transformers for cheap low current flyback supply, +/- 45V ?

[Ken Smith]

Does anyone have recommendations for a general purpose flyback
transformer? I want to get output voltages of +45 and -45 VDC, at only
10mA or so. Input will be 12V or so.

If it doesn't have to be isolated, consider making it a SEPIC and CUK,
perhaps with multiplier outputs. This way you can just use off the shelf
power inductors.



---)))))---+----------!!---+--->!---+----- +V
! ) !
! ) ---
! ) ---
! ! !
! GND GND
!
!
+----------!!---+---))))--+---- -V
! ! !
!!- V ---
---!!- --- ---
!!- ! !
! GND GND
GND



12V in and 48V out is near the pratical upper limit for the ratio. Adding
doubling increases the parts count quite a bit like this:


---)))))---+----------!!------------+--->!---+----- +V
! ) !
! ) ---
! ) ---
! ! !
! ! GND
! !
+----------!!---+--->!---+
! ) !
! ) ---
! ) ---
! ! !
! GND GND
.... etc ...

 

[mw]

If it doesn't have to be isolated, consider making it a SEPIC and CUK,

perhaps with multiplier outputs. This way you can just use off the shelf
power inductors.



---)))))---+----------!!---+--->!---+----- +V
! ) !
! ) ---
! ) ---
! ! !
! GND GND
!
!
+----------!!---+---))))--+---- -V
! ! !
!!- V ---
---!!- --- ---
!!- ! !
! GND GND
GND



12V in and 48V out is near the pratical upper limit for the ratio. Adding
doubling increases the parts count quite a bit like this:


---)))))---+----------!!------------+--->!---+----- +V
! ) !
! ) ---
! ) ---
! ! !
! ! GND
! !
+----------!!---+--->!---+
! ) !
! ) ---
! ) ---
! ! !
! GND GND
... etc ...

That's pretty clever. It doesn't need to be isolated from the mains, so
had thought about using a boost circuit. But I wasn't sure how to
derive the negative rail. It looks complicated to get the -45V to be
inside the PWM controller's control loop. But really, my requirements
aren't too critical (it doesn't matter if it is -42, or -46, as long as
it is close to -45V and reasonably stable).

What I could do is regulate using the +45V side as the feedback, then
let the negative run open-loop but derived from the same switching
waveform. I'd really like to avoid hand-winding of transformers, that's
a pain, and this approach would avoid it entirely.

Thanks to everyone that responded. There's lots of ways to get this to
work since my requirements are very loose.

mw

 

[Ken Smith]

That's pretty clever. It doesn't need to be isolated from the mains, so
had thought about using a boost circuit. But I wasn't sure how to
derive the negative rail. It looks complicated to get the -45V to be
inside the PWM controller's control loop.

You can only regulator one or the other in this case. To regulate the
minus supply, you most likely need an inverting op-amp circuit.

But really, my requirements
aren't too critical (it doesn't matter if it is -42, or -46, as long as
it is close to -45V and reasonably stable).

If the loads are fairly closely matched, the regulator can be quite good
on the open loop output. By its nature, the curcuit makes the drain
waveform have quite flat tops and bottoms, if the coupling capacitors are
large enough.

Be careful about the placement of the inductors. Chances are, you will be
using unshielded ones. If they are close together, there will be some
mutual coupling. Ideally, you want the mutual coupling to add to the
value not subtract and to be equal between the two supplies.

 

[Jim Thompson]

That's pretty clever. It doesn't need to be isolated from the mains, so
had thought about using a boost circuit. But I wasn't sure how to
derive the negative rail. It looks complicated to get the -45V to be
inside the PWM controller's control loop. But really, my requirements
aren't too critical (it doesn't matter if it is -42, or -46, as long as
it is close to -45V and reasonably stable).

What I could do is regulate using the +45V side as the feedback, then
let the negative run open-loop but derived from the same switching
waveform. I'd really like to avoid hand-winding of transformers, that's
a pain, and this approach would avoid it entirely.

Thanks to everyone that responded. There's lots of ways to get this to
work since my requirements are very loose.

mw

If you're coming directly off the mains, you could try something like
this...

OldStyleBuckSwitcherWithAddedNegativeOutput.pdf

on the S.E.D/Schematics page of my website.

