SG3252 SMPS 525W Problem with output power

[Tomek]

Hi i'm a Tom and it's my first topic,

I have problem with output power in my smps. In assumptions i wanted to get output voltage from 10V to 35V with regulated current from 1A to 15A.

Topology: Half Bridge
Core: ETD44 3C90
Frequency: 50kHz
Input Voltage: 320V

Windings calculated in ExcellentIT() but the lice I wound manually, wrongly cut the wire length and ran one turn at the primary winding and the secondary in relation to ExcellentIT. In the diagram that I posted red crosses, I marked the elements that were not soldered. The only difference with the schematic is the RC snubbers in parallel with the transistors. (Drain, Source) 2.7nF capacitor and 10 ohm resistor. Without them there were significant oscillations on the transistor drain, which, when powered by 320V, caused them to break. When the snubbers are used, the keys work fine, but the power dissipated on the resistors is so large that the inverter can operate for a maximum of 20 seconds and then the resistor is going to smoke.At the moment, the SG3525 works without feedback. The error amplifier on + has 3.6V on - it is 0V. Outputs PWM A and B work around 50%. During the test with a light bulb, the car 55W 12V light dim. The current is about 3.3A while the voltage as well I remember about 6V.I will add that the SG3525 is powered by an isolated 15V power supply

Please tell me why the output gets so little current? Why does the voltage so drop? Without load I have about 39V.


Schematic ->View attachment Schematic Prints.pdf

 

[Tomek]

There is a small bug in the title post when it comes to the name of the controller. It should be SG3525, but I do not really see the possibility of editing post for what I'm sorry.

 

[Silvio]

Hello Tomek,

Did you check the wave form without feedback and see what you are getting at the output? What I have learnt so far is that you have to start with a clean wave form before adding feedback and so on.
Regarding the output current the type of core material used in the output inductor depends a lot as it might block all the current. Make your initial tests without the feedback and the output choke and see what comes out. Do not forget that a cold bulb filament has a very low resistance.
You can load with a couple of bulbs in series first to limit current then gradually add in parallel more bulbs according to the output voltage to see what you can get.
Monitor the waveform at the low side gate. If you have ringing at the gates it will be reflected in the output.

Regarding snubbers you may be loading too much. Firstly see what are you getting at the primary of the trafo. (use an isolated supply otherwise you burn your scope) The transformer winding has to be tight coupled and the more leakage inductance you have in it the more ringing will be generated.

Show us some scope shots of your wave forms and we can talk from there.

Regards Silvio

 

[Tomek]

Hi Silvio

The oscilloscope mass is connected to the middle of the output winding. Probes to the beginning and the end of the winding. Power supply of 320VDC inverter
Outputs wave form without load

http://fotowrzut.pl/FG6BJUZZVN

Outputs waveform with load one bulp 12V 55W

http://fotowrzut.pl/9TXJ5L2J57

For all measurements I use an isolated transformer and a tuned transformer to power the inverter
Run on the primary winding of the transformer. Unfortunately, I do not have a HV probe of up to 300V so I had to investigate waveforms at a lower voltage than 320V

http://fotowrzut.pl/8OEDYUI4DO

In contrast, the work of a non-snubber converter is deadly for transistors.
Below are some waveforms from Dren Source and the transformer itself on the primary side.
Power supply of the inverter is only 30VDC.

Dren Source without snubbers:


Trafo without snubbers:


Dren Source with snubbers:


I will still check the output choke or sometimes does not make any problems.

I dont why some waveform are too small. I will add external links to images.

 

[Tomek]

After removing the output the inductor is much better. Already when the inverter power supply voltage is about 120V DC I reach 14V and close to 5A it is interesting what will be at 320V DC. I need to look for another load

I do not quite understand why this choke works so badly? Bad core? Too high inductance? There would be some theory for it.

Edit 20:34
I have checked the resistor 0.47 ohm and 10A with no problem can be obtained. Moments during the voltage regulation of the autotransformer terrible squeak inverter (transformer?)

 

[Silvio]

Hello Tomek

Some hints to help you out.

Transformer without snubbers.
1) Bring out the cursors and measure ringing frequency. Load a little the psu on the output so that you can see ringing.
2) Put a small ceramic or film capacitor across the element/s to be snubbed so that ringing frequency becomes 3 times more
than original. Example ringing frequency is 3.3Mhz new frequency with cap = 10Mhz.
3) The suitable capacitor that bring 3 times frequency was found to be for example 450pf

4) Your resistor is calculated as follows

R = 1/ 2pie FC where F= new frequency with capacitor and C is the capacitor used.

