Two switch forward based on UC3845

I've been contemplating at building a two switch forward converter for some time. Since I need a power supply for some powered speakers, I thought it's a good time to build one.
As far as I've learned about this topology, the advantages are:
- single ended drive (it can use any flyback controller)
- it can do current mode easily
- single positive DC rail needed (unlike half bridge)
- can do as high power as a half bridge
- no need for a reset winding
- seems more robust

The biggest disadvantage is the high side transistor drive circuitry, which can't be the same as in a half bridge since both transistors turn on at the same time, so there's no way to charge the bootstrap capacitor. There are integrated drivers that can work with this topology, but they're rare. The easiest method is by using a gate drive transformer.
Schematic is below. Feel free to discuss and suggest anything.
I chose a typical main transformer, but any big enough core can be used, even salvaged from an ATX power supply.
The gate drive transformer can also be salvaged from ATX power supplies or EE16, EE25 or even small green toroids can be used. The smallest one I tested could fit on a fingertip, and it was bifilar winded. The windings must have the same polarity.
 

Attachments

  • TSFW_Simplified.pdf
    45.8 KB · Views: 127
I only need about 200W at most.
I did some modifications and added things that I forgot.
I have a question regarding the output chokes. Each output will have its choke and the two power rails will be wound on it. But are they winded in the same direction or in oposite directions?
 

Attachments

  • TSFW_Simplified.pdf
    50.6 KB · Views: 76
PCB:



 

MicrosiM

Administrator
Staff member
I still believe that you can use the UC3845 or similar chip to drive a GDT and forget the regulation at this first test.

Make sure that your SMPS is working and try regulating it.

For 200W you don't need a regulated power supply I think.

Hope that helps
 

MicrosiM

Administrator
Staff member
No problem with regulation, use GDT, there are many circuits for the same chip with regulated output

try different core for the GDT
 

Silvio

Well-known member
I have a question regarding the output chokes. Each output will have its choke and the two power rails will be wound on it. But are they winded in the same direction or in oposite directions?
Regarding the output choke if you are going to use the same core for both rails then make sure that the winding sense is one opposite the other.
I had this issue a week ago and all sorts of things I was seeing in the output especially when trying to load one rail. If you have separate cores for each rail it is not important which direction you wind.
How to do it on one core?
You can wind both windings in the same direction (Bifilar) but on only one winding swap the start of the winding with the finish. This will give one of the winding a different sense (that is one opposing the other)

Coming to regulation.
As you might well know the load on the smps is varying all the time when used for audio. The value of inductance has to be good enough for light load and also when it comes to full load. The lighter the load the more inductance needed on the output inductor as the smps will have a very short dead time and may work in discontinuous conduction mode (DCM) Here the energy stored in the inductor has to make up for the lack of energy between pulses. When on load however it will switch to Continuous conduction mode (CCM). The output inductor value however has to be trimmed by trial and error to find the right value so that it will not be too high in inductance as this will surely heat up the core unnecessarily not to mention it may also saturate.

Some assumptions
200w at 60v = 200/60 = 3.3 amps. I guess 60-90uH for each rail is a good starting point assuming a minimum load of around 250mA. A 1mm ECW is enough to do the job on a 30-35mm yellow core from an old ATX computer smps. Your headroom voltage need to be around 40v to regulate for 30v at full load.

Non regulated smps.
If however you will opt for a non regulated smps like MicrosiM suggested life will be more simple. Say an IR2153 (HB topology) and a couple of IRF740 will do the job nicely at 200w power. Your ETD39 will give you 600w peak. Here you will not be needing a large inductor but a 6uH on each rail will filter any high switching noises and also limit the initial surge when charging up the secondary capacitors. As a rule of thumb use 1uF per watt of input capacitance across the 320v rail. Your 220uF/400v cap will suffice for 200w. All you need is 2 CBB capacitors of 0.47uF each to divide the voltage. About 2000uF on each rail for the secondary capacitors. Do not forget that music does not draw continuous power and only about 60% or so on a class AB amplifier if it is class D the power needed is less.

One last note. An unregulated smps working at near 50% duty cycle is more quieter than a regulated one. Spikes create noise and need more careful filtering.

Regards Silvio
 
Last edited:
Thank you, that's very useful advice!
I have built an IR2153 SMPS and had issues with it, never managed to solve them (there's a thread in here about it).
The TSFW topology seems easier to implement, and I had success with the GDTs I tested.
I will test both the unregulated and regulated versions. What about primary side regulation via the auxiliary winding? Is it less strict than secondary, reference based regulation?
 

Silvio

Well-known member
What about primary side regulation via the auxiliary winding? Is it less strict than secondary, reference based regulation?
I cannot give you an answer about this because I never experimented with regulation in this manner. What I can say is that it will only sense the voltage drop on the primary side of the transformer. I am assuming that when the voltage drops on the primary so does on the auxiliary and this will enlarge the pulse width to compensate, however all the other losses on the secondary will still be present. There will be some voltage drop in the diodes and capacitors and the secondary winding itself, these will not be compensated. It is usually used on low voltage supplies like 5 or 12v where the percentage of voltage drop will be insignificant, however when higher voltage is used the voltage drop will be more pronounced for the same percentage.

