Prev: Replacing floppy
Next: Gnats :-(
From: Hammy on 25 Apr 2010 10:42
I have some Flyback transformers (from CoilCraft) these were
originally intended for use in a 70W 60kHz application using the
NCP1200. If I increase the switching frequency to 100kHz is it
reasonable to expect 90-100W using the same core?

I've been running simulations and the rms and peak current for 100kHz
equalizes at 90W ,with the 70W 60kHz flyback. There is a higher DC
component of course (about 10-15%) due to a reduced switching cycle,
the converter is deeper in CCM.

I had a sheet giving approximation on core power throughput vs
frequency and it shows that for 70W 70kHz increasing the Frequency to
100kHz the same E-Core could handle a little over 100W. I know I could
expect higher DC winding losses; in a pinch I could add another
parallel winding there is already two it would be tight though.

Any magnetic experts out there?
From: legg on 25 Apr 2010 12:29
On Sun, 25 Apr 2010 10:42:36 -0400, Hammy <spam(a)spam.com> wrote:

>I have some Flyback transformers (from CoilCraft) these were
>originally intended for use in a 70W 60kHz application using the
>NCP1200. If I increase the switching frequency to 100kHz is it
>reasonable to expect 90-100W using the same core?
>
>I've been running simulations and the rms and peak current for 100kHz
>equalizes at 90W ,with the 70W 60kHz flyback. There is a higher DC
>component of course (about 10-15%) due to a reduced switching cycle,
>the converter is deeper in CCM.
>
>I had a sheet giving approximation on core power throughput vs
>frequency and it shows that for 70W 70kHz increasing the Frequency to
>100kHz the same E-Core could handle a little over 100W. I know I could
>expect higher DC winding losses; in a pinch I could add another
>parallel winding there is already two it would be tight though.
>
The spread-sheet assumes that you are reconfiguring turns, gap and
wire guage for the different operating frequency.

Simply increasing the frequency may reduce flux peaks, but core loss
will not reduce proportionally if the frequency is increased at the
same time. The net result is increased total loss even without an
increased current through-put.

Copper losses increase with the square of the current density.

What actually happens depends on how the original transformer was
designed. If you don't know the core shape, material, turns, wire
gauge, voltage ratios and frequency, there's little point in
speculating.

It'd be faster just to fire up the circuit and measure deltaT as the
frequency-setting component is altered - if the circuit is
pre-existing and you nkow no more about it.

If the circuit is not pre-existing, you'll get better results working
with fewer unknowns. Transformer loss is not the only consideration as
a flyback circuit's throughput power increases.

RL
From: Hammy on 25 Apr 2010 12:04
On Sun, 25 Apr 2010 11:29:52 -0500, legg <legg(a)nospam.magma.ca> wrote:


>>
>The spread-sheet assumes that you are reconfiguring turns, gap and
>wire guage for the different operating frequency.
>
>Simply increasing the frequency may reduce flux peaks, but core loss
>will not reduce proportionally if the frequency is increased at the
>same time. The net result is increased total loss even without an
>increased current through-put.
>
>Copper losses increase with the square of the current density.
>
>What actually happens depends on how the original transformer was
>designed. If you don't know the core shape, material, turns, wire
>gauge, voltage ratios and frequency, there's little point in
>speculating.

It is from this application note from On Semi.

http://www.onsemi.com/pub_link/Collateral/AND8076-D.PDF

This is pretty well all they have for the transformer.

"Transformer

Below are the key parameters you will pass to your
transformer manufacturer to help him select the right
winding size and tailor the internal gap:
Maximum peak primary current, including 160 ns
propagation delay: 1 / 0.33 + 374 � 160 n / 700 m = 3.2 A
Maximum primary RMS current at low line: 1.6 A
Maximum secondary RMS current: 6.9 A
Primary inductance: 700 mH
Turn-ratio, power section: Np:Ns = 1:0.166
Turn-ratio, auxiliary section: Np:Naux = 1:0.15"

I'm not exceeding any of those ratings upto 90W with a safety margin.
If I had to guess and I do it may be either a ETD34 its about 1/2 a cm
longer then an ETD29 core that I have or an EE32.

>It'd be faster just to fire up the circuit and measure deltaT as the
>frequency-setting component is altered - if the circuit is
>pre-existing and you nkow no more about it.

Yes your probably right. I will just hook it up and monitor for
saturation and temp rise.

>If the circuit is not pre-existing, you'll get better results working
>with fewer unknowns. Transformer loss is not the only consideration as
>a flyback circuit's throughput power increases.

Components have improved since they did the AN I'm using a FET with
1/5 the rdson with smaller gate charge then the one they used as well
as sync-rectification.That alone reduces the losses compared to
original design by about 4W.


>RL
From: Don Klipstein on 25 Apr 2010 20:45
In article <r1l8t5dcgfonup1r8drobfe13ad3tqts78(a)4ax.com>, Hammy wrote:

>I have some Flyback transformers (from CoilCraft) these were
>originally intended for use in a 70W 60kHz application using the
>NCP1200. If I increase the switching frequency to 100kHz is it
>reasonable to expect 90-100W using the same core?
>
>I've been running simulations and the rms and peak current for 100kHz
>equalizes at 90W ,with the 70W 60kHz flyback. There is a higher DC
>component of course (about 10-15%) due to a reduced switching cycle,
>the converter is deeper in CCM.
>
>I had a sheet giving approximation on core power throughput vs
>frequency and it shows that for 70W 70kHz increasing the Frequency to
>100kHz the same E-Core could handle a little over 100W. I know I could
>expect higher DC winding losses; in a pinch I could add another
>parallel winding there is already two it would be tight though.
>
>Any magnetic experts out there?

