A combination of theory and practicality- neither is a substitute for either..
Lived this for years while employed as automotive engineer in Detroit, MI-- Ford Rotunda and GM Flint
The meeting rooms valued both- so tier one suppliers submitted computer sims, then went out to eval the test fleets
Jim
Lived this for years while employed as automotive engineer in Detroit, MI-- Ford Rotunda and GM Flint
The meeting rooms valued both- so tier one suppliers submitted computer sims, then went out to eval the test fleets
Jim
As a start, let's say you start with a specific size of magnetic core and surface area, which means the core choice is your priority (size, weight and footprint priority)
1.The window of the core determines
a) How much volume of wire you will fit in.
-Higher wire gauge means lower losses, but fitting a lesser amount of turns
b) The dielectric thickness
-Thicker dielectric means lower capacitance in high potential difference regions and distributing dielectric in different thickness is a common technique to reduce capacitance, but also results in less usable window and potentially again, a lesser amount of turns.
2. The amount of primary turns and your core surface area determine your flux density potential. And dictates the amount of power for the lowest frequency you will be able to transfer from your OPT
Afe stands for the core surface area, T for the number of turns, f for the bottom frequency needed, L for primary inductance
Ba = Vpri / (4.44 x Afe x T x f)
Bdc = (L x Idc) / (T x Afe)
Ba applies to the alternating voltage across the primary, and Bdc applies to DC across the primary (for SE OPTs). For ideal PP topology OPTs, all flux density potential goes to alternating voltage, where for SE OPTs, some flux density needs to be sacrificed (usually 1/2 of the total) for magnetizing DC inherent in the SE stage.
For achieving maximum flux density potential, the more primary turns and the more core area, the better.
In a SE OPT, you can distribute flux density into the following trade-offs.
-Flux density into voltage swing (power)
-Flux density into DC vs primary inductance. More Idc or more primary inductance eat more of your voltage swing headroom, but give a higher operating tube current or lower LF cut-off.
3. However increasing the turn number increases the leakage inductance by a square factor. It also brings some HF resonances closer to the audio band. It also increases Rdc losses. Core size is also a limiting factor, because it increases the mean turn length of a winding. It is always easier to go with a bigger core, because the surface area is squared compaired to the increasing of mean turn length. Leakage inductance, capacitance and losses increase linearly with MLT, while leakage inductance increases by the square of primary turns.
4. After deciding on your power output, Idc vs L ratio and ohmic Rdc losses, one can start with choosing an interleaving geometry and calculate leakage inductance, capacitance and determine roll-offs in the HF region. Chance is, you'll have to often go back to the beginning. Choosing the interleaving schematic can take some time, due to the fact you need to tailor the leakage inductance, capacitance values to be optimal for your reflected primary load, driving impedance and secondary load.
5. A low interleaving schematic allows for some cheats that can make the frequency response better, like reversing the winding direction of a section, or the way you connect the secondaries. However the higher you go with interleaves, the less cheats become available. Another cheat is using low-epsilon dielectrics. How you connect secondaries is detrimental and is a region where I still study.
'Tis for now. This explanation is still vague, but if you're interested into expanding a specific part of the explanation, let me know.
1.The window of the core determines
a) How much volume of wire you will fit in.
-Higher wire gauge means lower losses, but fitting a lesser amount of turns
b) The dielectric thickness
-Thicker dielectric means lower capacitance in high potential difference regions and distributing dielectric in different thickness is a common technique to reduce capacitance, but also results in less usable window and potentially again, a lesser amount of turns.
2. The amount of primary turns and your core surface area determine your flux density potential. And dictates the amount of power for the lowest frequency you will be able to transfer from your OPT
Afe stands for the core surface area, T for the number of turns, f for the bottom frequency needed, L for primary inductance
Ba = Vpri / (4.44 x Afe x T x f)
Bdc = (L x Idc) / (T x Afe)
Ba applies to the alternating voltage across the primary, and Bdc applies to DC across the primary (for SE OPTs). For ideal PP topology OPTs, all flux density potential goes to alternating voltage, where for SE OPTs, some flux density needs to be sacrificed (usually 1/2 of the total) for magnetizing DC inherent in the SE stage.
For achieving maximum flux density potential, the more primary turns and the more core area, the better.
In a SE OPT, you can distribute flux density into the following trade-offs.
-Flux density into voltage swing (power)
-Flux density into DC vs primary inductance. More Idc or more primary inductance eat more of your voltage swing headroom, but give a higher operating tube current or lower LF cut-off.
