Linear PS vs. Switch-Mode PS

I have one small correction, but other than this you hit the nail on the head. (GA) Most commercial switch mode power supplies I've encountered have galvanic isolated outputs. You are correct though that many parts of the supply circuitry are not isolated from the input line voltages but a transformer operating at a frequency higher than line frequency does provide isolation. Because of this, a SWMPS should include in its design circuitry protection to incoming voltage spikes that would "vanish" in the core hysteresis and saturation of a linear supply's transformer.

I thought someone would mention that so I kept it as a generalised safety isolation point, especially as the level of electrical isolation present is SMPS specific and also depends on SMPS protection and failure modes (which can make some formally electrically safe SMPS unsafe in the event of a component failure).

Additionally some simpler low power, low component SMPS can be powered from capacitior voltage divider networks (which again are not electrically isolated).

You are both right on the money. I prefer Linear for bench work...

The comment you refer to is simply not true for most modern SMPS.....in fact, I may have been lucky, but I have never actually examined one that was not fully isolated.....

In my limited experience with over 30 years working with them, there is usually a small transformer working with the full current but at a high frequency, this isolates the input from the output effectively....

A check with the manufacturer before a purchase will find this out quickly......I cannot personally guarantee that a meter reading will be reliable enough or not.....

Regarding the efficiency and heat dissipation, I want o verify what wiki says (given that anyone can type in wiki). Under this link: http://en.wikipedia.org/wiki/Switched-mode_power_supply#SMPS_and_linear_power_supply_comparison

At the point of efficiency and heat it seems to say that SMPS have efficiency of just 30 to 40% and that (in "Notes") a good design may have 95%, whereas for a LPS it says 60 to 70%. Any comment on this?

You're misreading the Wiki article. The 30% to 40% efficiency text comes from a comparison chart of linear versus SMPS. This low efficiency is for a regulated linear supply not a SMPS.

Also these efficiency numbers are typically valid only at or near maximum power output conditions. A 500W power supply driving a 1/10W load of either topology will be very inefficient.

Here is the comments you referred to I believe:-

This is for an SMPS.

Switching losses in the transistors (especially in the short part of each cycle when the device is partially on), on-resistance of the switching transistors, equivalent series resistance in the inductor and capacitors, and core losses in the inductor, and rectifier voltage drop contribute to a typical efficiency of 60-70%. However, by optimizing SMPS design (such as choosing the optimal switching frequency, avoiding saturation of inductors, and active rectification), the amount of power loss and heat can be minimized; a good design can have an efficiency of 95%.

For a Linear supply, the following was mentioned:-

If regulated: efficiency largely depends on voltage difference between input and output; output voltage is regulated by dissipating excess power as heat resulting in a typical efficiency of 30-40%.^[1] If unregulated, transformer iron and copper losses may be the only significant sources of inefficiency.

I think that you mixed up the two.....

Usually an SMPS of a modern/reasonable design will have AT LEAST twice the efficiency of a normal Linear supply.....it will also run a lot cooler, wasting far less energy in heat.

actually i have some problem with passive infra red sensor (input 12v DC). while i am using smps it gives false signal to the system, some one suggest me about normal power supply.

Most likely this is due to insufficient filtering of the supply's output power. Without knowing anything of your system or what specific problem you're having, nobody can help you. Well at least I cannot help you. I recommend that you start a new thread on your problem. Give us as much information as you can so we can help you solve this. A schematic will help greatly.

A small cap across the supply right next to the sensor will most likely help. This is good practice anyway with many types of IC for example.

This may also be an indication of a poor quality SMPS or one that is going to be fully defective soon....Look at the voltage on a 'scope and see what electrical noise levels are apparent....

A quality SMPS should never make such problems when properly installed and if the main circuit follows normal basic design rules (like the cap I mentioned for example!) and noisy components are isolated from sensitive ones.....

When you say you plan to work on one what do you mean.

Have have you thought about converting a standard PC PSU: loads of current at 12V; lots of instructions on how to do it on the internet.

In general switched mode are much more economical now unless you are desperate to eradicate all switching noise.

I never trusted PC power supplies as bench-top power supplies, insufficient electrical isolation for my tastes and I never knew if the current-limiting circuitry would respond properly when I accidentally shorted the outputs.

Quite right about overloads. I used to repair Ground Power Units that had an SMPS to charge the batteries. If the batteries were dead flat, and you switched on the charger, you could guarantee to blow the SMPS FETs (which is why they came to me to be fixed). I overcame this by changing the FB resistor, so the SMPS only delivered half what it was designed to do. Problem fixed, but slower charging.

That is a design problem of those units, not a general problem for SMPSs....

SMPS' score on top for the three categories you mentioned. I would also add application to the list. Do you need high stability for voltage, current or both? Are you worried about safety (aren't we all)? Are you worried about RFI on your bench? EMI? Would you be willing to add a stabilization circuit post-power supply (this makes a the power supply a low level replacement).

