Single transistor flyback driver transistor selection

[John123]

You know that single transistor flyback driver which has been floating around on the internet since forever, does anyone know what determines a good transistor for it?

I thought it was just a crap circuit until recently when I tried an MJ15003 and wouldn't you know it works great with little to no heating and most of the input power ending up on the output.

But when I've tried 2n3055, MJE13007 and MJE13009 performance was poor, arcs were smaller and the transistor heated up a lot even with modest input powers (12v 2 amps for example), I can't seem to determine from the datasheets what makes the MJ15003 so special. Only thing I can see is:

The MJ15003 and MJ15004 are power transistors designed for high
power audio, disk head positioners and other linear applications.

Here is the circuit except I'm using 12v with the upper transistor being 220 ohms and extra protection on the base via a Zener diode in anti-series with a UF4007, although that doesn't seem to make any difference to performance.

The collector voltage is never allowed to exceed the c-e breakdown rating so I know its not breaking down and heating that way.

tl;dr What makes a good transistor for this circuit?  2n3055, MJE13007 and MJE13009 were all poor giving small arcs and got hot fast whilst MJ15003 hardly heated for the same power levels and the overall input>output efficiency seemed pretty good. I've read online the KD60x or KD50x transistor series work well for this driver too.

[dreamth]

If you aren't looking for more than 5 watts of power you may wanna look for tiny transistors like ztx1055, ztz 848, and usually the zetex have very , very low Vcesat transistors.ztx1055 is working down to  20...25mv sat voltage.Yet the 18v supply may be better to be lower...like 9..12v with ztx 1051a or ztx849.

[John123]

Thanks but it does need to be in the range of 20-40 watts. I've tried random junk-box transistors too for lower input voltages and currents but the same issues crop up just scaled down, even transistors which seem perfect on paper being advertised for SMPS use perform poorly (heat and small arcs). The MJ15003 is just magic (I'm using 12v btw so the upper resistor is 220 ohm).

I'm starting to think the internal structure of the transistor might be critical, or maybe the turn off and storage delay parameters.

Edit: I've also read online that the original 1960's 2n3055 models may of performed better with this circuit.

[SteveN87]

Maybe the magic of the MJ15003 is its decent current gain - perhaps the base drive is insufficient with the 2n3055 etc.

[John123]

I originally thought that but even transistors with really high gain don't perform very good, looking at the collector switching waveforms I get this difference.



The 2n3055 is a lot more delayed to turn off and "sticks" on whilst the MJ15003 looks fine.

[Twospoons]

I agree - current gain seems the most notable difference. With insufficient base drive the transistor wont saturate, and so would get hotter.  Try lowering the resistance of the divider, and/or adding a capacitor across the lower resistor to reduce the AC impedance.

Also - did you try adjusting the bias voltage from the divider? That sloppy turn off could also be from having the base bias too high.  You may also be getting a sort of thermal runaway with the fixed bias - Vbe drops as the device gets hotter - which may be exacerbating the issue you are seeing.

I'd be interested to see the collector current waveform too.

[SteveN87]

I'm starting to think the internal structure of the transistor might be critical, or maybe the turn off and storage delay parameters.

Yes - because you have the resonant rise capacitor, you'll also have a resonant fall on the base - i.e. slower than the typlical circuit without the capacitor; so a device with good turn-off characteristics should be better. More turns on the feedback winding may help.

[Twospoons]

Just for fun I ran some sims using a 2n3055 model. Bypassing the bias divider with 1uF made a huge difference to the power dissipation (1/4 !)  Obviously my sim is not going to be exactly like your circuit as I just picked a random transformer core, set up for Isat of ~0.5A.
 So I suspect a lack of drive current in the base is your problem, which would be why the MJ15000 works better (higher Hfe)

The sim was also showing the oscillator starting with no bias at all - feedback winding connected directly to ground. I guess the initial current through the primary to charge the collector capacitor is enough to get things going (in the sim at least).

[John123]

Ok finally back.

I'm starting to think the internal structure of the transistor might be critical, or maybe the turn off and storage delay parameters.

Yes - because you have the resonant rise capacitor, you'll also have a resonant fall on the base - i.e. slower than the typlical circuit without the capacitor; so a device with good turn-off characteristics should be better. More turns on the feedback winding may help.

During my tests the same slow turn off heating mechanism seems to take place regardless of the resonant capacitor size for some reason. With more feedback turns I get more current draw but also more transistor heating, the MJ15003 hardly heats at all with a small heatsink and pushes most of its input power to the output.

Just for fun I ran some sims using a 2n3055 model. Bypassing the bias divider with 1uF made a huge difference to the power dissipation (1/4 !)  Obviously my sim is not going to be exactly like your circuit as I just picked a random transformer core, set up for Isat of ~0.5A.
 So I suspect a lack of drive current in the base is your problem, which would be why the MJ15000 works better (higher Hfe)

The sim was also showing the oscillator starting with no bias at all - feedback winding connected directly to ground. I guess the initial current through the primary to charge the collector capacitor is enough to get things going (in the sim at least).

Thanks, I tried a 470nf capacitor across the lower 22 ohm resistor IRL and still the same, I even fiddled with the resistor value and had it down to 13 ohm at one point but it didn't really help except made the arcs a tad longer and the circuit consume more current. With a 33uF non-polarized electrolytic capacitor across it the current draw shot up to over 5 amps with only a moderate arc power increase and the capacitor got hot fast.

I also tried simulations too and whilst the 1uf across the lower resistor in the simulation worked it doesn't appear to help IRL.

Looking at the datasheets the 2n3055 appears to have a higher gain than the MJ15003 too so even more oddities there.

Unmodified blue=collector yellow=base


Same but with a 470nF capacitor across the 22 ohm resistor.


Would anyone be willing to lash one up IRL and run it with say 12v (220 ohm top resistor) and see what results they get and how much the transistor heats, power draw and arc size? Makes me think I'm going mad when people on youtube run these and get good results from the get go. Transformer is just one from a CRT TV, MJE13007 and 9 exhibit the same delayed turn off behavior as 2n3055.

I've also got a zener diode+uf4007 between base and emitter junction to protect it, but removing it doesn't change the slow turn off.

I know much better circuits to drive hv flyback transformers exist, but its intriguing non the less.

[Twospoons]

Looking at the datasheets the 2n3055 appears to have a higher gain than the MJ15003 too so even more oddities there.

Really? The datasheets I saw showed the opposite. Do you know who the manufacturers are for your devices?
Have you tried adding or subtracting one turn from the feedback winding? I still think you are running out of base drive, and the transistor is desaturating.

[John123]

Looking at the datasheets the 2n3055 appears to have a higher gain than the MJ15003 too so even more oddities there.

Really? The datasheets I saw showed the opposite. Do you know who the manufacturers are for your devices?
Have you tried adding or subtracting one turn from the feedback winding? I still think you are running out of base drive, and the transistor is desaturating.

I thought that's how this circuit operates, the cycle ending when the transistor comes out of saturation creating the primary pulse. Yep tried different combinations of primary and feedback turns, nothing comes close to the MJ15003.

Edit: Was literally going to get a collector current waveform by soldering a 0.1ohm resistor in there but my irons heating element just blew as I was heating it up to do so 

[SteveN87]

Your sim waveforms show that the capacitor has given you zero-voltage turn on, but not zero-voltage turn off. At turn off, there is high collector current - you don't want high collector voltage as well. Somehow the MJ15003 is doing a better job at turn off.

Was literally going to get a collector current waveform by soldering a 0.1ohm resistor in there but my irons heating element just blew as I was heating it up to do so

They pick their moments, don't they!