Interupting xenon flash current ?

[Mike]

I was reading about how autoexposure and speed flash units work and found out that in modern units IGBTs are
used to turn off the tube current at the proper time. I scrounged up the parts to flash a tube and measured
the current pulse during discharge. For my test setup I measured the the current peak to be about 1900A and am
fairly certain that number is pretty close. The current was above 1000A for about 6us and still at 300A after
200us then finally decayed to 0 after about 600us. How is it possible for any reasonably priced/sized IGBT to
survive under that kind of stress? I was going to try interupting the pulse, but the only IGBT I have is one I
took from a dead microwave oven inverter and it's only rated at 150A for 1ms. My test setup isn't all that
powerful, only about 37J (600uf flash cap charged to 350v). I'm also curious about why it takes 20us from the
time of the HV trigger to the time the tube starts to conduct. It may not be, but that just seems like and
awfully long time. Not enough trigger voltage maybe? Also, it takes another 15us for the current to reach
it's peak. Unfortunately I don't have any data for the flash tube. It is fairly large, 80mm long and 8mm dia.
The trigger is a thin wire spirally wrapped the entire length of the tube.

Mike


If there is no absolute truth then nothing can be known.

 

[Jan Panteltje]

I was reading about how autoexposure and speed flash units work and found out that in modern units IGBTs are
used to turn off the tube current at the proper time. I scrounged up the parts to flash a tube and measured
the current pulse during discharge. For my test setup I measured the the current peak to be about 1900A and am
fairly certain that number is pretty close. The current was above 1000A for about 6us and still at 300A after
200us then finally decayed to 0 after about 600us. How is it possible for any reasonably priced/sized IGBT to
survive under that kind of stress? I was going to try interupting the pulse, but the only IGBT I have is one I
took from a dead microwave oven inverter and it's only rated at 150A for 1ms. My test setup isn't all that
powerful, only about 37J (600uf flash cap charged to 350v). I'm also curious about why it takes 20us from the
time of the HV trigger to the time the tube starts to conduct. It may not be, but that just seems like and
awfully long time. Not enough trigger voltage maybe? Also, it takes another 15us for the current to reach
it's peak. Unfortunately I don't have any data for the flash tube. It is fairly large, 80mm long and 8mm dia.
The trigger is a thin wire spirally wrapped the entire length of the tube.

Mike

I always tought a SCR, not an IGBT, was fired parallel to the flash tube.
And the flash tube is connected via a small inductor to the cap,
the inductor limiting the current rise.
But it has been at least 20 years ago I looked....
Anyways SCRs can have huge peak currents.

 

[Jim Thompson]

I was reading about how autoexposure and speed flash units work and found out that in modern units IGBTs are
used to turn off the tube current at the proper time. I scrounged up the parts to flash a tube and measured
the current pulse during discharge. For my test setup I measured the the current peak to be about 1900A and am
fairly certain that number is pretty close. The current was above 1000A for about 6us and still at 300A after
200us then finally decayed to 0 after about 600us. How is it possible for any reasonably priced/sized IGBT to
survive under that kind of stress? I was going to try interupting the pulse, but the only IGBT I have is one I
took from a dead microwave oven inverter and it's only rated at 150A for 1ms. My test setup isn't all that
powerful, only about 37J (600uf flash cap charged to 350v). I'm also curious about why it takes 20us from the
time of the HV trigger to the time the tube starts to conduct. It may not be, but that just seems like and
awfully long time. Not enough trigger voltage maybe? Also, it takes another 15us for the current to reach
it's peak. Unfortunately I don't have any data for the flash tube. It is fairly large, 80mm long and 8mm dia.
The trigger is a thin wire spirally wrapped the entire length of the tube.

Mike


If there is no absolute truth then nothing can be known.

I did that in the '60's, but used an SCR, read my patent at...

http://analog-innovations.com/SED/Pat-3496411.pdf

...Jim Thompson

 

[Vladimir Vassilevsky]

I was reading about how autoexposure and speed flash units work and found out that in modern units IGBTs are
used to turn off the tube current at the proper time. I scrounged up the parts to flash a tube and measured
the current pulse during discharge. For my test setup I measured the the current peak to be about 1900A and am
fairly certain that number is pretty close. The current was above 1000A for about 6us and still at 300A after
200us then finally decayed to 0 after about 600us. How is it possible for any reasonably priced/sized IGBT to
survive under that kind of stress?

I believe there is a mistake with your measurement. In the common
cameras, the flash current is at the order of 100A and the duration is
~tens of milliseconds. They handle it with IGBT of SOT-223 form factor.

