# Power supply design approach

#### Blue_Electronx

Joined Jun 10, 2019
112
Hi,

I want to design a power supply with these specs for metering purposes:
Input voltage = 100-600VAC three phase (RMS line to line)
Output voltage = 48V and 3.3V/200mA

I'm looking into two approaches: A cap dropper and a flyback SMPS. Flyback has higher efficiency at the expense of higher cost and complexity. Cap dropper is kind of a cheaper option but capacitors at this high voltage are also expensive.

I'd appreciate any suggestions on trade-offs.

EDIT: Transformerless power supplies are no a permitted topic. So I will go for Isolated Flyback.

New question: For this high voltage, 600VAC or ~900VDC full rectified, what would the impact be in the transformer design in terms of cost and size?

Last edited:

#### wayneh

Joined Sep 9, 2010
17,176

#### MisterBill2

Joined Jan 23, 2018
13,185
First of all any supply with a ratio of input voltage to output voltage and a possible 600 volts in MUST have a transformer and a fuse on the input side. Doing it right is cheaper than doing it wrong and a whole lot safer as well.

#### MisterBill2

Joined Jan 23, 2018
13,185
In addition it is not at all clear what you mean by "for metering purposes". In addition there is no mention of the current required from the 48 volt part of the supply. Thus I can not even suggest the size of a suitable step down transformer or even begin discussion of a suitable circuit.
In order to get any useful suggestions you need to provide a listing of the input voltages that may be be used, the line frequency, and some description of the intended application of this supply. And also the degree of regulation for those two voltages. AND, is that 48 volt output supposed to be AC or DC?

#### ronsimpson

Joined Oct 7, 2019
2,404
I some times like capacitor/non-isolated supplies for metering, but I don't see how you can get 48V and 3.3V. I have done +5V and -5V but two outputs is not easy. 100V to 600Vac is hard to do that range and you then need to add 50/60hz. The range is (600/100)x(60/50)=7.2:1. That is just way to wide for a capacitor supply.

Doing a isolated supply, it won't care much about 50/60hz. So it only needs to do 6:1 voltage range. Not easy but can be done.

For the diodes: My last metering supply had to live off all three phases. If one or 2 legs went down it would still work. I used two-1n4007 diodes for each diode. ----edited---Now that I think about it, I also brought neutral in to the bridge with 2 more diodes so it would work if only one leg was hot. ---edited---

Next you will need a 1500 volt MOSFET in the supply. Your DC at the rectifiers will be very close to 1000 volts.

Looking forward to knowing what the total watts out is.

#### ronsimpson

Joined Oct 7, 2019
2,404
If you don't care about isolation and many meters don't care:
I found these don't regulate the best so I might go for 5V and use a linear to get 3.3V.
Just something to think about. Buck down no isolation. No transformer to design.
I remember the LinkSwitch is only 900V. Better check the max voltage.

#### ronsimpson

Joined Oct 7, 2019
2,404
If you need isolation:
I have one of these on my bench right now. I used a (TL431+2 resistors) for VR6. Better regulation.
I have 5V and 12V out. Could easy be changed to 3.3 and 48V. It can only regulate one output at a time. So the 3.3 is good but the 48 will be OK but not real good. (could regulate the 48 and use a linear regulator to go from 5V down to 3.3v)

Q1 is a very cleaver way to make the LNK150 work to 1500 volts.
Again I doubled up all the 1n4007s because we are too close to 1000V.
D9 is a UF4007 not a 1N4007!
Watch out for high voltage!
RonS.

#### Blue_Electronx

Joined Jun 10, 2019
112
Thanks @ronsimpson . I was looking into that design in the document I have attached below. It looks like a good approach, just that I'm not using neutral connection. I don't need isolation since the device is pretty well encapsulated. But I wanted to go for a SMPS for more safety and higher efficiency (although I don't care too much about efficiency in this case). The thing is that supplies like cap droppers have pretty low efficiency and current at such high voltage, furthermore high voltage caps are expensive! The 48V is needed for a solenoid with a DC resistance of 300 ohms. The 48V is a pulse applied briefly just to actuate the coil (say 30 ms), so no need of high current as I'll be cutting off the power asap. The 3.3V is for an ultra low power micro.

#### Attachments

• 278 KB Views: 3
Last edited:

Joined Jul 18, 2013
26,002
If you need isolation:
I have one of these on my bench right now. I used a (TL431+2 resistors) for VR6. Better regulation.
I have 5V and 12V out. Could easy be changed to 3.3 and 48V. It can only regulate one output at a time. So the 3.3 is good but the 48 will be OK but not real good. (could regulate the 48 and use a linear regulator to go from 5V down to 3.3v)
View attachment 198609
Q1 is a very cleaver way to make the LNK150 work to 1500 volts.
Again I doubled up all the 1n4007s because we are too close to 1000V.
D9 is a UF4007 not a 1N4007!
Watch out for high voltage!
RonS.
Why the rectifier in the (star?) neutral?
In a normal 3ph bridge, there will be no current flow in the neutral.
A standard 3ph bridge only has 6 rectifiers.
Max.

#### ronsimpson

Joined Oct 7, 2019
2,404
Why the rectifier in the (star?) neutral?
My specifications:
I was told, if one or two Ls are broken the meter must keep alive. The meter reports if wire(s) are down.
If I include N in the bridge the meter stays alive even with only one line hot.
---edited---
I think I went with 500V 15uF caps not 450V. I was worried that out on the very end of a long power line the voltages might inaccurate. There likely are large motors turning on/off.

Last edited:

#### Blue_Electronx

Joined Jun 10, 2019
112
My specifications:
I was told, if one or two Ls are broken the meter must keep alive. The meter reports if wire(s) are down.
If I include N in the bridge the meter stays alive even with only one line hot.
---edited---
I think I went with 500V 15uF caps not 450V. I was worried that out on the very end of a long power line the voltages might inaccurate. There likely are large motors turning on/off.
Do you know an estimate on the cost of this approach?

#### MisterBill2

Joined Jan 23, 2018
13,185
OK, so now we learn that it is a device to check for down wires, which will drop the voltage, or provide an alert about open phases causing the neutral point to shift from zero. So really what would be handy is a means to monitor the voltage of each phase relative to the neutral or ground. That totally alters the whole concept.