Design problem for scientific lamp current source.

[Geoff]

http://i251.photobucket.com/albums/gg302/triode101/sink.jpg

This circuit is to drive a scientific lamp which is low pressure
neon of strike volts 600V and typ run volts of about 200V. The
design intent is DC current from 0 to 60mA. This circuit
traditionally runs to about 30mA with no problem and is well
proven at these currents. Normally it is not darlington pair but
it is something I have tried..

The trouble I have with this circuit is that at above currents
of about 35-40mA the lamp starts to flicker indicating a higher
current than is dialed in by the pot. An oscilloscope at the
opamp output confirms that at the point where it starts to
flicker, the loop starts to try to lower the current since the
opamp output volts drops. The onset can be made to be a periodic
dip (tens of hz dip about a volt) before the lamp begins to
flicker. When the lamp flickers on and off more severely, the
opamp tries to go negative to compensate. This to me shows that
the output transistor is conducting, not the failure of the loop
control.. I have played around with the drive impedance, ie
with/without the 1k base resistor. The point of flicker can be
less severe with the resistor, or latches hard without. Maybe
the VCEO spec is being stretched here. I have tried a variac and
reduced the volts to about 580V, still does it but onset is
higher current (50mA). I have tried a BUL 416 transistor also. I
might just be applying too much voltage, but I can't see why,
might have to look at the op curves a bit harder. Maybe I could
use a MOSFET or IGBT.

 

[Bill Sloman]

http://i251.photobucket.com/albums/gg302/triode101/sink.jpg

This circuit is to drive a scientific lamp which is low pressure
neon of strike volts 600V and typ run volts of about 200V. The
design intent is DC current from 0 to 60mA. This circuit
traditionally runs to about 30mA with no problem and is well
proven at these currents. Normally it is not darlington pair but
it is something I have tried..

This is a slightly Russian circuit.

A more subtle approach is to use a transformer to apply the 600V
striking voltage to the lamp for long enough to get the discharge
going - less than a microsecond for a glow discharge, some
microseconds for an arc - which lets you get away with a supply
voltage a lot closer to the 200V running voltage of the lamp.

The trouble I have with this circuit is that at above currents
of about 35-40mA the lamp starts to flicker indicating a higher
current than is dialed in by the pot. An oscilloscope at the
opamp output confirms that at the point where it starts to
flicker, the loop starts to try to lower the current since the
opamp output volts drops. The onset can be made to be a periodic
dip (tens of hz dip about a volt) before the lamp begins to
flicker. When the lamp flickers on and off more severely, the
opamp tries to go negative to compensate. This to me shows that
the output transistor is conducting, not the failure of the loop
control.. I have played around with the drive impedance, ie
with/without the 1k base resistor. The point of flicker can be
less severe with the resistor, or latches hard without. Maybe
the VCEO spec is being stretched here. I have tried a variac and
reduced the volts to about 580V, still does it but onset is
higher current (50mA). I have tried a BUL 416 transistor also. I
might just be applying too much voltage, but I can't see why,
might have to look at the op curves a bit harder. Maybe I could
use a MOSFET or IGBT.

What you are saying is the circuit is unstable. This means that you
need to look at the frequency compensation. That 10nF capacitor is
decoupling the op amp from the output stage at higher frequencies, but
it may not be the right value to do what you want, and it could be
contributing extra phase shift where you don't actually want it.

Get hold of a text-book that talks about stabilising negative feedback
loops and read it carefully. "The Art of Electronics" by Horowitz and
Hill might be one such textbook - though it covers a lot more than
just negative feedback.

Some of our regular posters have published more specialised texts, and
may be prepared to stick their necks out further.

 

[Tim Williams]

SOA says you can only run 600V up to a hair over 10mA, maybe 12mA. BJTs
are a bad idea for dropping high voltages due to second breakdown.

Why not choose a tube like 6L6GC? Current production, doesn't take much
heater power, and takes the voltage easily. Okay, a MOSFET would be
smaller and cheaper (though probably about even if you include the
heatsink, come to think of it).

Note you're trying to carry a maximum of (800V - 200V) * 0.06A = 36W,
which is a moderate amount for a small transistor. Just in case, you'd be
better off with a TO-247 package, or two TO-220s in parallel.

Switching methods are much more efficient, even if you can't run the thing
directly on AC (I assume that's part of the design requirement).

