WPT Joule Thief low power reverse voltage base transistor protection issue?

[eneuro]

Hello,
Played a while with Joule Thief based circuit to provide low power (a few mA needed) output from input sources as low as possible but with possible input voltage increase up to around 5V and simple 12V 0.5W Zener in output stage as power source for simply (low power consumption 5kV insulated) mosfet gate driver.

In this simply simulation I noticed (maybe I fall into a trap since it is simple interactive circuit simulator) that when 100uH 1:1 trafo was used with classic Joule Thief (1k resistor to base, BC337 0.8A, 15turns of ethernet CAT 5 wires) in this setup, while in output stage is simply Zener shunt-regulator, I saw transistor base voltage lower than -5V which probably is above NPN transistor VBE reverse voltage limits.

Did you noticed in your classic Joule Thief designs such levels of reverse voltage on transistor base?
In simulation added 1N4148, but probably BAT85 diode could be better, to protect transitor base, but it will limit minimum input voltage source.

Need to check this in real Joule Thief circuit-for the moment made classic 3 elements with added 27R on its input and 1N4007 diode to lower down 1.24V deplated 1.5V battery by -0.65Vf  to around 0.6V and this dummy thing works even with random filter core 


Trying to figure out what happends when output load will be low and only Zener diode current limited by 100R,
however main concern is this quite high reverse voltage on transistor base I've found with 100uH 1:1 trafo.
How do you think this reverse voltage could be any issue there or I missed something (maybe too high inductor values)? 

[Zero999]

-5V across the emitter-base junction of a transistor is a non-issue. The BC337 is specified to at least -5V and other transistors can withstand higher reverse voltages: the PN2222A can take -6V.
https://www.fairchildsemi.com/datasheets/pn/pn2222a.pdf

[smjcuk]

I will say that I've blown up a few BC337-25 devices making joule theives. That may explain it but I suspect that it's not 100% conclusive. They're not very resilient devices at all. Even get hot in those circuits!

The old 2n2222 TO-18 canned transistors are far more resistant to such things.

[rob77]

-5V across the emitter-base junction of a transistor is a non-issue. The BC337 is specified to at least -5V and other transistors can withstand higher reverse voltages: the PN2222A can take -6V.
https://www.fairchildsemi.com/datasheets/pn/pn2222a.pdf

exactly, and furthermore the PN junction breakdown is not destructive if the current/power is limited (and it is limited in the circuit). a  PN junction after berakdown is not destroyed by the voltage, but heat. once the current/power is limited to a level the junction is not overheated - it's just ok and the PN junction will recover. you can consider the reverse biased BE junction as a zener.

[bktemp]

You can avoid the excessive negative voltage by designing the circuit correctly: First remove the 100 ohm resistor. It is useless and only reduces the efficiency. A flyback converter in DCM delivers a constant output power.
The generated voltage is limited by the zener diode. Without the resistor the voltage at the collector will be lower. Since you have a 1:1 transformer, the voltage generated in the feedback winding is the same as the voltage in the other one and will therefore be also lower.
If you need a higher output voltage, simply adjust the transformer ratio to get a feedback voltage of around 2-5Vpp.

[T3sl4co1l]

This is madness; you have values all over the place, and obviously have no rationale for their choice.

It is quite possible to design one of these circuits, from the ground up.

Tim

[LaserSteve]

You can remove the extra rectifier diode after the zener clamp, too... That is just a needless 0.6V drop. The zener will function as a reverse clamp at about  -0.7V anyways. 

Might be time to switch to  freeware LT Spice, the learning curve is steep, but its worth it.

Steve

[eneuro]

Thank you for some hints 
Yep, I use spice models too but for 3 elements just tried falstad first and hopefully nothing burned so far and basic version of Joule Thief concept with diode, BC337, 1K resistor, 1.24V battery and random core worked just after soldering everything together and I got bright light from 3mm 3V diode, so not so bad

Still unsure if going above absolute VEB BC337 ratings which is 5V reverse voltage is a good idea and it is difficult avoid this if we add shunt-regulator with 12V Zener diode, but it looks like I found a way to avoid this.

I saw @T3sl4co1l your Joule Thief device and @motocoder efficiency optimisation attempts in this thread https://www.eevblog.com/forum/projects/joule-thief-optimization-your-ideas-wanted/ , however efficiency is not critical in my case.

When "simulated" more classic basic circuit with transformer 10uH coils and 3V LED and transistor hfe 100-300 theres is NO this strange problem with transistor base reverse voltage at all, and LED current spikes are below its maximum ratings-everything looks fine:

However it looks like when we put higher inductance lets say 100uH, or higher hfe output LED current goes mad and there were above 100mA current spikes, so... added 330R in series with LED and now output curent has around 20mA spikes which I like, but... we get those strange close to -10VBE spikes 

Of course we can try choose proper inductor, etc, but... something what noticed also when tried add simple 12V Zener shunt-regulator.

