Re: ALF: why not DRSSTC?

["Tesla list" <tesla@xxxxxxxxxx>]

Original poster: Jimmy Hynes <jphynes@xxxxxxxxx>

Hey Greg,

Glad to hear your response, we were looking forward to hearing your comments.


On 9/22/05, Tesla list <<mailto:tesla@xxxxxxxxxx>tesla@xxxxxxxxxx> wrote:
Original poster: Greg Leyh <<mailto:lod@xxxxxxxxxxx>lod@xxxxxxxxxxx>

Hi Steve, All,

Not to stir the alphabet-soup further, but the ALF design approach is
clearly and fundamentally different from both the OLTC and DRSSTC
concepts.  This isn't surprising since the ALF design is driven by
unusual constraints that are largely unimportant to standard Tesla
Coil designs.  I'll attempt here to describe some of the key
operating issues and differences in the design topologies as I see them.

RTC
Currently, the ALF uses what I've termed an RTC (Regenerative Tesla
Coil) design approach, which I've outlined before briefly in a couple of posts:
<http://www.pupman.com/listarchives/1998/january/msg00059.html>http://www.pupman.com/listarchives/1998/january/msg00059.html 

http://www.pupman.com/listarchives/2004/October/msg00146.html


The key feature of an RTC is its ability to *quickly* (in less than 2
cycles) return the secondary energy back to the primary after leader
production in the arc channel drops off -- typically about 2-3 cycles
into the envelope.  Most standard coil designs do not have critical
efficiency requirements, and can allow this unused energy to simply
dissipate in the secondary. However in the ALF design, recovering
this unused energy can represent over 1MW in AC power savings.  The
extra work required for an RTC design is clearly worth it in this
case.  The main disadvantage of the RTC is that it requires a more
complicated charging system, since the returned energy will set the
primary capacitor to some arbitrary voltage at the start of each
charging cycle.


What's your trick for doing it "quickly"? The DRSSTC also sucks the 
energy back out of the secondary. If you don't think it's quick 
enough, we could probably implement the same sort of 'trick' you got 
to do it quickly.

"The coil itself would have a coupling between 0.25 and 0.35"

With a coupling coefficient so high, the energy would probably come 
back out just fine with a DRSSTC

OLTC
The OLTC uses an elegant passive resonant charging technique to
"Off-Line" charge the primary capacitor directly from the AC
mains.  The resonant charging swing in the OLTC depends on the fact
that the primary capacitor voltage always starts at zero.  Since the
ALF will routinely start each charge cycle at some non-zero voltage,
it cannot use the off-line charging method that the OLTC enjoys, as
this would produce incomplete and erratic charging
cycles.  Instead,  the ALF will require some form of active DC
command charging, most likely a polyphase buck regulator array,
between the 12kV prime power source and the primary cap banks.

The transformerless, off-line connection of the OLTC requires that
the primary circuit float at the common-mode voltage of the AC
mains.  By contrast, the ALF primary circuit must be referenced
symmetrically around earth ground to minimize the need for corona
control hardware and to ease the application of controls and
diagnostics.  The DC command charging system effectively isolates the
center-grounded ALF primary circuit from the AC mains.


We know that you aren't using the same capacitor charging method, but 
I was ignoring that part, and was asking about the 'spark gap 
replacement' part of it.

The ALF 1:12 prototype is actually more distant from the off-line
OLTC concept than the ALF, in that it uses 4 paralleled power
transformers to isolate the primary drive from the AC mains.  Tsk,
transformers... yes, I know.  This prototype unit will also operate
as an RTC to test out basic concepts and determine overall effectiveness.



I guess the term 'OLTC' doesn't convey the similarities very well, 
but I'd call them the same thing. Just as I call Marco's Thor a SGTC, 
even though it has a nice CCPS too.

DRSSTC
The DRSSTC, as I understand it, employs a full H-bridge of switching
elements to drive the primary of a Tesla Coil in a series resonant
configuration.  Secondary feedback controls the operating frequency
in addition to having a tuned primary, hence the term 'double resonant'.


The 'bleeding edge', as Terry would say, now uses primary feedback 
(safer on IGBTs is the main benifit here, although not the only one).

The quasi-CW nature of the DRSSTC results in lower stored primary
energy and RMS currents, easing the stresses on the primary capacitor
and IGBTs.
IMO there would be several key design challenges in extending the
DRSSTC to >1MW power levels:
A)  Managing the fast commutation dynamics between the IGBTs and
back-diodes, since zero-voltage switching is not possible in this
configuration.


ZVS should be possible in Class-DE operation, but unnecessary, since 
the operating frequency is so low. I'd be more worried about ZCS, 
which is easily obtainable.

Using the common method of primary feedback, the turn off is always 
at 0 current, but the turn on is slightly after current starts 
flowing, due to delay. At 6khz, this should be minimal. The only 
problem here is protecting the IGBTs from the reverse recovery spikes 
as the diodes turn back off.

B)  Effectively recovering unused secondary energy, since the
quasi-CW nature of the DRSSTC requires many cycles to transfer energy
between coils.  Typical DRSSTC envelopes appear to require about 10 cycles.


