Maker Pro
Maker Pro

Peak detector possible for 4-15 ns pulses?

C

Chris Carlen

Jan 1, 1970
0
Hi:

Actually, we could use both a peak detector and integrator, for
different situations.

What is the feasibility of producing a peak detector that can output a
voltage proportional to the peak value of 4 ns FWHM pulses? This is the
lower limit on the width of the pulses we ma wish to detect. The signal
is from a PMT looking at laser induced incandescence. More typical
pulse widths would be 4-10 ns risetime to the peak, followed by a slow
decay of about 1us.

It is also sometimes desired to integrate this signal. How about the
feasibility of an accurate integrator at this time scale? Oh, it might
have to be gated as well, with 10-50ns gate widths. I know this is
doable, since we have lots of gated boxcar integrators around here.

Right now we are splitting the signal into 4 Tek scope channels, with
different gains on each to get resolution over a wide dynamic range.
Then the scope is dumped over GPIB and software finds the
peaks/integrals. Trouble is, the GPIB isn't fast enough. We are
looking at the Gage digitizing cards as as solution. These would
certainly work for both aspects, and are quite a flexible solution, but
expensive.

We figure an analog method would be much cheaper once developed, if
possible.

Is this doable, or would you just digitize? What if you *had* to do it
analog?

Comments appreciated.


Good day!


--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
[email protected] -- NOTE: Remove "BOGUS" from email address to reply.
 
J

John Larkin

Jan 1, 1970
0
Hi:

Actually, we could use both a peak detector and integrator, for
different situations.

What is the feasibility of producing a peak detector that can output a
voltage proportional to the peak value of 4 ns FWHM pulses? This is the
lower limit on the width of the pulses we ma wish to detect. The signal
is from a PMT looking at laser induced incandescence. More typical
pulse widths would be 4-10 ns risetime to the peak, followed by a slow
decay of about 1us.

It is also sometimes desired to integrate this signal. How about the
feasibility of an accurate integrator at this time scale? Oh, it might
have to be gated as well, with 10-50ns gate widths. I know this is
doable, since we have lots of gated boxcar integrators around here.

Right now we are splitting the signal into 4 Tek scope channels, with
different gains on each to get resolution over a wide dynamic range.
Then the scope is dumped over GPIB and software finds the
peaks/integrals. Trouble is, the GPIB isn't fast enough. We are
looking at the Gage digitizing cards as as solution. These would
certainly work for both aspects, and are quite a flexible solution, but
expensive.

We figure an analog method would be much cheaper once developed, if
possible.

Is this doable, or would you just digitize? What if you *had* to do it
analog?

Comments appreciated.


Good day!

Questions:

How accurate?

What's the input voltage range?

What frames an event? Will the peak detector just reset itself some
time after a pulse is sensed?

Will you eventually digitize this anyhow?


The peak detector is sorta interesting; the gated integrator should be
fairly easy.

John
 
C

Chris Carlen

Jan 1, 1970
0
Questions:

How accurate?

What's the input voltage range?

What frames an event? Will the peak detector just reset itself some
time after a pulse is sensed?

Will you eventually digitize this anyhow?


The peak detector is sorta interesting; the gated integrator should be
fairly easy.


Uhm, good questions. Thanks for replying too!

1. Let's say you tell me about the feasibility of: 1%, 2%, 5% accuracy
levels?

2. Input voltage range from about 1mV to 500mV. Output voltage would
be preferred to be in the 20mV to 10V range, so some built-in gain would
be beneficial.

3. A trigger pulse for the laser is available, about 10-20ns before the
peak.

Perhaps the peak level can be simply held until the next one comes
along. Or if it is desired to reset before that, then hold for say,
10ms. I think if we could get it so the scope can record the peak
levels of several shots at 10Hz, then this would be beneficial. Is
there a reason why fast peak detectors can't have long hold times? If
so, then perhaps a second stage of S/H can be connected. Actually, if
it can hold for a few ms then we wouldn't need the scope, as medium
speed DAQ cards could digitize it.

It is possible that a true analog out system may be desired, to feed
signal to a data acquisition system in another facility. For our
purposes, we will ultimately digitize. The issue is how to get it done
cheaper than with an expensive scope or pair of $5000 Gage cards.


Good day!
 
T

Tom Bruhns

Jan 1, 1970
0
A long time ago, I did a peak detector that worked well on ~40nsec
laser pulses into a fast photodiode detector. Parts are so much
better now than they were then that I would think that 4nsec would not
be terribly difficult, though it depends on the accuracy you want,
too, which I didn't see mentioned.