Thanks @bktemp removeing 100R is fine and no strange high current spikes througth 12V 0.5V Zener diode, but 100R and another second diode was there for another reason-it was my attempt to integrate this Joule Thief with 1N4148 diode with already existing simple PC817 based module which is used to  triger thyristors using AC mosfets switch in spot welder, but no problem-I can by pass and short this resistor as well as not needed diode while with 100uH 1:1 trafo I've got in this simulation now nice close to 12Vgs 
BUT look at this !!! Close to -10VBE  ?


If we add BAT85 diode:


in series with transistor base it looks much better now and only (?) about -6VBE, but still above BC337 5VEB  datasheet absoulte ratings


Of course time to try the same in spice, while I've changed additional transistor base BAT85 diode to 300mVf which is at ~10mA IF, which I fought might have sense, but I've already this circuit soldered, so adding BAT85 is not a problem and can try it in reality right now, but always theoretical attempts are interesting.

The most interesting part is now.... 
What do you think what causes this VBE be still below -5VBE assuming we do not change at all output stage which now fits my requirements and even under load I have decent output voltage??? 


When we make the same trick with LED and 330R for current limiting we get the same quite nice VBE now 


In the case of LED probably better idea choose better values for transformer coil inductance to avoid any resistors if we are looking for efficiency, but in the case of shunt-regulator with 12V Zener for the moment I have no better idea, than increase slightly input voltage and add additional two schottky diodes; BAT85 on transistor base and at higher current ratings in circuit input, BUT after any caps and battery itself   

For the moment in soldered prototype with 3mm whit ediode will try add BAT85, remove 27R, add series resistance to LED and see what happends with BC337 VBE.

NOTE: Without 100uF inpuit capacitor and 1.24 battery only I've got also those quite big negative VBE without additional diodes in this falstad sim, so it will be interesting try the same in spice 
 

[bktemp]

Thanks @bktemp removeing 100R is fine and no strange high current spikes througth 12V 0.5V Zener diode, but 100R and another second diode was there for another reason-it was my attempt to integrate this Joule Thief with 1N4148 diode with already existing simple PC817 based module which is used to  triger thyristors using AC mosfets switch in spot welder, but no problem-I can by pass and short this resistor as well as not needed diode while with 100uH 1:1 trafo I've got in this simulation now nice close to 12Vgs 
BUT look at this !!! Close to -10VBE  ?

Since you use a 1:1 transformer, the BE voltage is around the same as the output voltage. For a single white LED this is ok (~ -3V VBE), but for 12V you need to adjust the transformer: Try 4:1 (4 for the collector winding, 1 for base winding). This should give around -2.5V at the base).

[eneuro]

For a single white LED this is ok (~ -3V VBE), but for 12V you need to adjust the transformer: Try 4:1 (4 for the collector winding, 1 for base winding). This should give around -2.5V at the base).

I will test it tomorow, but wiki somehow says the same I noticed in simulation-collapsing magnetic field creates voltage rise and mirrored to secondary probably can be much higher than 1.24Vbat if there is little load or 12Vz in my case, so -10VBE is close to it
http://en.wikipedia.org/wiki/Joule_thief
If the load on the circuit is very small the rate of rise and ultimate voltage at the collector is limited only by stray capacitances, and may rise to more than 100 times the supply voltage. For this reason, it is imperative that a load is always connected so that the transistor is not damaged. Note that, because VCE is mirrored back to the secondary, failure of the transistor due to a small load will occur through the reverse VBE limit for the transistor being exceeded (this occurs at a much lower value than VCEmax).

It makes sense for what they wrote above, however in simulation transformer with coil inductance 1uH or 10uH DIDN't worked at all, and output voltage was not increased at all with 12V Zenner shunt-resistor in circuit shown in previous posts, but when changed its inductance to 100uH I've got in falstad simulation close to 12V with 1:1 ratio 

We'll see-so far nothing  , but this toroid transformer might have not too big coils inductance, so it works with LED and first tried circuit above with two diodes added (BAT85 on BC337 base and 1N4007 0.65Vf on input with 27R bypassed), so we had ONLY around 0.6Vin and 3mm white diode output some decent light:


Tested also with two such deplated batteries in series so I had about 2.49Vin and tried bypass both input diode and 27R with 330R resistor in series with LED-nothing burned and got even warm for short time-but still transformer inductance might be different than 100uH coil in simulation, so futher investigation needed 


Unfortunatelly, I forgot... disconnect output load and put there 1N4148 with capacitor to see if output voltage will build up in the lack of any load and try to kill BC337 by VEB, so will try do it tomorow with different transformers including more bulky than this tiny 2cm OD 1cm height and less than 1cm ID-first with those additional protection diodes, than with bypassed, to see if this >5VEB really can kill BC337 or there is some  in simplified transformer models in falstad circuit simulator and it only looks bad in its ploted VBE wave forms.
 