However there are the special pulse tubes for use with lasers, etc. They
have higher rate of the current increase and the shorter flash duration.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

 

[John Larkin]

I was reading about how autoexposure and speed flash units work and found out that in modern units IGBTs are
used to turn off the tube current at the proper time. I scrounged up the parts to flash a tube and measured
the current pulse during discharge. For my test setup I measured the the current peak to be about 1900A and am
fairly certain that number is pretty close. The current was above 1000A for about 6us and still at 300A after
200us then finally decayed to 0 after about 600us. How is it possible for any reasonably priced/sized IGBT to
survive under that kind of stress? I was going to try interupting the pulse, but the only IGBT I have is one I
took from a dead microwave oven inverter and it's only rated at 150A for 1ms. My test setup isn't all that
powerful, only about 37J (600uf flash cap charged to 350v). I'm also curious about why it takes 20us from the
time of the HV trigger to the time the tube starts to conduct. It may not be, but that just seems like and
awfully long time. Not enough trigger voltage maybe? Also, it takes another 15us for the current to reach
it's peak. Unfortunately I don't have any data for the flash tube. It is fairly large, 80mm long and 8mm dia.
The trigger is a thin wire spirally wrapped the entire length of the tube.

Mike

The IGBT itself will limit the current. And waste some energy, I
assume.

John

 

[Mike]

I believe there is a mistake with your measurement. In the common
cameras, the flash current is at the order of 100A and the duration is
~tens of milliseconds. They handle it with IGBT of SOT-223 form factor.

However there are the special pulse tubes for use with lasers, etc. They
have higher rate of the current increase and the shorter flash duration.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

That was my frst reaction also since I expected to see something less than 400A, but I can't find any reason
for such a major error.
I have a 200mv 200A shunt in series with the tube and am simply grabbing the voltage waveform across the shunt
with an HP 54502A storage scope. So with 1mv/A and reading 1.9v peak that's 1900A. I connected the 1V scope
calibrator output from an analog scope to 54502s input and it reads right on amplitude and time.

Mike
If there is no absolute truth then nothing can be known.

 

[Mike]

I always tought a SCR, not an IGBT, was fired parallel to the flash tube.
And the flash tube is connected via a small inductor to the cap,
the inductor limiting the current rise.
But it has been at least 20 years ago I looked....
Anyways SCRs can have huge peak currents.

Yep, but the reason for opening the circuit instead of shunting the tube is to leave the unused charge in the
cap to allow much quicker recharge time.
If there is no absolute truth then nothing can be known.

 

[Mike]

The IGBT itself will limit the current. And waste some energy, I
assume.

John

and I fear, fusing to a blob of silicon in the process.

Mike

If there is no absolute truth then nothing can be known.

 

[John Tserkezis]

The IGBT itself will limit the current. And waste some energy, I
assume.

This isn't a case of "current limit". It actually does indeed interrupt the
flash tube current abruptly.

That's the only way that modern flash can work predictably, controllably and
repeatably.

I looked into this (for a non-commercial application) some time back and
couldn't find any off the shelf (one piece) semis that could do it. I
couldn't find any specific data on the way it was done, so suspected they used
two devices, an SCR as per normal, and a more traditional semi switch to turn
it off after the predetermined delay.

 

[Mark]

That was my frst reaction also since I expected to see something less than 400A, but I can't find any reason
for such a major error.
I have a 200mv 200A shunt in series with the tube and am simply grabbing the voltage waveform across the shunt
with an HP 54502A storage scope. So with 1mv/A and reading 1.9v peak that's 1900A. I connected the 1V scope
calibrator output from an analog scope to 54502s input and it reads right on amplitude and time.

Mike
If there is no absolute truth then nothing can be known.

two possible sources of error
1) if the shunt is inductive it will read higher for the pulse

2) you may have a hugh common mode problem, i assume you have the
shunt in the low voltage returm path...anyway, try connecting both
ends of the scope probe to the same side of the shunt, i.e. you should
expect to see 0V pulse when you fire the strobe...and check if you
do...
you may need to use a differential probe or other special technique to
get an accuragte small signal measurment in the presense of the large
discharge pulse..

or use a larger value shunt...

Mark

 

[Vladimir Vassilevsky]

That was my frst reaction also since I expected to see something less than 400A, but I can't find any reason
for such a major error.

I bet it is either inductive or capacitive parasitic pickup or a common
ground problem.

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

 

[Spehro Pefhany]

That was my frst reaction also since I expected to see something less than 400A, but I can't find any reason
for such a major error.
I have a 200mv 200A shunt in series with the tube and am simply grabbing the voltage waveform across the shunt
with an HP 54502A storage scope. So with 1mv/A and reading 1.9v peak that's 1900A. I connected the 1V scope
calibrator output from an analog scope to 54502s input and it reads right on amplitude and time.

Mike
If there is no absolute truth then nothing can be known.