Check it out, and buy. Off the shelf, engineered well. An lets you sleep well.

Anything else?

SMPSs are the only way to go, properly made you can achieve well over 86% efficiency. Also far more exact.....

I actually occasionally build my own power supplies and use a chip that is technically very similar to a SMPS, but uses a normal transformer for full and complete isolation.

The chip can be set up with two resistors to supply anything between 4 and 40 volts.....I built my fist one about 3 years ago, it has never gone wrong....The chip is the LT1074. A data sheet can be downloaded from here:-

http://www.datasheetcatalog.org/datasheet/lineartechnology/1074fd.pdf

Using a good toroidal transformer and good components, you can exceed 85% efficiency. If you need a circuit diagram and instructions for one able to supply up to 10 amps (2 chips), send me your private email address.

Technically, because of the transformer its called a "step-down switching regulator". A single chip is good for up to 5 amps and fully short circuit and over heat proof.....

Thanks Andy. I am not sure if I will get the chip in my area, but I will have a look.

It is available world wide, even from ebay occasionally via Hong Kong.

20 years ago I'd have said go Linear.
Today I say SMPS as the reliability has improved vastly.
Although I would guess that if you really wanted ultimate reliability, you would struggle to beat a linear as the components are less stressed.
Del

The Engineer is less stressed in designing a linear supply versus a SMPS. But by the higher efficiency of a well designed SMPS the parts themselves will be much less stressed in a SMPS.

Now an added complication to the complicated SMPS is failure modes of an SMPS. Since the energy is stored in many more and different parts in a SMPS as it gets converted to a different combination of voltage and amperage than the power distribution provides, when a component fails it is much more difficult for the designer to safely mitigate where and how the energy should dissipate in a failed unit.

Good point about repairability.
I'd still argue that switching (relatively) high currents and voltages at (relatively) high speed is more stressy on the C's than in a linear supply.
But as I've got a cold and am feelig sorry for myself I won't try to make that point.
Del _{(poorly poorly puss cat)}

Thanks Redfred, Andy for pointing out the mix up.

Since an LPS would be easier to create (right?), I would go with that.

I am making a 12 V dc supply using a step down transformer, bridge rectifier, LC filter and series transistorized voltage regulator (using the zener in reverse bias). Are there any specific values of the components, or points I need to keep in mind to maximize the efficiency?

Choose a transformer voltage so that the voltage drop across the regulating circuit is as small as possible. If you spend the money, you could have as low as 1.5V max, but 3V is more likely. However, if the unregulated DC tends to sag noticeably as the load increases, this voltage drop may need to be higher.

Filter the DC with a pi circuit, capacitor input with choke and capacitor to follow. Place a pot between input cap and ground. This will need to be a few watts rating, no more than about 100 to 200 Ohms. By adjusting the pot, you can get a variation of unregulated DC between approx 0.9U and 1.25U (theoretically 1.4U), where U is RMS Volts output from rectifier.

Allow 1000uF per amp of capacitor for good filtering. If you use a lot of capacity, up the current rating of the diodes. You can destroy a diode with the short-time high-amp surges that charge the input cap. I destroyed 25A diodes on switch-on because of this. I put in 100A diodes and had no further problems. Of course the diodes may have been suspect, but ... Some PSU makers place a small-value choke before the input cap to prevent the problem (slows the charge surge). Your choice.

When you still get hum from a DC PSU, and extra capacity won't get rid of it, a choke will fix it.

Put in a foldback circuit. I once modified a PSU. When I switched on, the meters showed zero Volts and maximum Amps (25A). It turned out I had shorted the output. The foldback circuit limited the current and saved the fuse, rectifiers, transistors, and my bacon.

Put rectifiers and voltage dropping components on adequate heatsinks. Despite what the data book says, the big power transistor will only dissipate about 30 Watts max comfortably, on a good heatsink, if you want long life. Limit yourself to that.

Have fun! Blow a few components and learn.

Nice post!!

For the best efficiency with linear designs, you have to design it so that it ONLY runs at full load all the time.......so you need a load that does not vary much.

Though not using Linear supplies for many years as I simply do not believe in them in this day and age, so I am not the best person to ask.

Linear supplies run very hot when NOT running at full load, from better sources I am told that 30 - 40 % efficiency is the best you can hope for.....whereas good SMPS designs run relatively cold/cool under all conditions of load and can achieve over 90% in some instances.

I think its probably safe to say at least 80% + with almost any modern SMPS design.....usually the modern ones are comprised of a few chips for the intelligence plus maybe a few MOSFETs or similar.

In the chip design I use, everything except for a few components is integrated in the chip(s), even the transistors......but a single chip is only good for upto 5 amps....for between 1.5 to 40 volts.