Maser 0.7 nsec jumps solved
Don't let the start of the simulation be the power on time. Best to
set up the AC mains volts at zero volts for a half second then go up
to 120 VAC. So you actually simulate the power switch. The time
before the start of the run is not defined
Also you should Google "spice transformer model" and see how others
have done it. You will need to add some extra inductors and series
resistance. As you found the Spice model does not have magnetics in
it. It is simply a pair of coupled inductors.
On Fri, Jun 3, 2016 at 10:40 AM, Attila Kinali attila@kinali.ch wrote:
On Fri, 03 Jun 2016 12:37:26 -0400
"Mike Monett" timenuts@binsamp.e4ward.com wrote:
I found a significant error in the LTspice analysis. I was wondering how
the current could jump instantaneously at zero when the voltage is applied
at the peak. That violates magnetism.
It turns out it doesn't. When LTspice starts an analysis, it first
calculates the operating point. For the Sine voltage source at 90
degrees, it applies the full voltage across the load. In this case,
it was 169.7V across 1 ohm, resulting in 169.7 Amps. That is what
was plotted, and is a significant error.
Actually, spice (the engine behind LTspice) does a DC analysis before
almost all modes of operation. This DC analysis has the intention to
start the circuit from a steady-state point and thus to reduce simulation
time. In order for this to work properly, you have to specify the DC voltage
and currents for all sources correctly. Spice messes this up at times
making the first part of a transient simulation worthless (it has even
worse problems when you do an AC analysis). Additionally LTspice hides
too much of these small complications for the problems to be visible to
the untrained eye and also at times makes it harder to provide the correct
values. Thus, caution is advised.
The general rule of "Never trust a simulation you haven't
forged yourself" applies.
Out of 13 examples I analyzed, I found only one that involves unloaded
transformers.
I found many references that discuss transformer inrush current caused by
core saturation. This is a serious problem as it puts stress on the
components and reduces operating life.
I only had a quick glance at your webpage, but it seems that you used
the standard LTspice transformer model. Unfortunately, this is not a
good model to study this kind of behaviour. For one, the only loss considered
in the model is the winding coupling, it doesn't even directly consider
resistive losses in the windings. In this case, the two most important effects
that you need to include are saturation and core losses, which are both
frequency dependent. The cores of electric machine transformers are very
poor when it comes to their "high" frequency behaviour. Where high frequency
starts somewhere closely above mains frequency. Ie 1kHz is already so far off
that somewhere around 90% of the energy would be dissipated in the core.
The sharp rise in voltage and the leading inrush current have frequency
components that are way higher than mains frequency. Hence the linear model
you used will give inaccurate results, to put it mildly.
Unfortunately, building an accurate transformer model in spice is not
easy and depends on higher order functions that might or might not be
available in the flavour you use. Not to mention that you will need
to have good (measured) numbers on the non-ideal behaviour of a transformer,
which are also not easy to get by.
Attila Kinali
--
It is upon moral qualities that a society is ultimately founded. All
the prosperity and technological sophistication in the world is of no
use without that foundation.
-- Miss Matheson, The Diamond Age, Neil Stephenson
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
--
Chris Albertson
Redondo Beach, California
On Fri, 03 Jun 2016 12:37:26 -0400 "Mike Monett"
timenuts@binsamp.e4ward.com wrote:
I found a significant error in the LTspice analysis. I was
wondering how the current could jump instantaneously at zero when
the voltage is applied at the peak. That violates magnetism.
It turns out it doesn't. When LTspice starts an analysis, it
first calculates the operating point. For the Sine voltage source
at 90 degrees, it applies the full voltage across the load. In
this case, it was 169.7V across 1 ohm, resulting in 169.7 Amps.
That is what was plotted, and is a significant error.
Actually, spice (the engine behind LTspice) does a DC analysis
before almost all modes of operation. This DC analysis has the
intention to start the circuit from a steady-state point and thus
to reduce simulation time. In order for this to work properly, you
have to specify the DC voltage and currents for all sources
correctly. Spice messes this up at times making the first part of
a transient simulation worthless (it has even worse problems when
you do an AC analysis). Additionally LTspice hides too much of
these small complications for the problems to be visible to the
untrained eye and also at times makes it harder to provide the
correct values. Thus, caution is advised.
After starting with Intusoft in 1985, moving to Microcap in 1991,
having a brief fling with PSpice around 1998, and switching to
LTspice in 2006, I can say LTspice has the easiest and fastest data
entry of any SPICE program I have tried.
