Implementation of the Switch Mode Theory

[hkBattousai]

Block diagram from Wikipedia:
**broken link removed**

This is my implementation of "Input rectifier and filter" stage:
**broken link removed**

From my understanding, the rest of the circuit must be something like this:
**broken link removed**

I'm stuck at the point "Inverter/Chopper". How do I create clock pulses from such a very high voltage level? Is there any IC for this purpose, or do we use a simple oscillator circuit using high voltage circuit elements?

Second question; can I design a simple ferrite core transformer on my own using the formula:
N1/N2 = V1/V2
N1: Number of turns on primary side
N2: Number of turns on secondary side
V1: Effective value of primary voltage
V2: Effective value of secondary voltage
Is this formula applicable to high frequency ferrite core transformers which has only a few turns on its primary and secondary coils?

Any help will be appreciated.

 

[MrAl]

Hi,


Most designs these days use MOSFETs as the main switching elements, you could look into them.

The design of a switching transformer is a little different than a sine transformer, the Faraday equation is a little different for one thing. The voltage ratios are about the same however. You also have to check for max power dissipation as the higher frequencies becomes more of a problem.
Another problem which could come up is DC saturation, which means you'll have to design some leakage inductance into the output transformer.

Im not sure if you want to be bothered by all this, so maybe you can use an input transformer that you can purchase somewhere. Yes, it will be bigger.
There are some good articles on the web about designing switch mode transformers, i think on the TI site.

 

[hkBattousai]

Thanks for the information, I will search more data about switch mode transformers.

What about this matter?

I'm stuck at the point "Inverter/Chopper". How do I create clock pulses from such a very high voltage level? Is there any IC for this purpose, or do we use a simple oscillator circuit using high voltage circuit elements?

Have any idea?

Thanks in advance.

 

[BrownOut]

I'm stuck at the point "Inverter/Chopper". How do I create clock pulses from such a very high voltage level?

You don't. The IC's that create the pulses consume very low power, and are operated from simple low current, low voltage power conversion networks, usually consisting of resistors, capacitors and zener diodes. Goolge "transformerless power supplies" to get an idea how simple these networks are.

 

[hkBattousai]

You don't. The IC's that create the pulses consume very low power, and are operated from simple low current, low voltage power conversion networks, usually consisting of resistors, capacitors and zener diodes.

Can you please name several of those ICs, they are what I was looking for.

Goolge "transformerless power supplies" to get an idea how simple these networks are.

I googled it, but the results are the circuits which are reducing voltage level by using capacitors, and rectifying by zeners. That's not what I'm trying to accomplish. I want to create clock pulse at 310V DC voltage level.

 

[BrownOut]

I googled it, but the results are the circuits which are reducing voltage level by using capacitors, and rectifying by zeners. That's not what I'm trying to accomplish. I want to create clock pulse at 310V DC voltage level.

I repeat for emphasis:

You don't.

 

[MrAl]

Thanks for the information, I will search more data about switch mode transformers.

What about this matter?Have any idea?

Thanks in advance.

Hi again,


Going from memory, check out the IR2184. That may do it.

In any case, the way you can make a buck circuit without a high voltage driver is to use a N chan MOSFET and drive it with say 10v gate pulse, which can be generated quite easily. The N MOSFET of course has to be rated for above the max operating voltage of 300 or more volts, so maybe a 400 or 500v device. The circuit is worked from ground rather than from the high voltage supply, but works the same really.
To work it from the positive rail, you would do almost the same thing except use another transistor to drive the gate of maybe a P MOSFET, or create an off line low voltage dc power supply (say 15v) and use that to drive the N MOSFET, with drain connected to the high voltage positive rail.
Dont forget you have to deal with spikes across the MOSFET or bipolar.

If you want a quick drawing i can post one that shows the basic layout.

 

[hkBattousai]

**broken link removed**

Output of 555 is a very clean square wave. But at the output of 6N60, the square wave attenuates to 2V, its shape is corrupted, has some spikes and also has an about 5V DC value.