3. However increasing the turn number increases the leakage inductance by a square factor. It also brings some HF resonances closer to the audio band. It also increases Rdc losses. Core size is also a limiting factor, because it increases the mean turn length of a winding. It is always easier to go with a bigger core, because the surface area is squared compaired to the increasing of mean turn length. Leakage inductance, capacitance and losses increase linearly with MLT, while leakage inductance increases by the square of primary turns.
4. After deciding on your power output, Idc vs L ratio and ohmic Rdc losses, one can start with choosing an interleaving geometry and calculate leakage inductance, capacitance and determine roll-offs in the HF region. Chance is, you'll have to often go back to the beginning. Choosing the interleaving schematic can take some time, due to the fact you need to tailor the leakage inductance, capacitance values to be optimal for your reflected primary load, driving impedance and secondary load.
5. A low interleaving schematic allows for some cheats that can make the frequency response better, like reversing the winding direction of a section, or the way you connect the secondaries. However the higher you go with interleaves, the less cheats become available. Another cheat is using low-epsilon dielectrics. How you connect secondaries is detrimental and is a region where I still study.
'Tis for now. This explanation is still vague, but if you're interested into expanding a specific part of the explanation, let me know.
Presumably a lot of this work is a one-off for a brand new design, and becomes your intellectual property when you have a successful commercial design that you can replicate for customers? Are there a finite number of transformers that cover 90% of requirements, or are there unlimited permutations?
Permutations as customer's requirements are all over the place, and despite my attempts trying to be minimalistic, they are many of them. I believe each transformer manufacturer in the end determines his philosophy and permutation limitation criteria, and they all differ from one another.
Throughout my journey, my hardest transformer fixed criteria were squeezing the most from a small footpring. The other challenge was sellecting a variety of core materials for the same footprint, as amorphous and nanocrystlaline cores are available in specific sizes only. Custom sizes need custom tooling for the core manufacturers and I cannot afford that for now.
As for my challenges towards high-frequency response optimization and squeezing more primary turns for large transformers with no trouble, it is a personal technical fetishism, but not really needed for widly said 90-95% transformers already available on the market.
Throughout my journey, my hardest transformer fixed criteria were squeezing the most from a small footpring. The other challenge was sellecting a variety of core materials for the same footprint, as amorphous and nanocrystlaline cores are available in specific sizes only. Custom sizes need custom tooling for the core manufacturers and I cannot afford that for now.
As for my challenges towards high-frequency response optimization and squeezing more primary turns for large transformers with no trouble, it is a personal technical fetishism, but not really needed for widly said 90-95% transformers already available on the market.
Looking good, compliment!
Would you take the challenge of winding on teflon? Its a pita, coldfowing under pressure, slippery as hell, static, almost impossible to tape or glue and certainly not possible to wind selfsupporting but if i where younger maybe doable with a bobbin. What do you thinck? I am asking because i am looking for a winder that would take up the callenge, i could supply the 0,1mm teflon and SU90 bobbins if needed, 12 bifilar wound layers with 0.5mm Cull on each bobbin, 4-8 in total.
I am to old for that, would you take the challange? Offcourse I would pay you properly, I know what it takes...
Would you take the challenge of winding on teflon? Its a pita, coldfowing under pressure, slippery as hell, static, almost impossible to tape or glue and certainly not possible to wind selfsupporting but if i where younger maybe doable with a bobbin. What do you thinck? I am asking because i am looking for a winder that would take up the callenge, i could supply the 0,1mm teflon and SU90 bobbins if needed, 12 bifilar wound layers with 0.5mm Cull on each bobbin, 4-8 in total.
I am to old for that, would you take the challange? Offcourse I would pay you properly, I know what it takes...
Last edited:
Hi, first, you don't need teflon, Nomex have very low dielectric coefficient and much more convinient for coil winding process. 2nd, if you are still keen to use teflon, there are adhesive teflon tapes with fabric-like surface, quite expensive but available of varius sizes without any problem. We used them in the past for industrial equipment repair.
A friend, colleague an winder teacher to an extent, told me a tale of his attempt of winding an OPT using teflon one time. It turned out to be a nightmare, he said. Not only due to softness, slipperiness and the inability to glue and hold the teflon well, but he also told me it would easily pick up electrostatic charges and attract all kind of dust around.
Here is a proper teflon adhesive tape if someone is still fanatically keen to use as insulation in transformer. Again, best overall insulation for transformer manufacturing is a Nomex paper. Period. Personally, I wouldn't mind to use teflon in this particular case.
Attachments
Hi Linuks Guru, I want teflon because it does not soak when impregnated and only polypropylene or polyethene would have comparable low loss factor.
