Early stage building a 800W SMPS

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Started looking at building an amplifier and discovered that transformers really got expensive, +/- $200 from RS Components for a 8A, 55V X 2 toroidal transformer.

So its time to start building an SMPS here are my specs.

Specs:
=====
Application: For a Home Theater Subwoofer Project
Power Rating: +/- 57V 8A, +/- 80v, 800W
Topology: Half Bridge Configuration
Unregulated
Common Sense Protection

Parts:
====

Transformer: ETD 49, EPCOS N87 ETD 49 Ferrite Core Transformer
MOSFETS: SIHG20N50C-E3 N-Channel MOSFET, 20 A
Frequency: 110 KHz (approx)
Flux Density: 0.17
Controller: FAN7387MX with 3A output buffer pairs ZXTN25100BFHTA/ZXTP25100BFHTA
Stand By Circuit: LNK304GN (HV BUCK)


I found rather lots of material on the internet lots of complete pro audio designs from service manuals, so I'm using mostly proven reverse engineered reference material instead of incomplete/poor designs I have no idea why so many never actually look at pro audio service manuals for inspiration.


I attempted to start with my transformer here is my first attempt.

Primary: 20T x2 1mm
Secondary: 8T X5 0.65mm

also shown is thick Kapton tape between layers and with 0.5mm heat shrink tubing for the wires exiting the core.
 

Attachments

  • transformer.PNG
    transformer.PNG
    250.9 KB · Views: 428
  • transformer_done.PNG
    transformer_done.PNG
    771.5 KB · Views: 413
Last edited:
If you want better coupling you could split the primary in 10T + 10T and place the secondary between the 2 split primaries. any plans to convert this design into a resonant one ?
True, but for now I'm not sure what the behavior of the transformer is so I just went low coupling. honestly I'm not sure what the effects are.

As for resonant converters, I tried looking at it but its to complex for me and risky to start at level 10, I will move to resonant converters after this attempt.
 
Have you considered creepage-distances from pri. to sec.? I'm no expert, but i think you need either triple-isolated wire or some distance (4-6mm) from pri. to sec.

Hop it helps a little.

Kind regards TroelsM

Input is always appreciated, I used 4-mm margins with 4 layers of Kapton tape, as for the wire, I don't know about the quality of the enamel copper wire.
 
18 and 22 Ga might be too large. 110Khz is pretty fast, are you getting 100% skin depth? Here is a chart that might be helpful. Perhaps multiple strands of thinner wire or Litz might be better at that switching frequency. I used 3 strands of 24Ga wire for offline transformer at 40Khz in a PQ32/30 core. 120V mains. Works great.:D. Coupling issues and leakage inductance can be minimized by splitting the primary and winding each half on either side of secondary(s) like a sandwhich. So: 1/2 primary, cover with proper thickness of insulation tape. Then secondary winding(s), then proper thickness of insulation tape. Then other half of primary on top, then insulate. Keep in mind more primary wire length will be on outside because of larger winding diameter. Make sure all windings are bi-filler. I use type 1350F-2 tape from 3M.
 
Last edited:
The photos seem to show no margin, or margin tape for the first winding (primary), and no margin tape for the secondary winding. Is there just 4mm margin provided nominally on each side of the secondary winding?

Are you just guessing at whether the coreset is meant to be ungapped, or a particular gap, or did you reverse engineer the required primary inductance, and cores size, and core material and losses based on no interleaving?
 
The photos seem to show no margin, or margin tape for the first winding (primary), and no margin tape for the secondary winding. Is there just 4mm margin provided nominally on each side of the secondary winding?

Are you just guessing at whether the coreset is meant to be ungapped, or a particular gap, or did you reverse engineer the required primary inductance, and cores size, and core material and losses based on no interleaving?

Thank you for the feedback.

I'm busy with another identical transformer interested in observing the effects of poor coupling vs good coupling (sandwiched) and observing the performance impacts I read about it but never seen it for myself.

