--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
I love to cook with wine Sometimes I even put it in the food
Changing the R2-3 divider doesn't really help, I'm afraid, as Q2(c) will still swing to 0v after the flyback pulse and re-fire Q1; no R2-3 combination prevents that.
Making sure timing cap C1 discharges below ~1v during the 'on' pulse would prevent double pulses, but that's not easily done, and depends heavily on the exact parameters (timing, desired peak current, etc).
Another possibility is not to discharge C1 fully at all, and instead deliberately design so that one pulse immediately follows the other, choosing R4-5 and C1 in combination to enforce the desired peak-current limiting.
So, the circuit is flexible, and several different modes of operation are possible just by changing a few values.
It isn't possible to optimize this more without knowing the actual specs: desired power output, output voltage needed, the transformer ratio, D.C.R., primary inductance, saturation current, etc.
There are a number of factors, and they all interact.
I'd been assuming 5uA or so at 500v (2.5mW output) would be enough to bias a Geiger tube, but that was a complete guess.
The original challenge was just to get the thing to start, and we've gone well beyond that!!
That 40 ohms was also a complete WAG. For the oscillator to start R1+R4 must drive Q2 so as to produce >=0.6v across the transformer's DCR, yet Q2 ought not saturate either.
I haven't looked at any real transformers--that's your job!
It's really neat what cleverness and simple parts can do.
With a 5uA load Wenzel's circuit draws about 2mA, so it'd run for 1,400 hours or so off a couple AA cells, or nearly a year continuous off a pair of 'D's.
The 150mH and 40 Ohms was actually pretty close. I found some info on real ones and for a regular 400mW voice/date its like 170mW and 65 ohms. I would say you did well for a guess.
Here's the simplified circuit I've been working with. I is practically exactly what Wenzel put forth. I found those values actually produce the greatest output. As you can I'm using the transformer now for the inductor.
formatting link
formatting link
This works well enough, although the efficiency at full load is like
16%. It also seems to function well enough without the voltage multiplier. I suppose I should try and see if I can get the feed back circuit to work too.
I noticed in the circuit here, the timing caps charges and discharges only partially., as you described. I have to look more into that. I have also found a transformer with a primary rated for 4 Watts. It should have a low DCR, 50 ohms was my guess :)
I have noticed the transformer version will work without the voltage doubler, but the voltage output of transformer rises to compensate for increased in output impedance. I figure Wenzel used the voltage doubler just to keep the voltage to a minimum in the transformer windings.
Now the I got this modeling like a champ,
formatting link
Red is primary, purple is secondary, light blue is both output volts and uamps, and green is the feedback signal. Does that ever work well! :)
The voltage across the primary hits 300V. I am using a low DCR transformer model. Without the voltage doubler it goes over 400. Also efficiency increases by about 4% with the voltage doubler.
Using a 1:1 provides plenty of boost for 1kV max generator. I tried modeling a transformer with dual 600ohm secondaries in series. Improvement in efficiency was less than 0.5%.
What made the biggest improvement in efficiency was lowering the DCR of the primary. I tried the dual secondary paralleled for the primary and the single primary as the secondary. The efficiency went up 5%, although so did the voltage across the primary. A transistor with higher CEO rating would have to be used.
After everything I tried, all I had to do was follow his directions verbatim.
have you considered using a tape wired toriode type xformer?
With conventional round conductors, a low DCR will still yield problems. AC electrical currents tend to push them self's to the surface of the conductor there by, making the majority of the material nothing more then a construction body and maybe a heat sink. (Skin Effect)
There are 2 methods I can think of at the moment which helps here.
Use insulated flat copper tape as the conductor. This can be thin and wide. SO, if you had 10 oz of copper in a round conductor verses 10 oz of copper in a flatten thin conductor of the same length. The flatten conductor will give you much better performance.
Pull in multiple leads of insulated conductors joined as one at the termination points. also, it is customary to twist these leads so that eddy currents can cancel each other, much like what you see in laminated plated cores.
You also have Litz wrapped wire with multiple insulated strands of small wire which are braided or weave in such a way that it allows for the complete set of strands to contribute to the job and not allow a eddy current effect to influence it as much.
I have considered that a autotransformer would be more appropriate than a multi-winding transformer here. It would have a better coupling coefficient and lower interwinding capacitance.
Also, since current flow is unidirectional in this application, a gapped E-I core would be an improvement, because it would hold off core saturation better during the current rampup.
1) make Q2, effectively, a current-controlled switch by just plowing a constant, limited base current into it. (e.g. high-value series base resistor.)
Ic(Q2) increases linearly during the 'on' time until Ib(Q2) * Hfe(Q2) < Ic(Q2), and then Q2 pops out of saturation.
The oscillator feedback circuitry then proceeds to cut off drive to Q2.
Meanwhile, the inductor flies back with Q2 still conducting full peak current for quite a long time, which is wasteful.
Or
2), you can choose the R-C constants to enforce a desired 'on' *time*, where Ib(Q2) falls exponentially while Ic(Q2) is rising linearly; that crossover can give pretty well- defined on-time and peak current. Much tighter than the Hfe(Q2) spread of #1. And since Q2's been starved, it gives up much faster once it desaturates, cutting switching loss.
Or lastly,
3) you can use Wenzel's emitter-follower feedback circuit to terminate the 'on' time by cutting off Ib(Q2). That's a notch more efficient than #2.
Many possibilities. Simple circuits aren't simple.
The advantage of the transformer is less stress on the the switching transistor, and generally, for voltage ratios this high the parasitic capacitance and resistance of inductor get unwieldy.
You could always turn that doubler into a tripler or more... depends on what output voltage you need.
I tried modeling a tripler, and it only added about 0.5% efficiency and the voltage across the transformers windings only dropped a little bit. I think the improvement is worth less than the added expense of the components.
On the other hand the doubler produces a good improvement in efficiency 4-5% and winding voltage over just the transformer feeding a RC filter stage.
Using a multiplier is mostly to ease component selection. An 80v transistor has higher gain and is much easier to find than an 800v transistor.
The greatest losses are i^2 * R in the inductor's d.c.r., and the switching and conduction losses in the power switching transistor Q2.
This circuit leaves a lot of time where energy is neither being stored nor delivered, so the power of each pulse, when its time comes, has to be increased accordingly to make up for all that dead time.
Smaller pulses applied more frequently would reduce peak currents, and cut i^2 * R losses proportionally.
Lowering switching losses in Q2 is mostly a matter of making it turn off faster. Ideally, an active turn-off. Finding a transistor with a lower Miller capacitance might not hurt either.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.