Two Switch Forward + PFC SMPS design in progress

[kkeeley]

Hi,

I've been working on designing a two switch forward converter with PFC, the design is still in the very early stages and I've been reading up on SMPS design.

So far I have found that a number of the different books out there provide formulas that are slightly different to each other and this has been making it harder to determine if I'm understanding things correctly, so I'd like to over time provide information here about my design with the hope to receive feedback to correct any errors in my understanding of what I've been learning and hopefully produce something that will work and not just produce smoke.

To start of this design is purely for personal use and I have no intention of it being mass produced, having said that it is my aim that it will be designed and built to meet any relevant safety standards as I don't wish to harm/kill myself or anybody else that may come into contact with it. I would also like to keep the cost at a reasonable level and therefor most of the work will be done by myself where possible, I will most likely only get the PCB manufactured. Any transformers and inductors used would hopefully be off the shelf designs or purchased as parts and wound myself as I feel that the cost to get them made as one off items will most likely be to high.

The basic design requirements so far are :

Input : 230Vac @ 50Hz
Outputs: 5V @ 15A, 3.3V @ 10A and 2.7V @ +/-2A

I've currently selected the following main componants:
PFC Controller : UCC28019
PWM controller : UCC3813-0
High Side/Low Side Driver : UCC27714
DDR Termination Regulator : LP2995 (used for the 2.7V output)

The basic design is to use the PFC controller to boost the 230Vac upto a 390V Dc Bus, The Power stage will use the UCC2813-0 to control the UCC27714 which in turn will drive the High and Low Side FETs. The main transformer will have a 400V primary, 12V auxiliary winding to power the on board Vcc rail also 5V and 3V secondary windings. The 2.7V output will be generated by the LP2995 connected to the 5V secondary which will also provide the 5V secondary with part of it's minimum load current. Feedback will be provided by a TL431 controlled by the 5V and 3V outputs passed back to the UCC2813-0 via an optocoupler.

Originally I was going to also have a 12V secondary that would be used to power the cooling fans and generate the 2.7V output, I have however been unable to find a supplier that has any DDR Termination Regulators that support 12V inputs. I may revisit this idea if I manage to find a suitable part/design.

Please feel free to provide feedback about the deign idea so far. In the next update I hope to provide some calculation details for the main transformer to confirm if the formulas used have produced a viable design.

Thanks
Ken

 

[Akanimo]

Hi Ken,

Your plan for the feedback loop has caught my attention. The 5V out is okay to provide feedback with TL431 but the 3V output is not. The TL431 has a 2.5V reference and the optocoupler diode has a 1V forward drop. This already is 2.5V+1V = 3.5V. You could have the 3V output without feedback.

 

[kkeeley]

Thank you for your feedback on my idea for the feedback loop.

I had based my idea for the feedback circuit on some others that I have seen, some show the fast lane powered by a 12V output and the slow lane using a resistor divider between the 5V and 3.3V outputs, I've also seen the fast lane powered by the 5V output. Based on that I had the idea to connect the fast lane to the 5V output and use a resistor divider on the 5V and 3.3V outputs to split it at say 70/30.

 

[Akanimo]

I do not get what your are saying. Can you post a schematic or a URL?

 

[kkeeley]

The attached image is a circuit that I'm looking to use to base my design for my main power stage on, this design is for a 150W 2 switch forward, so it's very close to my current requirements, as you can see the feedback circuit has the 12V output connected to the fast lane and both the 5V and 3.3V outputs connected to the slow lane.

I hope that explains the feedback design better, overall I'm looking to replace the main diode bridge in the attached circuit with a PFC circuit that also contains an EMI filter.

 

[Akanimo]

Okay. I've seen the schematic. So what's the value of R22 - the upper divider resistor at the 3.3V output?

 

[treez]

Well your output current is high, so therefore I hope you will design the 2TF txformer for low duty cycle………so as to reduce dissipation in the series diode at the output….but then the shunt diode will dissipate much, so maybe you can use a sync rect for this hopefully

 

[treez]

In fact, i reckon your design will be far easier and probably better if you ust do a 2TFC with say a 15V general output....then put some synchronous rectifier buck converters on the output to give you your 3v3, 10A and 5V, 15A.
Also, because you wont have to use opto regulation for the 3v3 and 5v bucks, you will get better feedback loop bandwidth (faster)
Linear.com do a lot of sync bucks...so do ti.com and manty others.
The sync rect bucks will be very efficient, if you dont use sync rects you will get much loss in your rect diodes.

