Hello,

I have a power supply problem that I've been working on for several months and haven't solved...I'm hoping someone on this forum can give me some suggestions. By the way--I'm a mechanical engineer (actually haven't even graduated yet), so my understanding of PWM and electronics in general is not too deep.

My senior project is retrofitting a six-axis robot with a new control system, which includes the power supplies. I need two voltages: 55VDC @ 10 Amps, and then 80VDC @ 20 Amps. These are weird voltages and high power requirements, and I can't afford to buy a one-off design. But, I'd still really like a switching power supply because it allows for variable input AC voltages (which is important for this robot), but still produces a constant DC voltage.

So, after a few hours of reading Power Supply Cookbook by Marty Brown our team decided we could build our own power supply. Even though we knew nothing about it. Haha, how young and arrogant we were! I wrote an excel spreadsheet to do the component calculations, we ordered the components, and soldered it up...

As I expected, the first time we plugged it in things blew up....all 4 of our MOSFETs cracked in half and jumped off the board. But I redesigned the gate driving section to include two isolation transformers and tried again, and this time SUCCESS! At least it appeared so... what is happening is that it will produce approximately 55 VDC, but under even the lightest load of like 0.5 amps the voltage drops like a rock down to 5-6 VDC. And yes, we do have a resistor across the output to maintain a slight load on the supply, though this isn't shown in my diagram.

I think something is wrong with our voltage feedback section. It contains a TL431 chip and I'm not sure how that is supposed to work. I believe it's a Silicon Controlled Rectifier (SCR?). Also, we are attempting voltage-mode control of this system using the UC3527AN driver chip. I'm becoming more convinced this was not the ideal choice for a switcher chip because it seems current mode control is more common...but this looked simpler at the time.

Perhaps I have confused the INV and NI inputs to the driver chip? Also, I'm using a MOC8102 optoisolator and we thought maybe the feedback signal was getting distorted in passing through it. Finally, what grounds should I be connecting to each other? I don't think the case ground should be connected to the output ground, right?, because this removes the isolation protection of the main transformer.

One other possibility...the output rectifier (right after the main transformer) is not an ultrafast diode type (the part number is GBPC3504), and it gets really hot. Is this somehow choking off our voltage? Oh, and the max duty cycle for this driving chip is 50%...how do I know if I need a higher duty cycle on the PWM of the driving signal to get the required output voltage?

The circuit diagram is attached...it's drawn by hand so if you have questions please ask! I realize this is a lot of stuff to consider but if anyone has suggestions or ideas on what to try I would really appreciate it. We're not sure what else to try, and if we don't solve this soon we'll have to go back to a transformer/rectifier/filter cap supply which will fluctuate depending on the input voltage It would work but not be as flexible.

If I can clarify anything, please ask.

Thank you,
Brennan

 

I am not an expert or much less and I am starting to read about SMPS, but I think the problem is the use of the optocoupler. Unless you found it being used like that , check the correct feedback polarity and try different scenarios.

You may try to avoid it at all and use a resistive network for the feedback. Yes, you will have your output not isolated from the PWM control, but that is usually not a problem , it is , in fact, one of the conditions of the full bridge SMPS. You can also try this for troubleshooting, then once the cause is found, you may consider going back to the optocoupler. Hope this helps.

 

That's going to be quite a load on the mains.

If your 80v 20a supply is 80% efficient (which would be quite good for a 1st go), you'll need 2,000 Watts of input power, or 9.1 Amperes.

The 55v supply will require another 688 Watts, or another 3.13A for a total load of 12.22A.

Does your mains supply support that much current on a circuit? If you have other loads on the same branch, you will need to figure those in as well.

The TL431 is a shunt regulator, sometimes called an "adjustable Zener". Texas Instruments and other manufacturers make this regulator. You should look at a datasheet to see how it functions.

I'm not quite sure why you're using the optoisolator either. I'd be using a comparator on the voltage divider to detect when the output voltage is high enough.

Your schematic really doesn't say much about the transformer you're using. Unless it's been very carefully designed, it will be very inefficient. For example, you couldn't simply use a transformer that was designed for mains power, as you would have terrific losses in the core. Mains transformers are designed for 50Hz/60Hz sine wave input. Your switcher is running at 100kHz, and the H-bridge is essentially placing a square wave on the primary ends of the windings.

[eta]
Your schematic shows a UC3527. Are you certain that is the correct part number? Or is it a UC3827, or something else?