Given your PMT output, it seems like you can integrate directly onto a
capacitor. You can use a set of (PNP) transistors whose emitters are
all tied together and to the PMT cathode to select just one path to
charge a capacitor, so you could that way have a set of integrators
sequentially in time. If more than two transistors, be careful how
you drive the bases to get clean switching from one to the next. The
PNPs can that way buffer the photocathode from voltage changes as
well. You can also use an op amp to transfer the charge out of the
cap that receives it directly from the PMT to one across the op amp,
and if the op amp input terminals are well-enough behaved, you will
transfer practically all the charge even if the op amp is not as fast
as the pulse.

There probably is lots of devil in the details, but I think both can
be done OK.

Cheers,
Tom
 
R

Robert Lacoste

Jan 1, 1970
0
My guess is that a pure analogic design will have some difficulties with
such small pulses. I've seen peak detectors (high speed diode+capacitor) up
to 20ns or so, but not 4ns. May be an idea : why not using a bank (say 10)
of high speed sample&hold circuits, clocked by delayed versions of a common
50MHz clock through a 10x2ns delay line, with a and then put a 20ns
analogic peak detector on the output of each sample & hold circuit. Taking
the maximum value of these 10 peak detectors you will get a 2ns peak
detector, no ?

Cheers,

Robert Lacoste - ALCIOM : The mixed signals experts
http://www.alciom.com
 
S

Sven Hegewisch

Jan 1, 1970
0
Chris said:
Hi:

Actually, we could use both a peak detector and integrator, for
different situations.

What is the feasibility of producing a peak detector that can output a
voltage proportional to the peak value of 4 ns FWHM pulses? This is the
lower limit on the width of the pulses we ma wish to detect. The signal
is from a PMT looking at laser induced incandescence. More typical
pulse widths would be 4-10 ns risetime to the peak, followed by a slow
decay of about 1us.

It is also sometimes desired to integrate this signal. How about the
feasibility of an accurate integrator at this time scale? Oh, it might
have to be gated as well, with 10-50ns gate widths. I know this is
doable, since we have lots of gated boxcar integrators around here.

Right now we are splitting the signal into 4 Tek scope channels, with
different gains on each to get resolution over a wide dynamic range.
Then the scope is dumped over GPIB and software finds the
peaks/integrals. Trouble is, the GPIB isn't fast enough. We are
looking at the Gage digitizing cards as as solution. These would
certainly work for both aspects, and are quite a flexible solution, but
expensive.

We figure an analog method would be much cheaper once developed, if
possible.

Is this doable, or would you just digitize? What if you *had* to do it
analog?

Isn't this a typical nuclear physics problem? Photomultiplier looking to
a fast plastic szintillator? Then preamp, main amplifier and
multichannel
analyser.
Perhaps you can find a solution at companies like Canberra, Lecroy,
Ortec,
Caen. Or google for nuclear physics electronics.

Regards

Sven Hegewisch
 
C

Chris Carlen

Jan 1, 1970
0
Chris said:
Hi:

Actually, we could use both a peak detector and integrator, for
different situations.

What is the feasibility of producing a peak detector that can output a
voltage proportional to the peak value of 4 ns FWHM pulses? This is the
lower limit on the width of the pulses we ma wish to detect. The signal
is from a PMT looking at laser induced incandescence. More typical
pulse widths would be 4-10 ns risetime to the peak, followed by a slow
decay of about 1us.

It is also sometimes desired to integrate this signal. How about the
feasibility of an accurate integrator at this time scale? Oh, it might
have to be gated as well, with 10-50ns gate widths. I know this is
doable, since we have lots of gated boxcar integrators around here.

Right now we are splitting the signal into 4 Tek scope channels, with
different gains on each to get resolution over a wide dynamic range.
Then the scope is dumped over GPIB and software finds the
peaks/integrals. Trouble is, the GPIB isn't fast enough. We are
looking at the Gage digitizing cards as as solution. These would
certainly work for both aspects, and are quite a flexible solution, but
expensive.

We figure an analog method would be much cheaper once developed, if
possible.

Is this doable, or would you just digitize? What if you *had* to do it
analog?



While eating breakfast, this occurred to me:

A bank of flash A/D converters, say 8 or 16 are connected to the signal,
all in parallel of course. Now it's just a matter of sequencing them.
At first I thought of just a transmission line, to sequence them at time
spacings of several 100ps up to a few ns. But then I recalled I have
seen some ECL delay generator chips, that are programmable with 10-20ps
resolutions and up to several ns delays. One of these per A/D converter
could do the sequencing. The trigger is the laser's trigger of course,
which would give several 10ns before the flash trigger sequence should
begin.