Tim

 

[George Herold]

http://i251.photobucket.com/albums/gg302/triode101/sink.jpg

This circuit is to drive a scientific lamp which is low pressure
neon of strike volts 600V and typ run volts of about 200V. The
design intent is DC current from 0 to 60mA. This circuit
traditionally runs to about 30mA with no problem and is well
proven at these currents. Normally it is not darlington pair but
it is something I have tried..

The trouble I have with this circuit is that at above currents
of about 35-40mA the lamp starts to flicker indicating a higher
current than is dialed in by the pot. An oscilloscope at the
opamp output confirms that at the point where it starts to
flicker, the loop starts to try to lower the current since the
opamp output volts drops. The onset can be made to be a periodic
dip (tens of hz dip about a volt) before the lamp begins to
flicker. When the lamp flickers on and off more severely, the
opamp tries to go negative to compensate. This to me shows that
the output transistor is conducting, not the failure of the loop
control.. I have played around with the drive impedance, ie
with/without the 1k base resistor. The point of flicker can be
less severe with the resistor, or latches hard without. Maybe
the VCEO spec is being stretched here. I have tried a variac and
reduced the volts to about 580V, still does it but onset is
higher current (50mA). I have tried a BUL 416 transistor also. I
might just be applying too much voltage, but I can't see why,
might have to look at the op curves a bit harder. Maybe I could
use a MOSFET or IGBT.

Why the Darlington?


If you have voltage to throw away you can always try putting in more
R.

And sure why not a high voltage FET.

George h.

 

[Geoff]

That's a honkin' big transistor for a few mA -- are you
just trying to be sure to dissipate the heat, or what?

Given that the data sheet says nothing about the transistor
cutoff frequency or input capacitance, I suspect that it's
not anything to write home about. That slow of oscillation
is surprising, but the 1N4007 is going to slow down the
speed that you can turn things off, because at that point
you'll be working the C-B capacitance of the predriver
against the 1k-ohm resistor.

I'd look at the base voltage of the drive transistor vs.
the opamp voltage, and see if you learn anything.

I'd also see if I could find some higher voltage devices.
I know they'll be thin on the ground at that voltage level,
but there should be something, and then you wouldn't have
to worry about your 700V transistor misbehaving with your
800V supply.

Assuming that you could stabilize it, a MOSFET or IGBT
should slide right in there.

I am looking at MOSFETS now.

 

[Geoff]

oups.com:

This is a slightly Russian circuit.

A more subtle approach is to use a transformer to apply the
600V striking voltage to the lamp for long enough to get
the discharge going - less than a microsecond for a glow
discharge, some microseconds for an arc - which lets you
get away with a supply voltage a lot closer to the 200V
running voltage of the lamp.


What you are saying is the circuit is unstable. This means
that you need to look at the frequency compensation. That
10nF capacitor is decoupling the op amp from the output
stage at higher frequencies, but it may not be the right
value to do what you want, and it could be contributing
extra phase shift where you don't actually want it.

Get hold of a text-book that talks about stabilising
negative feedback loops and read it carefully. "The Art of
Electronics" by Horowitz and Hill might be one such
textbook - though it covers a lot more than just negative
feedback.

Some of our regular posters have published more specialised
texts, and may be prepared to stick their necks out
further.

Your point about separating the run and strike ccts is well
taken. I might go that way if I can't find another solution
using the present transformer.

I don';t think it is a stability problem because at the point
of breaking down, the volts to the base are only dipping,
indicating the loop is trying to prevent the output from
conducting.

 

[Geoff]

SOA says you can only run 600V up to a hair over 10mA,
maybe 12mA. BJTs are a bad idea for dropping high voltages
due to second breakdown.

Why not choose a tube like 6L6GC? Current production,
doesn't take much heater power, and takes the voltage
easily. Okay, a MOSFET would be smaller and cheaper
(though probably about even if you include the heatsink,
come to think of it).

Note you're trying to carry a maximum of (800V - 200V) *
0.06A = 36W, which is a moderate amount for a small
transistor. Just in case, you'd be better off with a
TO-247 package, or two TO-220s in parallel.

Switching methods are much more efficient, even if you
can't run the thing directly on AC (I assume that's part of
the design requirement).

Tim

I agree I'm pushing the limits of the SOA. Now that this is
becoming clear that it is the problem, I can tackle it in the
best way I can. I had thought about a tube too. For the moment I
might try a MOSFET in a TO247 pack. The SOA is a bit more
friendly at the high voltage end.