R= 1 / 2 * pie *10^6 * 450 * 10^-9 = 35.3 ohms use nearest value example 33 ohms

Regulation

For regulation of the output you need to have a minimum load attached to the output. You also need an inductor but be careful try with less turns on the output inductor and see what current you get. The core could be absorbing a lot of current if inductance is too high. It could be that you are not using a powdered iron core. It needs to be air gapped. The core for output choke has air gap structure when manufacturing the material. It is specially made for that.

Deadly on transistors without snubbers.
Check dead time and see that you have around 1.5-2uS for a start until you settle out things then lessen accordingly to 1uS when you get ringing attenuated.

squeaking noise on transformer.
pulse width too small and feedback loop disturbing waveform at transformer primary. Transformer saturation.

You can measure the waveform on transformer primary if you isolate the scope with an isolating transformer. Or you can isolate your power supply instead. You only have around 155 volts across the primary of a half bridge. Use probe X 10.

You will experience a voltage drop at load of around 20% from the original un-regulated voltage due to the inductor and also the loaded voltage as with no load the voltage will be at peak, however Excellent IT calculated for that. You need a minimum load at the output to get good regulation and also reasonable pulse width.


Hope that helps.

Regards, Silvio

 

[Tomek]

Hello Silvio

The snubber method you wrote here was used here. As I remember correctly, it comes from the Maxim Integrated application note.

The oscillations on the transistors and the transformer I removed.
Dead times are OK, transistors do not heat up

I have also checked with other inductors. I soldered the choke from the old ATX power supply which had 12uH but the inverter still behaved like it would be choked. After unrolling a few turns I got 6uH, which when tested with 2 x 55W bulbs turned out to be the maximum power of the inverter. I have reached 320V DC on the power supply. I noticed that the core of the output choke is getting very hot.

Where should I look for problems?

I note that the inverter is still working without any feedback. So it is trying to reach maximum power now.

 

[Silvio]

Hello Tomek,
If you want to check what power the smps can give just bypass the output inductor but if you have high capacitance remove a part of it as the trafo will see a dead short without the inductor while initially charging them up and may trip your overload protection.
I also have been working on my 1000w smps and this time I added regulation. As you know well when adding regulation you will surely need an output inductor. I experienced the same problem and the inductor got very hot. My maximum output current is 10amp and what I noticed that the more power and current it carries the hotter it will get, Choosing a larger core will help out in this problem. I also tried a piece of ferrite rod around 10mm dia and 5cm long and wrapped my inductor on it These also got a bit hot but not as much as the ring. There is no fear of the rods getting saturated though. I have been playing quite a while with different number of turns but still even with a few turns on the ring core such as five the core still got hot if you keep the current flowing for a long time. I guess it all depends what are you going to do with the smps and what duty cycle is going to be worked at. Luckily I have my inductor on my smps is just at the outlet of the cpu heatsink that I used and the breeze passes by the inductor thus cooling it while working.
What is really needed is that the core will induce inductance at low current but when at full load it will saturate and seem like a dead short like a piece of wire. This will be the best situation I guess as it will no longer be needed as the wave form will be at full duty cycle. Voltage drop is always present when introducing inductors to the output. By the way my inductor value is 15uH on each side and the regulated output is 120v both rails in series. The peak no load unregulated voltage is 150v so you see I have around 22% headroom in voltage to maintain the regulated output. Keep also in mind that a compromise have to be chosen as the lower the output inductor the higher the minimum load need to be. In my case the minimum load is just over 1A below that when introducing high load the over current will trip due to pulse width opening suddenly and exceeding maximum current threshold trying to charge up the output capacitance and load simultaneously.

I hope this will help you understand better the situation

regards Silvio

 

[Tomek]

Hi Silvio
Something got me hooked up to test to connect the bulbs in parallel to get 10A load current. It was full of squeaks probably from transformer and burned fuse. Lower transistor and its driver burned.

I'm wondering if I could have done a good job with the transformer
Flux denesity I received 200mT, but by mistake with the number of windings this value increased. I wonder if I do not accept 120mT or 150mT eg?

Since the SG3525 is working properly, on the primary side the waveforms are correct and the transformer squeaks or the current is blocked in a strange way on the output choke it suggests some kind of transformer failure

Silvio Do you use snubber in your inverters?
What is the value of flux denesity in your transformers?

