design ideas needed

[mkogan]

Actually I am stuck in the bring-up stage...
It is pulsed current source using buck topology.

The load is a spark gap (modeled by ~40V MOV). Pulse duration
100-300us, current ~20A with ~10A ripple (exact values - TBD). Input
voltage source is ~50uF capacitor charged to ... whatever it needs to
be charged.
The feedback signal is taken from two current sense resistors (one - in
series with input energy storing capacitor and another - with diode)...
The problem I am facing is unstable operation and there is no easy way
to check feedback loop.

Any ideas/suggestions?
Thanks
Michael
[email protected], remove 999

 

[Tim Wescott]

Actually I am stuck in the bring-up stage...
It is pulsed current source using buck topology.

The load is a spark gap (modeled by ~40V MOV). Pulse duration
100-300us, current ~20A with ~10A ripple (exact values - TBD). Input
voltage source is ~50uF capacitor charged to ... whatever it needs to
be charged.
The feedback signal is taken from two current sense resistors (one - in
series with input energy storing capacitor and another - with diode)...
The problem I am facing is unstable operation and there is no easy way
to check feedback loop.

Any ideas/suggestions?
Thanks
Michael
[email protected], remove 999

I get a glimmer of what you're trying to say, but there are some
contradictions, too.

First: A 40V spark gap? I'd accuse you of missing a few zeros but you
say you need to buck the voltage down to make it work. I assume this is
a very _small_ spark gap? Can you tell us how it's built to operate at
just 40V?

So whatever it is, you want to give it constant current for 300us,
around 20A, and you want to accomplish this with a 50uF capacitor in
parallel with the spark gap?

You are currently monitoring the current going into the capacitor, and
also (presumably) the peak current in the gap -- yes?

If my assumptions so far are correct its no wonder that you're seeing
unstable operation. Spark gaps, at least in a gaseous environment, have
a negative resistance characteristic once they start. Driving one from
a capacitor would result in a current that's controlled more by the ESR
of the cap than anything else.

None the less, could you describe what is unstable, and in what way?
Could you explain why you can't check your feedback loop, and for that
matter expand on what the feedback is and what it's controlling?

I would be tempted to recommend that you use a different regulator
topology with an inductive output. Particularly since you can stand a
10A ripple you should be able to use a topology that puts an inductor in
series with the spark gap and controlls the average current to the
inductor. This should tend to null out the effects of the negative
resistance of the spark gap to good effect.

You may be able to achieve this simply by tossing your cap and driving
the gap straight from the inductor.

 

[Don Foreman]

Actually I am stuck in the bring-up stage...
It is pulsed current source using buck topology.

The load is a spark gap (modeled by ~40V MOV). Pulse duration
100-300us, current ~20A with ~10A ripple (exact values - TBD). Input
voltage source is ~50uF capacitor charged to ... whatever it needs to
be charged.
The feedback signal is taken from two current sense resistors (one - in
series with input energy storing capacitor and another - with diode)...
The problem I am facing is unstable operation and there is no easy way
to check feedback loop.

Any ideas/suggestions?
Thanks
Michael
[email protected], remove 999

Charged cap as source suggests an exponential decay. Why a buck
topology?

 

[[email protected]]

Actually I am stuck in the bring-up stage...
It is pulsed current source using buck topology.

The load is a spark gap (modeled by ~40V MOV). Pulse duration
100-300us, current ~20A with ~10A ripple (exact values - TBD). Input
voltage source is ~50uF capacitor charged to ... whatever it needs to
be charged.
The feedback signal is taken from two current sense resistors (one - in
series with input energy storing capacitor and another - with diode)...
The problem I am facing is unstable operation and there is no easy way
to check feedback loop.

I don't think that a 40V metal-oxide varistor is much of a model for a
spark-gap.

For a start there is a minimum in the Paschen curve of break-down
voltage versus gap length at around a few hundred volts, whcih depends
on the nature of the gas and the pressure.

Once your single electron has strayed into the spark-gap and initiated
the avalanche of ionisation that constitutes your spark, the discharge
is sustained by positive ion bombardment of the negative electrode, for
a while at least. With your 20A current, this would be followed by a
rapid (sub-microsecond) transition to the arc mode, where the discharge
is sustained by field emission from the spikes on the almost molten
surface of the negative electrode. Sustaining a glow discharge needs
something of the order of 100V, while an arc can get by with a lot
less.

