Testing the TK Tar Baby

[hoptoad]

A proposal for consideration:

I don't know if this will work and I don't have the components on hand to try it out. But I'm thinking that an optoisolator might work to isolate the FG's current path from the system, with minimum extra additions to the main circuit.

Here's what I'm thinking of, and I'd appreciate advice on whether or not it would even work, and also what effect it might have on oscillations.... and of course, inadvertent battery charging by the function generator. Well, at least it seems clear that it would eliminate that latter possibility.

I can't see one good reason why the opto-couplers wouldn't work. The only thing that might be a negative issue is dependent on the signal frequency. The rise time response of the opto couplers may be a little slow and introduce slur if your input signal is in the megahertz range. Other than that, I don't see any great issues.

If you want to retain the possibiltiy of parasitic oscillations arising, then the circuit as shown might still allow that to happen. If however you only want a faithful reproduction of the input signal to your mosfets, I'd be inclined to connect a resistor anywhere in the range value of 150 k to 1 meg between the gates of the Mosfets and ground. This will help to damp floating voltages, and help prevent stray capacitance from interfering with the operation of the Mosfets, and also from interfering with the true input signal.

Cheers

[TinselKoala]

Thanks, hoptoad. I think you are realizing what I want to do. I am trying to remove the FG's current path and power source from the switched power circuit in the TarBaby device. Whatever method I eventually use it must preserve the oscillations in the full 5-mosfet design. As soon as the stores open tomorrow I'll pick up a few optocouplers, plain and triac kind, to see what happens next. The required switch rate is absurdly slow.

The subject matter of the previous video, mosfet switching, seems so basic to you and me that we don't even think about it. However some other people working with similar circuits...you know who I mean, probably.... missed out on so many basics that they _can't_ even think about it. Believe me, years ago I spent weeks trying to convince...err... on..... somebody else that mosfets switched like this.

And now it's deja vu all over again.

[TinselKoala]

So.... as demonstrated, the oscillations in the mosfet switch demo circuit come out to 9 cycles in 9 major divisions of the scope's screen. The timebase is set to 0.2 microseconds PER major division. How do we determine the frequency from this data?

ANSWER: since frequency is measured in Cycles PER second, or Hertz.... and we have counted 9 Cycles PER 1.8 microseconds, what is the frequency?

We perform the operation 9 cycles DIVIDED BY 1.8 microseconds. Notice the units: the answer, whatever it is numerically, will have the "dimensions" or "units" of.... CYCLES PER SECOND. So let us calculate: 9 / 0.0000018 == 5 000 000 Cycles PER second, or 5 MegaHertz, to within rough reading accuracy. Actually it's a bit higher since there were a bit over 9 cycles in the 9 major divisions.

[TinselKoala]

The version of the Tar Baby that I'm working with here uses IRF830a mosfets in all 5 positions. The circuit that the Tar Baby is modeled after, the NERD device, uses the IRFPG50 mosfet, and it seems to be the claim of the NERD RATS that this part is critical to the "effect" they are talking and talking and talking about demonstrating. So I thought it might be interesting to compare some electrical parameters of the two power Hexfet, avalanche-rated mosfets.

The first thing one notices is that the PG50 is in the TO-247 package, an awe-inspiring slab of black plastic with beefy leads. The 830a is in the standard TO-220 case which looks tiny by comparison, and its leads are smaller and have the standard 0.100 inch spacing, so they fit nicely into breadboards and dip sockets and headers.

The most critical parameters for most uses are listed at the top of the first page of the data sheets.

Further parameters of interest are the various capacitances and avalanche energies, transconductances, and so on.

I've attached the data sheets. To my eye, if one does not absolutely need the 1000 volt standoff capacity of the PG50, one might be better off simply using the 830a, since they are somewhat cheaper, smaller, and seem to perform similarly at low power settings. In fact the 830 has lower Rds and better transconductance and will dissipate less power at the mosfet itself at power levels within its range.
In other words, it should run cooler than the NERD mosfets at similar power.

[MileHigh]

Hi TK:

You had a question about the capacitor test with respect to the capacitor size.

When I originally crunched the numbers I think the power dissipation in the other circuit was quoted as being about 30 watts.  I am now suspecting that was quite high.  I played with the number crunching and quickly realized with a single 25,000 uF capacitor it would drain very quickly assuming 30 watts being dissipated in the load resistor.  I also now realize that I wasn't even factoring in the fact that the total power dissipation would include the MOSFETs themselves and the 50-ohm resistor inside the function generator.

The bottom line is that I quadrupled the capacitance to 100,000 uF and crunched the numbers again and realized that they would still drain very quickly.   This is a real concern because if you let the caps drain to zero volts then they will continue past that and start to reverse-voltage.  Using polarized electrolytics this would be very very bad as you well know.

The simple crunching I did was to estimate the power dissipation.  Then calculate the energy loss in the 100,000 uF cap from 12 volts to about 6 volts (I think).  So then I could make an estimate how long that voltage drop would take based on the energy burn rate.   You have to keep in mind that the energy burn rate is not going to slow down because you are really talking about a voltage drop from 60 volts to 54 volts.   I had to base it on a 6-volt drop so the circuit could run for about 10 seconds.  I would have preferred something like a one-volt drop only but then the time got too short.  Again, assuming 30 watts power dissipation, you have to work around that if you only have a 100,000 uF cap.

So I would recommend that you make your biggest cap possible, and scope the voltage across the cap array when you switch on the circuit.  If it drops like a stone then disconnect right away so that you don't reverse-voltage the cap array.

If the total dissipation is more like a few watts as opposed to 30 watts then you should see a gentle voltage decrease across the cap and still see the 'magic' oscillations.  Then you can crunch the numbers and calculate the power dissipation in the circuit.   If you can time say 20 seconds and then do 10 runs like that and average your measurements, you will have a very decent number for the total power dissipation in the circuit.  You know the voltage so you know the average current flow.  That means you can calculate the power dissipation in the 50-ohm resistor inside the fucntion generator, and the power dissipation in in inductive resistor.  Look, you now get a very nice bonus, you can derive the power dissipation in the MOSFET array:  The power dissipated in the MOSFET array is the total power dissipation minus the 50-ohm resistor dissipation minus the inductive resistor dissipation.

Now, if you had a means to control the magic oscillation frequency, perhaps with an opto-isolator arrangement or something else, you could expect to see a trend.  The lower the oscillation frequency the less time the MOSFETs are in the linear region and dissipating power.  So if you lowered the oscillation frequency and ran the capacitor test again, you should be able to derive less power proportionally dissipated in the MOSFET array.

Finally, I tried to keep in simple for Rosie and I didn't want to mention the potential for the capacitors to reverse-voltage right away.  If anybody was going to go forward I was going to mention it.

MileHigh