Need to design for 380 VAC mains 240 W power supply

12 posts / 0 new

I am a new user to PI Expert and to PI chips. We need to design a 24 V @ 10 A power supply based upon TOP260YN chip. PI Expert calculated a design for the 220 VAC range, but I think it is impossible to achieve the input power range we need. I have seen that PI uses cascaded MOSFET on top of the MOSFET inside the chip, but I don't know how to design it. Could you help me modify the circuit that PI Expert suggested for the 220 VAC mains to work in an industrial environment with a single 380V AC mains 50 Hz ? Attached are a circuit snapshot and a PI Expert file containing the closest match to what we need except for AC input.

Thank you !

TOPSwitch-HX_PI_3280V.uds223.5 KB

have a look here:
And try typing "StackFET" in the search box in the upper right.

Dear PI-Tucker thank you for your reply.
I have read the complete article and understood the concept of StackFET.

Considering that PI Expert is not capable yet to handle this kind of circuit, I used PI Expert XLS as suggested in other posts. I have some doubts about the resulting parameters of the final design.

In my case, I need a tri-phase design 3 x 380 VAC input.
Attached is the file of the PI Expert XLS results. The input capacitor value has been exaggerated as adviced here, and I iterated some parameters to obtain just 1 warning regarding "REDUCE DRAIN VOLTAGE Vdrain<680 ... " that sounds reasonable ( the program is trying to protect the device ). Please kindly find the attachment with the program output.

For my tri-phase 24V 10 A output design, a transformer specification came out around an ETD39 core.

Regarding the schematic in ..

1. could you check if the results are reasonable ?
2. is TOP261YN correct for this power supply ?
3. the values of R6,R7,R8,R9,VR1,VR2,VR3,VR5,D9,R10 and VR4 15V zener are still OK for the needed rating ?
4. how is the input PI filter is calculated ?
5. can I obtain correct values of feedback TL431 + PC817 using PI Expert standard at 24V output ?

Thank you

Roberto Berner


1.Your spread sheet shows ETD29 core. But I saw in your post that you got EDT39. Try changing that to ETD39 by keeping the same secondary no. of turns so that you can reduce the peak flux density.

2.Yes, TOP261YN can be used to handle that much amount of power.

3.You have to redesign the clamp circuitry since the power dissipation will be higher in this case or else you can try using the clamp circuit which you got from PI Expert as a starting point and later you can optimize it.

4.There is no way to calculate that input EMI filter. Here is how it is done: based on your experience you start with a set of values and build one prototype. You get this thingy to the EMI test setup and discover the noise signature. Many circuits are involved in generating noise there. The main source is obviously the main power switch. Also, the OFF time ringing of reset circuits and output rectification diode ringing are major contributors. The shape and trace proximity of the PCB is also important. And not to ignore, the inductors used on the input AC EMI filters can oscillate when exposed to sudden current steps, hence the need in many cases to add a dumping resistor in parallel to these inductors. After you see the actual shape of the EMI signature of your first prototype you have a chance to identify probable sources and implement corrections, step by step. It is very much monkeying around with incremental changes to get the final compliance. If you do not have access to an EMI compliance test setup, you’re pretty much stuck with no other options. Most of our examples are EMI tested and is very much expected default regulatory compliance. The design tool gives you the original starting point based on our experience and you have to struggle with the final product compliance. There is a good chance to get the product right after the first trial but this is something you have to confirm. But for your power level a simple pi filter might not be good enough to pass the EMI. You can try using common mode and differential mode filters in order pass EMI.

5.Yes, you can get feedback component values by using PI Expert since StackFET configuration doesn’t affect the controller operation.