If you do want a constant current discharge you'd be better off with
energy stored in an inductor rather than a capacitor, as Tim Westcott
says. Because of the large voltage excursions as your discharge goes
from spark to glow to arc, you are going to have to worry about the
interwinding capacitance of your inductor which is going to be
self-resonant with an appreciable Q (if you don't do anything about
it).

The whole circuit will tend to ring at this resonant frequency as soon
as the gap has broken down.

Getting from there to a constant current could take some ingenuity.

 

[Tim Wescott]

The whole circuit will tend to ring at this resonant frequency as soon
as the gap has broken down.

In fact, in the very early days of radio there were no vacuum tubes and
the high-tech transmitters used the negative resistance of the arc to
create a sustained oscillation. AFAIK this was _not_ just a spark-gap
transmitter exciting a resonant filter; it was a true oscillation that
depended on the arc as an active element.

Hmm. Gotta research that. Sometime when its not past bed time...

 

[[email protected]]

In fact, in the very early days of radio there were no vacuum tubes and
the high-tech transmitters used the negative resistance of the arc to
create a sustained oscillation. AFAIK this was _not_ just a spark-gap
transmitter exciting a resonant filter; it was a true oscillation that
depended on the arc as an active element.

Arcs do tend to have a slightly negative incremental resistance - and
that could have been enough to offset the resistive losses in the coil.
There might have been some mildly odd effects as the arc current went
through zero, but your conducting channel is running at about 6000K and
won't cool off much while the current is changing polarity.

 

[Tony Williams]

Arcs do tend to have a slightly negative incremental resistance -
and that could have been enough to offset the resistive losses in
the coil. There might have been some mildly odd effects as the
arc current went through zero, but your conducting channel is
running at about 6000K and won't cool off much while the current
is changing polarity.

Hello Bill. Google for 'poulsen' and 'carbon arc'.

Eg, http://earlyradiohistory.us/1910poul.htm for a
good description of how Poulsen (and others) managed
to alter the resistance characteristics of a carbon arc.

 

[mkogan]

Thanks to everybody for the replies!

I will answer the questions I was asked.
The spark gap is ~0.75-1mm, filled with Argon.
I was told (by our chemical analyst, actually she referred me to an
article) then when spark is "on" (arc is present) the voltage across it
is ~48V, somewhat dependent on the current. The current polarity must
not be changed - no ringing allowed.
I cannot measure the spark characteristics as ~10kV spike is applied to
the gap to initiate the plazma (...this looks like really iteresting ee
challenge though).

I used 2.5Ohm resistor as the load until ordered TVS arrived yesterday.
In general, the system works (up to 250us wide output pulse), it is not
very stable (not shippable ) )

I modeled this thing (in Excel): switching frequency, and energy
storing capacitor voltage drop vs. time for different input/plasma
voltages, inductance, capacitance, min/max current values. I will be
more than happy to email the file if you want to play with it.
(mkogan1999 at rcn.com, remove 999) )

Until now we used ~100uF capacitor charged to 90-250V duscharged
directly into the gap through ~0.1-0.3 Ohm (resistor and whatever ESR
the capacitor has).
The problem: peak current exceeding 200A .
People use inductors to spread the pulse, but these things are BIG
(~100+uH for tens of Amps), we cannot use it.
I cannot say how soon the capacitor and HV diode (to isolate the
trigger pulse) will fail as 200A+ current is way outside any spec.

I was told that the amount of charge per pulse is what needs to be
delivered into spark gap, so controlling current by placing reasonably
sized inductor (~12uH@20A) in series with the gap seems to be the right
way to do it.
I think I found the suspect: one of (two) comparators in the feedback
circuit seems to be ringing (2mV built-in hysteresis does not seem to
be sufficient). This will be easy to do (cut some traces and glue
couple 0805 resistors)
Thanks again
Michael Kogan

 

[Fred Bloggs]

I will answer the questions I was asked.
The spark gap is ~0.75-1mm, filled with Argon.
I was told (by our chemical analyst, actually she referred me to an
article) then when spark is "on" (arc is present) the voltage across it
is ~48V, somewhat dependent on the current. The current polarity must
not be changed - no ringing allowed.
I cannot measure the spark characteristics as ~10kV spike is applied to
the gap to initiate the plazma (...this looks like really iteresting ee
challenge though).

