TL494 halbridge like ATX problem


New member
Hello. I'm not sure if section correct, if not please tell me.

Since my start with electronics ( I was 10 years old ) I used simply SMPS like LM2575 or so, to power microcontrollers or other stuff - these chips was simple, and resolving any problems took like minutes. Last year I wanted to start with mains-powered smps, so I took simply design from this page: Why this one? becauose it's almost identical to old ATX 200-300W supplies, which i can buy from scrapyard and get components. Everythink went well till i powered it up. there were some minor issues which were solved fast, but then I realised one big problem: In certain duty cycles converter is kind of buzzing or so(but overall regulation was good and also claimed power was easy achieved). And also when it's in one of these spots main transistors are heating extremely, Problem is not on full power, it's like in some points of regulation. I started digging in to this subject and simplified schematic to what's in attachment (I made it to ensure that wasn't feedback etc.). As 15V source I use lab bench power supply( it's factory one - 100% good ripple and things like that). Now I'm generating duty cycle from 494 (on separate board )and It looks ok, but as mentioned on some duty cycles it's going crazy. I'm using base-drive transformer from working ATX of exactly the same topology, main transformer is rewound, but exactly the same happened when I connected main transformer from ATX (on 12V windings). For test i've got 5 ohm load to prevent saturation of transfromer. Also when I increase auxiliary 15V to about 20V buzzing goes to higher freuencies, but transistors heating the same. And also, when it's not buzzing transistors are heating only a little bit( as expected ). And You know what is worst in all this: Every chinese stupid power supply works on exactly the same schematic (primary side) and I can't get it to work. I spent houndrets hours reading about this topology, I can tell what every component is doing in this supply, but I can't get it to work properly. :eek: Is this frustration?

I will provide You any kind of resources: photos, simulations, sreenshots from oscilloscope(digital or analog), but please HELP!!!!

Also sorry for my english - I'm from Poland.

when it's not visible - there is link:
Last edited:


Active member
@FiFi 22

Hello fifi I had just seen your post and I am going to try to help you solve your problem.
A few things to consider.
A transistor in an smps is usually switched hard in each duty cycle if its either 20% 50% or at full duty. However what really makes a transistor heats up is one of these conditions.

1) the transistor is not getting enough base drive on that particular pulse width thus it will run in linear mode.
2) spikes are generated from the BDT (base drive transformer) at that particular duty cycle thus giving false triggering and it could be the transistors tend to switch partly together.
3) The high side capacitor has not enough time to charge fully and giving a poor kick in the base of the transistor thus making it work in linear mode
4) high frequency oscillation out of the switching cycle thus confusing the actual pulse and also the transistor.

I would like to see some scope shots measured across R7 during this particular episode and when working normal. Be sure it is the low side transistor otherwise you will burn your scope. It is best to have an isolation transformer to isolate the smps. You can also isolate the scope if you want but be careful as the smps will still carry high voltages that can kill you if you mess up something.
You can also make a one or two turn coupling loop around the center core of the main transformer with a thin piece of insulated wire. Twist the ends and put apart the very ends and bare the wire to hook up your scope to them. This is the safest way to see what is happening and the scope will be fully isolated. In the picture below you can see the blue wire that makes a 2 turn coupling loop on my smps.
If you find a cure to this problem please tell us what you found so that others can learn.

Regards Silvio


  • 2000w smps 5.JPG
    2000w smps 5.JPG
    73.1 KB · Views: 10
Last edited:


New member
Thank You for quick help!

Power is currently connected in such a way: (bulb allows transistors to survive, when something goes wrong)

Now I can connect grounded scope enywhere i need, and converter has full 220V.

First waveforms without buzzing:
This is waveform on base of T1(337) or T2 - the same, but shifted of course.

Then two turns wound additionaly on Base Drive Transformer:

And on R7:

And finnaly on power transformer (also two added turns):
no_buzz tr2.jpg

Now waveforms with buzzing:

This is waveform on base of T1(337) or T2 - the same, but shifted of course.

Then two turns on BDT:

And on R7:

Finnaly on power transformer:

There is defenetly something wrong especially in last one, but what's going on? Also in next week i can get working ATX of this topology and post it's waveforms.


New member
Sorry for multiple posts in line, but I can't find edit button.

I forgot to mension that T1 and T2 are BC945 instead of BC337 and Q1 and Q2 are 2SC2335 instead of 2SC3039


Active member
As far as I can see the pulse is not right on all the scope shots. The pulse should show more like a square wave rather than a spike.
The configuration of the TL494 is rather odd, to my believe the opamp at pins 1 and 2 is there to control the output voltage (PWM) Why are you using the DTC to do this? Pins 15 and 16 are for the current control. From what I see here none of them are used for their purpose. For sure I can tell you that in ATX power supplies the control is totally different and they actually use the control pins 1 and 2 and also 15 and 16 for their purpose.
The DTC is to be set with a fixed resistor for say 1 or 2 uS and left as it is.

As you can see from the last picture the transformer is getting more than one pulse at the same time which is not right. I do not think that the switching arrangement or the BDT is bad so leave these as they are originally. I would like you to try the circuit that I am posting you which I have tried and worked for me. This is a modification to an existing ATX smps from a computer.




New member
I will try Your circuit, but why do You thinking the signal is not current at 494 output? To me it looks defenetly Square, but low duty cycle. As I said I'm learning - could You post screenshots from Your smps, for example at small and big duty cycle on base drive transformer? They will be helpfull. And this 494 arragment is pwm controll from datasheet, for now there is not feedback loop, becauose i wanted to check if it is not something That oscillate. And also I don't think transformer is getting multiple pulses, oscilloscope shows last few waveforms, so I believe it's diffrent pulses every time (30kHz). I will try Your circuit, but this will take some time (a week?)
Last edited: