Current sense transformer for half-bridge smps

[grizedale]

Hello,

We are doing a 330W offline half-bridge SMPS.

We must have primary current sensing, or else we won't have short-circuit protection for when we start-up into a shorted output.

However, the cheapest current sense transformer we can find is ....

**broken link removed**

Digi-Key - PE-63586NL-ND (Manufacturer - PE-63586NL)


(PE-63586 from pulseeng.com)

This is $1.76 for 400 pieces.

Its way too expensive.

There is nothing any cheaper anywhere, so we must wind our own.

Now, the turns ratio is 1:50

Getting good coupling from primary to sdecondary with that huge turns ratio is going to be challenging, meaning a multiple interleaved secondary, interleaving a multiple primary , with wide. flat wires.

...this is all going to be very expensive to manufacture.

-So does any reader know if we can get away with some cheap winding technique with such a current sense transformer?

 

[WimRFP]

Mostly when the primary wire goes through the core, the coupling is sufficient. So in my opinion, you don't need expensive interleaved technique. If you want a current output, the load resistor (connected to secundary) should have a value far below the inductance of the secundary winding at the lowest frequency.

If you measure secundary inductance, best is to remove the primary. Note that DC component through the primary may result in core saturation (especially in case of small diameter, hi-mu cores).

 

[mtwieg]

Coupling coefficient shouldn't matter at all for current tranformers. When used correctly, they output current, so the source impedance shouldn't matter. Even a K as low as 0.9 should work fine.

 

[grizedale]

-but with a turns ratio of 1:50 , it will be hard to get anywhere near 0.9?

one turn primary isnt going to couple well at all with a 50 turn secondary.

Good coupling is surely important in a current sense transformer, otherwise the outputted current will not accurately be a 50th of the primary current?

 

[WimRFP]

-but with a turns ratio of 1:50 , it will be hard to get anywhere near 0.9?

one turn primary isnt going to couple well at all with a 50 turn secondary.

Good coupling is surely important in a current sense transformer, otherwise the outputted current will not accurately be a 50th of the primary current?

Once your have chosen the core, the number of secondary turns on the core doesn't affect the coupling coefficient. Just wind them equally distributed to reduce stray effects.

Just try it yourself with a signal source. Run the source output through the core in series with a resistor. You can check the current input waveform via voltage measurement across the resistor (oscilloscope) and at another channel you can verify the waveform from the secondary coil (with resistive loading as mentioned in previous posting).

 

[grizedale]

thanks, ill try that.......though i think you hit the nail on the head when you say, "wind them equally distirbuted to reduce stray effects".


..of course, with a single turn primary, you canot wind equally distributed, it is lumped in one unfortunate place.

...remembering that perfect coupling is only achieved with an infinite length core too.....and because we dont have infinite core lengths , we really need the pri and sec to be occupying the same space as each other so to speak......with a 1:50 turns ratio we are far from that.

 

[WimRFP]

Best is to try it. I constructed various current transformers varying from Hz to MHz range and never experienced problems with primary to secondary coupling.

The problems were in reducing capacitive coupling from primary to secondary due to high dV/dt on the primary.

 

[grizedale]

Thankyou , i am going to try it, though i have to pick the core and make a winding scheme,

the peak bridge current is 9.1Amps and the Max RMS is 4.9 Amps.

i am wondering if i just pick a small core and just ensure i never go above 300mT?

Doing a single turn is a bit of a puzzler though.

Beause it could still be a single turn and not actually wrap fully around the core...because the current loop of the single turn goes round the rest of the circuit...so i am thinking of arranging it so that the single turn wraps the spindle of the core as fully as possible....or else i will not get good coupling.?

 

[WimRFP]

What is the value of the DC average current (also for a transient situation). If your core goes into strong saturation, the inductance reduces significantly. This increases the lowest frrequency where the CT works as a CT. You should select ur, core diameter and number of primary turns based on the highest DC current that occurs without loss of function.

Transients that fall within the bandwidth of the CT, you can ignore as there the field is counteracted by the secondary turn with resistive loading.

 

[FvM]

Doing a single turn is a bit of a puzzler though.

Beause it could still be a single turn and not actually wrap fully around the core...because the current loop of the single turn goes round the rest of the circuit...so i am thinking of arranging it so that the single turn wraps the spindle of the core as fully as possible....or else i will not get good coupling.?

I fear, you're making a thing out of it. A single turn is when the current loop goes once through the toroid core, don't matter which way it exactly takes. Everything else is about different amounts of leakage inductance respectively slightly changing K. It effectively won't change the current transmission.

