'd be easier(on the eyes) to write a script in js and use canvas in an HTML file to plot graphs than to deal with Excel, lol.

 
Thanks

Hi again,

Sounds good, i use C and C++ and an interpreted language but havent used C# yet.  I do mostly console applications for these tasks because they are so quick to write up, but i've gone to Windows applications for more versatile programs with interactive user interface and nice graphing.  The console programs go so quickly though that i usually run to that just to get something calculated where there will be a lot of repetitive calculations (funny integrals and the like).

What magnetic model are you using to do the BH curves?

[hanzdolo30@gmail.com]

Hi again,

Sounds good, i use C and C++ and an interpreted language but havent used C# yet.  I do mostly console applications for these tasks because they are so quick to write up, but i've gone to Windows applications for more versatile programs with interactive user interface and nice graphing.  The console programs go so quickly though that i usually run to that just to get something calculated where there will be a lot of repetitive calculations (funny integrals and the like).

Hi,

C# is nice if you're writing something where you'll need to spin up a quick UI, but for services, video processing or anything where managed code would slow you down, I prefer to go native with C++. Though in the case of services, if the task is light weight, you may want to consider using C#.

For console applications, as of last year, I tend to use nodeJS.  It's fast and if you're familiar with javascript, you'll know that objects are quite versatile.

Using node, I was able to put together a full featured MVC framework, similar to MVC.net, within about 45 days, using absolutely no foreign packages (I wrote all of the modules myself).

That was my introduction to node. I usually take on difficult professional level projects when learning something new (like power converters in EE, for example), rather than wasting time with little "Hello, world!" programs (In EE, I'm guessing blinky LED Circuits would be the equivalent).

I figure, why waste time, when I must learn everything necessary on the way (jump in and swim method). Otherwise, I get bored and lose interest.

What magnetic model are you using to do the BH curves?

I figured make it as broad as possible by creating a "MagneticCore" class with all of the usual core parameters, u_m, A_e, l_e, l_g,  etc., allowing me to just plug in values as needed.

[MrAl]

Hi again,

Sounds good, i use C and C++ and an interpreted language but havent used C# yet.  I do mostly console applications for these tasks because they are so quick to write up, but i've gone to Windows applications for more versatile programs with interactive user interface and nice graphing.  The console programs go so quickly though that i usually run to that just to get something calculated where there will be a lot of repetitive calculations (funny integrals and the like).

Hi,

C# is nice if you're writing something where you'll need to spin up a quick UI, but for services, video processing or anything where managed code would slow you down, I prefer to go native with C++. Though in the case of services, if the task is light weight, you may want to consider using C#.

For console applications, as of last year, I tend to use nodeJS.  It's fast and if you're familiar with javascript, you'll know that objects are quite versatile.

Using node, I was able to put together a full featured MVC framework, similar to MVC.net, within about 45 days, using absolutely no foreign packages (I wrote all of the modules myself).

That was my introduction to node. I usually take on difficult professional level projects when learning something new (like power converters in EE, for example), rather than wasting time with little "Hello, world!" programs (In EE, I'm guessing blinky LED Circuits would be the equivalent).

I figure, why waste time, when I must learn everything necessary on the way (jump in and swim method). Otherwise, I get bored and lose interest.

What magnetic model are you using to do the BH curves?

I figured make it as broad as possible by creating a "MagneticCore" class with all of the usual core parameters, u_m, A_e, l_e, l_g,  etc., allowing me to just plug in values as needed.

Hi,

Actually i was asking for the model you use for plotting the BH curve really.  Not sure if that is what you are doing now or not though.  That would be a formula or set of equations.

[hanzdolo30@gmail.com]

Yeah I'm not clear if you meant to wrap the gap, all the way around, with a solid (shorted) turn of foil -- if so, that's a big difference, but just that subtle change (leaving it overlapping but insulated) makes it workable.  (I mean, besides the problem that air gap fringing is just too vicious to shield.)

Placed between windings, this is also how an electrostatic shield is made.

Shields also enforce field uniformity, so they're useful in precision (balanced / bridge) transformers and RF.

