turns ratio

You used to be able to get power transformer "kits" from Maplin in the UK that had the primary pre-wound, with a stated volts per turn on the secondary. Low voltage secondaries don't need many turns, so hand- winding is quite practical for a few pieces of custom transformer.

Best regards, Spehro Pefhany

My father made a living designing transformers and motors with only a slip-stick, when he wasn't teaching how to do it.

you can get kits here:

-Lasse

If you had an ounce of brains, you'd know that the second derivative of the velocity (the rate of change of acceleration) is the jerk. Even elevator mechanics are quite familiar with it. ;-)

I too know such a person. The guy I know is a VERY smart engineer.

Is his name Carpenter, by chance?

Those are nice!

These are the ones I was thinking of:

The split bobbin gives you really good isolation (at the expense of some voltage regulation).

Best regards, Spehro Pefhany

Convince me. I want a transformer with a turns ratio of 1:1.000001. How many turns on the primary? How many turns on the secondary?

Ah, so you finally looked it up. Maybe you can learn.

You're certainly incapable.

Even if the transformers were wound with a turns ratio of exactly 1:1.1, you shouldn't expect the voltage transfer ratio (which is what you measured) to be

You minimize the error due to copper resistance by making your measurements with as small an excitation voltage as possible.

You can also get an approximation to the coupling coefficient like this:

k = SQRT(1.0943 * .905) = .99516

This would be quite good.

"Spehro Pefhany"

While "split bobbin" types do have more leakage inductance that others, there is not enough to affect performance into a RESISTIVE load that draws the nominal VA.

However, when feeding a rectifier and filter cap DC supply, regulation of the resulting DC can suffer due to leakage inductance increasing the impedance of the secondary at harmonics of the AC supply frequency.

... Phil

Something like:

Core #1: 10,000:10,001 with a tap at 10,000

Core #2: 1000:10 (adjust the 10 to get it just right: -0.1ppm/turn).

Best regards, Spehro Pefhany

Interesting.. never thought of it that way, but it makes sense.

Best regards, Spehro Pefhany

We have different definitions of "transformer". I would call that two transformers. Or maybe a subassembly. Or a module.

How does that work? Most iron core materials have lower permeability at low excitation. I've seen that effect in current transformers, much higher phase shift at low currents.

"John Larkin"

... Phil

Alright, then do it with fractional turns. Does that still count as a "transformer"?

The flux densities aren't equal, alas. But they're in the same physical hunk of core material. A transformer is a wad of iron and copper (or suitable substitutes), put together in a certain way so as to accomplish the desired ratio. Exactly how much quibble room is there in that?

E core:

Windings:

Use winding 3 = Primary

For 1V excitation, we get 1mV/turn total -- around the center limb, or around both limbs in series. A winding 4 would generate 1mV.

Connect winding 1 to winding 5 (in phase), so that the left leg has 999 times lower flux of the right leg. Total flux is 1 + 999 = 1000 parts, which is 1mV/turn worth. Thus, the right leg carries 0.999mV/turn, while the left leg carries the remaining 1uV/turn.

To achieve the +1ppm ratio, connect winding 2 in series with the primary to obtain an autoformer; a true transformer is easily arranged with another

Core capacity is reduced significantly -- by 49.95% or so -- so you'll need a somewhat larger core this way.

Note that a final addition step is performed, because your ratio is close to unity; we've been talking about ratios closer to zero, which is an equivalent problem give or take which end of the source you measure with respect to, though your ratio arguably isn't in the spirit of the conversation. So just understand that's where that extra step comes from.

If you look at the equation I wrote earlier for the difference, R = (R1 - R2) / (R1 * R2) Let R1 - R2 = delta. To minimize R1 and R2, we would like delta ~= 1 (i.e., R1 ~= R2, "as equal as possible without being equal"), so that R1 ~= R2 ~= sqrt(delta / R).

For a ratio of 1e-6, we'll get the smallest R1 and R2 with delta = 1, giving values around sqrt(1/1e-6) = 1000. Adjecent values, like 999 and 1000 (R =

A larger difference allows more fortuitous values, at the expense of more turns. We could pick 1998 and 2002, R = 1.000001..e-6, an absolute error of only 1ppt.

The intermediate steps (delta = 2 or 3) are akward because they involve roots of 2 and 3 respectively; rational approximations are required, likely without much better accuracy. 3 should be exceptionally poor as the odd number can't be divided evenly. The poor luck of odd numbers is tempered by the fact that not all roots are whole; we might fortuitously find a root with decimal close to 0.5, so that half the odd number rounds out very nicely.

If you were curious, 1413 and 1415 (delta = 2, R1,2 ~= sqrt(2e6)) give err =

I suspect there is no general purpose analytical formula to solve for the number of turns on all windings given some rational number, but I suspect there is a general purpose algorithm (probably related to continued fractions) which calculates increasingly accurate approximations to it.

Tim

Not true, k is smaller when mu is smaller.

Tim

that

measure

with

With lower excitation, the exciting current is less, hence the IR drop in the copper resistance is less.

I don't think there are any core materials whose permeability decreases enough at low excitation voltage to cause the excitation current to increase with decreasing excitation voltage, so the effect of IR drop is reduced with reduced excitation.

The permeability of regular silicon steel as used in power transformers does substantially decrease at low flux levels,

For example, I measure the inductance of a 12.6 volt winding of a small power transformer as 74 mH at 10 millivolt excitation, but 461 mH at 10 volt excitation, about a 6 to 1 change.

The high nickel core materials that are used in high quality audio transformers exhibit a much smaller change in permeability with excitation level.

Top quality current transformers with permalloy or supermalloy cores won't exhibit much change in performance at low currents. The low cost current transformers used for current sensing in power supplies don't use permalloy cores, and they are satisfactory for their intended purpose, but they will show a degraded performance at low currents.