I did this in a old GenRad portable tester.

Negative voltage IS regulated.

...Jim Thompson

 

[Terry Given]

mw wrote:




You omitted to say what frequency you want to use, what power level
you need and what the line input voltage is for starters.

Graham

look harder Graham. output +/-45V @ 10mA, input 12V(ish)....

Cheers
Terry

 

[Terry Given]

mw wrote:




WRONG !
MPP?

KoolMu?

#52?


You need an air gap with a flyback design.

Go learn some basics.

Go to www.powerint.com and read *all* the app notes. That's how I finally
learnt the multiple subtleties of flyback designs.




No you're not engaging your brain.

Do you think transformers are the same @ 20kHz and 1MHz ?




No it never was.

Graham

oops, Vishay had best remove the numerous product lines they have,
closely followed by coilcraft.

Philips (or whoever they became) make an Integrated Inductive Component
(IIC) that may also do the trick.

Cheers
Terry

 

[Terry Given]

mw wrote:




To answer your question directly.

There is no such thing as a universal flyback transformer.

Every one has to be designed individually taking into account operational
frequency, input voltage, power and desired output voltages.

isolation is really the important part, followed by turns ratio.

For a DCM flyback, Pin = (Vinmin*Dmax)^2/[2*Lmag*Fsmps]

so for a given Vinmin, Lmag you can twiddle Dmax and Fsmps to get Pin.
until things like minimum pulse widths and saturation rear up.

Not like your 50/60 Hz transformer at all.

Go to www.powerint.com and read their app notes and get their magnetics
designer program.

Graham

Cheers
Terry

 

[Phil Allison]

"Terry Given"

look harder Graham. output +/-45V @ 10mA, input 12V(ish)....

** Well, that is a very basic spec - but gives no clue as to the app.

The OP will not supply any clue, so has branded himself a TROLL.



.......... Phil

 

[Terry Given]

"Terry Given"




** Well, that is a very basic spec - but gives no clue as to the app.

The OP will not supply any clue, so has branded himself a TROLL.

......... Phil

Hi Phil,

its enough info to design a power supply....

Cheers
Terry

 

[Phil Allison]

"Terry Given"
Phil Allison

Hi Phil,

its enough info to design a power supply....

** Hardly the point at issue.

It is NOT enough info for me to recommend a suitable transformer.

Hiding the app plus not knowing that an app is not the same as a raw spec
makes the OP look like a time wasting troll and a fake.




......... Phil

 

[Terry Given]

"Terry Given"
Phil Allison




** Hardly the point at issue.

It is NOT enough info for me to recommend a suitable transformer.

Hiding the app plus not knowing that an app is not the same as a raw spec
makes the OP look like a time wasting troll and a fake.


........ Phil

Perhaps. The app may be confidential, who knows. More info on the
requirements would help optimise the xfmr, for sure.

I once got asked to peer-review a guys smps "design." His output 0V was
at the same potential as the -ve DC bus (rectified mains), so about
-160V. it was a battery charger.... he also used IF transformers as
little smps xfmrs for gatedrives

needless to say the review was less than favourable, and the word
"lethal" featured prominently in the first sentence.

Cheers
Terry

 

[Phil Allison]

"Terry Given"

Perhaps. The app may be confidential, who knows.

** Well, one thing is for sure - you don't.

Maybe it top secret 007 James Bond type stuff ???

More info on the requirements would help optimise the xfmr, for sure.

** Do you even know that the OP asked me for ??

Give even a tiny hoot ??

No need to reply.





........... Phil

 

[mw]

oops, Vishay had best remove the numerous product lines they have,
closely followed by coilcraft.

Philips (or whoever they became) make an Integrated Inductive Component
(IIC) that may also do the trick.

Cheers
Terry

Looking at your suggestions I found that Coilcraft does have a good
assortment of transformers, for example these:

http://www.coilcraft.com/pdfs/lm258x.pdf
http://www.coilcraft.com/pdfs/ifly0012.pdf
http://www.coilcraft.com/pdfs/a9967.pdf

I will have to look closely at them to see if any can be made to work,
and whether they can be obtained without a lot of hassle. I don't see
any of them listed using findpart.com , so that's not so good, could be
that you have to get personal with sales folks. ($$$).