I didn't have margin tape at hand, I'm waiting on delivery from fleebay and its taking very slow for delivery (any idea what I can use instead?, I used a 4mm heat resistant tape, I will try and do better with the margins, this is my first attempt.

As for guessing core data, no, I used ExcellentIT to determine the winding count (primary/secondary) , wire thickness, effects of switching, flux density ect The core data is EPCOS/TDK its N87 material, 1mm gapped.
 
Have you considered creepage-distances from pri. to sec.? I'm no expert, but i think you need either triple-isolated wire or some distance (4-6mm) from pri. to sec.

Hop it helps a little.

Kind regards TroelsM

It looks dangerous to me too.
An overheating transformer could short primary to secondary with disasterous results.

A split bobbin is the way to go if possible.
 
As for resonant converters, I tried looking at it but its to complex for me and risky to start at level 10, I will move to resonant converters after this attempt.

I had fun with my first LLC converter with exploding mosfets !
The key is measuring the LLC parts exactly so you dont go into the danger zone where the LLC becomes a short below a set frequency.

I used a PIC micro to control the frequency depending on feedback signal.
In the end it worked very well.
I ran the frequency very high on power up until capacitors got charged.
Then changed the frequency depending on feedback voltage to regulate.
 
I had fun with my first LLC converter with exploding mosfets !
The key is measuring the LLC parts exactly so you dont go into the danger zone where the LLC becomes a short below a set frequency.

I used a PIC micro to control the frequency depending on feedback signal.
In the end it worked very well.
I ran the frequency very high on power up until capacitors got charged.
Then changed the frequency depending on feedback voltage to regulate.

Interesting curious to know how you managed to optimize for ZVS and ZCS did you do any modeling ? Anyway I'm not ready for that journey I have more basic things to cover like optimizing magnetics, correct winding techniques, safety and other pressing matters for now I will keep things basic.
 
Guys, why not to help him in place of criticize only? Let him to try.

100KHz is pretty fine for modern MOSFET's available and most PWM drivers.

Which topology will you use? Current mode or voltage mode? Take into account, that, half bridge needs choke input filter and this, together to the output cap(s) will place a double pole in the voltage transfer function, making it slow to correct output voltage on loads demands.

Several years ago (+ 20) I made a ≈80WSPMS using current mode (UC3842) and STPS7N60 giving ±14V @ 3amps and a 12V @ 2A, for TDA2030 power stages, working at discontinuous mode and about 60KHz. Secondary side voltage sensing via TL431 and PC817. The thing has been designed to work into the 220VAC line, but once started it continues working OK down to only 24VAC (Twenty four volts) on line, with load. Marvelous.


It was very helpful for me the Abraham Pressman's Book.
 
Last edited:
I forgot to say that some years ago I made a small experimental SMPS only with tubes (No solid state at all) using the SEPIC +CUK topology, with good results. Final stage was a 33GY7A (My avatar from there) and a 35W4 for the Cuk side. I have pics and a witness, a professor of the university where I graduated. Running at 100KHz, it converted +275V into ±350VDC pretty fine.
 
Guys, why not to help him in place of criticize only? Let him to try.

100KHz is pretty fine for modern MOSFET's available and most PWM drivers.

Which topology will you use? Current mode or voltage mode? Take into account, that, half bridge needs choke input filter and this, together to the output cap(s) will place a double pole in the voltage transfer function, making it slow to correct output voltage on loads demands.

Several years ago (+ 20) I made a ≈80WSPMS using current mode (UC3842) and STPS7N60 giving ±14V @ 3amps and a 12V @ 2A, for TDA2030 power stages, working at discontinuous mode and about 60KHz. Secondary side voltage sensing via TL431 and PC817. The thing has been designed to work into the 220VAC line, but once started it continues working OK down to only 24VAC (Twenty four volts) on line, with load. Marvelous.


It was very helpful for me the Abraham Pressman's Book.

Thanks for the encouragement.

