Litz wire
in [Design]
|
From: Phil Hobbs on 22 Feb 2010 09:45 On 2/20/2010 1:04 PM, Tim Wescott wrote: > On Sat, 20 Feb 2010 05:38:23 -0800, dcaster(a)krl.org wrote: > >> On Feb 20, 5:47 am, life imitates life >> <pastic...(a)thebarattheendoftheuniverse.org> wrote: >>> On Fri, 19 Feb 2010 16:58:13 -0600, Tim Wescott<t...(a)seemywebsite.com> >>> wrote: >>> >>> >>> >>>> On Fri, 19 Feb 2010 16:18:49 -0600, Tim Williams wrote: >>> >>>>> "George Herold"<ggher...(a)gmail.com> wrote in message >>>>> news:550831a9-5935-4e3d-b37e- > c664ebd9d752(a)o3g2000yqb.googlegroups.com... >>>>>> We get Litz wire from MWS but nothing that big. If you don't need >>>>>> that much have you thought of 'rolling your own'? >>> >>>>> I do sometimes, but only for small things. I'm contemplating 10A at >>>>> 1MHz, so it needs to be pretty fine = way more strands than I'd want >>>>> to deal with. >>> >>>>> I actually have some copper rope, which is about 1/4" diameter and >>>>> looks to be made of 28AWG or so. I don't remember how many strands >>>>> it is, but if I guess the rope is wound from 7 strands of 31 strand >>>>> twist, that's 7*31 = 217. If 28AWG is good for ~200mA, 217 strands >>>>> should be good for 40A, which sounds about right, I'd call it 8 or >>>>> 10AWG equivalent. I salvaged this stuff from some old motor driver, >>>>> which used a spool of this stuff for air-core inductors. >>> >>>>> Tim >>> >>>> Isn't there some magic braiding pattern for Litz wire? >>> >>> No. The wire strands have to be mag wire, which segregates them >>> from >>> each other, allowing the skin effect to be taken advantage of. Without >>> strand segregation, it becomes a single strand, from the POV of the >>> current flowing in it, with only one skin for the entire mass. >> >> No again. The wire strands have to be segregated and also braided so >> that some of the time a strand is on the outside of the bundle and >> sometimes on the inside of the bundle. See attached from Wiki. >> >> Litz wire uses some different tricks. Instead of using one big >> conductor, it uses lots of little conductors (strands) in parallel >> (forming a bundle). Each little conductor is less than a skin-depth, so >> an individual strand does not suffer an appreciable skin effect loss. >> However, that is not the complete story. The strands must be insulated >> from each other -- otherwise all the wires in the bundle would short >> together, look like a single large wire, and still have skin effect >> problems. Furthermore, the strands cannot occupy the same radial >> position in the bundle: the electromagnetic effects that cause the skin >> effect would still disrupt conduction. The bundle is constructed so the >> individual strands are on the outside of the bundle (and see low >> resistance) for a time, but also reside in the interior of the bundle >> (where the EM field changes are the strongest and the resistance is >> higher). If each strand sees about the same average resistance, then >> each strand will contribute equally to the conduction of the entire >> cable. > > Why, then, does it work to make up multi-strand wire bundles for SMPS > service where the wires are just lightly twisted into a bundle? Are > these getting one part of the way there, but not all? > > I had heard about the Litz wire 'gotta be a magic braid pattern', but > then I've seen all these SMPS transformers that just have almost-parallel > strands. > The idea of the braiding is to try to equalize the effective resistance of all the strands by giving them all the same exposure to the outer layer of the bundle. A nice inclusive democratic idea, but all bat's wings and newt's eyes from the POV of electromagnetics--and even circuits. Here's the circuits argument. Suppose you take two series strings of resistors, each of which alternates 1k and 10k, like this: *----1k---10k---1k---10k---1k---10k---* | | 0--*---10k---1k---10k---1k---10k---1k----*--0 Total resistance: 1650 ohms (3.3k per string). Now suppose you put all the 10k resistors in one string and all the 100 ohm resistors in the other--like