Hello PI-NANO, thank you very much for your so valuable advice. I have fixed the core size to be a larger ETD39 instead of an ETD29, keeping the secondary no. of turns = 19 and it seemed to work (no warnings). I am attaching the resulting spreadsheet for you to kindly check. I am sorry, I am totally inexperienced with PI devices and I have little experience with switching devices in general. I have built some power supplies with the old TL494 and they worked fine but I am not able to fine tune some details. I understood what you explained in all points. I think you are right and the EMI issue needs some manual tailoring and I think I can manage it. My last concerns by now are the following, before setting on fire my prototype, and I hope that you guys will also give me a hand with this: considering the circuit in my feeling is that I don't have to alter the values of R6,R7,R8,R9,VR1,VR2,VR3 and VR4 because I think that all this parts are related to TOPswitch's device protection and the zener clamp for the gate of the StackFET @ 15V. Please correct me if I'm wrong. It seems that my design will use these same values. Instead, I am afraid VR5, D9 and R10 should change because they handle damping in off condition in a 240W power design. I believe this network could handle less than 0.5 A considering a tri-phase positive bus of ~535 VCC but I really don't know how to calculate these values for 240W output power. I tried PI Expert and will not help because it ends at 375V. For the rest of the feedback components I agree that they could be calculated with PI Expert because it is almost the same for any device knowing the output conditions. So could you please give some advice and put some light about this last step around these 3 components ? I really appreciate you time and patience.

Roberto Berner


You need not change values R6,R7,R8,R9,VR1,VR2,VR3 and VR4. You can calculate the snubber components based on the estimated leakage inductance and the peak primary current (By calculating leakage energy that needs to be dissipated). You can go through some of the documents which are available one the internet for the snubber circuit design. Here I am giving couple of links for those, but you can find so much of information in the web.


Thank you PI-NANO. I made some research about snubbers and I understood not only their function but also that the calculation is hard. It is based upon the ringing frequency of the turning off signal over the switch. To measure this I should use an unclamped device which is dangerous. Almost every paper speaks about a 3 element snubber network formed by a capacitor in parallel with a resistor and a diode. The basic circuit is what you can see here:

Please correct me if I am in error. The circuit that is seen in is more what I found is called a peak clamp instead of a snubber circuit. It does not seem to be a resonant circuit to me. It just uses a resistor and a peak clamp. Please find the attached ST datasheet for one of these devices. So I still don't have this point clear and the way to calculate the clamp. If it is a simple clamp, I think that it should clamp the maximum tolerable voltage of both MOSFET's forming the StackFET when the switch turns off. I could say that this clamp voltage for a bus line of almost 600 VCC could be as much as 700 or 800 V ( this is a question not an affirmation ). The role of the limiting resistor is also unclear to me if the device can manage the clamp by itself. I was thinking if we could consult the designer of this circuit about this particular issue. What do you think about this guys ? Thank you very much again for your help.

Roberto Berner


The purpose of having a snubber is to limit the peak drain voltage spike created by leakage inductance during turn-off transition of the MOSFET. The snubber can be a RCD network, it can be a zener clamp or it can have combination of both (RCDZ) based on power level. At higher power levels single zener might not be able to handle the power loss with acceptable thermal limits. Then you might have to use multiple zeners in series to distribute the power loss. In a RCD snubber network energy stored in the leakage inductance will be transferred to the capacitor. Some portion of the energy stored in the capacitor will get dissipated by the resistor and remaining portion of the energy will be retained by the capacitor. The reason for using RCDZ snubbers is to improve light load efficiency (At light loads the amount voltage spike on the drain and leakage energy that needs to be dissipated are smaller, so Zener will not conduct during that period (only RCD will conduct) if you size the snubber properly).

From the link specified below you can find some general design procedure for flyback converter which includes snubber design guidelines.

If you need any design help, you can contact one of our FAE’s in your geographical area. They will be able to help you with design. Please find the link here for our sales offices.


Thank you again PI-NANO, I printed a hardcopy of the webpage you suggested to study it. With all respect, please let me explain that my choice for PI was because of the simplicity of the chips and also because PI provided easy to use tools for people like me that have much less experience dealing with this kind of devices. By using these tools, my goal is to place a tri-phase power supply in 2 of our production equipments, before buying a complete power supply module to a third party, probably in Asia. Thanks to you and other people I could reach a point in which I have almost 90% of the circuit that is needed regarding StackFET and other tricks to make it work at a high voltage. What I need is somebody to suggest the values of these 2 or 3 missing components as a starting point for optimizing the design. I am really stucked at this point and I don't think that I would hire an engineer to calculate them. I need something faster the same way it is fast to normally implement a power supply with PI components. Notice that in my particular case there is no PC board with reference design available that I could happily buy from PI ( 24 V @ 10 A continuous output power ), and that when I called by phone to contact a support engineer, the answering machine redirected me to this forum. I don't want to bother with my questions but maybe somebody could do me a favor and calculate just these 2 tentative starting point values for me, as I can see that you all are experts that have the knowledge that unfortunately I don't. At this moment none of the software tools can deal with StackFET designs while anybody can obtain a 100% complete almost working schematic, PCB and transformer design for any other case. In this way, the spirit of ease of use that I felt when I first approached to PI would be accomplished. Thank you in advance.