I used 2.5Ohm resistor as the load until ordered TVS arrived yesterday.
In general, the system works (up to 250us wide output pulse), it is not
very stable (not shippable ) )

I modeled this thing (in Excel): switching frequency, and energy
storing capacitor voltage drop vs. time for different input/plasma
voltages, inductance, capacitance, min/max current values. I will be
more than happy to email the file if you want to play with it.
(mkogan1999 at rcn.com, remove 999) )

Until now we used ~100uF capacitor charged to 90-250V duscharged
directly into the gap through ~0.1-0.3 Ohm (resistor and whatever ESR
the capacitor has).
The problem: peak current exceeding 200A .
People use inductors to spread the pulse, but these things are BIG
(~100+uH for tens of Amps), we cannot use it.
I cannot say how soon the capacitor and HV diode (to isolate the
trigger pulse) will fail as 200A+ current is way outside any spec.

I was told that the amount of charge per pulse is what needs to be
delivered into spark gap, so controlling current by placing reasonably
sized inductor (~12uH@20A) in series with the gap seems to be the right
way to do it.
I think I found the suspect: one of (two) comparators in the feedback
circuit seems to be ringing (2mV built-in hysteresis does not seem to
be sufficient). This will be easy to do (cut some traces and glue
couple 0805 resistors)

If precision rate of charge delivery and no ringing is required then you
need a current source. The 20A by 250us is 5,000 uC of charge dump and
allowing for 10V overhead potential means that Vinit-5,000u/C=60V so
that for 100u capacitor this would make Vinit=110V or so, which is not
so bad. An arrangement like this should be only slightly more than a
Radio Shack project- nothing extraordinary about it:
View in a fixed-width font such as Courier.

..
..
..
.. 100U
.. -------------||------------------. -
.. | + | |
.. | | |
.. | --- |
.. ADJ HV IN>--[Rs]--+--------+----+------------. /// |
.. 100-150V | | | | |
.. | | | | |
.. | | | [Re] |
.. [RH3] | >| | |
.. C1 /| | [RB1] T3|----------+ |
.. /+|-----+ | /| | 20A |
.. .-< | | | | | | |
.. | \-|-. [RH4] | | |< v |
.. | \| | | | | .----| T1 |
.. | | | | | | |\ |
.. | REF1 | | | | | |
.. | --- | | |< pwr | |
.. | /// +----+----|T2 PNPs| | +
.. | | |\ | \ /
.. | | | | D2
.. | LOGIC TIMER [RB2] '-----+--|>|---> To Gap
.. | NEST + 250U | |
.. +----------+ | D1 |
.. | | |/ -|<|-'
.. TRIG>---|TRIG CSC|--[RB3]--|T4 |
.. | | |> [RH1] CLAMP
.. RDY <---|READY | | | | |\ C2
.. | DIV|--- | +-[RS]-+--|+\
.. |HVT CSO | | | | | >--
.. +----------+ | | [RH2] ----|-/ |
.. | | | | | | |/ |
.. | | | | |/ | |
.. | | -------|---[RB3]-|T5 REF2 |
.. | | | |> |
.. | | | | |
.. | | -----+----- |
.. | | | |
.. | | --- |
.. | | /// |
.. | '--------------------------------------------'
.. |
.. |
.. -----------------------------------------> HV TRIG IN
..


Operation is fairly obvious, the C1 gives logic nest indication of
capacitor full charge state, external trigger is received, CSC turns on
T4 enabling 20A current source T1+T2+T3, C2 senses full-on 20A and feeds
back to logic nest, logic nest triggers HV trigger circuit, HV trigger
applied to D2 cathode fires gap, since 20A*(RH1+RH2) is selected for
100V or so, negative resistance of gap diverts 20A into gap, C2 drops
out and nest DIV turns T5 off and starts pulse duration timer. When nest
timer expires, T4 is turned off, terminating pulse discharge.