[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Terry's DRSSTC - 6000 BPS >:o)



Original poster: Jimmy Hynes <jphynes@xxxxxxxxx>

Hey again,

On 5/3/05, Tesla list <tesla@xxxxxxxxxx> wrote:
> Original poster: Steve Conner <steve.conner@xxxxxxxxxxx>
>
> At 15:34 02/05/05 -0600, you wrote:
> >Original poster: Jimmy Hynes <jphynes@xxxxxxxxx>
> >
> >Hey guys,
> >
> >What is the point of this? Why don't you just have a longer on time,
> >and a reasonable bps?
>
> That's what I just said :-P

Sorry, I wasn't following too closely. Great minds think alike, right?
Or as your 'mum' would say, fools seldom differ :P

Seriously though, are we going to get a response from Terry about it?
Unless I missed that too...
>
> >If you take it to an extreme, and split the bang in enough pieces,
> >it's basically PWM, which is essentially lower peak power.
>
> I agree. That's how the current limiter on my PLL driver works- if the
> current gets too high it PWM's the drive by missing out cycles. So it's
> just the same idea of splitting the bang into pieces. I thought this would
> work better than just ending the bang at the first overcurrent event. It
> should put more power into a mismatched load like a ground arc.
>
> Here is a scope screenshot of the limiter under test at low power
>
> http://www.scopeboy.com/tesla/drsstc/driver2/111_1168.html
>
> You can see three places where it decided to miss a cycle. It will miss
> more or fewer as needed to keep the current to the level you set on the
> limit pot.

Your scope shot is missing a cycle when it's pretty much reached
equilibrium, and the secondary voltage isn't really climbing anymore.
When it cuts out for a cycle, the heavy streamer load pulls it down
pretty fast. Terry's is sorta the opposite; it is off more often than
off, and climbs much faster than the ringdown due to the low streamer
load.

In Terry's it seems that it might not hurt it too bad, since it climbs
up fast, but I still see no benifit. He's also cheating a bit by not
running it at full power :P

The optimum bps (defined as streamer length/silicon in 'burst' mode)
has got to be around the thermal time constant from the die to the
case. I lost all my thermal models, so I can't throw out a number
here. If it's much faster, then your IGBTs are keeping the heat
anyway, so you might as well do a big bang. If its much slower, then
you could bang the secondary again, and hit the streamers before they
cool. If the streamer's themal time constant is quicker than the
IGBTs, then you're SOL (note: streamer growing could still be 'worth
it' if you're looking for spark length to coil size, but as far as
spark length/Si, it doesn't make sense.)

What is the time constant of a streamer anyway? This could be found by
finding the 'knee' in the bps/spark length curve.

This next paragraph is just me rambling and overcomplicating things,
so you could just skip it, but i'll leave it anyway-
Another thing to note is that it isn't really a pure RC time constant,
as I'm sure there's some non linearity in the dielectric strength/temp
curve. Instead of defining the time constant as the time it it takes
until the heat has 'pretty much' returned to normal, you can define it
as the time it takes for the dielectric strength to 'pretty much'
return to normal. I know the time constant is really 1/e, but its hard
to be exact with streamers, especially since we don't really know what
function of bps for spark length is.

> I got my new H-bridge more or less finished too. It'll soon be time to
> crash'n'burn! I know, I forgot the heatsink compound so it all has to come
> apart again.
>
> http://www.scopeboy.com/tesla/drsstc/build4/
>
> Steve Conner

Hehe, cool. On a slightly off topic note, I'm working on making a
small one too. Nothing new, just for fun. I was gonna use the existing
uC controller, and existing secondary to make it easy. It turns out
that the little IGBTs are a bit more fussy about early switching than
the bricks. The hard switching causes huge spikes that kill them when
out of tune. Primary (or sec) feedback controller is on the way!