Any Switching Power Supply Experts???

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi,
I'm building a SMPS based off the SG3525 [with no feedback]. Without a load it outputs +/- 40 volts without any problems. I did some load testing and even while outputting 250 Watts the MOSFETS remain cool. As soon as I break the 300 Watt barrier the Mosfets start heating up very rapidly. I'm using a 5+5 turns 14ga "6"-FILAR primary in push-pull configuration. [3 mosfets in parallel per side--6 mosfets total]

I was finally able to get an output of 450 watts RMS for about 15 seconds before the mosfets blew. Right before the mosfets blew I heard this strange high pitch sound coming from the torroid.... Could that mean core saturation and if so how can I avoid it?

Here are the critical stats:

Controller Configuration:
SG3525
Push Pull [3 mosfets per side]
No feedback
80KHZ [40Khz to each mosfet]
40% Max Duty Cycle on each mosfet pair [it has plenty of dead-time]

Mosfet Parameters:
IRF1104
Vdss=40V
Rds(on) = 0.009OHM
Id = 100A
Pd = 170W
Qgs = 29nC
tr = 114nS
tf = 19nS

Torroid Core:
I obtained the torroid from www.mag-inc.com as a sample a few years back. It is a ferrite type with these physical dimensions:
~2" OD
~1.5" ID
~.5" Thick

I'm assuming the problem is with the torroid when in fact it might not be...

Any suggestions?

Thanks
-=Randy
 
Have you any means to view the current waveform in the primary winding (a low-value sense resistor perhaps?) If the core is going into saturation you'll see a distinct increase in di/dt as rthe incremental permeability crashes.

I'm more used to using ETD series ferrite cores rather than toroids, but the size does seem to be a bit on the small side.

Are you aware of the extremely useful site:

http://henry.fbe.fh-darmstadt.de/smps_e/#smps

which has on-line simulations and transformer/choke winding help for various SMPSU configurations.
 
Hey "Ouroboros",
Thanks for the reply! I went to that web site you posted and chose: "half-bridge push-pull converter". It's not exactly the same thing I'm doing but it should give me a good idea as to how big my xformer should be...

Parameters:
Vin(min) = 10V
Vin(max) = 16V
Vout = 100
Iout = 10
Frequency = 40 Khz

It outputs this:
L/H = 128E-6
DeltaIL/A = 4
N1/N2 46.84E-3
Vin/V = 14

It than gives me a suitable core type and here are the "Very good" ones.
E70/33/32
UI 93/104/16
PM74/59

Where can I buy these cores?

Thanks,
Randy
 
Sounds like saturation... A saturated transformer is effectively a short - which means you're effectively shorting the input voltage supply with your MOSFETs.

Do you have another identical toroid core? stack 2 and wind the same transformer on those, and see if you get more power. You'll know immediately if it's saturation you're up against...

As for where to get magnetics, I've ordered small quantites of parts from Magnetics Inc (mag-inc.com) directly. Get in touch with their applications folks and they'll help you out.
 
Your problem may be caused by the EMI radiation of the own circuit, whose magnitude depends on output current.

At some power levels, EMI may be strong enough to interfere with the control circuit and cause asymetric drive to the transformer [clock or duty cycle interferences]

Improper ground PCB layout of the control IC may also cause ground loops and asymetric duty cycle [I've experienced this sometimes in prototypes]

Any transformer that works with no load should never saturate under load. Core size is not directly related to output power. Output power is only directly related to the windings [resistance and leakage inductance], and the core just limits the maximum amount of copper and the minimum amount of turns before saturation
 
Eva said:
Any transformer that works with no load should never saturate under load. Core size is not directly related to output power. Output power is only directly related to the windings [resistance and leakage inductance], and the core just limits the maximum amount of copper and the minimum amount of turns before saturation
hang on now... (amps * turns) / (core cross-section area) has to be kept under a certain value. Amps increase with power, so yes, core size (or cross-sectional area, at least) is related to power. :D

But you're sort of right - if a transformer doesn't have enough magnetizing inductance for the circuit it's in (not enough windings) then you'll have trouble regardless of loading.

Another question for the thread starter - are you using output inductors on this supply? If you've got lots of output capacitance and a low-leakage transformer, and your output inductors begin to saturate, the peak current in your primary side MOSFETs will start increasing exponentially. I'd expect this to blow a rectifier first, but it's still a possiblity...
 
Schematic

Thanks for the replies!

Here is the Schematic:
 

Attachments

  • schematic.jpg
    schematic.jpg
    95.7 KB · Views: 1,354
Here is the board layout for the schematic.... there are a few discrepencies between the schematic and the board.... Input capacitors and Feedback network for one... Actually dont pay any attention to the feedback network in the schematic... it is totally wrong... Because of the file size limit I had to crop part of the image down on all 4 corners.... On the top and botton the groundplane continues for about another 3/4 inch... the part of the circuit that got cut off is the rest of the bridge rectifier and output capacitors
 

Attachments

  • board.jpg
    board.jpg
    89.9 KB · Views: 1,146
thoughts

Eva-- you bring up an interesting point... If you check my board layout the output of the torroid is in close proximity to the SG3525 controller IC.... Albeit when I start drawing 300+ watts the field gets pretty intense.... do you think i should lay out another circuit board or try some sort of a metal doghouse? I know in another post of yours you mentioned metal shielding is good for electrostatic fields but not nearly as good for magnetic fields...


Eva and Gmarsh -- Do you think this torroid core would be suitable ?
http://www.cwsbytemark.com/prices/toroidal.php?orderby=data4&orderdir=desc

F-240-77
Al = 2725
OD=2.40 inch
ID=1.400 inch
Length = 0.500 inch
Material Type = 77
U= 2000

What I really want to do is just buy something that I KNOW will work....