letting a couple of strands be the outside conductor the whole way. Then you have this: *---10k---10k--10k---10k--10k---10k---* | | 0--*----1k---1k----1k---1k----1k---1k----*--0 Which is 6K // 600 ohms, or 545 ohms. Which is better? The electromagnetic argument for Litz wire is also murky, because it relies on that dopey single-straight-wire-in-free-space derivation we all know and love. It's mathematically simple, but it has very little to do with the actual physical environment inside a transformer, where there's all that other copper running with similar dI/dt right nearby, plus the huge collective effect of the other turns and the core. We pretty well know what the dB/dt is inside the winding. For a single conductor, the B field at the surface is transverse, and is all due to the current in that one wire. That field configuration gives rise to zero eddy current in the wire, just a radial gradient of the current density, and it makes very little difference to the fields outside the wire. In a winding, however, there is a huge component of B perpendicular to the wire surface, due to the fields of all the other wires and the huge effect of the core. That will drive circulating currents inside the diameter of the wires (like turbulence in a pipe). There is thus the opportunity for a lot of eddy current loss in the copper if it isn't split up into very small transverse pieces (as in a laminated core). The net effect is that parallel multifilar windings ought to be at least as good as Litz wire in transformers. I've never seen a proper derivation of the eddy current loss in a transformer winding, but I'm sure it's been done. It sure doesn't have much to do with classical skin effect, though. Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
From: Jim Thompson on 22 Feb 2010 10:07 On Mon, 22 Feb 2010 09:45:48 -0500, Phil Hobbs <pcdhSpamMeSenseless(a)electrooptical.net> wrote: >On 2/20/2010 1:04 PM, Tim Wescott wrote: >> On Sat, 20 Feb 2010 05:38:23 -0800, dcaster(a)krl.org wrote: >> >>> On Feb 20, 5:47 am, life imitates life >>> <pastic...(a)thebarattheendoftheuniverse.org> wrote: >>>> On Fri, 19 Feb 2010 16:58:13 -0600, Tim Wescott<t...(a)seemywebsite.com> >>>> wrote: >>>> >>>> >>>> >>>>> On Fri, 19 Feb 2010 16:18:49 -0600, Tim Williams wrote: >>>> >>>>>> "George Herold"<ggher...(a)gmail.com> wrote in message >>>>>> news:550831a9-5935-4e3d-b37e- >> c664ebd9d752(a)o3g2000yqb.googlegroups.com... >>>>>>> We get Litz wire from MWS but nothing that big. If you don't need >>>>>>> that much have you thought of 'rolling your own'? >>>> >>>>>> I do sometimes, but only for small things. I'm contemplating 10A at >>>>>> 1MHz, so it needs to be pretty fine = way more strands than I'd want >>>>>> to deal with. >>>> >>>>>> I actually have some copper rope, which is about 1/4" diameter and >>>>>> looks to be made of 28AWG or so. I don't remember how many strands >>>>>> it is, but if I guess the rope is wound from 7 strands of 31 strand >>>>>> twist, that's 7*31 = 217. If 28AWG is good for ~200mA, 217 strands >>>>>> should be good for 40A, which sounds about right, I'd call it 8 or >>>>>> 10AWG equivalent. I salvaged this stuff from some old motor driver, >>>>>> which used a spool of this stuff for air-core inductors. >>>> >>>>>> Tim >>>> >>>>> Isn't there some magic braiding pattern for Litz wire? >>>> >>>> No. The wire strands have to be mag wire, which segregates them >>>> from >>>> each other, allowing the skin effect to be taken advantage of. Without >>>> strand segregation, it becomes a single strand, from the POV of the >>>> current flowing in it, with only one skin for the entire mass. >>> >>> No again. The wire strands have to be segregated and also braided so >>> that some of the time a strand is on the outside of the bundle and >>> sometimes on the inside of the bundle. See attached from Wiki. >>> >>> Litz wire uses some different tricks. Instead of using one big >>> conductor, it uses lots of little conductors (strands) in parallel >>> (forming a bundle). Each little conductor is less than a skin-depth, so >>> an