Beginning SMPS?


2004-08-01 8:03 pm
Hey everyone,

I have a good amount of experience with regulated and unregulated power supplies, and want to move into the SMPS arena.

I've done a good amount of research on the subject, and I understand the different topologies, and if you don't know what you're doing, you might not get another chance. :att'n:

So, with this in mind, where would be a good, safe place to start? What should I do if I eventually want to be able to build say a 400-500 watt SMPS off 120VAC?

Any and all help appreciated, including articles, schematics, parts, etc.

I've seen someone in here tweek out a K6 power supply. It is a kit you can purchase. It was designed for audio amps, although there are some improvements you can do to make it more robust. They are located in your neck of the woods!

You could start with the kit and study the wave forms, etc. You can download the schematics at the website first, if you wish. There's a legthy thread on it in the Class D forum here.
Yeah I am in a similar situation..

I had thought direct off line smps discussion was not allowed, or maybe the design of them (there were a few locked posts).

Anyway I can find a lot of 12V smps circuits and designs, but not a whole lot of 120V.

I would like to learn some board design as well as how to wind the transformer.

btw that k6 psu schematic, 2200uF seems like an insane amount for 200V input caps. Although they dont look too big in the pics.
I've seen some closed threads where people suggested directly connecting amplifiers to mains line, and I can't complain because that is quite an insane idea. However, I think I haven't seen any serious thread about SMPS closed or censored.

Anyway, there are some safety guidelines that anybody willing to experiment with off-line SMPS should follow:

- Never power your SMPS prototypes directly from mains line, use isolation transformers instead. A variac is also recommended in order to test circuits at different voltages. In case of money shortages, several transformers scrapped from audio amplifiers may be used with their secondaries connected in series to get the desired voltage. For example I routinely use a 500VA unit with two 40V secondaries and a 750VA unit with two 60V secondaries, this metod is quite smart because I can test my circuits at 40V, 60V, 80V, 100V, 120V, 140V, 160V, 200V and even 20V AC, depending on how I connect the windings. Lower voltages are safer and render less catastrophic results in case of failure.

- Keep 5mm or more clearance between 'mains-side' and secondary side PCB tracks. Note that mixing both sides in the same breadboard is not recomended at all.

- When winding transformers, pay great attention to output isolation. There should be three layers of isolating tape between primary side and secondary side windings. Magnet wires from different sides should never be in direct contact, there must be either isolating tape, proper plastic tubing or 3mm of air inbetween. There should be 5mm or more clearance between solder pins of different sides.

- Use fuses rated at the smallest possible current placed before and after the isolation transformers. Use light bulbs or even halogen lamps placed in series with isolation transformer output, they work fine as current limiters. Use primary-side peak-current limiting or shutdown. All these tips are very useful to prevent catastrophic failures.

And finally, remember that there are some mistakes that will hurt you, others will put your house on fire, and some others will kill you, so always act at your own risk.


2004-08-01 8:03 pm
Thanks for the tips!

My ultimate goal is a well-regulated, efficient supply (I guess that's nothing too unique;)). So however I can get that, assuming it's safe, reliable, and won't break the bank, is fine.

So, idea (I don't know if this has been said else where or not, but what you said just inspired me): get a standard toroidal transformer, say 30V (of good power rating), hook it up to the mains line and rectify/filter its output to DC, probably giving me around 40VDC. This will give me mains isolation and lower voltages to work with. Then feed this to a SMPS, which will regulate the voltage for the final output.

Good? Bad? Different? Too much work/cost for too little gain?
Yes, this is how I tried my first half-bridges and full-bridges some years ago. I was using IRFP460 MOSFETs with transformers for isolated gate drive, and a SG3525 PWM IC with a potentiometer to adjust duty cycle and powered with an auxiliary supply. At first I didn't even have got power transformer cores nor ultra-fast rectifiers to experiment with, so I used lightbulbs as a load for the switches. It was very frustrating until I got an oscilloscope...

Also, I tested my first PFC prototype with 40V AC input, and with the circuit temporarily patched for producing only 200V DC output. When it worked fine I removed the patches and tested it with 120V AC input and nominal 440V output. New problems appeared... :)
It depends on what you want. If you just want to learn and experiment with switching circuits then this kind of setup will be fine.

However, if you are more ambitious and have put your aims in efficient off-line SMPS design, then it will be just a begginer's setup for learning, training and testing. As you gain experience, your designs may start to be reliable with high (isolated) voltages like 120V or 230V AC and you may learn how to avoid risks and catastrophic failures.

After uncountable hours of testing, you might consider that your circuits are mature enough and met reasonable realiability, EMI and isolation requirements in order to be actually operated off-line.