I can't seem to locate any information on a UC3527.

 

Jofre- thanks for the suggestion on the optoisolator. I had heard from another engineer that that could also be the problem. So we tried removing it and jumping across the pins...obviously didn't know what we were doing because this caused all of our transistors to blow up.

How would I go about using a resistive network as feedback? It is basically a resistor voltage divider, which drops the output voltage to say between 0-5 volts for the control chip to sense? I could probably figure that out...

The way we have the optocoupler connected is suggested in my reference book, but you are the third person to suggest this could be the problem so maybe the book has a typo.



SgtWookie- thanks for the suggestions as well. If you go to mouser.com and type "UC3527AN" into the part# search, it is the first chip that comes up and the pdf datasheet is available there.

The mains circuit will support up to 20 Amps and is dedicated to this load, so that will not be a problem.

As for the transformer...we are indeed using one designed for 60 Hz. I wonder if this is the problem; even though it has the correct turns ratio I guess the high frequency will not work with that sort of core. I'm not sure about winding my own transformer...I doubt that I'd be very successful. We probably can't afford to buy one at that kVA level either, like you said it needs to be around 3kVA which would be 100s of $$$.

I believe we are going to give up on the SMPS and go with an unregulated transformer/rectifier/capacitor power supply for the time being. Time permitting, I may come back to this problem in a few months. It has been a terrific learning experience either way.

Thanks again for the suggestions.

 

First of all, what a PSU
Definitely not a first project for a beginner.
These kind of Power needs careful PCB layout, no matter what, you'll be lucky to get a regulated one.

Second is resistive feed back is a no no for SMPS's. Dangerous Live loops available to the User and chassis

Third rethink ur design and go for UC 3843 , UC 3844 PWM or SG3524.
and try winding SMPS ferrite core transformers.

Plus use a very heavy duty wiring.
and SMPS output should have a minimum constant load

 

Firstly, I think you are a brave and fearless person to dive into such a design, few are so fearless (I do not count myself as fearless in such areas either!).
I have the feeling (hopefully wrongly), that you have bitten off more than you can chew, sadly......
Firstly, I hope that someone here is clever enough to help you get your power supplies working properly.
Secondly, if that does not happen, I can send you a .pdf from Elektor that will show you at least how to build a SMPS for the 55 volts at 10 amps supply using some chips for all the intrecate stuff.
Using the same principles shown in this article (by the way I have already built one, it worked straight off!), you could also construct a 20 amp version quite easily......but with a max of 60 volts only.......if you could use a lower voltage (steppers?) then that would do it easily....
Just let me know if you need the help and I will mail you the details....
Regards
Andy

 

Hi Andy,

Haha yes I believe we have bitten off more than we can chew! But that was halfway the point of this project....I mean it would be awesome if we got it to work, but we can always fall back on the unregulated supply (just rectify the 208 VAC mains power). I attempted it for the challenge and so I could learn about electronics. To be honest, I either expected it to fail in a spectacular fashion or figure it out in a few weeks....but we have ended up stuck with a partially working supply for a few months now.

Because I have a limited amount of time to finish the project for now we will go with just transforming and rectifying the 208 VAC. We cannot go with a lower voltage than the 80 VDC for the servo motors because the speed would be decreased to an unacceptable level, or afford to switch them out for stepper motors. 80 VDC is already under their nameplate voltage but that is the max our servo driver can handle.

If you are willing to send me the schematic from Elektor, for curiosity's sake I would like to look over it. I do not expect we will have time to build and troubleshoot any more circuits, but I would like to compare it to what we have. You can email it to steelspin218@gmail.com and I appreciate the offer for help. Thanks

 

Firstly, I think you are a brave and fearless person to dive into such a design, few are so fearless (I do not count myself as fearless in such areas either!).
I have the feeling (hopefully wrongly), that you have bitten off more than you can chew, sadly......
Firstly, I hope that someone here is clever enough to help you get your power supplies working properly.
Secondly, if that does not happen, I can send you a .pdf from Elektor that will show you at least how to build a SMPS for the 55 volts at 10 amps supply using some chips for all the intrecate stuff.
Using the same principles shown in this article (by the way I have already built one, it worked straight off!), you could also construct a 20 amp version quite easily......but with a max of 60 volts only.......if you could use a lower voltage (steppers?) then that would do it easily....
Just let me know if you need the help and I will mail you the details....
Regards
Andy

If the schematic is published then I guess you can post it here, for one I like to see it