individual strand does not suffer an appreciable skin effect loss. >>> However, that is not the complete story. The strands must be insulated >>> from each other -- otherwise all the wires in the bundle would short >>> together, look like a single large wire, and still have skin effect >>> problems. Furthermore, the strands cannot occupy the same radial >>> position in the bundle: the electromagnetic effects that cause the skin >>> effect would still disrupt conduction. The bundle is constructed so the >>> individual strands are on the outside of the bundle (and see low >>> resistance) for a time, but also reside in the interior of the bundle >>> (where the EM field changes are the strongest and the resistance is >>> higher). If each strand sees about the same average resistance, then >>> each strand will contribute equally to the conduction of the entire >>> cable. >> >> Why, then, does it work to make up multi-strand wire bundles for SMPS >> service where the wires are just lightly twisted into a bundle? Are >> these getting one part of the way there, but not all? >> >> I had heard about the Litz wire 'gotta be a magic braid pattern', but >> then I've seen all these SMPS transformers that just have almost-parallel >> strands. >> > >The idea of the braiding is to try to equalize the effective resistance >of all the strands by giving them all the same exposure to the outer >layer of the bundle. A nice inclusive democratic idea, but all bat's >wings and newt's eyes from the POV of electromagnetics--and even circuits. > >Here's the circuits argument. Suppose you take two series strings of >resistors, each of which alternates 1k and 10k, like this: > > *----1k---10k---1k---10k---1k---10k---* > | | >0--*---10k---1k---10k---1k---10k---1k----*--0 > >Total resistance: 1650 ohms (3.3k per string). Now suppose you put all >the 10k resistors in one string and all the 100 ohm resistors in the >other--like letting a couple of strands be the outside conductor the >whole way. Then you have this: > > > *---10k---10k--10k---10k--10k---10k---* > | | >0--*----1k---1k----1k---1k----1k---1k----*--0 > >Which is 6K // 600 ohms, or 545 ohms. Which is better? > > >The electromagnetic argument for Litz wire is also murky, because it >relies on that dopey single-straight-wire-in-free-space derivation we >all know and love. It's mathematically simple, but it has very little >to do with the actual physical environment inside a transformer, where >there's all that other copper running with similar dI/dt right nearby, >plus the huge collective effect of the other turns and the core. > >We pretty well know what the dB/dt is inside the winding. For a single >conductor, the B field at the surface is transverse, and is all due to >the current in that one wire. That field configuration gives rise to >zero eddy current in the wire, just a radial gradient of the current >density, and it makes very little difference to the fields outside the wire. > >In a winding, however, there is a huge component of B perpendicular to >the wire surface, due to the fields of all the other wires and the huge >effect of the core. That will drive circulating currents inside the >diameter of the wires (like turbulence in a pipe). There is thus the >opportunity for a lot of eddy current loss in the copper if it isn't >split up into very small transverse pieces (as in a laminated core). > >The net effect is that parallel multifilar windings ought to be at least >as good as Litz wire in transformers. > >I've never seen a proper derivation of the eddy current loss in a >transformer winding, but I'm sure it's been done. It sure doesn't have >much to do with classical skin effect, though. > >Cheers > >Phil Hobbs It's been almost 25 years since I last worried about transformer windings. But I can recall at least one 5V 100A jobby where the output "winding" was 2" wide copper foil. (45� fold to make 90� connection tabs at the ends.) ...Jim Thompson -- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | I love to cook with wine. Sometimes I even put it in the food.