I figured a shorted foil was not the way to go, because would be the same as starting up a flyback (coupled inductor) with a full short, . My idea was actually to run leads from the foil winding to capacitors, being caps are an AC ground, but I wasn't certain if that would've been a good solution.

Coil area doesn't matter: very little flux flows in the air beneath the coil.  That's the point of the core.

That said...

The fact that the fringing extends out from the core, means, yes, the effective area is increased at that point!

So you're again not entirely wrong!

The effect is that, as you increase gap, the Ae of the gap increases as well (basically by adding gap length to all sides; if you assume it remains a square, that makes the new area,
Ae' = Ae + 4 * (l_g * H + l_g^2)
Where H is the height of the core.  (Or similarly for a round core, changing H for D (diameter), and 4 for pi, so that the diameter is increased the same way.)

That means you need slightly more gap than calculated (from a linear method), because the gap has lower reluctance (more area) than you were expecting.  And that also means this is only important when the gap is relatively large compared to H, say 5% or more.

Even further into the weeds, we can contemplate the core geometry itself; the fringing flux affects the core as well, so that flux is concentrated on the edges of the core faces (recall how metal filings hang off the edges of a bar magnet!), which makes the corners saturate earlier.  The core face could perhaps be rounded off, which would make the fringing volume even bigger (a downside), but would have the effect of sharpening saturation (i.e., the whole core saturates at once, more nearly).

But that's more of a digression, than anything you'd do; it's much cheaper to simply get the next larger core size, than to try and wring every last mT out of a core.

Sounds like more trouble than it's worth.
Onward to Litz...

Yes, though it depends on the Litz of course.  Regular Litz, for carrying currents in cables, let's say, won't work as well once it's inside a transformer: because of proximity effect.  Proximity effect is exaggerated when you have multiple layers carrying current in the same direction, so it's hardest on multilayer inductors.  Transformers (including coupled inductors like flyback transformers) can be wound with alternating primary and secondary layers, to minimize proximity effect.  (This is one advantage of forward converters: the primary and secondary currents oppose, so that a construction with alternating layers doesn't incur any extra proximity losses.  Flyback converters don't carry primary and secondary currents simultaneously, so the losses aren't reduced by current cancellation: but the reduction of leakage inductance is more critical for this, so the alternate construction is still important.)

Further, because the fringe field is intense, it might be that you need Litz made with several gauges finer strands, to avoid excessive loss near the fringe.

For sure, don't make a marginal design where the Litz is just barely good enough in the rest of the winding -- then the part deeper inside (subject to proximity effect or fringe eddy currents) will overheat.

But in general: absolutely, without a doubt, Litz is better, whether it's placed near the gap or not.  There may still be some advantage to avoiding the fringe area, but likely you'll be fine to put it there: don't worry.

You definitely have to worry if your strands are rather coarse.  Why coarse strands?  You'll find sometimes it's easier/cheaper to build a transformer with several individual strands, laid down in parallel, or twisted together.  Maybe you don't have Litz that size, or don't want to splurge for it.  An inductor designed for low current ripple will even be fine to make out of solid wire.  But if these materials are wound on a cut and gapped core, they may burn up near the gap, due to induced eddy currents (instead of load current).  This would be a situation where you'd add a spacer in the middle.

There are also good reasons to add a different kind of spacer, like this,



Which has more to do with reducing the capacitance between sections.  The insulation voltage also goes up.  These are commonly seen in high voltage transformers (including CCFL drivers).

And, apparently, whatever this transformer was doing.  It was removed from a CRT monitor, so a couple of those windings are mains voltage input (~350V, probably switched at ~140kHz -- flyback), a couple would've been ~100V output (for deflection and CRT driver), and a couple for logic supply (3.3 to 12V, say).

The single split bobbin design is also very common in mains transformers, where the huge isolation voltage is a plus, and the high leakage inductance is not-a-bug-it's-a-feature (crappier regulation, but reduced harmonics, fault current, and in the smaller sizes (< 20VA or so), tolerance of continuous short circuits).

Tim

Hey Tim,

 I thought the gap spacer would be the perfect solution, but it seems that using litz is the way to go .

I'm still a little confused about the litz though, but I'm sure you'll get me straight after this question.

It seems you're telling me to use multiple gauges in the same wire or is it a matter of the number of conductors.?

