Hello,
I would like to design the electronics to use the Eltec 4202 quadrant pyroelectric detector ( datasheet:
Yet I am unfortunately no expert in electronics, but willing to learn! I would greatly appreciate if someone could help me define the basic design to reach my goal, please.
Best regards, Mat
f) to detect the position of a
At 7-12 pF capacitance per segment in that package, you will most likely see the average of the waveshape, perhaps not the peak.
Take this over to sci.electronics.design and title you post Attention Phil Hobbs, detector question. dot one V is a wide specification as well.
Since the detector has huge source impedance, you get the RC product of something like 12 x10^-9 times somewhere between 10 ^ 7 to 10 ^10 as your time constant, and this is not good for fast response.
check with Dr Hobbs, as he wrote the major book on practical lab optics.
Expect to get fried in rhetoric by the engineers on Sci.electronics.design, but pay attention to what Dr Hobbs says.
Steve Roberts
f) to detect the position of a
Why the pyroelectric? (Are you in the IR?) What's the wavelength of the laser? (A Quad photodiode might be easier?)
The FIR pyroelectric detector I used in the past was very slow. Do you know the thermal time constant of the detector?
There are some sample circuits in the PDF file you posted did you try any of those? I'd try the opamp one first. Any good FET opamp will work.
What's the average laser power?
George H.
That's an interesting detector.
A pyroelectric is a temperature->charge transducer, which makes it unusual.
Photodiodes are optical power->current, which is easy unless you need to go fast with small signals. Thermocouples are temperature->voltage, strain gauges, thermistors and so on are temperature->resistance (so basically temperature->current with fixed bias voltage). Current and voltage-output transducers are straightforward to interface--generally you use an op amp to get a nice low impedance output voltage signal of a convenient size, make sure the offset voltage is going to be reasonably small and stable, and that the noise level is low enough.
All that takes a bit of algebra but is conceptually not at all difficult.
Charge transducers take a bit more care, because: (a) they're inherently AC--the available charge corresponds to the change in the measured quantity since the last time the charge was dumped out, and (b) there generally isn't much charge available.
The only other common transducer with a charge output is a piezoelectric (distance->charge), but people almost never try measuring displacement with piezos--it's almost always transient or sinusoidal acceleration, which is easier because the transient is what you care about, and that's what the piezo measures. So pyros are kind of strange birds.
If you hit a pyro with a temperature transient, e.g. a single-shot laser pulse, you get a step function change in sensor temperature, with a slow exponential falloff as the sensor cools down.
Well-insulated pyros usually follow a simple model using just the thermal mass (M_th) and thermal conductance (G_th). This is mathematically just like a parallel RC circuit, with M_th being C and G_th being 1/R, so that the pyro element's impulse response is a step function with a decaying exponential tail, whose time constant is M_th/G_th. (We're going to ignore the effects of nonuniform heating inside they pyro for now--we'll come back to it later.)
If you imagine turning on the 1-kHz rep rate pulse train at t=0, then with a very slow sensor with infinite shunt resistance, the sensor temperature vs. time will look like this:
|T ..... etc. | .---/ | .----/ | .----/ | .----/ |----------/
+----+-----+-----+-----+-----+-----+.... | -1ms 0 1ms 2ms 3ms 4ms...and the available charge output will look look like this
|q ..... etc. | .---/ | .----/ | .----/ | .----/ |----------/
+----+-----+-----+-----+-----+-----+.... | -1ms 0 1ms 2ms 3ms 4ms...The bad news is that the voltage depends on all the pulse energies in some complicated way, not just on the pulse you want to measure.
If you put a small resistor in parallel with the pyro, the additional charge on each pulse gets bled out right away, so you get pulse waveforms instead:
|V(t) | | | | |----------/\\----/\\----/\\----/\\----/\\....
+----+-----+-----+-----+-----+-----+..... -1ms 0 1ms 2ms 3ms 4ms....This is nicer from a measurement perspective, since the height of each output pulse corresponds to the laser pulse energy. However, that the pulses are very small, because the RC time constant is much shorter than the thermal time constant, and they have a quick exponential decay, which makes the measured value depend on the timing.
Reducing the thermal mass of the detector makes the pulses bigger, which helps a lot. Increasing the load resistor helps too, until the pulses start to overlap significantly, at which point you start getting back towards the staircase situation.