From: John Larkin on 22 Feb 2010 10:08 On Sat, 20 Feb 2010 21:22:51 -0800 (PST), George Herold <ggherold(a)gmail.com> wrote: >On Feb 20, 8:38�am, "dcas...(a)krl.org" <dcas...(a)krl.org> wrote: >> On Feb 20, 5:47�am, life imitates life >> >> >> >> >> >> <pastic...(a)thebarattheendoftheuniverse.org> wrote: >> > On Fri, 19 Feb 2010 16:58:13 -0600, Tim Wescott <t...(a)seemywebsite.com> >> > wrote: >> >> > >On Fri, 19 Feb 2010 16:18:49 -0600, Tim Williams wrote: >> >> > >> "George Herold" <ggher...(a)gmail.com> wrote in message >> > >>news:550831a9-5935-4e3d-b37e-c664ebd9d752(a)o3g2000yqb.googlegroups.com... >> > >>> We get Litz wire from MWS but nothing that big. �If you don't need that >> > >>> much have you thought of 'rolling your own'? >> >> > >> I do sometimes, but only for small things. �I'm contemplating 10A at >> > >> 1MHz, so it needs to be pretty fine = way more strands than I'd want to >> > >> deal with. >> >> > >> I actually have some copper rope, which is about 1/4" diameter and looks >> > >> to be made of 28AWG or so. �I don't remember how many strands it is, but >> > >> if I guess the rope is wound from 7 strands of 31 strand twist, that's >> > >> 7*31 = 217. �If 28AWG is good for ~200mA, 217 strands should be good for >> > >> 40A, which sounds about right, I'd call it 8 or 10AWG equivalent. �I >> > >> salvaged this stuff from some old motor driver, which used a spool of >> > >> this stuff for air-core inductors. >> >> > >> Tim >> >> > >Isn't there some magic braiding pattern for Litz wire? >> >> > � No. �The wire strands have to be mag wire, which segregates them from >> > each other, allowing the skin effect to be taken advantage of. �Without >> > strand segregation, it becomes a single strand, from the POV of the >> > current flowing in it, with only one skin for the entire mass. >> >> No again. �The wire strands have to be segregated and also braided so >> that some of the time a strand is on the outside of the bundle and >> sometimes on the inside of the bundle. �See attached from Wiki. >> >> Litz wire uses some different tricks. Instead of using one big >> conductor, it uses lots of little conductors (strands) in parallel >> (forming a bundle). Each little conductor is less than a skin-depth, >> so an individual strand does not suffer an appreciable skin effect >> loss. However, that is not the complete story. The strands must be >> insulated from each other -- otherwise all the wires in the bundle >> would short together, look like a single large wire, and still have >> skin effect problems. Furthermore, the strands cannot occupy the same >> radial position in the bundle: the electromagnetic effects that cause >> the skin effect would still disrupt conduction. The bundle is >> constructed so the individual strands are on the outside of the bundle >> (and see low resistance) for a time, but also reside in the interior >> of the bundle (where the EM field changes are the strongest and the >> resistance is higher). If each strand sees about the same average >> resistance, then each strand will contribute equally to the conduction >> of the entire cable. >> >> Dan- Hide quoted text - >> >> - Show quoted text - > >Cool, Thanks Dan. I'd never thought of that. > >A bit OT, but I remember seeing a video of high current experiments >done at the Magnet Lab, (then at MIT circa 1960's) Where they were >using several ~2-3" wide strips of copper. This is pretty good: http://www.newenglandwire.com/nepdfs/litz_Brochure.pdf John