This afternoon I spent some time working on my supply, using an old
board of mine (it uses a difficult-to-obtain custom flyback
transformer), and was able to get it working a little differently than
it did in its original +/-14V output configuration.

Some results:

Input (feeding the UCC PWM chip) : +17V through a 402 Ohm resistor. The
PWM chip and the FET together only draw 9.4mA!
Input (feeding the transformer) : +43V
Output (pos) : +56.2V
Output (neg) : -56.9V
That's today's set up, it will probably change.

The freq of the PWM is 57kHz. I find that PWM freq isn't too critical.
That's one thing nice about closed-loop controllers, they will work
reasonably well over a broad range of f, if you don't have tight req'ts.
I don't.

I'm getting a nice waveform on the current sense line that shows current
ramping up to about 1.36V during a 1.28us interval (while the FET is
on). I thought it should turn off at 1V, but maybe not. My current
sense resistor is 1.05 Ohms. So let's see, that would be peak current,

ipk = 1.36V/1.05 Ohm = 1.42A.

Let's assume that the current waveform is flat, rather than a ramp,
just to make things easier to calculate and to allow some slop for worst
case conditions. So current being drawn from the 43V supply is 1.42A
but only during the time 1.28 us interval (out of the 17.5 us PWM
period), or about 7.3%. So the worst case total power into the
transformer is about (43V * 1.42A) * 7.3% = 2.15W. That's not bad, and
that's about how hot things seem. Just noticeably warm to the touch.

OK. The "application" circuit that uses this draws about 4mA * (56V *
2) roughly, or about 448mW. So this supply will work pretty well, and
it does. I tried it out during heavy usage of the "application" and
found that the regulated +56.2V rail holds together very well, with <
50mV of noise. The unregulated negative rail does sag a little, sagging
toward 0 by about 0.5V. So that's not so good, but not a crisis.

One good thing is that the PWM isn't creating noise for my
"application", although it's probably radiating EMI big time. I have it
wired in using various 6" to 18" wires, but both boards (the PWM board
and the "application" board) have sweet ground planes.

Thanks to all,
mw

 

[Terry Given]

Looking at your suggestions I found that Coilcraft does have a good
assortment of transformers, for example these:

http://www.coilcraft.com/pdfs/lm258x.pdf
http://www.coilcraft.com/pdfs/ifly0012.pdf
http://www.coilcraft.com/pdfs/a9967.pdf

I will have to look closely at them to see if any can be made to work,
and whether they can be obtained without a lot of hassle. I don't see
any of them listed using findpart.com , so that's not so good, could be
that you have to get personal with sales folks. ($$$).

try digikey or mouser, ISTR one of them does coilcraft.

This afternoon I spent some time working on my supply, using an old
board of mine (it uses a difficult-to-obtain custom flyback
transformer), and was able to get it working a little differently than
it did in its original +/-14V output configuration.

Some results:

Input (feeding the UCC PWM chip) : +17V through a 402 Ohm resistor. The
PWM chip and the FET together only draw 9.4mA!
Input (feeding the transformer) : +43V
Output (pos) : +56.2V
Output (neg) : -56.9V
That's today's set up, it will probably change.

The freq of the PWM is 57kHz. I find that PWM freq isn't too critical.
That's one thing nice about closed-loop controllers, they will work
reasonably well over a broad range of f, if you don't have tight req'ts.
I don't.

I'm getting a nice waveform on the current sense line that shows current
ramping up to about 1.36V during a 1.28us interval (while the FET is
on). I thought it should turn off at 1V, but maybe not. My current
sense resistor is 1.05 Ohms. So let's see, that would be peak current,

it really should. you should have a nice triangular waveform, the slope
of which should be Rsense*[Vin/Lmag].

if the "slopey bit" isnt a straight line, and curves up, the core is
saturating. usually thats what causes Vsense to rise above 1V.

ipk = 1.36V/1.05 Ohm = 1.42A.