The loss in a ferrite core should be mostly hysteresis loss, which is
roughly proportional to frequency and square of magnetic field intensity.
Power throughput, at least in an oversimplified case, is proportional to
the squares of frequency and magnetic field intensity.

Ideally, ratio of throughput to core loss is proportional to frequency.

Meanwhile, there are other issues:

1. Resistance of the copper will be higher at the higher frequency,
approaching proportional to the square root of frequency once the "skin
depth" gets much smaller than the wire radius.

2. The transformer may have enough inter-layer capacitance to cause
a significant lowpass filter effect. Do you know that it will work at the
higher frequency?

Or are you planning to rewind it?

3. If you rewind it and use fewer turns of thicker wire, keep in mind
that the wire's resistance at the frequency in question may be closer to
inverse proportional to the wire's circumference than to its cross section
area due to the skin effect.

4. The switching transistor's switching loss, as a percentage of power
throughput, is proportional to frequency. The transistor's rise and fall
times will probably be in the tens of nanoseconds, possibly more.

Consider the energy dissipated assuming rise time times half the current
initially conducted by the transistor (possibly zero) times the input
voltage,
plus the fall time times half the current being conducted by the
transistor shortly before switch-off times the voltage that the transistor
experiences during switch-off (always more than the input voltage, usually
by a factor of more than 2, in flyback circuits).

Keep in mind that rise and fall times in transistor datasheets are at
junction temperature of 25 degrees C and with ideal or very good driving
of the transistor. In actual applications, these times are usually
slower (longer periods of time).

Multiply the switching losses by frequency, add conduction losses
(increases with temperature if the tyransistor is a power MOSFET), and
determine if that is going to be too much heat for the transistor to
dissipate. If the heatsinking is currently minimal, then it is *probably*
easy to hack additional or greater heatsinking onto the switching
transistor.

- Don Klipstein (don(a)misty.com)
From: legg on 26 Apr 2010 12:43
On Sun, 25 Apr 2010 12:04:08 -0400, Hammy <spam(a)spam.com> wrote:

>On Sun, 25 Apr 2010 11:29:52 -0500, legg <legg(a)nospam.magma.ca> wrote:
>
>
>>>
>>The spread-sheet assumes that you are reconfiguring turns, gap and
>>wire guage for the different operating frequency.
>>
>>Simply increasing the frequency may reduce flux peaks, but core loss
>>will not reduce proportionally if the frequency is increased at the
>>same time. The net result is increased total loss even without an
>>increased current through-put.
>>
>>Copper losses increase with the square of the current density.
>>
>>What actually happens depends on how the original transformer was
>>designed. If you don't know the core shape, material, turns, wire
>>gauge, voltage ratios and frequency, there's little point in
>>speculating.
>
>It is from this application note from On Semi.
>
>http://www.onsemi.com/pub_link/Collateral/AND8076-D.PDF
>
>This is pretty well all they have for the transformer.
>
>"Transformer
>
>Below are the key parameters you will pass to your
>transformer manufacturer to help him select the right
>winding size and tailor the internal gap:
>Maximum peak primary current, including 160 ns
>propagation delay: 1 / 0.33 + 374 � 160 n / 700 m = 3.2 A
>Maximum primary RMS current at low line: 1.6 A
>Maximum secondary RMS current: 6.9 A
>Primary inductance: 700 mH
>Turn-ratio, power section: Np:Ns = 1:0.166
>Turn-ratio, auxiliary section: Np:Naux = 1:0.15"
>
>I'm not exceeding any of those ratings upto 90W with a safety margin.
>If I had to guess and I do it may be either a ETD34 its about 1/2 a cm
>longer then an ETD29 core that I have or an EE32.
>
Quite frankly, I've never seen a power transformer described in this
manner, nor is the list actually complete without it's surrounding
article, in which topology, operating frequency, input voltage range,
and output voltage are actually mentioned. It has to be assumed that
the part must meet the requirements of some coordinated safety
standard ~ 60950.

The article itself seems a little quirky - focussing more on control
chip pecadiloes than power train. As the part is being made available
for this application, there is little emphasis on practical
transformer design issues.

One example; the choice of primary inductance is made arbitrarily on
the basis of full load transition from complete to incomplete energy
transfer mode (an irrelevant feature) at an arbitrary input voltage,
somewhere (the author hopes) the psu will never actually have to run.
Then this careful calculation is (equally arbitrarily) approximately
doubled.

Another interesting issue is the fact that, in the end, the author
somehow achieves the design spec output power only at a higher output
voltage than intended. Whether this indicates that the final iteration
was incapable of the design spec current and voltage, or that the
author was just to lazy to complete an accurate set of drawings for
his article - is a mystery to me.

You'd be better off haunting the old Unitrode seminar app notes, if
you're interested in flyback transformer design iteration.

http://focus.ti.com/docs/training/catalog/events/event.jhtml?sku=SEM401014

RL
 | 
Pages: 1
Prev: Replacing floppy
Next: Gnats :-(