Switch mode is out of the question at present, I know the
customer has a huge dislike, so I am doing it the way he wants.

 

[[email protected]]

SOA says you can only run 600V up to a hair over 10mA, maybe 12mA.  BJTs
are a bad idea for dropping high voltages due to second breakdown.

Why not choose a tube like 6L6GC?  Current production, doesn't take much
heater power, and takes the voltage easily.  Okay, a MOSFET would be
smaller and cheaper (though probably about even if you include the
heatsink, come to think of it).

Note you're trying to carry a maximum of (800V - 200V) * 0.06A = 36W,
which is a moderate amount for a small transistor.  Just in case, you'dbe
better off with a TO-247 package, or two TO-220s in parallel.

Switching methods are much more efficient, even if you can't run the thing
directly on AC (I assume that's part of the design requirement).

Tim

an upside down buck similar to this should be doable:
http://www.irf.com/product-info/datasheets/data/irs2980spbf.pdf

could make one with current sense on the low side


-Lasse

 

[Geoff]

Why the Darlington?

I had adapted the circuit from a similar one. I am not using the
darlington now. The problem persists, probably second breakdown.

If you have voltage to throw away you can always try
putting in more R.

That is worth doing if dissipation becomes a problem.

And sure why not a high voltage FET.

Trying that next.

 

[Jamie]

That's a honkin' big transistor for a few mA -- are you just trying to be
sure to dissipate the heat, or what?

Given that the data sheet says nothing about the transistor cutoff
frequency or input capacitance, I suspect that it's not anything to write
home about. That slow of oscillation is surprising, but the 1N4007 is
going to slow down the speed that you can turn things off, because at
that point you'll be working the C-B capacitance of the predriver against
the 1k-ohm resistor.

I'd look at the base voltage of the drive transistor vs. the opamp
voltage, and see if you learn anything.

I'd also see if I could find some higher voltage devices. I know they'll
be thin on the ground at that voltage level, but there should be
something, and then you wouldn't have to worry about your 700V transistor
misbehaving with your 800V supply.

Assuming that you could stabilize it, a MOSFET or IGBT should slide right
in there.

Trying to control the break over point on neon gas is going to be
tricky with so many slow components in line.

Using darlington adds to the problem with the storage time in the
trannies and the use of that diode in the base with nothing but a 1k
to pull it up isn't helping much.

That circuit needs the integrator/miller cap greatly reduced, diode
removed and possibly a Cap lead network from the feed back to give it
a little derivative in the - feed back. At least that will help with the
phase error and reduce os-kill-La-Trons!

MeSelf, I think a PWM would be nice but then again, we don't know what
the adverse effect of flicker would be..

Jamie

 

[Bill Sloman]

Well, it is by definition a stability problem because you're seeing
oscillation where you want to see steady operation.  The only question is
where the oscillation is coming from, which is why I was suggesting
looking at other points in the circuit than just the op-amp output.

The circuit topology is correct assuming that there's not a lot of
coupling between the collector and base of the transistors, that the
Darlington stage isn't too slow.  That 10nF cap is in the right place,
but Bill is correct that it may be too small.

I wasn't that specific. I just said that it wasn't doing what it had
presumably been put in there to do.

 

[Spehro Pefhany]

I agree I'm pushing the limits of the SOA. Now that this is
becoming clear that it is the problem, I can tackle it in the
best way I can. I had thought about a tube too. For the moment I
might try a MOSFET in a TO247 pack. The SOA is a bit more
friendly at the high voltage end.

Switch mode is out of the question at present, I know the
customer has a huge dislike, so I am doing it the way he wants.

Where does the 800V come from?

One thing I've done in the past is to make a weak voltage doubler that
drops down to about half when you start drawing serious current. The
effective series impedance would help stabilize the circuit, and it
would dissipate a lot less power.

With your circuit, you could try a 10K 100W resistor instead of the
330R. They even come in TO247 if that's what you really want.


Best regards,
Spehro Pefhany

 

[tm]

Where does the 800V come from?

One thing I've done in the past is to make a weak voltage doubler that
drops down to about half when you start drawing serious current. The
effective series impedance would help stabilize the circuit, and it
would dissipate a lot less power.

With your circuit, you could try a 10K 100W resistor instead of the
330R. They even come in TO247 if that's what you really want.