- - - Updated - - -

You can get 2tf's with sync rects, but it generally involves a mesh of extra circuitry that is often not worth it.
Remember with syc rects you must detect the onset of dcm...and turn off the sync rect (lo side) to stop current backflowing thru the buck inductor.

- - - Updated - - -

I also notice that you have a coupled inductor at the output...this is to improve cross regulation.....current mode 2TF's often need coupled inductors to assist with good cress reg.
But its a nuisance, and really, itd be easer and probably cheaper for you to just put bucks at the output of a 15v output 2tf.

 

[Akanimo]

Okay. I've seen the schematic. So what's the value of R22 - the upper divider resistor at the 3.3V output?

I hope you can see that there is actually no feedback of the 3.3V output. The 10k-10k voltage divider is for the 5V output, dividing it into two halves -- 2.5V for the TL431 reference. The resistor that seems to be connecting the 3.3V output to the feedback network is basically an open circuit.

 

[treez]

The value of your startup resistors is low ohms...and that will dissipate much all the time.

 

[Akanimo]

Hello treez,

The posted schematic is not his circuit. It is just an exemplary design that he posted to show his feedback plan for the 3v3 output.

 

[kkeeley]

Okay. I've seen the schematic. So what's the value of R22 - the upper divider resistor at the 3.3V output?

In the above schematic R22 is not installed, but if it was going to be used I've seen R20 with a value of 12K and R22 with a value 16K, these would need to be confirmed. My understanding is that you wish to produce about 2.5V across R21.

- - - Updated - - -

@treez, Thanks for your feedback, as stated above, the attached schematic is not my design, although it is what I'm partly basing some of my design on.

Over the weekend I have put some more thought into the design and I'm thinking of reducing the 5V output back to around 10-12 amps and then adding a 12V output @2-3 amps, the 12V output would be used to power a couple of 40mm fans and I'm also thinking of replacing the LP2995 DDR Termination Regulator with a circuit I've found that was used for the bus termination supply in some S100 backplane designs. This schematic used a 8V supply but I have seen it converted to easily run from a 12V, I would also remove the 7805 as it wouldn't be required.

I've also been doing some calculations on the main transformer construction requirements and I think I may have something that could work and I also did some calculations using PowerEsim's online Magnetic Designer which appears to show the design might also work. If I was to post details here on the design, what information would people recommend I post, would the design output from PowerEsim be ok, should I also supply a summary table of my results containing things like core shape and material details, number of turns, wire gauges, number of strands, copper and core losses, utilization or should I attach an excel file that contains this data and also allows people to review the formulas I've used.

Once again thanks for all the feedback.
Ken

 

[treez]

Name of core..(eg ETD59 whatever)
NP, NS1, NS2, Naux etc...
Hopefully you try and wind for full layers of turns across the bobbin....ie dont bunch all the turns of a layer (eg the primary) at one end of the bobbin only.....
Did you gap the core? if so how much (this has been debated on edabord before...in theory , a gap is not needed)

Name of ferrite used...eg "N87", or whatever

- - - Updated - - -

You know you can design a 2tf like a buck in a way...

You simply design it as a buck with an "effective vin"
You then make the transformer turns ratio suitable to give that effective vin from your actual vin
You obviously do this by selecting your max duty cycle...(i presume this will be less than 0.45 for you)
Remember to choose your max duty cycle, and do the calculation at your min vin, and obviously max load.

 

[kkeeley]

The contents of the attached zip file is a Transformer Report generated using the online PowerEsim Magnetic Builder app and also a second file that contains the calculations I put together based on the "Transformer and Inductor Design Handbook 3rd edition" that I have been reading along with others. Hopefully the calculations are on the right path, please feel free to provide feedback on the deigns and ask for any extra information that may not be present or clear.

Ken

 

[Akanimo]

Zip file? I'm afraid that a lot of people are very cautious with dealing with zip files. Can you post it in a picture format?

 

[treez]

HERE IS A CALC DOC for a 2 tran forward preceeded by a pfc...

you dont need that much in the way of calculations.

One page shows basic 2tf calcs...then theres a few calcs of doing it with various cores, and checking out the Bpk etc

- - - Updated - - -

also attached is the checking ltspice sim of it

- - - Updated - - -

also attached another 2tf sim in ltspice.

