TL494 Lab PSU

I've started building a lab PSU using the TL494. The schematic is attached below, it's missing a lot of things but I'm still testing stuff on the breadboard.
The problem I have is with the gate drive transformer. When I connect it, it pulls a lot of current through Q1, which quickly heats up. It's made of a green toroid with trifilar windings, 16 turns each (minimum calculated is 13 turns). Inductance is around 945uH. Frequency is 70KHz.
Is it something stupid that I've done or is it the core that's not suitable?



  • TL494_Lab_PSU.pdf
    50.5 KB · Views: 271


Hello Bogdan,

It could be few things that are making Q1 heat up, it is either there is not enough gate current to it and the fet is running in linear mode. The inductance of the gate trafo in not enough, you could try a series resistor to lessen the current through it and lastly Q1 cannot handle the current through it. I suggest also try to put a small heat sink to it and see how it performs.

I did not check what Q1 is and its parameters. Usually it needs 10v and over at the gate so that it can switch fully on quickly without entering the linear region.

Check your wave forms with the oscilloscope to see the rise time etc. Check also the wave form at the primary of the gate trafo to see what actually is arriving there, you need a good square wave so that it can produce a good wave form at the secondary windings.

I hope that helps

Regards Silvio
Last edited:
I've used the FQP50N06 transistor (60V/50A). What I've noticed is that in single ended mode the TL494 gives a 90% duty cycle square wave with the DTC pin grounded (image 1 shows the signal at the gate of the mosfet).
So I've added a trimpot and fixed the duty cycle to around 50% (image 2). That didn't change much, maybe the current was slightly lower but still too high.
Then I put the transistor on a heatsink and used a small GDT from an old power supply. This one doesn't pull that much current, but the heatsink still gets considerably hot after a short while. Image 3 shows the primary and secondary waveforms (the ringing is mainly from the wires). Anything higher than 50% just pulls the supply into protection.


  • NewFile1.jpg
    20 KB · Views: 41
  • NewFile2.jpg
    19.9 KB · Views: 30
  • NewFile4.jpg
    20.4 KB · Views: 35
  • IMG_20180317_142351.jpg
    1.5 MB · Views: 44


Hi Bogdan, Try to take a shot between drain and source to see the wave form there. I still cannot believe that mosfet is getting hot. That should run cool without a heat sink. I think it is still the fact that its not driven hard enough. Try to make amplifier with a couple of transistors feeding the gate of the fet. The TL494 is not so hefty when it comes to drive current

When you are talking of current you did not mention how much. I believe that it should run in the region of 100-300 mA without load.

You can also try to put a small resistor in series and measure across it to see the current peaks with the scope. This will give you a better idea of what is going on. It is also the norm to put a series resistor with the gate trafo primary to limit current. Try some values and see what are the results.

Lastly if all looks well there might be an issue with switching losses and core material.

I guess the trafo from the old psu was meant to work at around 20Khz

Try lowering or increasing the frequency slowly and see how it performs.

I hope this helps.
I made a simple buffer and connected one of the windings of the toroid to it. I can't believe how well it works. Only around 20mA with not load on the secondary.
Primary waveform:

Secondary waveform:


On the other circuit, I tried messing with the frequency and duty cycle and burned the transistor (short circuit between drain and source). So something was definitely wrong with that one.


I am glad you found a solution to your problem. It is now a question of drive power from that buffer stage. I do not know what power are you seeking from the lab psu. Well that I leave to your judgement. You can take some ideas for old computer power supplies and you can judge well from some comparisons.

Good luck
I'm looking at 10A limit, or even 5A will do. I'm more interested in higher voltages, 50-60V.
Don't know about the driving power. But maybe bigger transistors such as BD139/BD140 would work if needed.


Hi Bogdan,
Well at that voltage I would rather go for 7 amps at 60v at the most (420w) this is what a medium power fet can go. Try to be modest at first as the more power the more problems will arise. You will be needing an ETD39 more or less or something equivalent for that power.

I would divide the voltage if I where you thus having 30-0-30 on the output. It will be useful when testing audio equipment. It will also be a bit difficult to go down to low voltages with high power at a high input voltage as the fets tend to switch in the linear mode down there. Another option is to switch between with more tappings on the traffo so at low voltages you can still maintain a good output and the pulse width will not be to narrow. I am talking through experience here as this happened to me.

Well with the buffer stages I think the BD transistors will be a better choice however this depends on the fets used.
Regarding fets try your tests with low cost fets such as IRF740 these are quite capable for the load wanted if not held at high loads for long periods. These fets are easily driven with around 200mA. Use a separate supply to drive the pcb as at low voltages an auxiliary feed from the main trafo will not supply adequate voltage. Try to buy genuine fets as the ones on ebay are mostly fakes and cannot withstand even half the load. Take a look at the video on Youtube by me there is also a blog post here about the subject.