With the low voltage converters, I usually use recycled toroids from old PSUs,   
I spin the litz on a couple of 9/0 fish hooks, one end stationary and the other slowly on my drill, adding as many conductors as necessary to handle the greatest potential current going through the inductor. Can I use the same method?

I'm planning on using 26 AWG being the max skin depth is 107kHz and my target frequency is 100kHz, but I do have about 30' of some premade 50/32 AWG litz I bought last year, should I use that instead?

[hanzdolo30@gmail.com]

Hi,

Actually i was asking for the model you use for plotting the BH curve really.  Not sure if that is what you are doing now or not though.  That would be a formula or set of equations.

Hello again,

Oh, I see, you were referring to the set of equations I was planning to use. I haven't started putting those together yet, but I was planning on using the equations provided to me by the textbooks.
Is that how I should be doing it?

I read that the hysteresis curve changes based on u_e (ideally, wanting a perfect slope from -B => B), where; u_e =  u_m/1+u_m(l_g/l_m) and the level of excitation of the inductor. I was planning on writing a program that'll provide me with the ideal, u_e, based on my desired power throughput, which (if I'm not mistaken) will be able to compute the necessary l_g, along with other necessary factors, then graphically plot the hysteresis curve. I.e., a program that'll compute the exact design parameters of an inductor that cannot be saturated based on the necessary power output, input & output voltages, core geometry and material properties.

Does that sound about right?
Please correct me where I'm wrong.

Thank you very much.

[MrAl]

Hi,

Actually i was asking for the model you use for plotting the BH curve really.  Not sure if that is what you are doing now or not though.  That would be a formula or set of equations.

Hello again,

Oh, I see, you were referring to the set of equations I was planning to use. I haven't started putting those together yet, but I was planning on using the equations provided to me by the textbooks.
Is that how I should be doing it?

I read that the hysteresis curve changes based on u_e (ideally, wanting a perfect slope from -B => B), where; u_e =  u_m/1+u_m(l_g/l_m) and the level of excitation of the inductor. I was planning on writing a program that'll provide me with the ideal, u_e, based on my desired power throughput, which (if I'm not mistaken) will be able to compute the necessary l_g, along with other necessary factors, then graphically plot the hysteresis curve. I.e., a program that'll compute the exact design parameters of an inductor that cannot be saturated based on the necessary power output, input & output voltages, core geometry and material properties.

Does that sound about right?
Please correct me where I'm wrong.

Thank you very much.

Hi,

Oh ok, i just thought you had some on hand that you could show here so i could take a look.  There are different equations used in magnetics so i always welcome the chance to look at what might be  new ones or something.
If you got them from a text book then i guess they include all the necessary features.  If you get around to showing one we could take a look.

[hanzdolo30@gmail.com]

Hi,

Oh ok, i just thought you had some on hand that you could show here so i could take a look.  There are different equations used in magnetics so i always welcome the chance to look at what might be  new ones or something.
If you got them from a text book then i guess they include all the necessary features.  If you get around to showing one we could take a look.

I didn't realize how rusty I was, it took me 4 hours to get back into the swing of things. I haven;t touched C# in about 4 Years. I was oblivious to the fact that C# uses Math.Exp(value) to handle exponents and apparently javascript does the same so I wasted a few hours (trying to be a perfectionist) trying either create a new operator or add the ability to the variable type like 3.toString() . No luck there so I guess, I'll have to go the long way;

Visual Basic has the ^ operator but I haven't touched VB in so many years, it was just weird being in there. I'm not sure if you're aware of this, but MS said they were going to  drop VB back in 2011
so it was much to my surprise when 2017 had VB in it's list of languages. Basic was my first language,but after working with braced languages, It's difficult to turn back. I may go Javascript. I can build all of the math into an object. The beauty of JS is that you can write functions into the objects. I'm trying to decide. I guess I'll have all three compute a core of the same value and see which one I can build the best UI for.
This is a controller from My Framework Just to give you an example:

API Engine

Controllers

Now that I’ve gotten to the part where I’ll have to write the controllers, I’ve found that the best approach to forming API methods is to form them the same as you would if you declared any other java script class with the exception that there should be guidelines followed in order to use the helper methods written in to the controller module.