Let's assume that the current waveform is flat, rather than a ramp,
just to make things easier to calculate and to allow some slop for worst
case conditions. So current being drawn from the 43V supply is 1.42A
but only during the time 1.28 us interval (out of the 17.5 us PWM
period), or about 7.3%. So the worst case total power into the
transformer is about (43V * 1.42A) * 7.3% = 2.15W. That's not bad, and
that's about how hot things seem. Just noticeably warm to the touch.

dreadful approximation.

hows about this:

firstly, the average input current * Vin = input power.

energy stored in core at end of on-time is:

Ecore_end = 0.5*Lmag*Ipeak^2

if the current sense waveform is a triangle (complete energy transfer,
aka Discontinuous Conduction Mode), then the energy stored in the core
at the beginning of the on time is:

Ecore_start = 0.5*Lmag*0^2 = 0

otherwise the waveform is a trapezoid, so

Ecore_start = 0.5*Lmag*I_min^2

energy transferred to load is Eload = Ecore_end - Ecore_start

power transferred to load is Eload*Fsmps.

For a DCM flyback, if you do the maths you get:

Pin = (Vinmin*Dmax)^2/[2*Lmag*Fsmps]
Ipeak = (Vinmin*Dmax)/[Lmag*Fsmps]

OK. The "application" circuit that uses this draws about 4mA * (56V *
2) roughly, or about 448mW. So this supply will work pretty well, and
it does. I tried it out during heavy usage of the "application" and
found that the regulated +56.2V rail holds together very well, with <
50mV of noise. The unregulated negative rail does sag a little, sagging
toward 0 by about 0.5V. So that's not so good, but not a crisis.

its pretty good. did you bifilar wind the two secondaries - that'll help
the cross-regulation.

One good thing is that the PWM isn't creating noise for my
"application", although it's probably radiating EMI big time. I have it
wired in using various 6" to 18" wires, but both boards (the PWM board
and the "application" board) have sweet ground planes.
urgh!


Thanks to all,
mw

NWD

Cheers
Terry

 

[Ken Smith]

OK. The "application" circuit that uses this draws about 4mA * (56V *
2) roughly, or about 448mW. So this supply will work pretty well, and
it does. I tried it out during heavy usage of the "application" and
found that the regulated +56.2V rail holds together very well, with <
50mV of noise. The unregulated negative rail does sag a little, sagging
toward 0 by about 0.5V. So that's not so good, but not a crisis.

Have you considered this:

---------+------->!----- +V
! !
)!!( ---
)!!( ---
)!!( !
)!! ! !
)!! ---------------------+---- GND
)!! ! !
)!! ---------+------->!--
)!! !
)!!(
)!!(
)!!(
!
-------------------------- -V


You can get light load to full load tracking that is well under a diode
drop this way. It does increase the ripple current in the -V's filter cap
a bit but usually not enough to matter.

 

[Ken Smith]

mw wrote: [...]
try digikey or mouser, ISTR one of them does coilcraft.

Coilcraft will sample parts. I believe RencoUSA does too.

J.W. Miller is known mostly for their RF parts but they also make power
inductors. These are carried by digikey.

it really should. you should have a nice triangular waveform, the slope
of which should be Rsense*[Vin/Lmag].

.... for some values of "nice". There is a spike at the turn on due to
interwinding capacitances etc. Sometimes the spike will have a bit of
ringing after it.

if the "slopey bit" isnt a straight line, and curves up, the core is
saturating. usually thats what causes Vsense to rise above 1V.

If the slope decreases with time(enough to see on a scope), you have way
too much resistance in the primary side unless:

Many years back I did a supply that made about 1KW running at 4KHz. On
one prototype inductor, the slope slightly decreased. Grabbing the
transformer with my hand changed the shape. We just happened to line a
mechanical resonance up with the running frequency. Doing up the bolt
fixed it.

its pretty good. did you bifilar wind the two secondaries - that'll help
the cross-regulation.

I second the vote for bifilar if the option is open to you.

 

[mw]

it really should. you should have a nice triangular waveform, the slope
of which should be Rsense*[Vin/Lmag].

if the "slopey bit" isnt a straight line, and curves up, the core is
saturating. usually thats what causes Vsense to rise above 1V.