There is no problem with specifying the sources in LTspice. Nothing
is hidden. The setup menus are extremely easy to view and
understand. If you wish, you can have the input parameters displayed
on screen, as I have done with two of the functions.
LTspice checks all the information given, and if it detects an error
it generates an error message and won't run.
My original problem was not the setup menus. It was picking the
wrong model.
The general rule of "Never trust a simulation you haven't forged
yourself" applies.
Most people would be hard put to do a hand simulation of a wideband
op amp in a closed-loop feedback network. That is what SPICE is for.
Out of 13 examples I analyzed, I found only one that involves
unloaded transformers.
I found many references that discuss transformer inrush current
caused by core saturation. This is a serious problem as it puts
stress on the components and reduces operating life.
I only had a quick glance at your webpage, but it seems that you
used the standard LTspice transformer model. Unfortunately, this
is not a good model to study this kind of behaviour. For one, the
only loss considered in the model is the winding coupling, it
doesn't even directly consider resistive losses in the windings.
You would be advised to learn LTspice as it would save you a great
deal of misconceptions about how it works.
The winding resistances are included in the inductor model. You
specify them as needed. I usually used a series resistance of 1 Ohm,
but changed it in some examples to suit the application. You can
also specify the parallel winding capacitance and resistance.
In this case, the two most important effects that you need to
include are saturation and core losses, which are both frequency
dependent. The cores of electric machine transformers are very
poor when it comes to their "high" frequency behaviour. Where high
frequency starts somewhere closely above mains frequency. Ie 1kHz
is already so far off that somewhere around 90% of the energy
would be dissipated in the core.
I was not interested in examining the frequency response, saturation
effect or core losses. These are only important after the core goes
into saturation.
I was only interested in the result of switching at the peak or at
the zero crossing. This is clearly defined at the beginning of the
document.
The sharp rise in voltage and the leading inrush current have
frequency components that are way higher than mains frequency.
Hence the linear model you used will give inaccurate results, to
put it mildly.
Unfortunately, building an accurate transformer model in spice is
not easy and depends on higher order functions that might or might
not be available in the flavour you use. Not to mention that you
will need to have good (measured) numbers on the non-ideal
behaviour of a transformer, which are also not easy to get by.
The saturation and core losses are outside the scope of the
investigation. The investigation was only to examine the effect of
switching at the peak or at the zero crossing. This was clearly
stated at the beginning of the paper.
My analysis correctly defined an unloaded transformer as the only
case where switching at the peak or the zero crossing made any
difference. This was the goal, and it was met.
I also showed that very few solid state switches were available that
switched at the peak, that most vendors simply supply devices that
switch at the zero crossing and state to get a model that will
accept the surge currents, that switching at the peak could cause
severe surge currents with capacitive loads, and that I could not
find any reference that stated switching at the peak would not cause
core saturation.
Your comments offer no additional information regarding the
advisability of switching at the peak or the zero crossing. The
information you do supply is irrelevant to the problem, and mostly
irrelevant to LTspice.
Attila Kinali
It is upon moral qualities that a society is ultimately founded.
All the prosperity and technological sophistication in the world
is of no use without that foundation.
- Miss Matheson, The Diamond Age, Neil Stephenson
You need to consider getting new sigs. The two you post have little
or nothing to do with timenuts, and I'm sure everyone has them
memorized by now.
MRM
Am 08.06.2016 um 02:31 schrieb Mike Monett:
I was not interested in examining the frequency response, saturation
effect or core losses. These are only important after the core goes
into saturation.
I was only interested in the result of switching at the peak or at
the zero crossing. This is clearly defined at the beginning of the
document.
...
The saturation and core losses are outside the scope of the
investigation. The investigation was only to examine the effect of
switching at the peak or at the zero crossing. This was clearly
stated at the beginning of the paper.
My analysis correctly defined an unloaded transformer as the only
case where switching at the peak or the zero crossing made any
difference. This was the goal, and it was met.
Saturation is not outside the scope. It is the very heart of the problem.
You need to build up a voltage opposite to the grid voltage to keep the
current small.
That requires an inductance and that requires a core that can be magnetized.
If the core is already magnetized to the limit from a previous session,
it is as good
as simply not there at all. What remains is some meters of copper wire
without an
appreciable L and that is not enough.
I'm haunted by that effect myself on a regular base in that I have a fat
class A Krell
audio amplifier and it pops the fuse of my living room once in about 5
times of
switching it on.