The teflon tape you mention is glass fiber filled, expensive and not nearly as good as teflon, I would rather etch teflon to make it glue able if needed, but best would be to wind without it. In my youth I did both, etching, without, and also used teflon tape when winding wideband transformer for measuring equipment (0,5-500khz), I am very experienced when it comes to transformers from low level and up to hundreds of kVA for Rf heating. I was specialized in Rf heating. So I know about all there is to know regarding teflon, but now I am 75+ and a bit shaky and this job needs nerves f steel, steady hands and good eyes.
The teflon tape you mention is glass fiber filled, expensive and not nearly as good as teflon, I would rather etch teflon to make it glue able if needed, but best would be to wind without it. In my youth I did both, etching, without, and also used teflon tape when winding wideband transformer for measuring equipment (0,5-500khz), I am very experienced when it comes to transformers from low level and up to hundreds of kVA for Rf heating. I was specialized in Rf heating. So I know about all there is to know regarding teflon, but now I am 75+ and a bit shaky and this job needs nerves f steel, steady hands and good eyes.
I need the coils to be with the lowest possible dielectric losses in order to be able to check my calculations regarding proximity losses. The 500hz to 500ooohz transformer was for a EG meter i did between 1967 and 68. The coils i need now are for a different purpose. A multilayer coil with 0,5mm wire would be a total no no at 500khz
Last edited:
Yes, proximity losses depend a.o.t, on the thickness of the wire. What we call skin losses are the self-induced eddy current losses of a single wire with no other wire nearby. Proximity losses increase sharply when you put layer over layer. In Litzwirehe wire the wire has to be MUCH much thinner than the skindepth and INTERPOSITIONED. Litzwire has to be looked upon as MULTILAYER with interpositioned wires.
As to multifilar, (4 wires of halve diam) here are for example the calculated Rac/Rdc ratios for close spaced single isolated copper wire.
As temperatur goes up skin depth increases, the table below is for 40degC
0,6mm 1layer 18kc Rac/Rdc 1,101
0,3mm 2layer 18kc Rac/Rdc 1,031
0,6mm 1layer 24kc Rac/Rdc 1,173
0,3mm 2layer 24kc Rac/Rdc 1,055
0,6mm 1layer 36kc Rac/Rdc 1,358
0,3mm 2layer 36kc Rac/Rdc 1,124
0,6mm 1layer 72kc Rac/Rdc 1,999
0,3mm 2layer 72kc Rac/Rdc 1,478
0,6mm 1layer 144kc Rac/Rdc 2,96
0,3mm 2layer 144kc Rac/Rdc 2,684
As to multifilar, (4 wires of halve diam) here are for example the calculated Rac/Rdc ratios for close spaced single isolated copper wire.
As temperatur goes up skin depth increases, the table below is for 40degC
0,6mm 1layer 18kc Rac/Rdc 1,101
0,3mm 2layer 18kc Rac/Rdc 1,031
0,6mm 1layer 24kc Rac/Rdc 1,173
0,3mm 2layer 24kc Rac/Rdc 1,055
0,6mm 1layer 36kc Rac/Rdc 1,358
0,3mm 2layer 36kc Rac/Rdc 1,124
0,6mm 1layer 72kc Rac/Rdc 1,999
0,3mm 2layer 72kc Rac/Rdc 1,478
0,6mm 1layer 144kc Rac/Rdc 2,96
0,3mm 2layer 144kc Rac/Rdc 2,684
Last edited:
Please observe, the given values of Rac/Rdc are for layers of equal currents, so in order to get a paralell connection of 4 wires with equal current sharing you have to wind halve a bifilar layer 1 first and then interpositioned wind the rest in layer 2, so each layer consits of one halve beneath and halve on top of each other. Also, at the extreme ends of the layer proximity losses increase and therefore the above given average Rac/Rdc ratio will be higher.
Last edited:
Nice thread... How important is the laminate thickness ? Is it absolutely necessary to use 0.3mm laminate or If i need to work with 0.5mm, what parameters do i need to take into consideration? I do understand in yesteryears most OPTs were made up of regular iron cores and not too thin,but still achieved good bandwidth...
If you have a choice, take grain orientated 0,3mm lams. 0,5mm has usually higher losses and less permeability. In the old days 0,4 - 0,6T max was suggested for 0,5mm lams (contained at least 4% Si). Mind you, now days silicium content is considerably lower (probable to save tool costs). High silicium content made the lams very hard and brittle but increased resistance wich in turn lowered eddy currents. Furthermore, the lams isolation (paint or paper) was much thicker and that had a much bigger negativ effect on the stacking factor with thinner lams, so 0.5mm lams where not so bad a compromise. Also, in the old days sound sources where much more frequency limited, and wide BW was not so important. Altogether, the use of thinner lams is nowadays much more rewarding than it once was.
Last edited: