SSTC, Modes and Soft Switching Discussion
Began as "Mode Splitting" Discussion. Start there.
Edited/Updated: October 21, 2004
Subject: SSTC, Modes and soft switching. Date: Fri, 20 Aug 2004 13:01:58 -0600
Original poster: "Bob (R.A.) Jones"
Hi all, As described in my message subject: Mode splitting. By offsetting
the primary and secondary frequencies one of the modes of the split fundamental
mode disappears.
This may be a useful configuration for an SSTC because both the initial
transient and the steady state current may be in phase with the drive voltage if
its at the frequency of the remaining mode. Hence soft switching even at
the start of a burst may be achievable (assuming no break out)
I have not yet done the transfer function analysis or transient analysis.
Which I will try to do unless someone can point out a problem with this approach
or error in my thinking. As always it may well have already been
considered.
I believe it has been suggested that from observation of the performance that a
system with an offset works better. Has anyone quantified the offset and or got
a frequency response of such a system? Comments anyone? Bob.
Subject: Re: SSTC, Modes and soft switching. Date: Sat, 21 Aug 2004 07:57:15 -0600
A very loose observation from practice on my part suggests that best performance is obtained when the primary of a standard sparkgap coil is tuned to what would have been the lower sideband in a system with equal pri-sec tuning with the same coupling constant. A frequency sweep shows the LSB enhanced at the expense of the upper. Malcolm.
Subject: Re: SSTC, Modes and soft switching. Date: Sat, 21 Aug 2004 07:56:34 -0600
Original poster: "Antonio Carlos M. de Queiroz"
Tesla list wrote:
> Original poster: "Bob (R.A.) Jones"
> As described in my message subject: Mode splitting. By offsetting
the
> primary and secondary frequencies one of the modes of the split
fundamental
> mode disappears.
I don't see how this could happen...
> This may be a useful configuration for an SSTC because both the initial
> transient and the steady state current may be in phase with the drive
> voltage if its at the frequency of the remaining mode. Hence soft
> switching even at the start of a burst may be achievable (assuming no
break out)
This is possible, at least in good approximation, even with all the modes
present.
> I have not yet done the transfer function analysis or transient
analysis.
> Which I will try to do unless someone can point out a problem with this
> approach or error in my thinking. As always it may well have
already been considered.
I have worked out a design approach that appears to work, at least in
simulation:
http://www.coe.ufrj.br/~acmq/tesla/sstc.html
I have written a simulator too:
http://www.coe.ufrj.br/~acmq/programs
I don't like much the consideration that there is a load connected to the LC
network all the time, but the waveforms obtained from the designed networks
without load are not very different during the
rising transient.
> I believe it has been suggested that from observation of the performance
> that a system with an offset works better. Has anyone quantified the
offset
> and or got a frequency response of such a system?
What I think that happens is:
- The system always has two different natural oscillation frequencies.
- If the excitation is at one of them, the output voltage can rise to very large
values, limited only by the Q of the natural frequency, but rise to the maximum
is initially linear with the time, and takes about Q cycles to reach the
maximum, what may be too much.
- Fastest rise, to arbitrarily high values in a system with a transformer, can
be obtained by excitation somewhere between the natural frequencies.
- The reason is that this excitation approximates the excitation of a system
with two identical natural frequencies, at this same frequency. This would
result in a rise proportional to the square of the time instead of just to the
time.
- The best excitation frequency is at the geometrical mean of the two natural
frequencies, if it is assumed that a load is always present at the secondary
end. The system can then be designed as a band-pass filter, or an impedance
matching network.
- Without load, the best excitation frequency would be one that puts the three
frequencies in a ratio of three successive integers, as 1:2:3, 3:4:5, etc. The
output voltage would than be identical to the
one that can be obtained with an optimally designed magnifier. (I have
already figured out how to make the design for this case, but am still
organizing my notes.)
Antonio Carlos M. de Queiroz
Subject: RE: SSTC, Modes and soft switching. Date: Mon, 23 Aug 2004 07:41:05 -0600
Original poster: "Steve Conner"
Hi all, Here is a little view from the lab on the mode-splitting and tuning
properties of DRSSTCs. Well my DRSSTC at least.. I'm not sure how it applies to
systems with higher coupling and lower primary surge impedance.