That's exactly how many HeNe gas laser power supplies work. The first stage
of the CW multiplier have large capacitors. Then it has enough stages with
smaller capacitors to make the start voltage. A proper current limiting
ballast resistor establishes the run current.

 

[Owen Roberts]

Take a look at the I-V curve for the lamp. You may be entering a
region of oscillation in the plasma.

Steve

 

[Geoff]

Where does the 800V come from?

One thing I've done in the past is to make a weak voltage
doubler that drops down to about half when you start
drawing serious current. The effective series impedance
would help stabilize the circuit, and it would dissipate a
lot less power.

With your circuit, you could try a 10K 100W resistor
instead of the 330R. They even come in TO247 if that's what
you really want.


Best regards,
Spehro Pefhany

800V is from volatge double from a 270 volt secondary winding at
50VA. The caps are a pair of 450V 47uF. I think that is a good
idea too, effectively reducing the supply regulation.

 

[Geoff]

roups.com:

Take a look at the I-V curve for the lamp. You may be
entering a region of oscillation in the plasma.

Steve

I might just do that too. There is a chance it acts like a
relaxation oscillator. The lamp does not misbehave with a simple
R in series with a HV supply at 50mA or so.

 

[Robert Baer]

http://i251.photobucket.com/albums/gg302/triode101/sink.jpg

This circuit is to drive a scientific lamp which is low pressure
neon of strike volts 600V and typ run volts of about 200V. The
design intent is DC current from 0 to 60mA. This circuit
traditionally runs to about 30mA with no problem and is well
proven at these currents. Normally it is not darlington pair but
it is something I have tried..

The trouble I have with this circuit is that at above currents
of about 35-40mA the lamp starts to flicker indicating a higher
current than is dialed in by the pot. An oscilloscope at the
opamp output confirms that at the point where it starts to
flicker, the loop starts to try to lower the current since the
opamp output volts drops. The onset can be made to be a periodic
dip (tens of hz dip about a volt) before the lamp begins to
flicker. When the lamp flickers on and off more severely, the
opamp tries to go negative to compensate. This to me shows that
the output transistor is conducting, not the failure of the loop
control.. I have played around with the drive impedance, ie
with/without the 1k base resistor. The point of flicker can be
less severe with the resistor, or latches hard without. Maybe
the VCEO spec is being stretched here. I have tried a variac and
reduced the volts to about 580V, still does it but onset is
higher current (50mA). I have tried a BUL 416 transistor also. I
might just be applying too much voltage, but I can't see why,
might have to look at the op curves a bit harder. Maybe I could
use a MOSFET or IGBT.

Try a simple resistor in series with the lamp and adjust the resistor
and/or supply for that current range.
Very possible the lamp is in a negative resistance region.
BTW, the secondary ionization voltage for neon is 40.9V and the third
ionization voltage for neon is 63.2V - meaning you better see about 63V
across it when operating properly.

 

[Bill Sloman]

   Try a simple resistor in series with the lamp and adjust the resistor
and/or supply for that current range.
   Very possible the lamp is in a negative resistance region.
   BTW, the secondary ionization voltage for neon is 40.9V and the third
ionization voltage for neon is 63.2V - meaning you better see about 63V
across it when operating properly.

Not if it's operating as a glow discharge. Getting electrons out of
the cathode by positive ion bombardment needs quite a large voltage
drop across the anode glow region.

 

[[email protected]]

http://i251.photobucket.com/albums/gg302/triode101/sink.jpg

You seem to be missing the obvious for the circuit and that is you need an isolation resistor between the OA output and transistor base such that the rolloff pole formed by that resistor and the worst case transistor Miller capacitance is at least a decade higher in frequency than the RC feedback ofthe 10n and 1K (really). There's nothing exotic about a crummy VCS for a neon lamp, scientific or not, it is hack simple. A lot of your circuit values are screwy.

 

[Martin Brown]

200 V at 60 mA is 12 W! Must be a big neon lamp!

That's about the rating of a few elemental reference low pressure
emission lamps. A low pressure sodium lamp is pretty bright even at 12W.

My guess is that the lamp exhibits negative resistance at this
current. You will find it pretty difficult to control the current
through a negative resistor. I think the only hope is to add
series resistance greater than the value of the negative resistance.

Seems reasonable to me. Replacing the 330R with ~3k9 30W ought to do it
by dropping about half the HT voltage across the external resistor after
the lamp strikes and the extra series resistance should then discourage
it from becoming a negative resistance relaxation oscillator.