Oscope - Roman Black's Two Transistor Switcher Design

[Mr RB]

Here's the resistor divider thing I was talking about, assuming Vin is fairly constant this gives an extra voltage on Q2 base because of the R4:R3 voltage divider. Now R2 only needs to make about 0.15 to 0.2v for Q2 to turn on.

I changed the pos feedback R to a C, using two 330pF caps for simplicity, the cap values and ratio have not been tweaked but should oscillate well.

This is only 1 extra component compared to the previous circuit, but should give a lot less R2 energy loss.

It probably needs a resistor in series with Q2 base for best "RC delay" type action to slow down freq and increase efficiency, and a resistor in series with D2 zener will increase eff there too as prev discussed (to reduce current peak through zener+Q2 base clamping).

 

[()blivion]

Sorry all, I'm still alive. Have a laptop to work on is all. Something about "working for an income" or whatever.

I'm a little concerned by that 0.75 sec delay before startup!

That's what happens when you start it with a load very near the max load. If you drop the load from 610 Ohms to 590 Ohms the thing won't turn on at all. With loads like 1K and above it snaps right on as you can see here...

Blue = R1 current. Green = Vout
View attachment 66472

I believe this slow start happens because the load is stealing current from the inductor, which causes Q2b voltage to stay below the required 0.6v to turn on.

This is the best part so far though...
View attachment 66473

Edit: WOW! with your new configuration, and those much much better performing transistors. I get less that 100uV ripple in the sim!!

 

[Mr RB]

That's cool you are getting the low Vout ripple, is that with the series resistor on the output (RC output filter as per your last circuit) or with the exact circuit I provided above?

That circuit is far from proven, the real "show me the money" will be in the efficiency stats. Or at least some charts of Q1 Vc, Ve and Q2 Vb which should show if Q1 is switching fast and clean, and saturating well, and showing the voltage on R2 is only needing about 0.2v now to initiate the turnoff.

Also the big problem with that last circuit is it relies on Vin for the biasing (not just on R2 as before) so Vin regulation is likely to be suspect, although the zener may do well towards fixing that.

 

[()blivion]

That's cool you are getting the low Vout ripple, is that with the series resistor on the output (RC output filter as per your last circuit) or with the exact circuit I provided above?

That reading was including the output low pass filter, yes. It makes it processor grade power to be sure. I have seen as low as 10uV ripple, up 1mV under worst load conditions. I suspect most projects can omit the output filter, but it's nice to know it works well if and when you need it. It drops ripple by one or two orders of magnitude from the raw. If you don't need that much filtering, you could configure it a lot of different ways.

That circuit is far from proven, the real "show me the money" will be in the efficiency stats.

Agreed, and the sim isn't going to account for all real life things either. It will give us a very good idea though. I'm also having problems trying to get it to preform well with the other transistors now. Kinda painted myself in a corner playing with them. As tvtech said (in different words) on page 4. If this circuit relies on very specific parts to function, then it's kind of selling the car for gas money. Low parts count implies simple, which invariably implies commodity parts. So I would prefer to make it work 100% with generic small signal transistors. Figure around 500mA and hFE ~200. But it's hard to want to do that now with how well the ZTX1048A works with this circuit (T.T)

Or at least some charts of Q1 Vc, Ve and Q2 Vb which should show if Q1 is switching fast and clean, and saturating well, and showing the voltage on R2 is only needing about 0.2v now to initiate the turnoff.

Q1 pretty much SLAMS on and off right now. But this is with the better ZTX1048A transistors. Here is the voltages on Q1 collector...

View attachment 66512

And here is Q1 emitter...

View attachment 66513

Here is Q2 Vb...

View attachment 66514

And this is the circuit I'm getting this with...

View attachment 66515

A few notes:
I took out one of the caps and it works almost exactly the same either way. I also am using a different Zener diode model now, one that lets you edit the revers breakdown voltage. It's set to output almost exactly 5v as is. I was playing with the idea of using a TL431 and trim pot for selectable output voltages instead, but that adds 2 more parts plus replacing the zener when just getting a different zener does mostly the same thing, if not a bit less convenient. The TL431 has better regulation and thermal stability too though. In either case, I have no LTspice model for the Tl431 or similar.

Also the big problem with that last circuit is it relies on Vin for the biasing (not just on R2 as before) so Vin regulation is likely to be suspect, although the zener may do well towards fixing that.

Yeah, with the zener connected it doesn't seem to care what the input voltage is. As I have said many times though, I have seen certain configurations where there is a major global circuit characteristics change with Vin at around or above 1/2 target Vout. I'm not seeing that as much with this though, but it could still happen in a real device.

 

[ericgibbs]

In either case, I have no LTspice model for the Tl431 or similar.

hi ob1,
You have now, plus some other goodies..
E

EDIT2:

Hi Obi Wan

Added the 3 asy's

Added the TL431.mod [ change the .txt extension back to .mod] had to do that to upload.

 

[()blivion]

Cool thanks, When I'm not busy I'll play with that.

 

[Mr RB]

There's a number of issues.

When I said about Q1 having fast turnon and turnoff times, they should be vertical. Your first diagram shows Q1 turnoff (0v-5v transistion) is very slow, taking maybe 12% or more of the total cycle time. That's really poor in terms of efficiency.

I think this is partly because freq is too high, looks like about 11uS per cycle or 91kHz, if you can drop this to 20kHz with the same imperfect turnoff time then you have 4.5 times less switching losses. The freq is probably high because of removing C2, which is the main RC delay component that slows the circuit down and improves efficiency.

In your second chart (Q1 Ve) you can see the R2 voltage, which should rise until it reaches about 200mV and causes the turnoff to begin. This is not happening, because turnoff is starting about 4mV I think you are getting the old problem of oscillating based on Q1 saturation prblems instead of the desired oscillation based on R2 current >setpoint.

I would start by removing the zener, and setting a higher value R4 until it oscillates at a point where R2 (Q1 Ve) reaches 0.2v. And also with a large enough value for C2 so that oscillation is 20-30kHz. At that point you will be putting max current to the load resistor so choose R2 value to give the max desired current into the load.

You might also need to replace the resistor that was originally where Cfb is now, that resistor keeps Q2 turned on for the duration of the flyback pulse, which decreases frequency.

Have you worked out a desired max output current yet? Your R2 value of 1 ohm , switched at 200mV will set output current of maybe 100mA, which seems high to me.

 

[()blivion]

There's a number of issues.

Work in progress