1. It looks like it is possible to get soft switching all the way through a
burst, no matter what the tuning and coupling of your DRSSTC. You use a phase
locked loop or self-resonant driver circuit, sensing the primary current, to
force the phase angle betwen inverter voltage and primary
current to zero. The system then operates at a frequency where this phase
condition holds true.
2. With a driver like this, that forces zero phase shift, you then have to tune
your primary to find the point of maximum primary current. This also seems to be
the point of highest output voltage, and largest and fiercest sparks :)
3. When I checked my coil with a frequency sweep after tuning it like this, I
found the primary was tuned surprisingly low (lower than the optimum tuning when
I used the same coil with spark-gap excitation) and the lower mode was
emphasized compared to the upper one. Steve C.
Date : Sat, 21 Aug 2004 20:07:47 -0600. Subject : Re: SSTC, Modes and soft switching
Original
poster: "Steven Ward"
> >
> > I believe it has been suggested that from observation of the
> > performance that a system with an offset works better. Has anyone
> > quantified the offset and or got a frequency response of such a system?
> >
> > Comments anyone? Bob.
>
>A very loose observation from practice on my part suggests that best
>performance is obtained when the primary of a standard sparkgap coil
>is tuned to what would have been the lower sideband in a system with
>equal pri-sec tuning with the same coupling constant. A frequency
>sweep shows the LSB enhanced at the expense of the upper. Malcolm.
Im not sure i understand all of the theory yet, but all of my SSTCs (dual
resonant type) all tune to the lower frequency of the split. This seems to
produce the best spark lengths. I prefer using feedback from the secondary
to control the whole system and compared with using a VCO or feedback from
the primary, the secondary feedback produces the best output. Also note
that i must tune the primary considerably lower than the operating
frequency. If i remember correctly, in my largest SSTC system the primary
is tuned to around 40khz while the secondary's fundamental frequency is
about 50khz. The whole thing actually operates at 46khz. Steve.
Date : Mon, 23 Aug 2004 07:41:05 -0600. Subject : RE: SSTC, Modes and soft switching
Original
poster: "Steve Conner"
Hi all,
Here is a little view from the lab on the mode-splitting and tuning
properties of DRSSTCs. Well my DRSSTC at least.. I'm not sure how it applies
to systems with higher coupling and lower primary surge impedance.
1. It looks like it is possible to get soft switching all the way through a
burst, no matter what the tuning and coupling of your DRSSTC. You use a
phase locked loop or self-resonant driver circuit, sensing the primary
current, to force the phase angle betwen inverter voltage and primary
current to zero. The system then operates at a frequency where this phase
condition holds true.
2. With a driver like this, that forces zero phase shift, you then have to
tune your primary to find the point of maximum primary current. This also
seems to be the point of highest output voltage, and largest and fiercest
sparks :)
3. When I checked my coil with a frequency sweep after tuning it like this,
I found the primary was tuned surprisingly low (lower than the optimum
tuning when I used the same coil with spark-gap excitation) and the lower
mode was emphasized compared to the upper one.
Steve C.
Date : Mon, 23 Aug 2004 07:41:14 -0600. Subject : RE: SSTC, Modes and soft switching
Original
poster: "Steve Conner"
>Without load, the best excitation frequency would be one that puts
>the three frequencies in a ratio of three successive integers
Woo, this sounds interesting... But wouldn't this cause terribly hard
switching as the primary current would be a bizarre waveform?
On a related question, what happens to the third harmonic of the inverter
output voltage in a 1:2:3 system? Could we use it to give the "3" mode an
extra kick?
Alternatively, could the 3f mode be tuned to flatten the primary current
down to zero near the zero crossings? This would reduce switching losses in
IGBTs at the expense of higher peak currents.
Steve C.
Date : Mon, 23 Aug 2004 19:35:56 -0600. Subject : Re: SSTC, Modes and soft switching
Original
poster: "Antonio Carlos M. de Queiroz"
Tesla list wrote:
> Original poster: "Steve Conner"
> 1. It looks like it is possible to get soft switching all the way through a
> burst, no matter what the tuning and coupling of your DRSSTC. You use a
> phase locked loop or self-resonant driver circuit, sensing the primary
> current, to force the phase angle betwen inverter voltage and primary
> current to zero. The system then operates at a frequency where this phase condition holds true.