Basic construction of a standard controller class should be as follows

//This is an example of a root folder controller
function defaultController() {
    // this is routing and security informatiom
    this.aliases = ['/'];
    this.authorize = true;
    this.methods = {
       
        index: {
            aliases: ['/', 'default', 'index'],
            GET: function () {
                returnView('default');
            }
        },
       
        about: {
            aliases: ['/', 'about', 'aboutus', 'about_us', 'about-us'],
            action: function () {
                returnView('about');
            }
        },
       
        contact: {    // aliases which call the get method
            aliases: ['contact', 'contactus', 'contact_us', 'contact-us'],
        // GET request sends the default
            GET: function () {
                returnView('contact');
            },
            PUT: function () {
                // Do things to send email from fields upon contact form submission
            }

        }

    }

}

//Controllers should pass url parameters by dividing them into Controller = Param[0], action = param[1], other named parameters will be defined by the user while being defined in an object local to the controller by name.

[T3sl4co1l]

I figured a shorted foil was not the way to go, because would be the same as starting up a flyback (coupled inductor) with a full short, . My idea was actually to run leads from the foil winding to capacitors, being caps are an AC ground, but I wasn't certain if that would've been a good solution.

Electrostatic grounding has no effect on the magnetic field, of course.

I'm still a little confused about the litz though, but I'm sure you'll get me straight after this question.

It seems you're telling me to use multiple gauges in the same wire or is it a matter of the number of conductors.?

Hmm, I suppose you could use multiple sizes in the same wire, but that would be annoying.  And hard to arrange optimally, because you'd want the finer strands in the middle.  But you want the strands always circulating in and out, hence the woven construction of Litz, so that's impossible!

Just a matter of the number of strands, what size strand you start with in the construction of an equivalent size cable.

With the low voltage converters, I usually use recycled toroids from old PSUs,

Hmm, ferrite or powdered iron?  You usually see the latter (yellow-white or blue-green colored), which is very lossy material (by 100kHz, it's practically a better resistor than inductor!).

I spin the litz on a couple of 9/0 fish hooks, one end stationary and the other slowly on my drill, adding as many conductors as necessary to handle the greatest potential current going through the inductor. Can I use the same method?

This is a poor way to do it, because it strains the wires, and doesn't make a tight cable anyway (it springs back a little).

Unfortunately, the proper way to do it isn't very easy (you need one spool for each strand, and the spools need to rotate freely, so as the wire is drawn off and twisted into a bundle, the spools rotate at the same rate).

If you only need a few meters, it's not terrifically annoying to weave or twist loose strands together in the same way (no spool needed if they're loose... just don't let them tangle up..).  Otherwise... it's one of those cases were "poor" is about all that's feasible to do.

I'm planning on using 26 AWG being the max skin depth is 107kHz and my target frequency is 100kHz, but I do have about 30' of some premade 50/32 AWG litz I bought last year, should I use that instead?

That sounds good.  Maybe save it for something fancier, though?

Note that, because of proximity effect between turns, and within the cable itself, the skin depth is considerably shallower: 26 AWG is one skin depth (approximately) for a single solid wire at 107kHz, but buried under layers, the skin depth may be more like 30 AWG equivalent.  Using a Litz cable built with that strand size would help, but still wouldn't be optimal; you might need to use 38 or 42 AWG strands to get AC resistance down to nearly the DC resistance.

So, of course, it's a tradeoff; if you don't need ACR being as low as DCR, who cares.  If you need maximum efficiency, you need surprisingly fine Litz.

For small converters, and where you aren't terribly worried about efficiency, losses can be quite high, and solid wire is fine.

You can also double it up, so you're using more wire than you should need (based on DC ampacity), and it won't be optimal at all (because it's made of few, large, strands), but it's still better than just one (even if not strictly N times better, with N strands -- because of proximity effect).

Tim

[hanzdolo30@gmail.com]

Hey Tim, Good to hear from you,

Electrostatic grounding has no effect on the magnetic field, of course.

Yeah I know. I just figured, I'd run it by you anyway. The Idea was that it would ground the current in the shielding keeping it from burning up cooler, but that only works in biased amplifiers. Even if that worked there would be a power loss. 