Note that off-line operation is a potentially harmful practice where mistakes are not allowed. Shall you decide to do it, you will be following your own criteria and acting at your own risk.
If you haven't done it already, you may try classic 12V push-pull car-audio SMPS stuff as an starting point. First you may try a simple unregulated topology. Then you may try to add regulation and mess with filter and transformer design, current and voltage feedback loops, etc.. There is a lot to learn here and this knowledge will be useful later on for off-line designs.

Do you own a car?

Also, don't you have old transformers, heatsinks and filter capacitors scrapped from audio gear? This kind of stuff will be very useful.

Currently I'm experimenting with average current control. It allows for precise current limiting, current sharing between several independent SMPS and very fast transient response. The power switching section of this prototype (full-bridge transformer-coupled buck converter) is mature, reliable and protected by cycle-by-cycle current limiting, so I'm using 200V AC (isolated) for testing (about 280V DC). Actually most of the trouble I'm currently dealing with comes from the control loop and the requirement for an isolated DC current sensing mechanism. I have just discovered zero-flux current transformer operation (invented by someone in 1969). Note that my control circuit is also isolated from the primary switching side, so I have double insulation in my breadboards.
This is a simple and small unregulated push-pull power supply. It takes 12V DC as input and produces two independent, perfectly isolated and carefully filtered 12V DC 1.5A max. outputs. The layout is somewhat optimized for EMI reduction.

I developed it to break ground loops in multi-media car systems where you have at the same time a DVD player, a TV tuner, one or more TFT screens, etc..., thus removing unwanted audio and video parasitistics caused by badly designed equipment.

Board layout:

Three boards just after being etched:

More pictures:

Finished unit:

Parts list:

1n C9,C10,C17
1N4148 D9,D10,D11,D12
1u 63V C7,C8,C13
2k2 R11,R12,R15
6k8 R10
10n C14,C15
10u 35V C18,C19
10uH L1,L2
11DQ6 D2,D3,D6,D7
11DQ6 D1,D4,D5,D8
22 R5,R6
22k R13,R14
22u 35V C16
47 R7,R8
68 R3,R4
68uH 4.7A L3
100 R9
100p C5,C6
220 R1,R2
470u 35V C1,C2,C3,C4,C11,C12
SG3525 IC1
BC546B Q3
BC556B Q4

switching freq.: 120Khz aprox.

winding notes for magnetic components:

T1 :
NTF 16 (16mm o.d. high permeability ferrite core, probably 3E25 mat.)
pri : 12T + 12T 0,5mm 28cm
sec : 13T + 13T 0,5mm 30cm

T2 y T3 :
NTF 16 (16mm o.d. high permeability ferrite core, probably 3E25 mat.)
20T + 20T 0,6mm 48cm [21T actually]

L3 :
grey-white 16x9x7mm iron-powder core (unknown mat.)
68uH 4.7A 26T 0,6mm 55cm 0,036ohm
Thanks Eva. Unfortunately, your pictures don't appear to be coming up for me.:( Don't know if it's me or the server.

Sounds like a great use for a SMPS though. Still trying to learn, so let me ask a question: you have an extra winding on the secondaries of your transformer. Is that extra winding there so you can still get 12 volts out while leaving some down time between switches to not create a short circuit on the primary side?

Thanks for the part list! I will be referring to it for sure! BTW, where do you typically go to get cores?
The lack of images was my fault.

There are no additional windings, there are just a center-tapped push-pull primary and two independent isolated secondaries.

I get these ring cores in a local store, they are somewhat cheap but their manufacturer and model are unknown so there are almost no specs for them. I had to learn to measure by myself permeability and energy storage for iron powder cores, and volts*seconds/turns product before saturation for ferrite ones, so I can design with them.

In Europe I have also bought magnetic components from Farnell, Conrad Electronic (Germany) and RS-Amidata.


2004-08-01 8:03 pm
Unless I'm mis-reading it (very good possibility FYI), you listed the primaries as having 12 turns, and the secondaries as having 13?

Sounds like learning to measure the permeability of the cores I use is something I'd like to do (especially so I don't design a PSU that saturates the core I'm using, right?).
rjon17469 said:
Unless I'm mis-reading it (very good possibility FYI), you listed the primaries as having 12 turns, and the secondaries as having 13?

This is because the output of each secondary winding has to be rectified and there is 500mV approx. voltage drop on each schottky rectifier under maximum load (1V in total), and this voltage drop has to be compensated in order to get back 12V output.

rjon17469 said:
Sounds like learning to measure the permeability of the cores I use is something I'd like to do (especially so I don't design a PSU that saturates the core I'm using, right?).