The pll idea is conceptually elegant, but there is a problem:
How can the pll lock immediately, considering that the bursts are of
just a few cycles, and that the spacing between the zero crossings may
be not uniform? With any conventional pll, the first cycles would be
severely out of phase. Maybe better to use just a comparator sensing
the sign of the input current, that would act instantaneously.
> 2. With a driver like this, that forces zero phase shift, you then have to
> tune your primary to find the point of maximum primary current. This also
> seems to be the point of highest output voltage, and largest and fiercest sparks :)
Makes sense.
> 3. When I checked my coil with a frequency sweep after tuning it like this,
> I found the primary was tuned surprisingly low (lower than the optimum
> tuning when I used the same coil with spark-gap excitation) and the lower
> mode was emphasized compared to the upper one.
My idealized analysis resulted in a driving frequency identical to the
resonance frequency of the secondary + topload alone, and a tuning
relation 1/(L1*C1)=(1-k^2)/(L2*C2). The primary shall really be tuned
to a lower frequency, specially for high coupling. Of course there is
the question of streamer loading too.
Antonio Carlos M. de Queiroz.
Date : Mon, 23 Aug 2004 19:36:22 -0600. Subject : Re: SSTC, Modes and soft switching
Original
poster: "Antonio Carlos M. de Queiroz"
Tesla list wrote:
>
> Original poster: "Steve Conner"
> >Without load, the best excitation frequency would be one that puts
> >the three frequencies in a ratio of three successive integers
>
> Woo, this sounds interesting... But wouldn't this cause terribly hard
> switching as the primary current would be a bizarre waveform?
Really, without load and with an output waveform that grows only to
a limit and then starts to fall, soft switching is not possible. In
the first cycles, the input voltage and current are quite in phase,
but close to the maximum they are at almost 90 degrees. This is more
dramatic in fast modes, as 1:2:3.
> On a related question, what happens to the third harmonic of the inverter
> output voltage in a 1:2:3 system? Could we use it to give the "3" mode an extra kick?
So far, I could find the solution with maximum efficiency with the "2"
being the excitation frequency, assumed a cosinusoid. It's possible then
to have all the energy in the system at the output capacitance after a
single input cycle.
The solution departs from the normalized circuit:
o---C1---L1---+-----+--o
| |
L2 C2
| |
o-------------+-----+--o
C1=5/3; L1=8/30; L2=1/4; C2=1
The transformation into an SSTC structure is as in:
http://www.coe.ufrj.br/~acmq/tesla/sstc.html
A solution with the "1" being the excitation frequency exists too, but
apparently doesn't result in total energy transfer (or I didn't find
how to obtain it).
Square wave input would change the waveforms significantly in such a
low mode.
A detail: The best square wave input would start with a pulse with only
half of the normal width. A square wave version of a cosinusoid.
> Alternatively, could the 3f mode be tuned to flatten the primary current
> down to zero near the zero crossings? This would reduce switching losses in
> IGBTs at the expense of higher peak currents.
I see the idea. I am not sure if it is possible. Something to
investigate.
Antonio Carlos M. de Queiroz.
Date : Mon, 23 Aug 2004 19:37:48 -0600. Subject : RE: SSTC, Modes and soft switching
Original
poster: "Malcolm Watts"
On 23 Aug 2004,
at 7:41, Tesla list
wrote:
> Original poster: "Steve Conner"
>
> Hi all,
>
> Here is a little view from the lab on the mode-splitting and tuning
> properties of DRSSTCs. Well my DRSSTC at least.. I'm not sure how it
> applies to systems with higher coupling and lower primary surge
> impedance.
>
>
> 1. It looks like it is possible to get soft switching all the way
> through a burst, no matter what the tuning and coupling of your
> DRSSTC. You use a phase locked loop or self-resonant driver circuit,
> sensing the primary current, to force the phase angle betwen inverter
> voltage and primary current to zero. The system then operates at a
> frequency where this phase condition holds true.