Hmm, I suppose you could use multiple sizes in the same wire, but that would be annoying.  And hard to arrange optimally, because you'd want the finer strands in the middle.  But you want the strands always circulating in and out, hence the woven construction of Litz, so that's impossible!

Do you mean like desoldering braid?
My cats would have a field day,  .

Just a matter of the number of strands, what size strand you start with in the construction of an equivalent size cable.

Hmm, ferrite or powdered iron?  You usually see the latter (yellow-white or blue-green colored), which is very lossy material (by 100kHz, it's practically a better resistor than inductor!).

The black powdered iron boost inductor toroids. The green ones are just for filtration, right?
The yellow ones get crazy hot at 12V 75kHz

This is a poor way to do it, because it strains the wires, and doesn't make a tight cable anyway (it springs back a little).

Unfortunately, the proper way to do it isn't very easy (you need one spool for each strand, and the spools need to rotate freely, so as the wire is drawn off and twisted into a bundle, the spools rotate at the same rate).

If you only need a few meters, it's not terrifically annoying to weave or twist loose strands together in the same way (no spool needed if they're loose... just don't let them tangle up..).  Otherwise... it's one of those cases were "poor" is about all that's feasible to do.

Yeah I know about the wheel, but I do a pretty good job with my drill (i've had my share of birds nests, ), I just wrap the strands around the hooks with tension, then I put one of em in my drill maintaining tension, spinning as slow as possible following the cable as it gets shorter, being sure not to let it stretch, being the twisting flattens it a bit increasing the skin depth.  It's usually perfect when I'm done. I'll upload a pic when I'm done for you to see.
But this conversation is making me wanna use that 50/32. Is kinda thick(1.7mm).
I was planning on using it in the transformers.

That sounds good.  Maybe save it for something fancier, though?

Note that, because of proximity effect between turns, and within the cable itself, the skin depth is considerably shallower: 26 AWG is one skin depth (approximately) for a single solid wire at 107kHz, but buried under layers, the skin depth may be more like 30 AWG equivalent.  Using a Litz cable built with that strand size would help, but still wouldn't be optimal; you might need to use 38 or 42 AWG strands to get AC resistance down to nearly the DC resistance.

So, of course, it's a tradeoff; if you don't need ACR being as low as DCR, who cares.  If you need maximum efficiency, you need surprisingly fine Litz.

I'd rather add more strands and have less resistance.

I do have a spool of 30 awg, would  that be much better?
I've spun 30 before and it has less memory, but you need 3 times the strands to carry the current.

For small converters, and where you aren't terribly worried about efficiency, losses can be quite high, and solid wire is fine.

You can also double it up, so you're using more wire than you should need (based on DC ampacity), and it won't be optimal at all (because it's made of few, large, strands), but it's still better than just one (even if not strictly N times better, with N strands -- because of proximity effect).

Tim

Tim, I'm sanding inductors for a dual phase PFC stage, don't you notice I'm being kinda OCD about this,   efficiency is the goal!

Seriously though I do have like 50 EE25/13/7s should I make smaller ones to play with first?
Though I didn't think they'd be good enough for 400V boost inductors. I was planning on using of them to make flybacks, for powering the primary.

I've never used an optocoupler...Nevermind, we can talk about that later if you want (I Must be driving you crazy) . 

[MrAl]

Hi,

Oh ok, i just thought you had some on hand that you could show here so i could take a look.  There are different equations used in magnetics so i always welcome the chance to look at what might be  new ones or something.
If you got them from a text book then i guess they include all the necessary features.  If you get around to showing one we could take a look.