Roberto Berner


I am covering for PI-Nanao since he is away. I have studied your post and all the subsequent posts on the subject.

The StackFET is an arrangement that can be used for designing power supplies that can work from high input voltage so that the V-DS rating of the MOSFET is not exceeded. The VDS rating of the MOSFET inside our parts is typically 725V.

The StackFET arrangement though results in a slightly lower efficiency and is probably not a good solution for a power level as high as 240W. We typically use it up to 20 or 30W at the most.

I believe at this high a power level ( 24V , 10A) you are better off considerinbg other power supply topologies such as the half bridge configuration.

If you choose to use a Flyback, the snubber loss would be considerable.

Unfortunately we do not make any parts that may be suitable for your use and you may be able to design using discrete power MOSFETs and a controller.

Befroe you move further, it is important for you to dfetermine the correct lowest and highest operating voltage which will set the MOSFET rating that you will need for your design.

I hope you find this information useful. I am sorry if I dissapointed you regarding StackFET, but the reality is that the StackFET arrangement while ssimple, results in lower efficiency which may be OK for low power applications but is certainly not recommended for this high a power level. We have not implemented a StackFET design at 240W and hence do not have a reference design.



According to my experience, TOPxxx chips cannot able to get so much powerful output. Maybe 60W-80W only.

Thank you PI-Sarek for your return and for keeping track of the forum thread for my case. I understand what you are explaining but I would like to contribute to this forum, if you allow me to express a couple of thoughts. The spirit of these comments are positive and respectful, always trying to help improving such a powerful tool as this forum is. So please don't take me bad. First of all and despite your very much appreciated explanation, and considering a single phase 175 to 265 VAC mains range design, the 240 W output is something that the PI Expert software itself allows with no complaints. All the related bibliography and available material do not discourage users from such a 240 W power output and more than that, the datasheet for the TOP261 family talks about more than 300 W output in open frame devices. Even if the topology is that of a flyback converter, which is the standard design topology chosen by PI, it makes me conclude in that PI is trying to make big efforts to allow people like me (with years of electronics experience in many fields) to feel that designing SMPS with these chips is really easy, even if these users don't have much experience in this particular SMPS subject, considering the large amount of design details to take care of. So this fantastic PI Expert software is even producing output of a complete luxurious schematic with a complete BOM that the user can trim to reach a perfect optimization. I could never think that these output results are wrong, even if a half-bridge topology could be a better option as you explained and I agree. Using this PI Expert design schematic is still a good option for me in terms of simplicity, parts count, design effort, etc. In second place, PI-nano interacted and helped me obtain almost 90% of a working project without never mentioning this problem of the StackFet inefficiency on top of the 240 W output power inefficiency itself. I really don't understand this point, specially considering that I am overseas and have spended several phone calls to get PI's help in the form of a design support engineer. All what I've got is an answering machine deriving me to this forum as the only way to get advice on PI's products design aid. The other suggested way is a field sales engineer in Georgia ( I will not mention any contact name ) that is not returning my calls after a couple of recorded messages. So this is my question (respectfully): what's the use of going into a deep conversation full of details and illustrated with uploaded and downloaded excellent sample designs to end up with this final recommendation of not using 240 W power ? To me it is a little confusing. I want to remark that I understand that I asked for a tri-phase design with the PI suggested StackFet configuration. Again, the StackFet was recommended in this forum by a PI expert but it seems that the needed target specification was skipped if the StackFet is inefficient for my case ( 240 W , 600 VCC ). To finish my comments, the missing 10% of my project was about calculating the snubber network for my design. I asked for practical values that someone expert could calculate for me and you can read from the posts what I received in return. I suggest to limit the PI software to discourage designs out of the safe operating area or the reasonable efficient area. Also, I suggest PI forum experts to read what customers really need before giving advice on impractical implementations. Finally I suggest to improve technical support live contacts even if they are payed including the possibility of checking an user's design. As an example IR ( International Rectifier ) has an excellent free consulting service that helped us produce devices using their line of products. I want to thank all of the PI forum components for their collaboration in any way.

Roberto Berner