>
> 2. With a driver like this, that forces zero phase shift, you then have to tune your primary to find
>the point of maximum primary current. This also seems to be the point of highest output voltage,
>and largest and fiercest sparks :)
>
> 3. When I checked my coil with a frequency sweep after tuning it like this, I found the primary was
>tuned surprisingly low (lower than the optimum tuning when I used the same coil with spark-gap
>excitation) and the lower mode was emphasized compared to the upper one.
>
> Steve C.
Referring to point 3., I suspect without proof that the lower-than-
sparkgap optimum tuning may be a result of the lower primary losses.
I would be interested to know whether the final frequency came close
to matching the LSB condition I mentioned earlier.
Malcolm.
Date : Tue, 24 Aug 2004 07:45:55 -0600. Subject : RE: SSTC, Modes and soft switching
Original
poster: "Steve Conner"
>Really, without load and with an output waveform that grows only to
>a limit and then starts to fall, soft switching is not possible.
I guess the corollary of this is, that if you force soft switching, the
output waveform must grow without limit :) I have observed this in
simulations and experiments. And it makes sense in theory too- if the
inverter voltage is held in phase with the primary current, then real power
must be leaving the inverter at all times- the flow can't reverse.
It really does happen, and is one of the things that makes DRSSTCs such
"fun" (the primary current grows without limit too)
>The pll idea is conceptually elegant, but there is a problem:
>How can the pll lock immediately, considering that the bursts are of
>just a few cycles, and that the spacing between the zero crossings may
>be not uniform? With any conventional pll, the first cycles would be severely out of phase.
In the (relatively loose coupled) coils I've played with so far, the rate of
change of frequency is slow enough that the PLL can track it easily. The
rate of change of frequency seems to be a property of the coil system, not
the PLL.
I set the PLL up so its unlocked frequency is equal to the resonant
frequency of the primary alone. So it is practically in lock to start with.
I found experimentally that this setting gave the cleanest switching
overall. As the burst progresses I can see the frequency fall until it
reaches the lower split frequency. This is accompanied by a slight phase
error since the PLL needs an error signal to perform a frequency change.
Because of the loose coupling and small tank capacitance I have been using
bursts of around 50 cycles to get enough bang energy. I'm going to try
tighter coupled coils next and will report on how it goes.
>It's possible then to have all the energy in the system at the output capacitance after a
>single input cycle.
The problem then is how to deliver a large amount of energy in such a short
time- it leads to immense peak currents in the transistors. The DRSSTC's
built so far use 10 to 30 cycles and peak currents of 200 to 1500A.
Steve C.
Date : Tue, 24 Aug 2004 17:17:26 -0600. Subject : Re: SSTC, Modes and soft switching
Original
poster: "Antonio Carlos M. de Queiroz"
Tesla list wrote:
>
> Original poster: "Steve Conner"
> I guess the corollary of this is, that if you force soft switching, the
> output waveform must grow without limit :) I have observed this in
> simulations and experiments. And it makes sense in theory too- if the
> inverter voltage is held in phase with the primary current, then real power
> must be leaving the inverter at all times- the flow can't reverse.
Certainly. Only losses can limit the output voltage (and the input
current) in this case. Equivalent to excite a fixed system at one
of its resonances.
> I set the PLL up so its unlocked frequency is equal to the resonant
> frequency of the primary alone. So it is practically in lock to start with.
> I found experimentally that this setting gave the cleanest switching
> overall. As the burst progresses I can see the frequency fall until it
> reaches the lower split frequency. This is accompanied by a slight phase
> error since the PLL needs an error signal to perform a frequency change.
>
> Because of the loose coupling and small tank capacitance I have been using
> bursts of around 50 cycles to get enough bang energy. I'm going to try
> tighter coupled coils next and will report on how it goes.
Ok. Why not to use a simple comparator sensing the sign of the input
current? Noise, I imagine?
> The problem then is how to deliver a large amount of energy in such a short
> time- it leads to immense peak currents in the transistors. The DRSSTCs
> built so far use 10 to 30 cycles and peak currents of 200 to 1500A.
I don't have yet current comparisons for the several possible modes.
In the capacitor discharge system, the 1:2:3 mode is the one that
requires the smallest maximum input current for a given output energy.