From: George Herold on 22 Feb 2010 11:00 On Feb 22, 10:08 am, John Larkin <jjlar...(a)highNOTlandTHIStechnologyPART.com> wrote: > On Sat, 20 Feb 2010 21:22:51 -0800 (PST), George Herold > > > > > > <ggher...(a)gmail.com> wrote: > >On Feb 20, 8:38 am, "dcas...(a)krl.org" <dcas...(a)krl.org> wrote: > >> On Feb 20, 5:47 am, life imitates life > > >> <pastic...(a)thebarattheendoftheuniverse.org> wrote: > >> > On Fri, 19 Feb 2010 16:58:13 -0600, Tim Wescott <t...(a)seemywebsite.com> > >> > wrote: > > >> > >On Fri, 19 Feb 2010 16:18:49 -0600, Tim Williams wrote: > > >> > >> "George Herold" <ggher...(a)gmail.com> wrote in message > >> > >>news:550831a9-5935-4e3d-b37e-c664ebd9d752(a)o3g2000yqb.googlegroups.com... > >> > >>> We get Litz wire from MWS but nothing that big. If you don't need that > >> > >>> much have you thought of 'rolling your own'? > > >> > >> I do sometimes, but only for small things. I'm contemplating 10A at > >> > >> 1MHz, so it needs to be pretty fine = way more strands than I'd want to > >> > >> deal with. > > >> > >> I actually have some copper rope, which is about 1/4" diameter and looks > >> > >> to be made of 28AWG or so. I don't remember how many strands it is, but > >> > >> if I guess the rope is wound from 7 strands of 31 strand twist, that's > >> > >> 7*31 = 217. If 28AWG is good for ~200mA, 217 strands should be good for > >> > >> 40A, which sounds about right, I'd call it 8 or 10AWG equivalent. I > >> > >> salvaged this stuff from some old motor driver, which used a spool of > >> > >> this stuff for air-core inductors. > > >> > >> Tim > > >> > >Isn't there some magic braiding pattern for Litz wire? > > >> > No. The wire strands have to be mag wire, which segregates them from > >> > each other, allowing the skin effect to be taken advantage of. Without > >> > strand segregation, it becomes a single strand, from the POV of the > >> > current flowing in it, with only one skin for the entire mass. > > >> No again. The wire strands have to be segregated and also braided so > >> that some of the time a strand is on the outside of the bundle and > >> sometimes on the inside of the bundle. See attached from Wiki. > > >> Litz wire uses some different tricks. Instead of using one big > >> conductor, it uses lots of little conductors (strands) in parallel > >> (forming a bundle). Each little conductor is less than a skin-depth, > >> so an individual strand does not suffer an appreciable skin effect > >> loss. However, that is not the complete story. The strands must be > >> insulated from each other -- otherwise all the wires in the bundle > >> would short together, look like a single large wire, and still have > >> skin effect problems. Furthermore, the strands cannot occupy the same > >> radial position in the bundle: the electromagnetic effects that cause > >> the skin effect would still disrupt conduction. The bundle is > >> constructed so the individual strands are on the outside of the bundle > >> (and see low resistance) for a time, but also reside in the interior > >> of the bundle (where the EM field changes are the strongest and the > >> resistance is higher). If each strand sees about the same average > >> resistance, then each strand will contribute equally to the conduction > >> of the entire cable. > > >> Dan- Hide quoted text - > > >> - Show quoted text - > > >Cool, Thanks Dan. I'd never thought of that. > > >A bit OT, but I remember seeing a video of high current experiments > >done at the Magnet Lab, (then at MIT circa 1960's) Where they were > >using several ~2-3" wide strips of copper. > > This is pretty good: > > http://www.newenglandwire.com/nepdfs/litz_Brochure.pdf > > John- Hide quoted text - > > - Show quoted text - Wow.. lotsa different configurations there. Here's the magnet lab video for anyone interested. http://www.archive.org/details/magnet_laboratory_1959 Enjoy! George H.
From: qrk on 22 Feb 2010 13:41
On Sun, 21 Feb 2010 16:09:55 -0600, "Tim Williams" <tmoranwms(a)charter.net> wrote: >Let's say 200kHz-2MHz. > >Tim As a gross rule of thumb, proximity effect dominates under 1MHz. Above 1MHz, skin effect will come in to play. Most of my work designing power pulse transformers is in the 50kHz to 800kHz range. Checks with an impedance analyzer (HP 4195 and 4194) have shown that the equations given in E.C. Snellings "Soft Ferrites" book are pretty good. E.C. Snelling and Giles have another book on transformers. I've forgotten the title, but it also covers proximity effect. -- Mark |