In your case, the pulse rep rate is 1 kHz, and the thermal time constant is 5 ms. You aren't the detector designer, so you can't do anything about the thermal properties, so the problem is how to optimize the circuit. If you replace the resistor with a MOSFET switch, synchronized with the pulses, you can get the best of both worlds--a flat topped waveform, maximum output, and no staircase.
I assume you have some data acquisition system you want to use for this, and that you're comfortable programming it. If so, the easiest thing to do is to put an op amp buffer (or a noninverting amplifier if you like) on each quadrant, with a small MOSFET such as a 2N7000 connected with its drain to the op amp input, source to ground, and gate to a digital output of your data acq board. Synchronize the data acq system so that the op amps' outputs are sampled before and after each pulse, e.g. t_pulse-50 us and t_pulse+250 us, and turn on the MOSFET (digital output = 1) from t_pulse+500 us to t_pulse+900 us, and subtract the values you get. That way you'll get the maximum available output from each pixel, and each reading will correspond to exactly one pulse.
This is called correlated double sampling, and is the key to good pyroelectric or CCD measurements.
If your data acq system isn't flexible enough to do that, you'll need a delay generator. Resist the temptation to use a bunch of monostables for this--you want the measurement to be nice and stable, and monostables and stability don't go together. A digital delay box, e.g. one from SRS or Highland Technology (John Larkin's company) will give you all the timing control you need. If you have the time and want to learn some electronics (highly desirable for a scientist!), you could build your own small special delay box. You might use a 10x frequency multiplier based on a phase-locked loop. This one gives you 10 equally-spaced timing pulses spread out through the 1 ms pulse period:
47k GND--RRRR-+ 1nF | +-CCCC-+ | | | Rep rate +------+---+------+-+ +-------------------+ Clock 10k | C C | | CD4017B | 0--------RRRR-----|Sig In CD4046B | | | | | | | | VCO Out|----|>Clock | | PD2 VCO | | In | | Out In Comp In | | Q0 Q1 Q2 Q3...| +-------------------+ +-------------------+ | | | | | | | | +-----------+ V etc | | MOSFET R | gate 1 | R | 1 | | | +-----+----R2R2---C1C1---+ | GNDHere you could start with R1 = 1M, R2=100k, C1=100nF, and use a 5V supply, which will (unless I've made a blunder somewhere) give the VCO a centre frequency of 8000 Hz, a loop bandwidth of about 30 Hz, and nice stable frequency compensation. You have to figure out the rest of the connections by reading the CD4046B and CD4017B datasheets carefully--there are supplies, bypass capacitors, VCO inhibit, count inhibit, and reset inputs to worry about. If you turn it on and nothing whatsoever happens, check the power supply connections, then check the rest of the connections, then look carefully at the enables, inhibits, resets, and so on.
This circuit will give you 10 timing pulses for every laser pulse, with the rising edge of pulse 0 coinciding nearly with the sync pulse. Assuming the positive edge of the sync pulse is when the laser pulse arrives, you can use pulse 8 for the pre-pulse sample clock, pulse 1 for the post-sample clock, and pulse 6 for the MOSFET gate.
If your data acq system doesn't have simultaneous sampling (Bad system! Bad! No biscuit!), you'll need to add four sample/hold chips, e.g. LF398s, and connect their sample/hold lines together.
Cobbling together CMOS parts like this is a great way to get into building your own apparatus, which is a key skill for experimentalists--if you design it, you don't have to read the stupid manual, and you can get a lot done while the other guy is waiting for his capital equipment money to be released.
Let us know what you decide to do, and how it comes out!
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
Mathieu G writes
Hi Mat,
one of the key problems you will have is: light travels just a fraction of a mm in 100fs. This could make it very difficult to design - there will be many layout and component problems. For example, if 0.03mm is significant, thermal expansion of the components. Most people find it hard to design things that work well at just a few hundred MHz, ie where the signals are about the same size as a PCB, because you get unexpected reflections and interactions at this size scale. So, you need really expert advice here. (I see you have already been sent to Phil Hobbs - I asked him some questions myself recently.)
Some things to consider:
- Is this a one-off item, or do you want to make many of them? If more than one, the assembly needs to be designed well enough to be repeatable.
- What frequency is the laser light? This may affect choice of materials in your optical assembly.
- What sources of optical and electrical interference will be present? A shielded (ie, metal) box would be best against electrical interference... you will probably want an optical filter to allow only the laser light through and cut out daylight.
- A pyroelectric detector is not ideal for detecting fast signals. It has too much capacitance.
- What is your development budget? You may end up needing some expensive specialist items.
