Re: [TCML] DRSSTC tuning at high power

[Michael Twieg <mdt24@xxxxxxxx>]

On Thu, May 26, 2011 at 12:41 PM, Steve Ward <steve.ward@xxxxxxxxx> wrote:

> Mike,
>
> On Thu, May 26, 2011 at 10:06 AM, Michael Twieg <mdt24@xxxxxxxx> wrote:
>
> > Yeah I've always been on the watch for detuning as power level changes.
> >  The
> > weird thing is I observe that as power and streamer length increases, our
> > measured operating frequency also increases.  Completely
> counterintuitive;
> > it should decrease as streamers grow.
> >
>
> This is peculiar, now that you mention this I will have to see if i get
> similar results on my own systems.  Is your measurement just what the
> oscilloscope displays for a frequency measurement? Or are you looking at
> the
> period between zero crossings?
>
> I cant think of an explanation for it yet, either way.
>
We normally do frequency measurements with a MCU on the gate driver board.
It looks at primary current zero crossings during the driving period and
takes an average period over that.  It's generally as accurate as a
microcontroller should be (essentially perfect).  I also double check it on
the scope sometimes and it always agrees.

>
>
> I used to pay more attention to my "Freau Factor" which was part of John
> Freau's spark length (in inches) vs input power (in watts) relationship:
>
> Spark_length = K*sqrt(wall_power)
>
> I think John claimed that his best performing spark gap coils would have a
> K
> = 1.7 or so, at 120BPS.  I found some of my DRSSTCs to get nearer to K =
> 2.0, but i cant recall if that was with power factor correction or not.
> Well lets see, I seem to recall my coil makes approximately 10' sparks with
> 3500W (DC bus power) at 120BPS, so my K = 2.0.

I'd estimate that at 100BPS we're getting 90" streamers at about 3.3KW in.
So not quite 1.7.

>  Of course K goes down as BPS
> goes up, since i dont gain enough spark length at higher BPS, but input
> power is roughly linear with BPS.

 I've noticed that with our system it's not really linear, but more like a
x^3/2 kind of curve.  I always attributed this to the dual affect of more
bangs per second and more energy per bang (since streamers tend to lengthen
at higher BPS).

>   But i personally would say if you get a K
> > 1.7 (at 120BPS), then you are probably getting near the best performance
> that most tesla coils have achieved.  So until more theories are worked out
> as to how to better optimize things, this seems to be the most simple
> benchmark to use.
>
> As a side note, whenever i play music through DRSSTCs, i apply a pulse
> width
> truncation vs frequency (the curve was determined mostly experimentally but
> follows a sqrt(freq) trend mostly).  That way you can squeak out the high
> notes at reduced pulse width (which does reduce the spark length) and save
> your power controller from going into current limit and dropping the output
> voltage.  Id imagine its more efficient to run shorter pulse widths at full
> bus voltage than it would be to run full pulse widths at reduced bus
> voltage, but that might vary from system to system.

In our midi controller we have the output transition from constant pulse
width to constant duty cycle (around 600Hz or something), like you say.  But
before then we usually hit the power limit on our supplies.  This is fine
since they're meant to limit power softly and gracefully, so this also acts
as a good control at high frequencies.

>  And on top of this, i
> actually boost the pulse width for the lowest notes (anything below 100hz
> or
> so) as they seem to have less benefit from ions/hot air left over from
> previous sparks, and could stand to use some more bang energy.  Simple
> controls, but extremely useful for this application.
>
That sounds like it might be worth trying, but only if I can get the tuning
under control.  As shown in those waveforms I linked, at the time the driver
stops the current is racing upwards.  Any longer and the Ip will probably
get out of control.  I'll definitely keep that in mind though.

Also, we've come to the conclusion that our primary tank capacitors aren't
nearly as capable as they should be (we're already probably pushing them to
twice their ratings), so we're going to eventually make new ones, probably
out of arrays of smaller ones.  Would it be worth trying to shoot for a
higher/lower capacitance?  I've read a lot of conflicting explanations of
what it does to the tuning of the system (besides needing to adjust Lp).  I
assume that making the ratio Lp/Cp lower makes the thing a lower impedance
and gives greater voltage gain and faster rise time.  But would that make it
harder to tune?  Seems like it would, since smaller deviations in Lp will
have a greater affect.

Thanks,
-Mike
_______________________________________________
Tesla mailing list
Tesla@xxxxxxxxxx
http://www.pupman.com/mailman/listinfo/tesla