Regards Silvio
 
Hi

what pcb software do you use>?

thanks
Kicad

Some updates on this project: the power supply works but the output voltage is too low and it drops even more under load. It seems that I'm having some issues on building transformers, every project that I tried lately has failed. I eventually built myself a small core test rig where I can test them at lower voltages, but they all seem unsuitable, so something's wrong in what I'm doing.
 

kees54

kees57
Thanks, I do also.

I did use eagle, but since it is changed with that internet trouble I do not use it anymore, kicad do come close to that
as learning curve.

regardsScreenHunter 504.jpg
 
I haven't yet built the power supply I posted, but I built a similar one. The problem I'm having is that the transistors are getting really hot even under the minimum load. There's also a ringing noise on the transformer.
First waveform is with no input, second with input. I suspect the transformer, but I'm not sure.
 

Attachments

  • TSFW_EI_GDT.pdf
    30.1 KB · Views: 58
  • DS1Z_QuickPrint5.png
    DS1Z_QuickPrint5.png
    38.7 KB · Views: 9
  • DS1Z_QuickPrint4.png
    DS1Z_QuickPrint4.png
    42.4 KB · Views: 9

Silvio

Well-known member
From what I am seeing is that the drive to the gate is not strong enough. If you look closely you will see a rather slow rise time. This will put the mosfets in the linear mode hence they are getting hot. You either make this part stronger or you will still get problems. The best cure is to make a separate auxiliary supply to feed the low voltage with.
It also seems that the auxiliary winding is not producing enough voltage when the smps is without load. Do not forget that the pulse width will be very narrow when the smps is not loaded. If you will add more turns to the auxiliary you will get the needed voltage but surely not to be regulated with a zener. My guess is using an LM317t to regulate the auxiliary voltage. This can take an input voltage of around 40v. You will expect that the auxiliary winding will swing its voltage with load according to the pulse width. With this in mind you need to have adequate voltage at minimum load and keep it stable when the auxiliary winding voltage swings high when the smps is on load.
 
Last edited:
From what I am seeing is that the drive to the gate is not strong enough. If you look closely you will see a rather slow rise time. This will put the mosfets in the linear mode hence they are getting hot. You either make this part stronger or you will still get problems. The best cure is to make a separate auxiliary supply to feed the low voltage with.
It also seems that the auxiliary winding is not producing enough voltage when the smps is without load. Do not forget that the pulse width will be very narrow when the smps is not loaded. If you will add more turns to the auxiliary you will get the needed voltage but surely not to be regulated with a zener. My guess is using an LM317t to regulate the auxiliary voltage. This can take an input voltage of around 40v. You will expect that the auxiliary winding will swing its voltage with load according to the pulse width. With this in mind you need to have adequate voltage at minimum load and keep it stable when the auxiliary winding voltage swings high when the smps is on load.
This was tested using a separate power supply for the chip. The slow rise time comes from the drive circuitry around Q1. I can maybe try driving the GDT directly through a series cap, see if that helps.
 

Silvio

Well-known member
Taking a closer look at your schematic it seems rather odd when it comes to the switching the external mosfet. I cannot understand why the primary of the GDT is also attached to the drive Pin 6. Who is doing the switching to the GDT? Is it the mosfet or the totem pole in the IC?
Is there some kind of mistake here?
Pin 2 of the GDT should be connected to the 12v supply I guess and not to the drive pin 6. The mosfet then will switch the winding to the full voltage. The drive pin 6 will only drive the gate of the mosfet. D2 and the snubber will shunt down any spikes generated during switch off of the GDT winding.
 
Last edited:

Silvio

Well-known member
What mosfet are you using? I hope its not a large one as it may absorb all the switching current from the drive IC.
Taking a look also at the drive IC in the blog schematic (UCC28730) This IC has only one mosfet driving the output. In the uc3845 however has a totem pole in its drive output. Practically saying that the mosfet in your previous arrangement is not really needed. The totem pole within UC3845 will do the same thing. I will give it a try if I were you without the mosfet. If you have the right inductance in the GDT you can drive it directly from the IC.
 
Last edited:
What mosfet are you using? I hope its not a large one as it may absorb all the switching current from the drive IC.
Taking a look also at the drive IC in the blog schematic (UCC28730) This IC has only one mosfet driving the output. In the uc3845 however has a totem pole in its drive output. Practically saying that the mosfet in your previous arrangement is not really needed. The totem pole within UC3845 will do the same thing. I will give it a try if I were you without the mosfet. If you have the right inductance in the GDT you can drive it directly from the IC.
Q1 is 2N7000. I did remove it along with D2 and placed a cap between pins 1 and 3 of Q1 and the waveform looks better. But nothing changed, and eventually one of the mosfets blew up along with some tracks. I'll re-do the PCB.
 
Top