Design Considerations for SMPS Transformer

Thread Starter

reynolds087

Joined Jul 22, 2020
35
Can you have too many turns in the winding of a primary on a high-frequency ferrite transformer? If so, what problem does it cause? What happens if you don't have enough windings?

I'm trying to build a stepdown power supply from mains voltage down to 32vdc at 150KHz, and the problem is that the voltage drops on the secondary even with a relatively small load and cause a lot of current through the primary. Not sure what is causing it.
 

MaxHeadRoom

Joined Jul 18, 2013
25,229
The HF ferrite core transformer is typically used at a higher switching frequency than 50/60hz.
You need iron core for LF. mains Transformer !
 

Thread Starter

reynolds087

Joined Jul 22, 2020
35
With no load, the output voltage is 32V. So it's at the correct voltage. It just can't drive a load without dropping the voltage. Also, when the load is connected, the unit is drawing about 20-30 watts of power, but only 1 watt is going through the load resistor.
 

Thread Starter

reynolds087

Joined Jul 22, 2020
35
The HF ferrite core transformer is typically used at a higher switching frequency than 50/60hz.
You need iron core for LF. mains Transformer !
Thanks, I know that. I should have included some pictures and schematics. The 60 Hz transformer is only for VCC for the control chip. The switching and power transfer is happening in the high frequency ferrite transformer at 150KHz. I have added a schematic and picture.
 

Thread Starter

reynolds087

Joined Jul 22, 2020
35
I made a mistake on the schematic. There is no bridge rectifier on the low frequency transformer for the chip VCC. It's just a half-wave rectifier using a single 1n4148. Also, the transformer is only rated for 1 watt. Not sure if it matters. I'm not seeing any voltage drop on the supply voltage. I know the mosfets want at least 15v on the gate, so maybe that is affecting the RDS on?
 

Thread Starter

reynolds087

Joined Jul 22, 2020
35
Where is the 19 to 29 watts going? Something must be hot!

I do not know if the feed back is working. I think "fb" needs a 10k pull to ground resistor. The opto pulls up but nothing pulls down.
The mosfets definitely get hot. I don't let it run like that long enough for anything to start smoking. I am also leary of feeling around for hot components because I have a 150K bleed resistor on the mains cap, and it takes about a minute for the voltage to drain after shutdown. I'll try pulling down the inverting input to ground. Voltage is very stable with no load connected and draws about 4 watts quiescent current. Once the load is connected, then I see the power spike even though the load only sees about 1 watt.
 

Thread Starter

reynolds087

Joined Jul 22, 2020
35
If the core was saturating, would that cause the voltage on the secondary to drop, and the mosfets to dissipate tons of power in the primary? It kind of seems like the power isn't transferring to the secondary the way it should be. Like I said, the voltage can get to 32V on the output side, but any load just kills it.
 

Ian0

Joined Aug 7, 2020
5,185
1. Are you sure that's a ferrite core, not an iron powder core?
2. There only seems to be one layer of windings on half the core. You need the secondary to cover the entire primary, otherwise you will have a lot of leakage inductance, and the coupling between primary and secondary will be poor resulting in the voltage dropping when a load is applied.
 

Thread Starter

reynolds087

Joined Jul 22, 2020
35
1. Are you sure that's a ferrite core, not an iron powder core?
2. There only seems to be one layer of windings on half the core. You need the secondary to cover the entire primary, otherwise you will have a lot of leakage inductance, and the coupling between primary and secondary will be poor resulting in the voltage dropping when a load is applied.
It could be an iron powder core. Is there any way to differentiate the two? What would the result be of using an iron powder core as a transformer? Is manganese zinc ok for a transformer, because I think I have some of those.

If I have much less windings for the secondary, do I just need to approximate the coverage, like space the windings farther apart so they traverse the entire toroid? Also, with a center tapped transformer would the center tap be right next to two endpoints in order to make the windings both do a full rotation around the circumference, or should each winding go halfway around?
 

LowQCab

Joined Nov 6, 2012
2,060
With the 100's of things that could be off, my first guess would be insufficient Gate-Drive.

Your Low-Voltage-Supply just ain't gonna get it.

100-Ohm Gate-Resistors at 150khz is probably a good portion of the problem.

The FETs are also getting ~90% of what ever the Low-Voltage-Supply is drooping-down to.
You don't know that that Supply is adequate without looking at it with a Scope while it's running,
or at the very least, by doing the most realistic simulation that You can in Software.

Also, building a Transformer is an "Art-Form", not something that You can do by-guess-and-by-golly.
I would suggest calling a Custom-Transformer-Manufacturer and letting their Computer-Software
design exactly what You need.

You never mentioned how much Secondary-Current You were expecting at ~30-Volts.
.
.
.
 

Thread Starter

reynolds087

Joined Jul 22, 2020
35
With the 100's of things that could be off, my first guess would be insufficient Gate-Drive.