I would not be surprised by something similar in the forced response
system, but didn't verify yet.
Antonio Carlos M. de Queiroz.
Date : Tue, 24 Aug 2004 17:17:54 -0600. Subject : Re: SSTC, Modes and soft switching
Original
poster: "Bob (R.A.) Jones"
Hi all,
I had prepared responses to some issues but most have been already been
answered so I will just sum up then add any questions I still have.
First if the input has a series C the input current must go to zero at some
point so soft switching is always possible.
Initially the softswitching frequency follows the transient current. The
transient current is the transient current for each mode so the drive
frequency will be approximately the geometric mean (as AQ puts it) of the
split frequencies.
As time progresses the transients decay but one decays faster than the other
so the soft-switching frequency shifts to maintain soft-switching gradually
moving to the lower frequency mode which then exponential grows to a very
high value. I assume all this is with out break out.
Unfortunately as the frequency is shifting any input power during the shift
may be wasted as it may not constructively sum at the output. However this
may be the price you have to pay for soft-switching.
Yes there may be configurations that can peak in a few cycles but that
requires more peak current. The way to reduce the peak current is to operate
at a lower frequency and or use more cycles getting to the peak.
There appears to be a consensus growing that the primary must be tuned lower
than the secondary for best performance..
Steve:
Are you certain the current grows in the primary or secondary with out
limit. I would expect it to level off as the input power equals the losses
i.e. only finite Q or do the input power grow too?
How quickly does the frequency fall is that compatible with the Q's? I
assume that was without break out. What happens with break out.
Third Harmonic:
The third harmonic of the drive has the opposite polarity at the peak of the
fundamental so 180deg phase shift is required if they are to add at the
output. However that is the relationship between split modes. Achieving
the coupling could be impractical. Perhaps the 3/4 wave mode of the
secondary could be used but its difficult for me to see how it could be
arranged to be accurately three times the fundamental 1/4 wave mode given
that the 3/4 wave mode would be plus a little bit and 1/4 wave mode
truncated.
Bob.
Date : Wed, 25 Aug 2004 07:47:25 -0600. Subject : RE: SSTC, Modes and soft switching
Original
poster: "Steve Conner"
>Steve: Are you certain the current grows in the primary or secondary with out
>limit. I would expect it to level off as the input power equals the losses
>i.e. only finite Q or do the input power grow too?
It will level off as the input power becomes equal to the losses. However
the losses are very low in a well-designed DRSSTC, so to a first
approximation the current is unlimited. It would certainly grow to a value
far in excess of the safe rating of the primary components, and blow the
apparatus to pieces.
In practice though, you design the coil so that it breaks out before the
primary current reaches an unsafe level. Once it has broken out the streamer
loading spoils the "Q" and stops the current from growing much further.
Streamer loading always seems to have this self-limiting effect* except in
the case of arcs to ground. These can practically short the secondary out,
and the primary then resonates alone with a high Q. This can cause
dangerously high currents during ground strikes. One approach for dealing
with this is to add a little extra resistance to the primary circuit.
*You can prove this by considering impedance matching. As the streamer
grows, its input impedance decreases, so the power delivered to it increases
and it continues to grow. However, once the streamer impedance gets equal to
the output impedance of the coil, any further growth would DECREASE the
power delivered to it. Hence no further growth is possible.
Steve C.
Date : Wed, 25 Aug 2004 21:02:49 -0600. Subject : Re: SSTC, Modes and soft switching
Original
poster: "Antonio Carlos M. de Queiroz"
Tesla list wrote:
>
> Original poster: "Steve Conner"
> *You can prove this by considering impedance matching. As the streamer
> grows, its input impedance decreases, so the power delivered to it increases
> and it continues to grow. However, once the streamer impedance gets equal to
> the output impedance of the coil, any further growth would DECREASE the
> power delivered to it. Hence no further growth is possible.
If the system is designed to behave as a pure transformer at the driving
frequency, and if the reactive elements don't change, the load sees the
impedance of the driver only (in parallel with a resonant circuit), and
there is no current limiting.
This happens with the designs shown in my web site.
(http://www.coe.ufrj.br/~acmq/tesla/sstc.html)
Antonio Carlos M. de Queiroz