Now it's just a matter of collecting the data. The first requirement is
that the A/D converters latch their data indefinitely, but keep it
latched behind a 3-state bus buffer. They could be put on a bus, with
simply a line for each A/D to tell it when to assert it's data. Now a
CPU can poll each converter at it's leisure to collect the data.

We want to work with perhaps up to 20-50Hz laser pulses, so this gives
at least 20ms of collection and processing time per cycle. The data
rate is so low that the thing could collect data for long periods,
holding in RAM or even streaming to a hard disk if desired.

In the case that an actual analog peak level output was desired per
shot, such as for interfacing to existing analog DAQ equipment, then the
CPU could actually perform a curve fit and peak determination on the
data, and spit out an answer in analog or digital format each shot.
This shouldn't be difficult within a 20ms time frame.

This is even more interesting when you consider the possibility for real
time data analysis related directly to the theory of LII. For instance,
instead of fitting a general curve just for the purpose of finding the
peak, the CPU could fit to the theoretical model of LII signal shape,
outputting whatever parameters are involved in that model in real time
on a shot to shot basis. This might allow dynamic adjustment and
optimization of combusion parameters.

What do you think?


Good day!




--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
[email protected] -- NOTE: Remove "BOGUS" from email address to reply.
 
C

Chris Carlen

Jan 1, 1970
0
Robert said:
My guess is that a pure analogic design will have some difficulties with
such small pulses. I've seen peak detectors (high speed diode+capacitor) up
to 20ns or so, but not 4ns. May be an idea : why not using a bank (say 10)
of high speed sample&hold circuits, clocked by delayed versions of a common
50MHz clock through a 10x2ns delay line, with a and then put a 20ns
analogic peak detector on the output of each sample & hold circuit. Taking
the maximum value of these 10 peak detectors you will get a 2ns peak
detector, no ?

Cheers,

Robert Lacoste - ALCIOM : The mixed signals experts
http://www.alciom.com


Interesting. I just posted an idea to use a sequence of flash A/D
converters. Perhaps it is better to just digitize, then do whatever is
desired. The difference is that I am just collecting 8 or 16 or some
small number of points, rather than the 500 or whatever our scope is
collecting, limited by the data throughput from the scope's GPIB bus.

Please see my new post "-- An Idea."


Thanks. Good day!




--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
[email protected] -- NOTE: Remove "BOGUS" from email address to reply.
 
C

Chris Carlen

Jan 1, 1970
0
Sven said:
Isn't this a typical nuclear physics problem? Photomultiplier looking to
a fast plastic szintillator? Then preamp, main amplifier and
multichannel
analyser.
Perhaps you can find a solution at companies like Canberra, Lecroy,
Ortec,
Caen. Or google for nuclear physics electronics.


Thanks for the input. I'll check out this avenue.


Good day!




--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
[email protected] -- NOTE: Remove "BOGUS" from email address to reply.
 
R

Rene Tschaggelar

Jan 1, 1970
0
Chris said:
Hi:

Actually, we could use both a peak detector and integrator, for
different situations.

What is the feasibility of producing a peak detector that can output a
voltage proportional to the peak value of 4 ns FWHM pulses? This is the
lower limit on the width of the pulses we ma wish to detect. The signal
is from a PMT looking at laser induced incandescence. More typical
pulse widths would be 4-10 ns risetime to the peak, followed by a slow
decay of about 1us.

It is also sometimes desired to integrate this signal. How about the
feasibility of an accurate integrator at this time scale? Oh, it might
have to be gated as well, with 10-50ns gate widths. I know this is
doable, since we have lots of gated boxcar integrators around here.

Right now we are splitting the signal into 4 Tek scope channels, with
different gains on each to get resolution over a wide dynamic range.
Then the scope is dumped over GPIB and software finds the
peaks/integrals. Trouble is, the GPIB isn't fast enough. We are
looking at the Gage digitizing cards as as solution. These would
certainly work for both aspects, and are quite a flexible solution, but
expensive.

We figure an analog method would be much cheaper once developed, if
possible.

Is this doable, or would you just digitize? What if you *had* to do it
analog?

Directly digitize doesn't offer that many bits. Rather have a logamp
to compress the range. A possible integrator, ahem a cap for the
detector requires an own logamp, as inegrating after the logamp doesn't
work anymore (wrong result).
Logamps are available with 60 to 100dB range and 500MHz bandwidth,
at least the last time I looked.

Rene
 
Top