Your Low-Voltage-Supply just ain't gonna get it.

100-Ohm Gate-Resistors at 150khz is probably a good portion of the problem.

The FETs are also getting ~90% of what ever the Low-Voltage-Supply is drooping-down to.
You don't know that that Supply is adequate without looking at it with a Scope while it's running,
or at the very least, by doing the most realistic simulation that You can in Software.

Also, building a Transformer is an "Art-Form", not something that You can do by-guess-and-by-golly.
I would suggest calling a Custom-Transformer-Manufacturer and letting their Computer-Software
design exactly what You need.

You never mentioned how much Secondary-Current You were expecting at ~30-Volts.
.
.
.
Thank you for the suggestions. I have a higher voltage 60 Hz transformer I can try for VCC. What gate resistor value would work at 150KHz? Is there a formula to calculate it?

I definitely am not going to have a transformer designed. I'm literally just breadboarding this power supply as a learning experience. I've simulated the circuit and it worked fine, but of course that doesn't account for the parasitics in my breadboard setup, "less than ideal" transformer, and any other errors I've introduced by mistake.

Ideally I'd like to get to around 100 watts or so. I tried a ferrite core I bought from the local electronics surplus, and was able to drive a 100 ohm load without any voltage drop. They sell a 50mm core with a 6mm cross section for about $1 a piece. Very good prices there. When I tried 60 ohms I started getting voltage drop.

How do I calculate the ideal inductance of the primary? Can I use this formula?
Primary_Turns = Voltage_Input * 10^8 / 4 * frequency * Max_Flux_Density * Cross_Section(Centimeters)
 

LowQCab

Joined Nov 6, 2012
2,060
I haven't tried your Transformer Formula so I can't comment on it.

My personal rules, if not using a Custom-made-Transformer,
are that You select the Core-Material based on Frequency-Range first,
then, the Core can't be too big, bigger is always better.

Core-Size / Material will determine the Maximum-Power that the Transformer can handle.

Power-Handling usually goes up with Frequency, but
Switching and Filtering losses start to become critical over about ~10khz.

Stacking 2-Cores works just fine to increase Core-Volume.
You want to stay well away from Core-Saturation under full-Load,
otherwise, the wasted Power starts going through the roof.

Stop adding Primary-Windings when your "No-Load", or "Light-Load" Current is acceptably low,
then make sure that the Primary-Wire-Gauge is adequate for your expected Primary-Current,
without too much resistance, which generates excessive Heat.

After You get the Quiescent-Current down to an acceptable level,
( by effectively increasing the Inductance with more turns of Wire ),
You can start adding Turns to the Secondary until You get the Voltage that You want under Full-Load.

If your Core-Material choice is "close" to acceptable, and
You have an adequate, ( or excess ), Core-Mass for your calculated Maximum-Power,
You will then have a reasonably acceptable Transformer.

I would also drop the Frequency back down to ~50khz or so to reduce Switching-Losses,
and make all the other parts less critical.

If You are Powering a DC-Load, especially with Electronics involved,
You will need a properly designed "Choke-Input" Filter on the Output, preferably 2-Stages.

What are You Powering, and why does it require ~100-Watts of Power ???
What's the part-number of your Switching Controller ?
What's the part-number of your FETs ?
.
.
.
 

Ian0

Joined Aug 7, 2020
5,185
It could be an iron powder core. Is there any way to differentiate the two?
Is it painted two colours? If so, definitely iron powder. Any other markings printed on it?
Otherwise, wind a few turns on it an measure the inductance, and work out the permeance (inductance/turns^2)
Iron powder is in the nH range, ferrite is in the μH range.

What would the result be of using an iron powder core as a transformer? Is manganese zinc ok for a transformer, because I think I have some of those.
Low permeance means low inductance, which means high off-load current.
If I have much less windings for the secondary, do I just need to approximate the coverage, like space the windings farther apart so they traverse the entire toroid? Also, with a center tapped transformer would the center tap be right next to two endpoints in order to make the windings both do a full rotation around the circumference, or should each winding go halfway around?
Spread the winding over the entire core. Better still, wind thin wires in parallel and spread them out.
Each half of the centre-tapped winding should cover the entire circumference.
Poor coupling between the two halves of the primary causes big spikes when the MOSFETs turn off.

And 100Ω gate resistors are much too high for that frequency, and 1k pull-downs are much much too high. I would not recommend the 3524 because it is not designed for driving MOSFETs. The 3525 has proper MOSFET drivers.

A final point - and you are a long way from this - but your feedback will not be stable with no capacitor on COMP.
When using an opto, connect the + input of the error amp to REF, connect the -input to ground, and connect the opto between COMP and ground.
The error amp won't mind because it is a transconductance amplifier with limited output current.
You will need a capacitor across the opto.
 
Top