SMPS experiments

[dominicbeesley]

Hello all,

I'd now like to make my 32-line, neon timebase, monitor into a self-contained 12V battery/mains adaptor powered set. To this end I'd like to build a small SMPS using salvaged computer parts. I'd like to see if I can squeeze the whole shebang into a computer PSU case including 3/4 valves + CRT and SMPS.

I've had a go at building an SMPS in the past - a battery eliminator that gives a nice stable 4A 2V or 1.4V output for LT. With enough shielding etc to not upset the '30s battery sets I've tried it on.

That circuit was a step-down, for this next one I'd like to make a step-up converter. What I'm not sure about is how to calculate the number of turns for the primary and secondary of the transformer.

I'd like the primary to be push-pull powered from something like a TL494 or similar (I've got a number of older computer PSUs to scrap) and use the core of a computer PSU tranny.

I'm guessing that there is some sort of flux-density per square inch that I should aim to be below for the transformer.

Then for the secondaries just wind something up that will give me about double what I really want and let the chip regulate things?

For secondaries I'll want something like 24.5V for the heaters and then 250V at 10 mA for the B+ and 750V at a few uA (from a tripler) for the EHT. I'd probably use the heater voltage as the feedback voltage and leave the other two unregulated. Something like the attached sketch, but with better filtering.

So for say a 10V input voltage I'm guessing I'll want P turns on each half of the primary and the P*5 for the heater winding and P*50 for the B+ winding. But how to calculate P?

Any help suggestions as always gratefully received.

Cheers

Dom

[julie_m]

There's a good article on transformer winding here: http://www.ludens.cl/Electron/Magnet.html

For cheap and nasty ferrite cores, you will need to reduce the working flux density to about 0.2 Tesla.

[dominicbeesley]

Excellent site, I think I now understand, better at least, how to do this!

Dom

[kalee20]

Hi Dom,

With the circuit as you've drawn it, the chip will not regulate the output because all the diodes will be peak voltage rectifying. All that happens as the pulse width is reduced, is that the diodes will pass more peak current, for a shorter time. Component stress will increase, efficiency will suffer, RFI will increase, but the output voltage will hardly change! Pulse-width modulated push-pull converters need an averaging choke between the rectifiers and the output capacitor, and a 'flywheel' path for choke current (which is automatically provided with full-wave rectification).

PNP transistors are really not best choice for switching devices. NPN switch faster and there tend to be many more types to choose from. Most people these days use power MOSFET's, but bipolar NPN's can be used with a suitable base drive circuit. Push-pull is harder withbecause you need to match storage times of the two devices - an imbalance of 1usec can severely unbalance the transformer leading to saturation.

With any push-pull SMPS, you also ewant to aim for symmetry, which this circuit doesn't do. Can you let us have more details of power requirements? Assuming that it's less than 30W total, I'd honestly try and steer you towards a flyback converter.

[Ed_Dinning]

Hi Dom, transformers for SMPS can be complex to design because of the waveforms and frequencies involved.
There are some good primers on SMPS design on the Nat semi "Simple Switchers" and the Fairchild site. Also look at the Power Integrations range of chip controllers.
See how you go and PM me as I may be able to find some bits for you.
If you are up in the Newcastle area I have some notes on SMPS design that came from a course I arranged for my engineers when I was working for a large Japanese company.

Ed

[dominicbeesley]

Hi Pete,

I put PNPs in as these are what were on the TL494 datasheet, I'll just rob whatever parts are in the PSUs I'm using as donors. Indeed I know one has FETs in it as one of them is running the PSU in the Philips Discoverer set that I recently fixed!

Re the chokes, I'd not realised that they were part of the actual regulation. I'd thought the inductance of the winding did that job, but of course it makes a lot more sense now you've mentioned it, I'm not really sure how to rate the chokes, but if I get chance tomorrow I'll try winding up a test rig, I spent half an hour prepping a core, which involved smashing up the original and then glueing the core back together...

If I do the thing with a full-wave and choke for the LT and use that for feedback can I keep the trippler for the EHT or would I be better adding a seperate winding for the EHT (well another 500V in series with the HT)?

I just wanted to try and keep the high voltages away from the transformer and use a tripler to simplify the transformer construction, though this may be an unrealisable idea!

Ed,

I'll have a read through of the Nat-Semi and Fairchild sites tomorrow as a break from the report I'm writing

I keep meaning to come up to Newcastle at some point, unfortunately my new car died, ironically when I'd driven over to the storage where I'm keeping my old car while I repair it, so I now have two knackered cars. So progress may be slow....

Once again, thanks both for the advice and help, always appreciated and a pleasure to learn new things.