I have a few a books (Abraham Pressman's Book in my list) also many good datasheets and app notes, I try to keep questions focused on things I'm struggling with even after exhausting internet research.

Nope. Voltage and current modes are the two "regulating conditions" that control the output of the supply, I'm using nether, as most audio power supplies are unregulated, the chip I went with is self oscillating (mostly marketed as a ballast controller for lighting applications) but may be used for SMPS applications, I'm tired of looking at TL494, UC3842 and SG3525 diy internet designs (they almost always used without regulation in mind with their respective error amplifiers almost always defeated) no problem with these chips... they always need a GDT transformer or a driver to achieve high side bootstrap floating.

In my case I'm using a FAN7387MX it has interesting cost effective gains, like its internal high side bootstrap driver and the ability to driver heavy Qg older fets with totem poles ZTX buffers. External comparators like a 0.5$ LM393 are used to supervise input rails, act as schmitt trigger with OPTO couplers signalling shutdown on the HOT side.

Its strange, I keep seeing LLC suggestion claims but so far eva has probably presented most practical and theoretical data, others hasn't mastered traditional SMPS topologies yet keep dumping primitive brute force claims with knowledge gaps. ouch.
 
Last edited:
Voltage and current modes are the two "regulating conditions" that control the output of the supply.

No guy. Voltage or current mode designs are different in the behavior and the performance of the supply, not only a regulation mode. Voltage mode uses the ramp in a oscillating cap to compare with the amplified error signal from opamp, while current mode uses the inductor's (trafo or fly back inductor) to make the comparation, which in turns eliminates one pole and one zero from the transfer loop, making it quicklier than voltage mode and also it corrects for line variations cycle by cycle.
 
Voltage mode uses the ramp in a oscillating cap to compare with the amplified error signal from opamp

Thanks for the feedback Osvaldo, I assume you referring to the internal mechanics of the controller chip then yes? or are you saying the "voltage feedback" node taken from outside the converter feeding in and comparing that reference to the internal voltage ramp.


I started with the controller module last night it looks like this, component values has not yet been assigned as I'm still very much busy with the redoing my calculations.

Converter Topology : Half-Bridge
Nominal input voltage (Vin) : 220VAC
Nominal output power (Pout) : 1000 W
Nominal output voltage (Vout) : 150V
Target efficiency (η) : > 90%
Switching frequency (f) : 100 kHz

Assuming 90% efficiency the input power is
Pin = Pout/0.9= 1111W, 1.1Kw

Maximum average input current
Iin = Pin / Vin(min) = 1111 / 220 = 5.05A(avg)

Since dead-time has to be provided in order to avoid
simultaneous device conduction, it is better to choose
the maximum duty cycle of each phase as:
Dmax = 0.9 * (t*on/T) = 0.45, 45%

Maximum input RMS current
Iin(rms) = Ipft * sqrt(Dmax) = 5.31A(rms)

Maximum MOSFET RMS current
MosRMS = Ipft * sqrt(Dmax) = 5.31 * sqrt(0.9) = 5A

Vmax = 250ac * 1,414 = 353VDC, 353 / 2 = 457V (per device)
VBrkMos = 1.3 * 2 * Vin(max) = 1.3 * 2 * 176 = 457V

Transformer turns ratio
N = N2/N1 = Vout/2Vin(min)Dmax = 53/2 * 220 * 0.9 = 0.26
 

Attachments

  • controller.PNG
    controller.PNG
    49.8 KB · Views: 174
  • volatge_feedback.PNG
    volatge_feedback.PNG
    104.8 KB · Views: 173
Last edited:
Why ETD49 N87 110KHz 170mT 20 turns and 500V Mosfets?

Hi MorbidFractal, attached are initial transformer calculations, the MOSFET calculations are supplied in the previous post.

The ETD49 N87, I had a complete brand new unused core with bobbin, it would be a sin not to use it. :)
 

Attachments

  • transformer_data.PNG
    transformer_data.PNG
    67.8 KB · Views: 162
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.