I didn't realize how rusty I was, it took me 4 hours to get back into the swing of things. I haven;t touched C# in about 4 Years. I was oblivious to the fact that C# uses Math.Exp(value) to handle exponents and apparently javascript does the same so I wasted a few hours (trying to be a perfectionist) trying either create a new operator or add the ability to the variable type like 3.toString() . No luck there so I guess, I'll have to go the long way;

Visual Basic has the ^ operator but I haven't touched VB in so many years, it was just weird being in there. I'm not sure if you're aware of this, but MS said they were going to  drop VB back in 2011
so it was much to my surprise when 2017 had VB in it's list of languages. Basic was my first language,but after working with braced languages, It's difficult to turn back. I may go Javascript. I can build all of the math into an object. The beauty of JS is that you can write functions into the objects. I'm trying to decide. I guess I'll have all three compute a core of the same value and see which one I can build the best UI for.
This is a controller from My Framework Just to give you an example:

API Engine

Controllers

Now that I’ve gotten to the part where I’ll have to write the controllers, I’ve found that the best approach to forming API methods is to form them the same as you would if you declared any other java script class with the exception that there should be guidelines followed in order to use the helper methods written in to the controller module.

Basic construction of a standard controller class should be as follows

//This is an example of a root folder controller
function defaultController() {
    // this is routing and security informatiom
    this.aliases = ['/'];
    this.authorize = true;
    this.methods = {
       
        index: {
            aliases: ['/', 'default', 'index'],
            GET: function () {
                returnView('default');
            }
        },
       
        about: {
            aliases: ['/', 'about', 'aboutus', 'about_us', 'about-us'],
            action: function () {
                returnView('about');
            }
        },
       
        contact: {    // aliases which call the get method
            aliases: ['contact', 'contactus', 'contact_us', 'contact-us'],
        // GET request sends the default
            GET: function () {
                returnView('contact');
            },
            PUT: function () {
                // Do things to send email from fields upon contact form submission
            }

        }

    }

}

//Controllers should pass url parameters by dividing them into Controller = Param[0], action = param[1], other named parameters will be defined by the user while being defined in an object local to the controller by name.

Hi,

Dont feel bad i havent used Java yet at all.  I meant to get to it, but never found any good tutorials on it on the web.

A magnetic model would be a formula or set of equations like this, which could be written in any language:
B=f(H)*u

where for example f(H) for a very very simple example (and this whole thing is over simplified anyway) for a linear core:
f(H)=H*u

or in terms of current we have a function B such that:
B(i)=i*K

where 'i' is the current and K is a constant.  Obviously there's no hysteresis in this over simplified model.

Some of the sites i found for Java seem to be broken too.  Know any good sites for starting in Java?

[hanzdolo30@gmail.com]

Hi,

Dont feel bad i havent used Java yet at all.  I meant to get to it, but never found any good tutorials on it on the web.

A magnetic model would be a formula or set of equations like this, which could be written in any language:
B=f(H)*u

where for example f(H) for a very very simple example (and this whole thing is over simplified anyway) for a linear core:
f(H)=H*u

or in terms of current we have a function B such that:
B(i)=i*K

where 'i' is the current and K is a constant.  Obviously there's no hysteresis in this over simplified model.

Some of the sites i found for Java seem to be broken too.  Know any good sites for starting in Java?

You know, I haven't gotten around to java either, but whenever take a glance at java I always assume it's c# until I really look at it. I'll Find you a good PDF, it's a PROMISE!

I got back into the swing of things, it's like riding a bike,lol.  However I did choose VB because it's much easier to deal with exponents, which as you know, we deal with all the time and it has the same user interface as c#. nodeJs is awesome, if you're writing network based applications.
That code I showed you earlier is one of the help pages from the framework I wrote. There are 3. The controllers are actually that simple to write. No routes configuration necessary, upon server startup it:
1. Reads the main configuration file
2. Locates all of your controllers based on the directories specified, notifies you of any route     
              conflicts and autocreates your routeconfig.
3. Serves your index file as specified in the routeConfig
That's all. There's Serverside Javascript support in the Views, but the models can be messy, so I leave that to the discretion of the developer, everyone has their own style and favorite ORM so, I decided not to stand in the way of that by cluttering the framework.

After looking at it today I realized "Ah the broken monitor!", I't not 100% done yet. I'd day 85%
 
Check out the attached PDF, It explains it all in 3 pages with full examples.
Give me your honest opinion as a developer. What do you think it could use? I can provide you with a copy of the configuration folder if you'd like so you can get the full gist of how it works.

Please bear in mind this is my very first node JS project, ever. I didn't use any external packages.
I wrote all of the modules. There's just 1 very important thing that I didn't get to finish and that's POSTs.