It's mostly like a straight line. Not perfect, but I think my probe is
picking up a lot of EMI.

dreadful approximation.

hows about this:

firstly, the average input current * Vin = input power.

energy stored in core at end of on-time is:

Ecore_end = 0.5*Lmag*Ipeak^2

if the current sense waveform is a triangle (complete energy transfer,
aka Discontinuous Conduction Mode), then the energy stored in the core
at the beginning of the on time is:

It's in discontinuous mode.

Ecore_start = 0.5*Lmag*0^2 = 0

otherwise the waveform is a trapezoid, so

Ecore_start = 0.5*Lmag*I_min^2

energy transferred to load is Eload = Ecore_end - Ecore_start

power transferred to load is Eload*Fsmps.

For a DCM flyback, if you do the maths you get:

Pin = (Vinmin*Dmax)^2/[2*Lmag*Fsmps]
Ipeak = (Vinmin*Dmax)/[Lmag*Fsmps]

Thanks for the math. I don't have the Lmag available to me here at
home, but I will keep this in mind when I look at the other
transformers. I know that this transformer works well in its original
application (it supplies about 8W, and doesn't saturate).

its pretty good. did you bifilar wind the two secondaries - that'll help
the cross-regulation.

I am not sure how it is wound (don't have the data sheet here at home),
but it's a recent design by a custom house, and they do pretty thorough
work.


Thanks for your help Terry,
mw

 

[Terry Given]

it really should. you should have a nice triangular waveform, the
slope of which should be Rsense*[Vin/Lmag].

if the "slopey bit" isnt a straight line, and curves up, the core is
saturating. usually thats what causes Vsense to rise above 1V.

It's mostly like a straight line. Not perfect, but I think my probe is
picking up a lot of EMI.

dreadful approximation.

hows about this:

firstly, the average input current * Vin = input power.

energy stored in core at end of on-time is:

Ecore_end = 0.5*Lmag*Ipeak^2

if the current sense waveform is a triangle (complete energy transfer,
aka Discontinuous Conduction Mode), then the energy stored in the core
at the beginning of the on time is:

It's in discontinuous mode.

Ecore_start = 0.5*Lmag*0^2 = 0

otherwise the waveform is a trapezoid, so

Ecore_start = 0.5*Lmag*I_min^2

energy transferred to load is Eload = Ecore_end - Ecore_start

power transferred to load is Eload*Fsmps.

For a DCM flyback, if you do the maths you get:

Pin = (Vinmin*Dmax)^2/[2*Lmag*Fsmps]
Ipeak = (Vinmin*Dmax)/[Lmag*Fsmps]

Thanks for the math. I don't have the Lmag available to me here at
home, but I will keep this in mind when I look at the other
transformers. I know that this transformer works well in its original
application (it supplies about 8W, and doesn't saturate).

if current slope is a straight line and Vdc is constant, you have
already measured it.... V = LdI/dt => L = Vdt/dI

I am not sure how it is wound (don't have the data sheet here at home),
but it's a recent design by a custom house, and they do pretty thorough
work.

If you have a spare, tear one to pieces. I have long since lost count of
how many transformers my leatherman and I have destroyed

 

[Terry Given]

mw wrote:
[...]

try digikey or mouser, ISTR one of them does coilcraft.

Coilcraft will sample parts. I believe RencoUSA does too.

J.W. Miller is known mostly for their RF parts but they also make power
inductors. These are carried by digikey.

it really should. you should have a nice triangular waveform, the slope
of which should be Rsense*[Vin/Lmag].

... for some values of "nice". There is a spike at the turn on due to
interwinding capacitances etc. Sometimes the spike will have a bit of
ringing after it.

I just did a bit of consulting work, on excactly this. customer was
concerned they had a wiggly line not a straight line. turns out Ton was
about 200ns at min line.... I suspect the wiggle is undamped gate current.

 

[mw]

Have you considered this:

---------+------->!----- +V
! !
)!!( ---
)!!( ---
)!!( !
)!! ! !
)!! ---------------------+---- GND
)!! ! !
)!! ---------+------->!--
)!! !
)!!(
)!!(
)!!(
!
-------------------------- -V


You can get light load to full load tracking that is well under a diode
drop this way. It does increase the ripple current in the -V's filter cap
a bit but usually not enough to matter.

I'll take a look using my simulator. It would be nice to avoid
regulating the neg rail.

Thanks,
mw