Is it always the case that the inverter without output choke achieves more power? Is this maybe just a problem with badly designed inverters?

 

[Silvio]

Hello Tomek,
The maximum flux depends on what Excellent IT gives and this depend on the frequency and core material. In my case at 60Khz I go for 0.16 T for N87 material But if you see clearly it will be somewhat a bit less this is due that the input voltage is not always the same. In your calculations you should have used AC input and wrote for example 210 -230- 245 this is a more realistic input voltage range as main voltage fluctuates, ETD 44 should have given you ample power in excess of 500 watts, I got 1200 watts peak from an etd 49. As the frequency goes up the flux goes down

snubbers, well it depends on transformer outcome and also if the smps is regulated or not. If I see that the wave form is clear without any spikes I will not use snubber but if the smps is to be regulated then I use snubber as due to short pulse width it tends to spike a little.

I suggest you see the article here on the blog post regarding transformer winding practices for smps. It will show some hints and what to do and not do, as making a bad transformer could kill you if it is used for offline.

I also suggest that you see my video on youtube regarding winding transformer for smps by Silvio De Leonardo. It will show a proper way to do it.

If you done your transformer the way described in the article and video than it is good. if not then I suggest you follow what I have explained and you will have a good trafo.

A sqeaking noise coming out of the transformer is a sign of saturation.

Regarding the last question No not actually more power but less voltage drop. The output inductor also serves to store some charge between cycles and smooth out the ripple. It also helps out a bit the trafo during heavy load not to let it struggle alone, it also limits inrush current to the output.

Hope that helps

Silvio

 

[Tomek]

Hi Silvio

I watched your video about transformer winding, but the question is, are you always sharing the primary winding in two parts?

What if the winding is in one layer?

If you see the film well, you do not cut off the wire in the middle of the winding, right?
You wind up half the windings and then you wire out the wire. Then you wind up the next layer and you release the wire again. Then the remaining wire from the first layer wind up on the other until you reach the boobin solder leads?


I also tried to wind up the transformer so that the primary windings fit into the width of the boob. This gave me the first layer. Insulating tape. Then I wound up half of the secondary winding. The insulation tape and the last layer were the other half of the winding.

I rolled 40 turns on the original and after 11 on each of the secondary
The inductance should be 140mT.

In addition, I used the core from the original output choke that had 47uH. I have added an additional 10uH winding.

Did something give? Little. Transformer still likes to squeal at 5A. After all, with a large choke (core of about 45mm diameter), there is no problem with getting high output current.

For a moment I had a very strange situation. After connecting the oscilloscope (1 channel) to the output of the transformer, I saw here a waweform:
View attachment 6215

The load was two 55W car bulbs, according to ammeter 9.5A, and the whole worked without any squeaks. Only the waveform on the transformer is very distorted.

I connected the second oscilloscope probe to the second secondary winding and received this waveform:
View attachment 6216
Suddenly such an improvement? Trying again to get 9.5A and here surprise, with 5A squeaks from the transformer. I detached one probe and returned to the state just before the minute. Interestingly, the passage does not contain only a normal rectangle. Of course, at 5 A there are still squeaks from the transformer.

So when the waveform was bad I got no problem with 9.5A and with normal output waveforms at 5A there are strange things happening.

I'm still wondering how the inverter power supply regulates through the autotransformer. This is the voltage across the capacitor inputs and the transistors themselves.

It is much safer to slowly increase the supply voltage than suddenly connect the 320V. Only does it cause any problems?

 

[Tomek]

I can not edit the post above so I will write a new one.
I found out why the transformer squeaks, so that the cores that hold the cores are too tight. When the transformer starts to squeak, just press your finger on the upper clasp and the squeak will disappear.

A 100% transformer will have to be rewound. I made a winding error and the secondary winding is not equal in length. This is visible on the oscilloscope under load and one diode is much warmer.

I think I will try to do a similar wrapper to transformers like yours.

If he was to estimate the power then more like 120W would not get in the current configuration. Should the transformer be guilty?

 

[MicrosiM]

Post editing is time based, and its limited by 30 minutes.

 

[Silvio]

I have worked the trafo details for you.
The bobbin width is 29.5mm less 8mm for margin tape = 21.5mm

Your transformer will have margin tape on the primary but it may not need margin tape for secondaries. You already have 4 mm on each end for the primary so the minimum clearance is already there. See how the secondary will fit in each layer and add margin tape as necessary if needed.