Dom

[ppppenguin]

Originally Posted by ajs_derby

There's a good article on transformer winding here: http://www.ludens.cl/Electron/Magnet.html

That's about the simplest exposition of inductor design that I've seen. He's also a vintage radio fan.

On the few occasions I've played with DC-DC conversion I've used flyback topology. Very simple at low power. Think of it as transferring a fixed amount of energy per cycle so load regulation is inherently poor. I once did a HV supply for a vidicon camera this way. Output was (I think) 900V at a very few mA with a string of zeners to supply the other voltages.

[dominicbeesley]

Hi Lads,

I've done a bit more experimenting and have set up the test rig below, this allows me to try it out with various loads, duty cycles etc before I setup the feedback loop (which looks relatively straightforward once I've got the tranny working). It runs at roughly 100kHz this can be adjusted using the R/C combination.

I set up a spreadsheet using the formulas on the Ludens site, and this gave me 4 turns for a 12V primary. So I wound a centre tapped 8T primary and a centre tapped 24T secondary and gave it a belt. Results were disappointing...

I tried this with both NPN and PNP (as shown) drivers but both gave poor results, drawing lots of power and burning my poor fingerses. (ok up to about 10% duty cycle then quickly drawing a large current that went up to about 3A at 80%)

I tried swapping the primary and secondary (to see if I was too few turns per volt) and this gave more promising results, about 0.5A with no load at full 90% duty cycle.

I'm using the core from a 300W computer PSU so this should be up to the job?

This gives quite a weedy output though that drops by about half when a 25R load is placed across it. Is this normal?

Is there a special way to calculate the inductor or do I just go with bigger is better?

I really don't want to go with anything I can't rob from PSUs, or many more components if I can help it as I'm working to a tight budget and a tight space constraint!

As always, help and advice would be very gratefully received.

Dom

[PJL]

I am not an expert and have only designed low power flyback designs. I think your are looking for around 15W? Too many turns and you will saturate the core (not a 300W one mind you).

Even a 15W inverter needs fast-switching low Vce sat transistors. My guess is the peak current is going to be 5Aish.

[julie_m]

Originally Posted by dominicbeesley

I tried this with both NPN and PNP (as shown) drivers but both gave poor results, drawing lots of power and burning my poor fingerses. (ok up to about 10% duty cycle then quickly drawing a large current that went up to about 3A at 80%)

If increasing the duty cycle just makes it get hot, then you could well be saturating the core.

(Apologies if you already knew this: )
Every molecule of magnetic material is like a tiny compass needle, which will align itself with the magnetic field being applied by the windings. When all the "compass needles" in the core have become aligned, the core is said to be saturated: no more energy can be stored in the magnetic field, and the coil just behaves as a normal piece of wire.

What did you calculate the flux density to be? Remember, cheap Chinese ferrite cores probably are only good for about 0.25 Tesla.

[kalee20]

Hi Dom,

The circuit in principle looks OK apart from one aspect. However, it's really difficult (almost impossible) to develop SMPS's without an oscilloscope - do you have one?

Looking at the circuit, the aspect I'm not happy about is the base drive to the switching transistors. First, your 47R resistors should be connected bas-emitter; connected as shown, you're certain to bias the transistors on all the time. No wonder they get hot!

Second, connect a capacitor (start at 1uF) in parallel with your 150R resistors. Bipolar transistors do show big charge storage effects when driven to saturation, and you need to suck this out of the base to get them to turn off fast. Otherwise, you can easily get several microseconds overlap of conduction, especially at 100kHz!

Are your secondary diodes fast-recovery types? If salvaged from an old computer power supply, then they should be (unless you've used the AC input rectifiers!)

I'd honestly cut your teeth by running at a lower frequency, 25kHz, and when this is OK redesign the transformer and change the control chip RC values to go to 100kHz.

[kalee20]

Hi again Dom,

I've just worked out from the pin labelling that the TL494 has uncommitted output transistors, not totem-pole outputs. Now I can understand the use of PNP transistors, because the open-collector output from the TL494 needs to turn the external switching devices 'On' when the colector is pulled down.

NPN's tend to be faster, but you will need an intermediate inverting stage. I can sketch out a circuit for you if interested. I have designed SMPS's professionally, and although not with the TL494, I'm familiar with several similar push-pull control chips.

What types of PNP switching transistors are you using?

[dominicbeesley]

Thanks for the replies...

Originally Posted by PJL

Even a 15W inverter needs fast-switching low Vce sat transistors. My guess is the peak current is going to be 5Aish.