No more time today, for spice simulation, but I've BAT85 spice model so it will be fun compare in other more advanced sim this tricky Joule Thief modified circuit.

[bktemp]

It makes sense for what they wrote above, however in simulation transformer with coil inductance 1uH or 10uH DIDN't worked at all, and output voltage was not increased at all with 12V Zenner shunt-resistor in circuit shown in previous posts, but when changed its inductance to 100uH I've got in falstad simulation close to 12V with 1:1 ratio 

The inductance of the primary winding (between collector and input voltage) must stay the same, only the feedback winding (between the resistor to input voltage and the base) should have less turns and therefore less inductance. 1/4 turns ratio means 1/16 inductance -> 6.25uH.

[eneuro]

The inductance of the primary winding (between collector and input voltage) must stay the same, only the feedback winding (between the resistor to input voltage and the base) should have less turns and therefore less inductance. 1/4 turns ratio means 1/16 inductance -> 6.25uH.

Tested my AC mosfets switch gate driver low power PSU concept and it works as expected and fits into those simulations above and below.


Easy get in real soldered circuit from 2.49Vin (2x deplated 1.24V  AAA batteries, after 1N4007 about 2V) close to 12V WITH 1:1 ratio and I'm not  since saw this in simulation before when decent inductance 1:1 transformer was used

Note, that I've required load at >11V even WITH 100R resistor in series with 1N4148 before 12V Zener diode and output voltage under required 4.7k load didn't drop to much and is within design requirements -using random filter core and 15 turn 1:1 ratio  When 27R input resistor is not bypassed, than this voltage drops to around 5V @ 4.7k, but 27R is not needed, but I use BAT85 in series with BC337 base and it was about 250mV average voltage drop on this diode, so close to what I assumed in this simple falstad simulation.

So, yep it could be easier sometimes add a few more turns to this Joule Thief transformer, depending on toroid core we have, or choose transformer with different ratio, but whatever we do we must be carefull to do not kill BC337 by too high VCE, especially in no load conditions, with additional resistance in series with Zener, etc.

Since this circuit fits my needs, now it is time to make small PCB and prototype more such circuits and then lets try to choose some bulky transformers, by pass protection diodes (especially BAT85 there) and try kill BC337 by >5VBE, however it can be killed by VCE too, if too high inductance transformer will be used in a no load conditions, but probably the first test and easiest-disconnect any load or add potentiometer only after 1N4148 diode and we'll try to lower load to the point where BC337 will fail 
However, with BAT85 on base and 1N4007 on battery input It will be probably difficult, while in simulation only using a huge inductance I had high VCE (45VCEmax for this transistor), so probably we'll need bulky transformer to kill this BC337 in diodes protected Joule Thief circuit

@T3sl4co1l Mad scientists have sometimes advantage-they are creative, but still use  to verify their crazy ideas 

[eneuro]

Now it is time explain WPT in this topic - wireless power transfer sucessfully  tested in this experiment between... Sun and again I've got easy with one BC337 transistor voltage levels needed for my AC mosfets switches

At decent sun crappy garden lamp small PV panel at 1.34V with my Joule Thief explained above, so input voltage was ONLY 0.6V (after 1N4007 input diode voltage drop) with 100k  output load (simulated mosfet gate ground resistor for another project) I've got.... 12.5V  with this thing

This is with 1:1 transformer 15 turns 

Time for more chalenging experiment and ... RF wireless energy transfer in room using huge high frequency coil around it 
Of course in this case 1N4007 diode will be replaced with BAT85 and input... resonant LC cicruit like in old school crystal radios, where germanium diodes were used, but BAT85 has only ~200mV voltage drop at 1mA current, so maybe it will work 

[eneuro]

Now Joule Thief with protection diodes and full wave rectifier for split transformer secondary input on more compact 1inch2 PCB with custom transformer prototype 

This version above easy boosts from ~1.5V to output 12V needed for AC mosfets switch low current gate driver 

I think we're ready for more spectacular room WPT (wireless power transfer) with copper pipes hidden in ground heating system reused, so now designing high frequency powerfull radio transmiter running @ a few MHz