Primary winding 22 turns 11+11 3 wires 0.6mm leaving 4mm margin at the ends wire is to be layed side by side. do not cut wire leave remaining wire to wind the other half of the primary. Like shown on video. do not forget to make sleeve on ends.

secondary winding 7+7 6 wires 0.6mm (2 bundles of 3 wires twisted together)

winding sequence half primary 11 turns ( make 4mm margin at both bobbin cheeks with height of 0,6mm) sleeve wire ends with good high temp sleeving like shown on the video.

3 layers of maylar tape or similar covering full width of bobbin.

half secondary 2 bundles of 3 wires twisted to make 7 turns. Start from center tap from top of transformer like shown in video. no margin is needed. insulate with 2 layers of tape join on pins below

half secondary 2 bundles of 3 wires twisted. start from center tap from top of the transformer as above

cover with 3 layers of maylar tape covering the whole width

second half primary. make 4 mm margin on the bobbin ends to be 0,6mm high
Return the remaining wire to the bobbin and continue the last half 11 turns and finish on the pin/s below.

attach wire to the bobbin pin and cover the whole winding with maylar tape.

Regarding your questions.
1) A good wave form has to be established before you get good current output. If the wave shape is disturbed in some way due to other factors then you have to correct that first.
2) be aware that when adding feedback the wave will change shape. I will take a few pics on my smps and will show you what it looks like when the pulse width changes. I take my measurements across the primary winding. Please give me some time as I am very busy at the moment.
3) one other practice I use a 100watt bulb in series with the supply line so if there is something wrong the bulb will limit current. you can hook it up instead of the fuse on your pcb. however if you intend to draw more current then the bulb you have to remove it as it will light up and drops all your input voltage.
4) transformer squeaks if the cores try to vibrate. put some contact glue on the faces where the cores meet. (evostick or similar)do not use super glue otherwise you will never take them apart if the need comes.
5) My outcome on Excellent IT is attached. The core selected is with N87 material but it is similar to 3C90.
6) put some vaseline or petrolium jelly before twisting wires together to help them find their place while twisting them. Wash wire with white spirit and dry well after twisting them to help the tape to adhere to it.
7) your attachments could not be seen please post them again
8) What auto transformer are you referring to? is the feedback done with an opto coupler or not?
9) You are building a SMPS and not an inverter.

Regards Silvio

 

[Silvio]

Here is a better view of the calculation on Excellent IT

I suggest you read the blog post regarding making Litz wire at home on diysmps you will find some tips how to twist wires and more information.

 

[Silvio]

What wave forms on smps looks like

The first pic shows the output inductor and the added feedback pcb with a pc 817


In the next set of pictures you will see the wave form taken across the drain source of the low side fet at various loads


Pic 2 shows the wave form with minimal load of 200mA


pic 3 shows wave form at 1.4A load


pic 4 shows wave form at 3A load

Pic 5 at the bottom shows wave from at 6A load

You will notice from the pics above that ringing is induced and this could be due to having wires treading along for the feedback as the smps was not designed to have feedback and this was added later. The un-regulated waveform however was very nice and not disturbed in any way.

I will post some more pics at the gate wave form in the next post

Silvio

 

[Silvio]

The next set of pictures shows the wave form at the gate of the low side fet

The pics are as follows 1) at 200mA 2) at 1,5 amp 3) at 7 amp load and the last one at 9 amps load.

You will also notice that as the load is increased noise is also increasing on the gates and this may be reflected on the output wave form.

 

[Tomek]

Hi Silvio

Strange that earlier attachments do not work.
In that case, I'll include again:

Attachment 6215:


http://fotowrzut.pl/TQV28CBBQ7

Attachment 6216

http://fotowrzut.pl/REKEBX9S2A


I have a question about winding:
1. "Half the secondary bundles of 3 wires twisted" - Of these two litz I do one and so I wind? Or can I twist parallel two litz like in a film where you wound up with 6 wires at the same time?
2. What better way to wind up?
A) Litz wire from shop for example 6 wires x 0.6mm?
B) Two litz wire in parallel each 3 wires x 0.6mm?
C) Two litz wire twisted each 3 wires x 0.6mm?

Thank you for the photos. I will have a comparison.

 

[Silvio]

Hi Tomek, For primary 3 wires side by side (parallel) 0.6 mm

for secondary Two litz wire in parallel each 3 wires x 0.6mm

 

[Tomek]

Hi Silvio
In the days I will wind up the transformer and when I will be after the first test I will write what are the effects.
Thx for help.