The current I was getting in was about 3A average, there's no real easy way of measuring the peaks at the transistors with the big 3300uF cap at the front but I'm guessing it's quite high. The transistors I've been using should be fast and low Vce (they're all robbed from doing similar jobs).

Originally Posted by ajs_derby

What did you calculate the flux density to be? Remember, cheap Chinese ferrite cores probably are only good for about 0.25 Tesla.

I was aiming at keeping below 0.1T (which a similar core here: https://www1.elfa.se/data1/wwwroot/w...19bbdf5d02.pdf seems to be good for at 100kHz, mine's about 11mm in diameter).

Originally Posted by kalee20

The circuit in principle looks OK apart from one aspect. However, it's really difficult (almost impossible) to develop SMPS's without an oscilloscope - do you have one?

I have several! (mostly broken or half working) is there any particular waveform you'd like me to post up? The waveforms look pretty ropey, but then again they always do to me when I've poked about in SMPS's!

Originally Posted by kalee20

Looking at the circuit, the aspect I'm not happy about is the base drive to the switching transistors. First, your 47R resistors should be connected bas-emitter; connected as shown, you're certain to bias the transistors on all the time. No wonder they get hot!

What an idiot I am, the actual test circuit is correct, I just forgot to redraw the 47R's to Vcc when I changed from NPNs to PNPs.

Originally Posted by kalee20

Second, connect a capacitor (start at 1uF) in parallel with your 150R resistors. Bipolar transistors do show big charge storage effects when driven to saturation, and you need to suck this out of the base to get them to turn off fast. Otherwise, you can easily get several microseconds overlap of conduction, especially at 100kHz!

I'll try that, if I do that can I use something less than 47R as the pull down?

Originally Posted by kalee20

Are your secondary diodes fast-recovery types? If salvaged from an old computer power supply, then they should be (unless you've used the AC input rectifiers!)

Yes, I stole the rectifier from a PC PSU, should be good enough.

Originally Posted by kalee20

I'd honestly cut your teeth by running at a lower frequency, 25kHz, and when this is OK redesign the transformer and change the control chip RC values to go to 100kHz.

[/quote]

I really just wanted to try and get as few turns on the coil so I can use thicker wire and because I always loose count

Originally Posted by kalee20

I've just worked out from the pin labelling that the TL494 has uncommitted output transistors, not totem-pole outputs. Now I can understand the use of PNP transistors, because the open-collector output from the TL494 needs to turn the external switching devices 'On' when the colector is pulled down.

NPN's tend to be faster, but you will need an intermediate inverting stage. I can sketch out a circuit for you if interested. I have designed SMPS's professionally, and although not with the TL494, I'm familiar with several similar push-pull control chips.

What types of PNP switching transistors are you using?

I can connect the output transistors the other way up but then the chip's drivers are running as emitter followers, I did this and there didn't seem to be much difference to be honest. Will using them in emitter follower mode be a bad or good thing? I'd prefer to use NPNs mainly because I've got lots including FETs.

I tried a few different PNP and NPN transistors all seemed to be much the same - the ones I tried (successfully sort of) with 12t per half were BD712s but they were just picked out of the junk box.

I think I'm doing something wrong with the calculations of the number of turns per volt, on that Ludens site there are the 2*2*11% correction factors, I took out the 11% as this is a square wave (well when its at 100% duty) but left in the other 2*2 is this where I'm going wrong? Am I just a factor of 2 out?

Once again thankyou to everyone for your much appreciated time and help!

Off to the pub now, finally got my car in bits and the big end is definitely gone

Dom

[dominicbeesley]

Just another quick observation, I'm pretty sure the transistors are switching fast enough the turn on is pretty sharp then the voltage at the collector rises during the on time the turn off is all a bit messy though but I'm guessing that that is the back emf of the coil.

I'm also still not sure how to calculate the correct size for the inductor on the central leg of the secondary

Dom

[PJL]

Hi Dom, interesting project so I did a bit of reading...a PP design which does almost what you want can be found in this datasheet with all values on page 9 http://www.datasheetcatalog.org/data...rola/TL494.pdf.

The TL494 is fixed fequency PWM and directs the pulse to each transistor in turn. Your setup doesn't seem to have any feedback so it is running flat out and it's not designed to do that. It's not an AC application as the input is not a nicely balanced sine wave and I think the magnetic field drifts to saturate the core.

I don't understand how it all works but it seems as if the PP design transfers energy during both the conducting and flyback phases. The design has a slow start through the dead time control, and a current limit.

[dominicbeesley]

Thanks PJL,

That's where I'd cribbed my circuit from originally, I'm running with the feedback loop broken so I can control the duty cycle with the two potentiometers. With it set up like this I can check that it's going to work under all load conditions before I hook up the feedback loop. The feedback loop will, as in the datasheet, contain a current limiting element and this would distort any test results.