But surprisingly, the ferrite core, even in its this imperfect form, works like a charm. Input wave form is directly transferred to output, it even transfers the spikes!

About the corruption of square wave, is there any alternative way to drive ferrite-core transformer with the MOSFET? What am I doing wrong?

 

[BrownOut]

There isn't enough impeadance on your transformer's primary to effect a decent voltage change on the transistors's drain connection. Transformer drivers aren't typically constructed that way. You need to decide what "mode" you want to drive the transformer with (flyback, forward) and then construct the driver properly, when a transformer constructed to work with the driver, frequency, current, etc. Just winding a few turns on a core and running it isn't going to do much for you.

 

[MrAl]

Hi,


Judging by the size of the core you could possibly need more turns on the primary. Try another 5 turns and see if the performance improves.
Do you have a way to post scope pictures in this thread?

 

[hkBattousai]

There isn't enough impeadance on your transformer's primary to effect a decent voltage change on the transistors's drain connection. Transformer drivers aren't typically constructed that way. You need to decide what "mode" you want to drive the transformer with (flyback, forward) and then construct the driver properly, when a transformer constructed to work with the driver, frequency, current, etc. Just winding a few turns on a core and running it isn't going to do much for you.

Ok, I will search more about flyback and forward driving techniques.

Judging by the size of the core you could possibly need more turns on the primary. Try another 5 turns and see if the performance improves.
Do you have a way to post scope pictures in this thread?

I will try winding more turns tomorrow.


These are the scope pictures of the current circuit:

Output of 555
**broken link removed**
1 V/square, 20kHz

Gate of 6N60
**broken link removed**
1 V/square, 20kHz

Drain of 6N60
**broken link removed**
2 V/square, 20kHz

Secondary side of transformer
**broken link removed**
1 V/square, 20kHz

 

[hkBattousai]

Same circuit, but now the core has 25 turns on each side:

**broken link removed**


Output of 555
**broken link removed**
20kHz, 1V/sqr

Gate of 6N60
**broken link removed**
1V/sqr

Drain of 6N60 (Primary windings)
**broken link removed**
1V/sqr

Secondary windings
**broken link removed**
1V/sqr (Notice that the wave form is same as at primary side, but it is inverted, I should have connected oscillator probes reversed)

Secondary windings when Vcc = 12V (I used 5V all the time)
**broken link removed**
2V/sqr (The step becomes wider as I increas Vcc voltage)

 

[hkBattousai]

All flyback driver circuits I found require that the primary windings has to be center tapped.

The most fitting one appears to be this one:
**broken link removed**

The circuit in the image looks quite similar to mine. The only difference is the BJTs that drive the MOSFET. Do we really need those BJTs? MOSFET drains no current from 555, isn't it?


EDIT: There is also a direct driving version:
**broken link removed**
Source: https://sites.google.com/site/uzzors2k/flybacktransformer

 

[audioguru]

I think you have a 10k resistor feeding the gate of the Mosfet. It should be only 68 ohms so that the high capacitance of the gate can be quickly charged and discharged.

 

[hkBattousai]

I think you have a 10k resistor feeding the gate of the Mosfet. It should be only 68 ohms so that the high capacitance of the gate can be quickly charged and discharged.

That thing took my attention, and I was just about to try it...
I will post the result here.

 

[MrAl]

Hi again,

When i first looked at this i didnt realize you were working at such a low frequency like 20kHz. You probably have to boost up to 50kHz at least because 20kHz will probably require too many turns for that small core. To know for sure however, post your core dimensions including ID,OD,HT, which is inside diameter, outside diameter, and height of the core when it is laying flat. Dimensions in mm would be good if you can.
It would be good to know the material the core is made up of too.

Another informative scope pic would be of the power supply itself, right across the two rails on your plugboard. It looks like this design, as is, may be pulling the dc voltage down quite low for a time and that will screw up the timing quite a bit.

The two bipolar transistors are used as a simple drive current boost, so that the MOSFET can switch on and off faster. If you have some small transistors laying around that would be a good idea too.