Good luck and happy experimenting :)

regards Silvio
Thank you for the really good ideas! I will consider the 30-0-30 configuration as I will definitely need it for power amps. That's about the only power-hungry application I can think of.
I do have the IRF740 and the IRF840. For 2-switch forward I think the 740 will do fine. I'll obviously get genuine parts once I'm confident with the design!

For center tapped windings the main issue that I can think of is current sensing. I will have to investigate how that could be implemented.
Last edited:


Hi Bogdan, There are a few options for you to choose, Make a stationary switching psu at the back with center tapped windings, the two separate outputs can be adjusted with a couple of buck converters these have adjustable voltage and current. There are quite a few to choose from on ebay as its not worth building them as they are rather cheap being around 3 pounds sterling.

The other option is to build two linear regulators with current limit. If you want to take an idea you can take a look on the link and I will explain what I did for myself

stand by aux supply

Buck convertor

Regards Silvio
I've studied some of the options before:
- LTC3080 buck/boost regulator with current limit - really nice but limited output voltage (can only reach around 30V as most regulators)
- LM2576-ADJ - goes up to 40V or even higher with the HV version; no negative counterpart (I know it can be hooked as a negative regulator but there are drawbacks). Also when I tried it it seemed kind of "slow"
- Other commercial buck regulators - limited output voltage, no dual output
- Linear regulators - easy to use but need transformer, heatsink and are very limited in current capability (unless you parallel them)

I hadn't had the time to experiment with this design but I'll come back with results. Even if it won't work, I'll use it as you said and try and build a buck regulator.


Hi Bogdan, You can use two positively regulated buck converters in series on the output one for each rail. I think they still work. Its like putting two power supplies in series.

Let us consider this. So rail A has its regulation on the positive and the return path will be the center tap. Let us consider that electrons move from Positive to negative so the current path is from positive to negative.(center tap)

Rail B however sees the center tap as the positive and the current path is now from center tap to the negative rail. well considering this it seems a bit confusing but is should work just the same. The rectifier bridge will allow only current flow in the right direction according what is coming or going through the center.

Try to hook up a couple of batteries in series and take out the center point between them. If you try to load them with a couple of bulbs they still work fine even if you try to load one side more than the other so why the buck converters should not work as well.

Regarding the current limit this has to be set separately for each buck converter or you can use a dual potentiometer to control them simultaneously.

Regarding the linear regulators well what I did in my dual supply was that I made the smps follow the the output voltage of the linear regulator with around 5v more using a zener and transistor thus controlling pin one of TL494. This way heat dissipation is kept to a minimum at all times.

Regards Silvio


New member
A 2-Switch Forward converter is a good choice for a lab power supply pre-regulator. You should set it up to regulate just a few volts higher than the output, then do the final output regulation with linear regulators to eliminate any switching noise between the SMPS and the final output terminals. The lower the voltage dropped across the linear regulators equals significantly less heat than a single stage linear supply. A TSFC(Two-Switch Forward Converter) is very easy to setup to use constant current output (output current limiting) since it uses current mode regulation and since the secondary rectifier is essentially a buck converter. Now, on to the problems...

1.) The TL494 is a PWM controller based on an old design. While this chip has been used in millions of different types of power supplies, I've never seen one used in a TSFC. The TL494 cannot do true current-mode control, so it is almost always driving a half-bridge supply in voltage-mode control. I have seen them used in low-voltage push-pull power supplies, such as those used in car audio sub-woofer amplifiers.

2.) As you found out, the TL494 cannot directly drive a MOSFET. The outputs of the TL494 are uncommitted transistors. They can either be configured to pull-up or pull-down, but they are single ended. This is the reason you need the 2-transistor totem-pole to drive your GDT MOSFET properly, like you showed in the image in post #5. Even then, you should have a 1K to 2.2K resistor going to ground and connected at the junction where R1, D4, and D5 all connect together to give a good solid pull-down for when the transistor in the TL494 turns off. You can then use the output from the totem pole to drive a MOSFET gate which can be used to drive the GDT.

3.) You also showed another issue regarding the output of the TL494. When in single-ended mode, both outputs have the same signal and the maximum duty-cycle is over 95%. This is completely useless for a TSFC as the duty-cycle must be limited to less than 50% to ensure the power transformer can reset during its off period. You said you used a POT to adjust the DTC pin to limit the duty-cycle to <50%, but this isn't ideal. You should set the TL494 to operate in push-pull mode so the outputs are limited to <50% by default. Then you would just use one of the TL494's outputs to drive your GDT and leave the other output unconnected.

4.) Finally, I would recommend dropping the TL494 all together. While it is a simple chip to use, although it is often difficult to get right, it really isn't meant for the type of power-supply control that you need for a TSFC design. I'm sure that you could get a working supply using the TL494, but it is a poor choice for the TSFC topology and it would take a lot of effort to get it to work properly. I would recommend that you use something like the UC3844/UC3845. These are very cheap 8-pin chips. They have a single output that can directly drive a MOSFET for your GDT. They feature true current-mode control regulation and the 3844/3845 are internally limited to a 47-49% duty-cycle, both of which are ideal for a TSFC. (The UC3842/3843 go up to 98% for other types of supplies.) Examples of TSFC supplies that use the UC3844/3845 can be found all over the internet, many are proven designs. I did have application notes that show good examples of these, but I can't seem to find the files right now. If I locate them I will post them for you.