What I'm trying to work out isn't really the circuit itself, I'm fairly happy I've more or less got the gist of what's going on from reading and re-reading that and other data sheets, the difficult bits (or at least what I'm stuck on) are:

a) working out minimum turns per volt
b) working out inductance of the middle leg in the secondary

Once I've got that working for my particular core then I'll need to make further experiments to produce another secondary for 250Vdc and one for 500Vdc. I just want to keep the number of turns to a minimum if possible so that there's room and so I don't have to do too much counting!

Cheers

Dom

[julie_m]

Originally Posted by dominicbeesley

I think I'm doing something wrong with the calculations of the number of turns per volt, on that Ludens site there are the 2*2*11% correction factors, I took out the 11% as this is a square wave (well when its at 100% duty) but left in the other 2*2 is this where I'm going wrong? Am I just a factor of 2 out?

Quite likely.

One of the twos comes from the fact that the magnetic excursion is not from zero to full-on, but from full-on one way to full-on the other way. The other two comes from the fact that this excursion happens twice in each cycle. (Draw a rough sketch of y = cos(x) for x = 0 to x = 2*π and you'll see what I mean.)

But you have a centre-tapped primary where each half is being driven from zero to full-on, then from full-on to zero, then staying at zero for the rest of the cycle. So although there are still 2 excursions per cycle, they are only half the size they were before. Each half of the winding needs to be the same as what the whole winding would be if you were driving it full-wave.

Think of it this way: You have n turns going from zero to full positive flux and back to zero; then a different n turns going from zero to full negative flux (because the current is now flowing in the opposite direction) and back to zero. But the core doesn't care which turns they are; just how many turns and what voltage is being applied to them.

[PJL]

I am still not convinced you can run it without the feedback and expect it to behave properly. It would have been nice to have an explanation of exactly what is going on, my guess is:-

During the conducting phase, the turns ratio of 1:6 develops an output voltage which is rectified by the diode.
During the off period the transformer behaves like a flyback converter and again the diode transfers the stored energy into the capacitor via the diode.
Now there is current flowing in L1 things start to change as L1 is relatively high impedance and is trying to maintain a constant current flow so it's terminal voltage is going to vary to maintain that current.
L1 also determines the rate of change of current and therefore how much energy is transferred in the time allowed.

During the conduction phase, the primary current will comprise two components, the energy being stored in the transformer determined by the primary inductance and the energy being transferred to L1/capacitor determined by the turns ratio and L1.

Very complicated. If L1 is too small, the output voltage would not be regulated as the turns ratio would determine the voltage and the peak primary current would be horrible. I am not so sure when it is high...Peter

[dominicbeesley]

PJL, its not meant to work correctly (as in regulated) yet, but is meant to test whether the transformer will ever saturate under full load conditions.

I think I understand your reasoning about the inductor, indeed Peter (Kalee20) mentioned this previously. Indeed I've tried it without the inductor and there is little regulation, with the 1.4mH inductor there is quite a bit, I suspect I need a bigger inductor though as it only works up to a point when drawing a large current, will have to hunt round for one big enough!

ajs_derby, thanks for the mental picture, it makes sense a bit more now - so can I keep one of the factors of 2 but not the other.

In fact if I added two more transistors (so each leg pushed and pulled) I could keep just half the turns on the primary as the same half would work like both (see attached sketch missing final filter cap!).

I'll probably not do this for the primary but it if it would work for the secondaries I'll try it. I'm just not sure it will work as it would be trying to send current both ways through the inductor. Could I have the inductor _after_ the bridge perhaps?

Cheers

Dom

[kalee20]

Hi Dom,

A couple of tips:

First, connect a diode between collector and emitter of each of the switching transistors. That will stop any inductive swing from trying to drive the transistors backwards - the diode will clamp the swing. Again, you need fast-recovery diodes here.

You can easily check whether the transformer core is saturating. Just disconnect the secondary completely, and run the primary at full duty cycle. If everything is benign, there is no saturation problem.

For the inductor at the secondary, too little inductance causes a change in regulation characteristics at low load currents. Excessive inductance doesn't matter, although it will make the whole thing slower to recover from load changes (called a poor transient response).

Your new circuit in principle would work for the primary (but not with the bases connected as shown!). However, the secondary MUST have the choke connected AFTER the bridge rectifier. Otherwise you'll get a small output which collapses under load.

The primary and the secondary can be reconfigured quite independently - centre-tap either, or bridge them, as you like!