What is the problem with the amount of cycles? It means less stored 
energy that need to be taken out. Steve's DRSSTC-3 uses about 25 
cyles, but it doesn't break out for the last 3-4 cycles.

There's no rule saying DRSSTCs have to use 10 cycles either. Using 
one half cycle essentially makes it an OLTC, but you could use one or 
two to keep it short if you'd like.

C)  Managing circuit reliability and total parts cost, with the
larger number of IGBTs that a full DRSSTC H-bridge requires.


With less stress on the IGBTs and primary capacitors, you can use 
*less* IGBTs. I think you're gonna be using more IGBTs in your 
topology than the fundamental minimum of 2 for a DRSSTC anyway :P I 
don't see any reason a DRSSTC should be less reliable, or more costly.

Of these, the slower energy transfer characteristic would likely be
the biggest obstacle.  Since the Fo for ALF is unusually low (~6kHz)
there is not sufficient time to operate in a quasi-CW DRSSTC mode, as
10 cycles of 6kHz would use up most of the time available between
pulses at 300PPS.


300PPS!?!? Why so high? Even much smaller coils find better 
efficiency at half that. Bigger coils seem to need less PPS (which 
would make sense with the square/cube law thing going on with the 
streamers), so a huge coil should need rather low repitition rate. 
With your concern with the RTC thing, that was a real suprise to me.


Ideally, we would like to spend at least 80% of
that time charging, to minimize impulse loading on the generators.
To maximize charging time, the RTC design strives to transfer energy
between coils at much higher impulse -- typically in less than two
cycles.


With a DRSSTC, you could use the whole time to charge the capacitors, 
since its just a DC rail.

 This impulsive transfer requires the primary capacitor to
store the full shot energy, and support higher RMS currents than in a
DRSSTC design.  In fact at PPS rates over 200Hz, the DC command
charging circuit will start its normal charging cycle *before* the
energy transfer from the secondary has fully completed.

Regarding IGBT's, 1200 volts is not a magical number.  The 4500V and
6500V IGBTs are just as easy to use, although the 6500V parts have
somewhat longer turn-off tails.  The best performing silicon I've
seen by far is the Mitsubishi 4500V PT.  Higher voltage operation can
be readily attained by stacking IGBT's in series, with little
degradation in dynamic performance.  Here's some performance data for
a homebrew 12kV IGBT stack, consisting of five Mitsubishi 4500V IGBTs
in series:
<http://www-group.slac.stanford.edu/esd/ILCMarx/12kVDatasheet.gif>http://www-group.slac.stanford.edu/esd/ILCMarx/12kVDatasheet.gif
The prototype looks like this:
<http://www-group.slac.stanford.edu/esd/ILCMarx/12kVassemInset.jpg>http://www-group.slac.stanford.edu/esd/ILCMarx/12kVassemInset.jpg 

Typical switching waveforms at 12kV, 180A:
<http://www-group.slac.stanford.edu/esd/ILCMarx/12kV50kHzBurstTest.gif>http://www-group.slac.stanford.edu/esd/ILCMarx/12kV50kHzBurstTest.gif

To summarize, the ALF design does not fall into either the OLTC or
the DRSSTC category, by any stretch of the terminology.  The
parameter space for resonant transformer designs is vast, where any
one design topology occupies only a relatively small area.


-GL


>Original poster: Steve Ward 
<<mailto:steve.ward@xxxxxxxxx>steve.ward@xxxxxxxxx >
>
>Hello all,
>
>This message is particularly aimed at Greg Leyh, but I would like
>comments from others as well.
>
>As far as i know (and i might be wrong) Greg is currently working on
>a scale model of his ALF towers.  This prototype uses the OLTC
>topology to drive the Tesla resonator.  Since silicon appears to be
>the weapon of choice already, I'm curious as to why not DRSSTC
>instead of OLTC?
>It seems (at least on our hobbyist level) that the DRSSTC can
>outperform an OLTC for similar amount of silicon used.  The DRSSTC
>also does not have the difficulties that the OLTC intruduces as far
>as primary coils are concerned (many OLTCs are just 1 or 2 turn
>primaries).  The DRSSTC also does not have to store the entire bang
>energy in the tank cap (another benefit)
>
>One possible issue i could see is this:  1200V devices will only get
>you so far until you are looking at using single turn primaries and
>giant tank capacitors (resembling the OLTC, but this is even more
>problem for OLTCs as they scale up as well).  So you might be forced
>to look at 1700V or 3300V devices.  But I'm aware that these devices
>also have their limitations (they are slower and have greater
>losses, but i think these are not much to overcome).  Ive heard that
>the real problem is from cosmic rays causing the devices to turn on
>or avalanche (what is the exact mechanism?) when you don't want them to.
>But, wouldn't this also be a problem with using higher voltage
>silicon in the OLTC?
>
>So for each problem I see with scaling a DRSSTC to ALF size, it
>seems an OLTC would have the same problems.  As I (and others) see
>it, the DRSSTC is overall a better topology.  So to summarize: why
>OLTC over DRSSTC?  I'm guessing Greg has thought about this more
>than i have, so i would really like to hear his response.
>
>Thanks,
>
>Steve Ward