 

[mtwieg]

grizedale I think you're way overthinking the importance of coupling and leakage. Go ahead and draw up the equivalent transformer model and see how it's performance is actually affected and I think you'll find that it's not bad at all.

Are you planning on using the CT directly in series with the primary, or using one on either (or both) of the switches?

 

[grizedale]

mtwieg:
we are using one in series with the primary

Third page "choosing a Core" of this article states that its necessary to get good coupling with a current sense transformer....
http://www.mmgca.com/apps/MMG-ctdesign.pdf

They also say to use a toroidal core, but i imagine that thats very expensive in labour costs to assemble...because someones got to sit there and wrap turns through the core...no machine can do that.

i would prefer to select a small ETD core with vertical bobbin for it.

 

[tsvetelin_velkov]

I would go and buy a ready made current sense transformer for SMPS that is rated for the current you want to use in your design, no need to home-brew everything. Or I will buy an old / damaged SMPS and will take out the current sensing trafo, last PC supply blocks that I managed to find and open were using such a topology ... besides there are lots of other good parts inside that you can also use ... than you can adjust the sensitivity by adjusting the resistor and capacitor on the output - respectively input of your controller

 

[mtwieg]

mtwieg:
we are using one in series with the primary

Third page "choosing a Core" of this article states that its necessary to get good coupling with a current sense transformer....
http://www.mmgca.com/apps/MMG-ctdesign.pdf

Notice they don't actually explain what qualifies as "good" coupling or why it's important. Don't just believe some offhand, unsubstantiated comments in an app note. Try the analysis yourself!

They also say to use a toroidal core, but i imagine that thats very expensive in labour costs to assemble...because someones got to sit there and wrap turns through the core...no machine can do that.

Of course there are toroid winding machines. They're actually really cool things.

Toroids are nice because you can make them as completely independent sensors, often with no built in primary. Just thread the primary conductor through and you're done. But if you're okay with having the primary built in, then I can't see any reason a why a split core with a bobbin wouldn't work. Just beware that if you wind on the same bobbin to watch out for too much capacitance between the primary and secondary.

 

[treez]

Coupling coefficient shouldn't matter at all for current tranformers. When used correctly, they output current, so the source impedance shouldn't matter. Even a K as low as 0.9 should work fine.

.Are you sure?...please see the attached simulation which compares two current sense transformers, one with k=0.9, and the other with k = 1.
The one with k=0.9 suffers more than 10 percent loss of signal.
Surely this is not acceptable?

(its an LTspice simulation)

 

[mtwieg]

.Are you sure?...please see the attached simulation which compares two current sense transformers, one with k=0.9, and the other with k = 1.
The one with k=0.9 suffers more than 10 percent loss of signal.
Surely this is not acceptable?

(its an LTspice simulation)

Naturally their is going to be signal loss, as expected from simple models of coupled inductors. This isn't a problem as long as the coupling is known and accounted for (which is way easier than just insisting on having a perfect transformer).

 

[Orson Cart]

For a 5 x 5mm core (25mm2) dB/dt = V/(N.Ae), so allowing for 2 x schottkies in the full wave rect of the CT output plus 2V peak of signal at 9.1 amps peak in the bridge, then at 100kHz (5uS each way) dB = 200mT (-100mT to +100mT), dt=5uS, V = 3V, Ae = 25mm2, and CT rating = 75uVsec, for lower power in the burden, use a lower Bmax and hence more sec turns.
To convert 9.1 amp to 91mA the turns ratio is 1: 100, and you will need a 22 ohm (burden) res to give 2V from 9.1 amp pri, 91mA sec, (after the 4 diode rect on the CT), the Bmax will be 3mT (6mT traverse, i.e.. -3mT to +3mT), and the worst case power in the 22 ohm is < 0.18W, in this example.

From the above you can see that you can go to a smaller area core (if the 100 turns will fit and the pri wire!), you could go 30x smaller to 25/30 mm2, and the flux would go up 30x to 90mT each way.

Something in the middle of these will be a good compromise...


Regards,
Orson

 

[treez]

Sorry I didn't follow...in your example, dB = ([3 * 5e-6] / [100 * 25e-6] ) =6000mT
...not 200mT?

 

[Anna Conda]

Sigh, dB= V. dt / (N.Ae) = 3 x 5 / (100 x 25) = 15 / 2500 = 0.006 = 6mT.

 

[treez]

yes my mistake, but where did the 200mT come from in the post above yours?