Oh dear:
Well the dust had settled and the smoke cleared, and i sat and looked at a scene of carnage where once a tidy little converter had stood. The fets had gone to where fets go after they have been dealt with unfairly..... this bit I had half expected. The totempole transistors had been demolished and the tl494 was still looking alright. The circuitboard had blown a track, but apart from that all was chipper......egad.
Looking across past the PWM to the coil, the ferrites, so surprised by the massive currents of the dying fets passing through my super coil, both decided to try and take control of the middle ground at the same time. My variable gap went from 3/16th of an inch to zero in a fraction of a second. The deceleration was so severe that the ferrites destroyed themselves. this I hadn't expected.
Well, 2 more buk556 60A and some more bc 139 bc140 trannies later, the PWM was happy and functional (soldered up the vaporised track also) not connected to the "super coil"
Time to be not so cavalier...pull out the trusty scope. The square waves were picture perfect, control was perfect.... what happened. After looking at the circuit decided the constant voltage control (samples the output and adjusts the pulse width to compensate) was causing runaway with an effective short of the coil across the output, and so it opens up the pulse width to compensate for the zero output voltage and sends everything into orbit.
Curses magazine editors, writers, deliveryboys etc.
Time for an interesting trick with computer power supplies I designed years ago (the trick that is).
Most AT supplies are useless for anything because they trigger scr shutdown on current overload. As soon as you try and drive a low impedance load, zip.. nothing. (sometimes flipflop or some other shutdown system).
For some reason, all the supplies I had worked on had a similar trait of NOT using the second internal error amp. They always tie it off so it doesn't play any role in the PWM. This is very useful, as pin 16 is always earthed, 15 is connected to 14 (vref) and so is ready for expolitation. simply lift pin 15 out of the board, an rc network to pin 3 (feedback... say 10n 330R) a 47k from 14 to pin 15 and suddenly we have an active second control point. If we lift the secondary center tap of the power tranny out of the board, put 2-3" of wire in series to the board, we create a current resistor out of the piece of wire. We can then run a wire from the tap/wire interface to the pin 15 (via 1k trimpot), and we have a current regulator. Computer supply will now run like a normal low impedance power supply, fully current controlled. Will not trip on overcurrent as over current cannot occur(if it did due to fault, overcurrent in supply will load up deadtime to max and shut off)
So using that knowledge, i ran a 3" piece of wire from the PWM to neg. a wire from neg to (pull pin 15 as above), and suddenly I had control of my output. (so I think)
It is time. I hook it up, and tentatively prod the wire to the pos terminal and await the explosion.....all quiet on the home front...where is the whine? Oscillator freq is only 3k. should be able to hear it.....bloody ferrites forgot to replace them.
This time, replace the ferrites...and place 3/16" pieces of cardboard between the two E's so they cannot hit or move inwards and implode. Luckily have an abundance of these as well.
Try again,.... this time hear a whine ..hard then soft.....? Now what??
The soft start is coming on then the current reg pulls it up...check the scope, hash everywhere, drawing 2amps, output 1 amp. Doesn't exactly glisten with rectitude.
Damn, the freewheel diode is still connected....grab the side cutters...exit freewheel diode.
It is testament to a crude current control that now I can do anything I please and no damage can occur even a dead short.
Try again, this time, I have full control of the current. scope shows a little ringing....no snubber but who cares(later..later), I'm charging the second battery now at 6.5A...now were cooking.
Some preliminary figures:
input 3.0A 36w 5.0A 60w 10A 120w
output 2.5A 12.8v 32w 3.6A 13v 46.8w 6.5A 13.5v 87w
Note :looking back I should have measured the voltage of the donor battery as well. It's voltage probably dropped under increasing load and the efficiency may not be as bad as it appears. I used nominal 12v for calc, but it is a 7ah batt, so at 10A draw, it may have fallen to 11.7 or less so. I should have checked. The recieving battery was 17ah a much heavier low inpedance device. damn and damn again.
So with a thrown together boost, 50T edt39 ferrite and 12v batt to 12v batt this is what I have gotten so far.
2 fets sent to hell, 2 npn pnp trannies in the same direction, two Ecore shattered, and three hours later we have built our first boost converter, and so just as Flux said, a handful of components, a coil and presto a converter.
Now what to do to improve performance?
1. stop the ringing with a snubber network
2. May have to do the math and work out the correct coil turns and gap. It is just possible that "hunch coils" aint what theyre cracked up to be.
3. Some heatsinks on the fets so we can turn it up a bit more (thats the only reason I stopped at 10A input no heatsinks). The efficiency dropped from 91 down to 67 over that range... that seems a bit ordinary I think the ringing may help explain this.
Any ideas welcome before testing tomorrow night.
will the fets meet their maker again?, or can we improve our "rail gun" ferrites.
............oztules