Well, if that's what the 'scope shows, it's *got* to be true.


Best regards,
Spehro Pefhany

 

[John Larkin]

This isn't a case of "current limit". It actually does indeed interrupt the
flash tube current abruptly.

That's the only way that modern flash can work predictably, controllably and
repeatably.

I looked into this (for a non-commercial application) some time back and
couldn't find any off the shelf (one piece) semis that could do it. I
couldn't find any specific data on the way it was done, so suspected they used
two devices, an SCR as per normal, and a more traditional semi switch to turn
it off after the predetermined delay.

A couple of people make ICs that do the whole function: flyback
converter, trigger stuff, IGBT gate driver with timing.


Shunt igbt, wastes energy:

http://focus.ti.com/lit/an/slva197/slva197.pdf


Series:

http://www.onsemi.com/pub/Collateral/NCP5080-D.PDF


There are several others.

John

 

[John Larkin]

That was my frst reaction also since I expected to see something less than 400A, but I can't find any reason
for such a major error.
I have a 200mv 200A shunt in series with the tube and am simply grabbing the voltage waveform across the shunt
with an HP 54502A storage scope. So with 1mv/A and reading 1.9v peak that's 1900A. I connected the 1V scope
calibrator output from an analog scope to 54502s input and it reads right on amplitude and time.

Current shunt inductance can be a serious error in this time domain. A
current transformer would probably be more accurate, and avoid ground
loops too.

1-turn primary, maybe 100 turn secondary on a ferrite toroid, 1 ohm or
less non-inductive burden resistor. We've done this in the 100 amp, 50
ns sort of range, on a node that kicks a kilovolt, and it worked well.

John

 

[John Tserkezis]

A couple of people make ICs that do the whole function: flyback
converter, trigger stuff, IGBT gate driver with timing.

Yes, but where were they when I needed them back then?

 

[John Larkin]

Yes, but where were they when I needed them back then?

Maybe they didn't exist?

The older controlled-flash units had two flashtubes, one to actually
flash, and a second shunt tube, inside, that shorted out the cap when
enough light had been generated.

John

 

[Mike]

A couple of people make ICs that do the whole function: flyback
converter, trigger stuff, IGBT gate driver with timing.


Shunt igbt, wastes energy:

http://focus.ti.com/lit/an/slva197/slva197.pdf


Series:

http://www.onsemi.com/pub/Collateral/NCP5080-D.PDF


There are several others.

John

Thanks, I had no idea anyone made a controller like that.
In looking at the data sheet and their demo board info, I see they are dealing with an 80A pulse with only 1J
and small tube. So maybe my measurment isn't all that far off.


Mike
If there is no absolute truth then nothing can be known.

 

[Winfield]

http://www.onsemi.com/pub/Collateral/NCP5080-D.PDF

They show a flash-tube current of 60A peak (with
a short little extra bit to 80A) decaying with an
approximately 70us time constant. They suggest a
CY25BAH-8F IGBT, which is a 400V "150A" Renesas
part in a tiny tssop-8 package.

 

[Mike]

Current shunt inductance can be a serious error in this time domain. A
current transformer would probably be more accurate, and avoid ground
loops too.

1-turn primary, maybe 100 turn secondary on a ferrite toroid, 1 ohm or
less non-inductive burden resistor. We've done this in the 100 amp, 50
ns sort of range, on a node that kicks a kilovolt, and it worked well.

John

I wonder how I might calibrate such a homebrew current transformer?
Shouldn't I see some serious ringing in the current waveform if my shunt is that inductive?

If there is no absolute truth then nothing can be known.

 

[Mike]

two possible sources of error
1) if the shunt is inductive it will read higher for the pulse

2) you may have a hugh common mode problem, i assume you have the
shunt in the low voltage returm path...anyway, try connecting both
ends of the scope probe to the same side of the shunt, i.e. you should
expect to see 0V pulse when you fire the strobe...and check if you
do...
you may need to use a differential probe or other special technique to
get an accuragte small signal measurment in the presense of the large
discharge pulse..

or use a larger value shunt...

Mark

This is a very low inductance shunt. It consists of 2 straps of resistance material about 1" wide and 5" long
stacked on one above the other in brass blocks with Kelvin sense terminals. I just measured the inductance to
be 0.2 to 0.4uH. It wants to drift around a bit. I have used this same shunt to measure a 3000A pulse in an
impulse magnitizer that I built, but that was a much longer pulse. Somewhere around 8ms if I remember
correctly.
Yes the shunt is located between the capacitor neg terminal and the flash tube. The scope gnd, the charge
power neg, capacitor neg, and shunt are all connected to the same node. I connected the scope probe tip to the
that same node and the scope showed zero voltage.

Mike
If there is no absolute truth then nothing can be known.