Thanks,
-=Randy
 
In the multi-kilowatt designs I've built using ETD cores, i've never run into any big magnetic field problems. As long as your control IC and any other sensitive analog stuff is grounded well and isn't adjacent to high power magnetics and high power traces, you're fine.

You're running the 3525 open loop anyway, so I doubt that's a problem.

That's a big core - I doubt you'll have any trouble winding it :D I'm personally fond of F material (i've yet to use anything else in a toroid supply, and I've never had problems) but 77 material should work just fine.

One comment though - i'd go with a heavier transistor for your gate drive. IIRC, the 2n390x parts can only do 250mA collector current - moving up to 4401's or 5551's will give you a bit more gate drive current. Personally, I tend to go overkill and put UC3710's everywhere...
 
Your layout is pretty inadequate

1- The output of the transformer forms a big loop antenna and the control IC is inside this loop

2- The control IC is inside a big ground loop as the small signal parts are grounded between the transformer and the output capacitors

3- The primary and the secondary side share the same ground, so you will have to use balanced inputs in the amplifier to get rid of alternator whine [et all] if you use such a circuit in a car

4- You are taking feedback from the output in a topology that is not suitable for regulation [output inductors would be required for that]

To make this prototype work, remove feedback and see how to rewire the transformer and separate the grounds on the attached picture....

Altough I suggest a complete redesign :

- Output capacitors and rectifiers should be near the transformer

- Use independent grounds for primary and secondary side, couple them with something like a 10nF capacitor and a 1K resistor in paralell [using feedback then would require an optocoupler or differential techniques]

- Use two independent grounds for the control IC : A signal ground and a power ground. Route both to a common point

- Place the control circuit away from the transformer, try to place some filter capacitors inbetween for shielding

I recommend placing the control circuit in a daughter PCB soldered vertically over the main PCB, this causes some inmunity to magnetic fields radiated by tracks in paralell to the daugther PCB
 

Attachments

  • board.jpg
    board.jpg
    94.1 KB · Views: 708
summary of changes

Hey,
I took your advice and redid the circuit board. Here is a basic summary of the changes:
1. There are now 2 grounds: Primary Gnd "PGND" and Secondary GND "SGND"
2. I placed the rectifier diode and caps right next to output of xformer.
3. A Mosfet with a higher breakdown voltage has been chosen
VBRdss = 55V
Package = T0-247
rds(on) = 12mOHM
Id @ 25C = 72 Amps
Qg = 86.7nC
Pd = 130Watts
4. Control circuit is as far away from xformer as i can get it...

Eva-- In your last post you said: "couple [SGND and PGND] with a 10nF and 1K resistor in parallel" why would I want need to couple the grounds if i went through the trouble of seperating them?
 
The secondary side layout is now much better but you are still making some mistakes in the primary side

Never assume a ground plane to have the same potential on all its points. Never ground signal devices in a ground plane that is carryng current from power devices, this will allways cause voltage gradients across the plane and errors in the signal

All the signal components have to be grounded in a single track that connects to the ground plane in a single point. This ensures that all the signal components are grounded to the same potential [assuming negligible signal currents]

Also, the drains from the mosfets must be routed following the shortest path straight to the ground of the input filter capacitors, and this involves using wire bridges or double sided PCB

So your current layout still involves two big current loops in the primary side and ground loops on the IC. Try to fix them and post the new layout

And be aware that ground planes are no longer effective when PCB traces are dividing them in several isolated sub-planes. In these circumstances, they only effective in terms of reducing copper related pollution when etching the boards :D

EDIT: Coupling grounds at RF reduces potential EMI problems related to the capacitance between the windings of the transformer. If you don't couple them at RF, there will be AC potentials between them [and the metal case, remebmer to AC couple it to the primary ground] and this may cause RF currents to flow throug audio signal lines. Try to couple them through the shortest path. The resistor in paralell with the capacitor is optional and adds some DC coupling to prevent excessive DC potentials. Some designers use strings of back-to-back clamping diodes to allow for a limited swing, this allows to pass remote and shutdown signals without optocouplers

Also, I don't know how much power you are expecting to transfer reliably, but I've used a pair of IRFZ44V TO-220 devices with suitable heatsinking and 18V gate drive to transfer up to 1KW continuous at 14.4V, so you may be a bit on the overkill side

A simple trick to reduce Rds-on is to get 24..28V by rectifying the push-pull sides of the primary, regulate it to 20V using a 7820 and power the control IC with that voltage. This produces about 18V Vgs drive. An additional diode from +12V to the output of the 7820 ensures proper startup [use also a diode from intput to output of the 7820 to prevent reverse bias at startup]
 
Summary Of Changes

Summary Of Changes:

1. Its now 2 layers.
2. I put the driver transistors near the mosfets
3. Controller IC has its own ground: "IC-GND" and this trace connects to Primary GND "PGND" at one point like.
4. For heatsinking purposes I moved rectifiers to the walls.
5. Xformer Secondary ground connects straight to cap gnds

What do you think?
 
need help on designing new SMPS.

planing control
TL494(mb TL4941), using Half-Bridge Push-Pull Converter [Vin=260V-240V, Vout=(in calculations 50V actualy for 4x[URL=http://www.tubecad.com/march2001/page26.html]amp[/URL] ), Iuot=10A(I think this would be enought), f= 100KHz].
now there is few problems:
1: what R & C to use on TL494 pin 5&6 to get 100KHz :confused:
2: how to do feed back(this is my first designing SMPS :bawling: )?
3: what C to use in converter?

Lauris
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.