I also showed that very few solid state switches were available that
switched at the peak, that most vendors simply supply devices that
switch at the zero crossing and state to get a model that will
accept the surge currents, that switching at the peak could cause
severe surge currents with capacitive loads,
Nobody uses large transformers anymore, everybody has a diode bridge ,
capacitor
and a DC/DC behind it. Then zero voltage switching makes sense.
and that I could not
find any reference that stated switching at the peak would not cause
core saturation.
I provided references that zero voltage switching leads to saturation,
and so did others.
Your comments offer no additional information regarding the
advisability of switching at the peak or the zero crossing. The
information you do supply is irrelevant to the problem, and mostly
irrelevant to LTspice.
you are right. This is not a LTspice problem but your modelling problem.
Attila Kinali
It is upon moral qualities that a society is ultimately founded.
All the prosperity and technological sophistication in the world
is of no use without that foundation.
You need to consider getting new sigs. The two you post have little
or nothing to do with timenuts, and I'm sure everyone has them
memorized by now.
OMG , I'm not Attila, but I may need a special time nuts .sig!
regards, Gerhard
--
Es ist schon alles gesagt worden, aber noch nicht von jedem. (Valentin)
Hold your horses folks.
There is more on this tale!
To recap we put the SSR on the aircon (at zero crossing) and the jumps got
very very worse. So we turned the aircon off again (winter here - so not
really needed) and the jumps dropped, but didn't go away. :-(
So we also have a heater in the room (simple 1200W column heater) and a
temperature monitor that turns the cooling or heating on as appropriate. So
we also replaced the heating relay with an SSR and it all now seems to have
gone away.
We are now thinking that the aircon AND heating relays had started to pit
after years of use and so give off radio transients which managed to get in
and interfere with the extremely low (-100 dBm) signal coming from the
physics package and going into the maser electronics.
We have now run for four days with no clock jumps with both aircon and
heater on and with 0V crossing SSR relays.
"Welcome to the jungle, we've got fun and games..."
Jim Palfreyman
On 3 June 2016 at 15:00, Jim Palfreyman jim77742@gmail.com wrote:
Hi All,
Thanks so much for your input and thoughts. It has really proved helpful
here at the observatory.
As it turned out we easily obtained a zero-crossing solid state relay so
we thought we'd try it.
And, drumroll......
It made things so much terribly worse than ever before. (As predicted by
many of you above.)
We are going to try a SSR that switches at the peak - but we need to order
one. So stay tuned on those results.
There is of course the "move the bloody thing far away from the maser"
solution which could end up being a serious option. These air conditioning
units are small and cheap (window-type), so we are trying to find the
cheapest solution - and if that ends up being some ducting - so be it!
Jim Palfreyman
On 26 May 2016 at 13:13, Andy AI.egrps+tn@gmail.com wrote:
On Wed, May 25, 2016 at 12:59 PM, Mike Monett <
timenuts@binsamp.e4ward.com>
wrote:
LTspice shows switching at 0V is the best point in time. ...
Bzzzt! Your simulation is seriously flawed, and your conclusions are
wrong. What you forgot, or may not have realized, is that SPICE's initial
transient solution is obtained by having the signal sources already turned
on (at the moment of the Big Bang) and set to their initial value, so the
current through L2 is limited by DC conditions. That is not anything
close
to switching the driving voltages on. It is having one waveform sit at
+169.7V DC for a very long time ('forever'), and then letting it follow a
cosine wave.
Re-run the simulation with "UIC" added to the .tran statement (.tran 50ms
uic) and see what it shows. Using UIC forces the initial voltage to be 0V
at time=0, the start of the simulation. That's like having the switch
initially open.
Or if you don't like that, multiply the sources by a PWL waveform that
starts both voltages at 0V and then switches them on, a few milliseconds
into the simulation, with the appropriate phase.
Or use an actual switch. LTspice has a switch element you could use.
I guarantee you, the case with the voltage switching on at the 0V point in
the voltage waveform, causes greater currents.
The smaller surge current happens when the source is connected at the
moment when the current i(t) would be 0A if it were a continuous waveform.
For an inductive load, this happens when the voltage v(t) would be +/-
peak
(or near peak, for a real load which has both inductance and a little
resistance). This condition also results in no surge, thus no L/R decay.
All of this might not be relevant to a mechanical system, where surge
current is caused by rotational inertia, rather than anything electrical.
Regards,
Andy
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.