I am sure you know that a 2tf is just a full bridge with the diagonally opposite fets just done as diodes.

- - - Updated - - -

Also, a full bridge smps, is just two interleave switched 2tf's.

- - - Updated - - -

all the maths you need for a 2tf trxformer is in the attached doc (Theres more than that in there too)

 

[kkeeley]

Zip file? I'm afraid that a lot of people are very cautious with dealing with zip files. Can you post it in a picture format?

Ok, I've converted the two Excel files from within the above attached zip file to PDF's for easier viewing, if you wish to see how the calculations where done you will have to download the zip file as this forum doesn't allow xlsx files to be attached.

- - - Updated - - -

HERE IS A CALC DOC for a 2 tran forward preceeded by a pfc...

you dont need that much in the way of calculations.

One page shows basic 2tf calcs...then theres a few calcs of doing it with various cores, and checking out the Bpk etc

- - - Updated - - -

also attached is the checking ltspice sim of it

- - - Updated - - -

also attached another 2tf sim in ltspice.

I am sure you know that a 2tf is just a full bridge with the diagonally opposite fets just done as diodes.

- - - Updated - - -

Also, a full bridge smps, is just two interleave switched 2tf's.

- - - Updated - - -

all the maths you need for a 2tf trxformer is in the attached doc (Theres more than that in there too)

Thanks for the documents, I'll compare them to the ones I generated and see what I missed.

 

[kkeeley]

HERE IS A CALC DOC for a 2 tran forward preceeded by a pfc...

Do you have any further information that explains what each cell in you spreadsheet is for, the formulas look like some I have seen before but I'm finding it difficult to determine what each cell is calculating as not all the labels are familiar.

Ken

 

[treez]

ok ill do you a fresh one, and make it clearer, ill add a sim with it too...

I am taking your spec is..

5v at 12a = 60w
3v3 at 10a = 33w
+/-2v7 at 2a = 10.8w
12v at 1a = 12w

so you have 115.8w

..back soon...

- - - Updated - - -

...also your vin is the pfc output, which is 390v

- - - Updated - - -

ok, attached is the ltspice simulation, and design document, including the transformer, for the two transistor forward converter
i made it 12vout...then you can use that for the fans, and to feed the secondary side sync bucks

- - - Updated - - -

woops sorry, i made the current sense resistor in the ltspice sim too high ohmic value...please find the corrected simulation attached...called "two transistor forward"

- - - Updated - - -

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
KKeeley…one point to watch……in the sim I gave it a coupling factor of 0.993, you will see.
…Now change the sim so that this is 0.98…….then watch the primary magnetising current…it goes above i(sat).
This is a little known thing about two transistor forwards , and other similar converters…..if the leakage inductance is too high, then you may go into saturation.
So this is one thing to watch…….in the lab, if you look at the primary current, you can actually see the peak of the magnetising current as it falls down, just after the fet switches off…check that it is not too high for your transformer core.
You know that to calculate the i(sat) you use i(sat) = B(sat)*A*N/L
This is ultimately derived from Faradays law and Lenzs law
…N.(d(phi)/dt )= L.(di/dt)
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
…If you do find that you end up with the magnetising current wandering off into saturation, then you can either re-design to get your duty cycle down lower(this allows the magnetising current more time to die down to zero)…or you can wind the transformer to reduce leakage..eg interleaving etc etc.
..What is happeing, is that if you have too much leakage inductance, then it prevents the magnetising current from die-ing down to zero.
..But i think if you have full layers of turns in both pri and sec, then generally your leakage will be low enough.
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Please note from the simulation how a certain amount of leakage inductance actually helps you…because it acts as a turn-on snubber, and prevents high voltage.current overlap at fet-switch-ON

 

[treez]

+++++++++++++++++++++++++++++++++++++++
Also, I have also attached for you here, a 2 transistor forward converter with a bootstrap high side drive, which means you don’t need a gate drive transformer. So cheaper.
When using a bootstrap hi side fet drive with a 2tf, you are better off if you use a full bridge driver IC, because then you can use one of the outputs to turn on the “refresh” fet, which refreshs the charge in the bootstrap capacitor……remember, a 2 tran forward is just a full bridge with two diagonally opposite fets replaced with diodes.