Please note, I am not an expert in SMPS design or engineering. My suggestions here are based solely upon my own experience in building power supplies and from other designs that I have studied.


Edit: These UC384x chips can often be found in newer(since 2000) ATX computer power supplies and/or ones that are higher powered(>400W). They're sometimes even configured in the TSFC topology so if you found one it would not only give you the chip, but a good, proven example design to help you with your own.
Last edited:
Thanks again for the replies.
The reason I've used TL494 is because I can play with the error amplifiers. And it's got two.
I might sound stupid, but the threshold voltage for the UC3845's current sense input is 1.1V (say 1V to simplify everything). So that means 10A through a 0.1 ohm resistor. But what if I want to limit the current to 5A through the same resistor? That would be 0.5V. How do you "lower" the threshold so it would start limiting at that value? A comparator can be used, but it 1V is too low for a supply voltage.

Apart from that, I thought about what Silvio said and I had the idea of building a single rail 50V supply and then adding another push-pull stage with a 1:1:1 transformer that outputs +50V and -50V. All the regulation made at the "primary" of this transformer will reflect at both of the secondaries, and being 1:1 makes everything easier. Of course, that means double the losses and half the power.


Thanks again for the replies.

Apart from that, I thought about what Silvio said and I had the idea of building a single rail 50V supply and then adding another push-pull stage with a 1:1:1 transformer that outputs +50V and -50V. All the regulation made at the "primary" of this transformer will reflect at both of the secondaries, and being 1:1 makes everything easier. Of course, that means double the losses and half the power.

Hi Bogdan, I think I was misunderstood in my explanation. What I meant was the same trafo driven by the TL494 will have 2 symmetrical secondary windings (eg 30-0 and 0-30) but these if not separately controlled will move together, The option I told you about is that you either have 2 separate buck converters controlling each secondary this way you will have a dual supply which can have two separate different voltages on the output. These can also have different current limiting as well. When the need comes for audio you can switch them in series thus having a double output.

The second stage can also be done with a linear regulator, with current limit. I used the schematic on the datasheet (National semiconductor)of the LM317 chip followed by a MJ2955 and for current limit it uses the LM301A chip, this circuit can handle a current of 5 or 6 amps and a current limit of 0-5 amps. It is very simple to make the TL494 to follow the output of the linear regulator with a few volts higher. This minimizes dissipation very drastically. I am assuming that you are going to use a pulse transformer with the TL494 and the circuitry will be totally isolated from the grid.

Regards Silvio
Silvio - sorry for the misunderstanding.
What I meant was that I tried making the GDT work properly and I'm not quite happy with it. So I'll try one of the secondary-side options we've discussed here. We'll see which one works better.
I've also looked at UC3845/UC3843 and SG3525 (for push-pull mode). I'm not sure if 2-switch forward can be run on duty cycle greater than 50%.


Hi bogdan here is what I did in my dual psu. It was actually a converted computer psu with some modifications. You could build your own if you want. As far as the pulse transformer you can always find one from an old computer psu they all have them. You can use switching transistors instead of mosfets you will still get good power.

The circuit I used from the data sheet is slightly modified as I made a negative reference of around 1.2v so that the linear section can go down to 0 volts otherwise it will start from a minimum voltage of 1.2v.

You will also find in the 2 pdf files what I did to control the TL494 to follow the linear regulator. The current is set to around 6 amps on the smps and this will reduce the voltage in case something goes wrong in the linear regulator.

Although the LM317 can go up to 36v max you can go further to 50v as the regulator will only see a few volts between
input and output

Here is the link to see my PSU

Here are the files for you to download and see

View attachment Dual PSU.pdf View attachment Dual PSU 2.pdf View attachment data sheet LM317.pdf

Good luck regards Silvio
Thank you, it looks very interesting.
Can I ask how the isolation between the primary and secondary is made? I see the current feedback path directly tied to the error amp of the 494. Does it have a gate drive transformer?


The isolation is made through the pulse transformer. Do not forget it is a modified ATX power supply from a computer which uses the tl494. All the low voltage circuitry including the pwm chip are all on the low voltage end including the primary of the pulse transformer. Only the secondary of the pulse trafo is tied to the bases of the switching transistors.

That is why I told you to use BJTs as switchers as you can easily find a ready made pulse trafo from an old atx computer These are usually EE16 or EE19 cores. Do not mistake it with the stand by supply trafo in the ATX as this has a different role. You will know which one it is as you will follow its traces leading to the base of the switching transistors.

Here is a link to see how to modify one which will have constant current and constant voltage

You can also take an idea of the circuit of an ATX power supply

Use google translate to change the website language as it is in Italian


Regards Silvio