Another very important scope pic would be of the primary current in the transformer. This is what really tells the story about if we are saturating the core or not. To get this kind of scope pic, you can insert a small resistor like 0.1 ohms in series with the primary and measure the voltage across that resistor and scale it so we can see a decent waveform. Of course if you own a good current probe that would be even better. This test is like the number 1 test to be performed and is very very important, especially since this kind of design is going to cause net DC in the primary which can easily saturate the core all by itself.
Do you by any chance have a way to gap the core if need be? Probably not, but thought i would ask anyway.

As audioguru says, if you have a 10k resistor in series with the gate this thing may never work right
Must be a lower value. Thanks to audioguru for noticing that.

 

[hkBattousai]

Sorry for the late reply, I needed to buy some equipment to go on my work.

When i first looked at this i didnt realize you were working at such a low frequency like 20kHz. You probably have to boost up to 50kHz at least because 20kHz will probably require too many turns for that small core. To know for sure however, post your core dimensions including ID,OD,HT, which is inside diameter, outside diameter, and height of the core when it is laying flat. Dimensions in mm would be good if you can.
It would be good to know the material the core is made up of too.

I increased the circuit frequency, new frequency is 57kHz.
ID = 12.5mm (ID value may not be accurate)
OD = 27.5mm
HT = 8.8mm
The above is the core I'm currently using in the circuit. I have one another core, its dimensions are:
ID = 7.3mm
OD = 20.1mm
HT = 13.5mm
Forgive my curiosity, what formula are you going to use these dimensions in, and what are you aiming to calculate?

Another informative scope pic would be of the power supply itself, right across the two rails on your plugboard. It looks like this design, as is, may be pulling the dc voltage down quite low for a time and that will screw up the timing quite a bit.

**broken link removed**
2V/sqr, 57kHz

Another very important scope pic would be of the primary current in the transformer. This is what really tells the story about if we are saturating the core or not. To get this kind of scope pic, you can insert a small resistor like 0.1 ohms in series with the primary and measure the voltage across that resistor and scale it so we can see a decent waveform. Of course if you own a good current probe that would be even better. This test is like the number 1 test to be performed and is very very important, especially since this kind of design is going to cause net DC in the primary which can easily saturate the core all by itself.
Do you by any chance have a way to gap the core if need be? Probably not, but thought i would ask anyway.

Primary winding current appears to saturate if I rise supply voltage higher than 7.5V (when Vcc < 7V overall circuit draws less than 0.1A, when Vcc > 8V, it draws 1A. Current drawing monotonically increases by Vcc, but there is a huge increment at Vcc = 7.5V), or run the circuit in 100kHz.
Here are some scope pictures:
(I connected a 0.1Ω resistor in series with the primary winding, oscilloscope is at 10mV/sqr, that makes the screen reading 100mA/sqr)
Vcc = 7.5V
**broken link removed**
100mA/sqr, 57kHz

Vcc = 8.0V
**broken link removed**
100mA/sqr, 57kHz

As audioguru says, if you have a 10k resistor in series with the gate this thing may never work right
Must be a lower value. Thanks to audioguru for noticing that.

This time I connected a 100Ω resister, what if I remove that resistor and just short circuit output of 555 and gate of 6N60? Gate of a MOSFET has super high input impedance, isn't it?

 

[MrAl]

Hi again,


Some of those current waveforms dont look too bad really, but you'll have to zoom in a little more. That high stuff might just be some noise or ringing.

The reason i asked is because if we know more about the core we can apply the Faraday equation. That equation relates frequency, voltage, number of turns, and core area to the induction. In short though, if you are getting this to work at some voltage then you can reduce that equation. The equation at a given frequency with a given core area looks simply like this:
B=E*K1/N
where B is the induction and E is the voltage and K1 is a constant we dont have to know right now and N is the number of turns. Because of that simple equation we can call your circuit, the way it exists now, as:
B1=E*K1/N
and since E=7 we can insert that too:
B1=7*K1/N

Now looking at that little equation and realizing that everything works when E=7, if we want to go up to E=14 that would be multiplying that 7 times 2 (2*7=14 of course) and that means we also multiply B1 by 2:
B1*2=2*7*K1/N

Now looking at that we can see that the induction went up by 2 times, and that's not good because it saturates. More to the point, when we increase the voltage to 7.5 volts that increases B1 by 7.5/7 and we know that doesnt work, so we can assume for now that B has to be less than or equal to B1.

We also notice that N is in the denominator, and if we increase N to match the increase in voltage, we keep the same B, which is B1:

B1=E*x*K1/(N*x)

It's quite plain to see that if we increase the voltage by a multiplier x that if we increase the number of turns by x also we keep the same B1 and everything still works, even though we are now higher in voltage.

This means that to get higher in voltage by a factor M we have to increase the number of turns by the same factor M.

Of course what this means is if you want to get up higher in voltage, you have to increase the number of turns by the same factor that you raise the voltage. Sometimes this is not practical with a given core size so you may have to go to a core with a bigger cross sectional area (area A is in the denominator too).

Also, to help the circuit operate a little better you should connect a Schottky to the output with a capacitor to form a feed forward circuit, assuming that's what you are after (and i think you are). The polarity of the transformer must be observed to get the right output. If you draw a dot at the input winding that connects to +Vcc and the winding sense is such that you draw the secondary polarity dot at a given output terminal, that dotted output terminal is connected to the anode of the diode, and the cathode connects to the cap, and a small load resistor would be nice too, say 100 ohms. This means we will have a somewhat realistic load to work into which is a good thing for this circuit.
We may have to work with a snubber later too.

You might be able to remove that gate resistor, as long as the 555 output transistors survive. The faster the transistor switches, the less losses due to switching. The slower it switches, the less primary ringing.
The gate has relatively high impedance, but any impedance draws current lets face it, and that current goes up with frequency because the input is basically a capacitance. Since a fast rising pulse wave has harmonics that are far above the fundamental, the gate is almost a short circuit to those harmonics and thus for pulses the gate looks like a low impedance during the rise and fall and high impedance otherwise. Thus, when the pulse first rises it has to either:
1. Supply a lot of current to get the gate capacitance to charge up fast to get the transistor to switch fast,
or
2. Supply a lower current and allow the gate to charge slower and thus have the transistor switch slower.

Since the choice here is limited by the driver that already exists, trying to get the most current out of the 555 might work ok but you should check the data sheet about that where it may talk about short circuit current duration and related.

 

[indulis]

You need to decide what "mode" you want to drive the transformer with (flyback, forward) and then construct the driver properly…

Quite right... a flyback and a forward converter are two completely different animals. In one you want to store energy and in the other you don’t. A isolated flyback doesn’t have a transformer, it has a coupled inductor, and the inductance factor/core material you choose for a transformer vs. coupled inductor are completely different. There is no real difference in how the primary "switch" is driven… it always comes down to how much current is needed to drive a "total gate charge" of X to a certain voltage in a certain amount of time. There are driver circuits that control "turn-on" time as well as "turn-off" time or both. A common circuit is a gate resistor which has in parallel a series RC, in parallel with a series diode (anode to gate) and resistor.

It would be good to know the material the core is made up of too.

I would say that knowing this is probably more important than the core dimensions. A core meant to be used as an inductor isn't going to work very well as a transformer.

Looking at your current waveform, I'd say the spikes/ringing isn't real. It's probably pick-up from the loop on your scope probe formed by the tip and ground lead. A "tip jack" placed directly across your current sensing resistor will go a long way in cleaning up the waveforms. As you can see the rise and fall of the current is linear (as it should be)… when you see the rising current waveform starting to "tail-up" when it nears the peak, that is a sure sign that the core is starting to saturate, although there are SMPS that are designed to run their core into saturation.

You'll also want a snubber either from the drain to gnd or the input rail (more efficient) to kill some of the spikes/ringing irregardless of which topology you end up deciding on.

 

[MrAl]

Hi,


I assumed he was going with forward converter because that's why he seemed to want. We'll have to wait for his reply i guess to find out for sure.