This does not sound like a beginner's project.
-- Nemo
n
=A0etc.
=A0etc.
d o----+
| =A0CD4017B =A0 =A0 =A0 =A0 =A0|
=A0 =A0 =A0 =A0 =A0 |
=A0 | =A0 =A0| =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 |
---|>Clock =A0 =A0 =A0 =A0 =A0 =A0 |
=A0 =A0| In =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0|
Q0 =A0 Q1 =A0Q2 =A0Q3...|
---------------+
=A0 =A0 =A0 =A0 | =A0 =A0| =A0 =A0 =A0 =A0|
-----+ =A0 =A0V =A0 =A0 =A0 =A0etc
=A0 =A0 =A0 =A0 =A0 =A0 MOSFET
=A0 =A0 =A0 =A0 =A0 =A0 gate
=A0 =A0 =A0 =A0 =A0 |
=A0 =A0 =A0 =A0 =A0GND
gt -
Excellent!!! I wish I was building a Quad pyro detector. Phil, at wavelenghts above 1um or so, (where photodiodes turn on) are pyro's better at integrating the light intensity than photodiodes?
George H.
Nah, I'd rather have a PD any day, except when trying to do wide wavelength band pulse measurements. Pyros are easier to calibrate because their response is flatter.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
f) to detect the position of a
If you just want to center the laser, i.e. make equal amounts of light fall on the sensors, I'd use the I to V circuit shown at the bottom right of the data sheet. When the output off all the op amps is equal, then the beam is centered.
The time constant of the sensor is a factor of 1000 compared to your pulse. (I can't even comprehend femtoseconds in anything but the abstract.). I doubt you can sense individual hits.
to detect the position of a
The pulses themselves would look like delta-functions to any circuit whatsoever, but each one deposits some finite amount of energy, and it's the sensor's impulse response that matters. AIUI, the OP's question was how to measure the energy of each individual pulse, so as to be able to normalize some other measurement. This is a common thing to want to do, and as long as the beam profile doesn't change with the pulse energy (which it does in some systems, as I once learned to my cost) it works pretty well. The quad cell approach allows him to use the same sensor for aiming and energy normalization, which is cute.
Each femtosecond pulse will make a step change in the sensor temperature, but it won't be that large. The problem is not the 100 fs or whatever, it's the 1 kHz rep rate--the sensor has a 5 ms time constant, so the actual sensor temperature is a running average over many previous pulses. If the rep rate were 10 Hz instead, the sensor temperature would have returned to a steady state between pulses, and the problem would be easier.
With a 5 ms time constant, the sensor temperature signal is a little irregular sawtooth waveform sitting on top of a big drifty pedestal. The falling slope of the sawtooth is proportional to the instantaneous sensor temperature, so it's important to do a correlated double sampling measurement with the sampling pulses as close as possible to the step.
There will be some thermal diffusion inside the pyroelectric, which will round off the rising edge of the pulse, so you there's a limit to how close you can go. The little CMOS circuit I posted will do a pretty good job, and a digital delay generator can do a better one.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
Hello,
Thank you all for commenting on my issue! And my apologize for the delay replying, my newsgroup reader played me a bad trick!
Phil Hobbs scripsit: > The quad cell approach allows him to use the same sensor > for aiming and energy normalization, which is cute. Exactly! :-)
1E8 Ohms, so that the time constant should be of the order of R*C = 1E8I lack time to develop the DIY part right now, but I will come back to it ASAP, and might then come back here for further explanations.
Thank you very much for your precious and detailled replies, they are highly valuable!
Best regards,
Mathieu
Regarding this issue, I would like to mention, it is not necessary to be able to resolve each pulse. We would be happy to integrate over, say, a second (1000 pulses) since anyway the motor used for the beam stabilization can not offer instantaneous response... But what is the best option to do this while keeping the highest signal-to-noise ratio?
Well, that's a completely different problem. At that point you can reset the pyro after N pulses instead of 1.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
es
eYup... and now you will need to worry about the thermal time constant. I also read it as 5 Hz from your posted data sheet. (I don't know where Phil got the 5ms from?) You certainly don't want to integrate for 1000 pulses... 1 second. but something closer to the thermal time constant... 30 ms or so... OK if you just want the biggest signal you can integrate for a few thermal time constants. Say 100ms.
George H.
Ok, this is now clear as for beam stabilization. Thank you for the comments!
What if in the future I want to improve the electronic to allow for pulse to pulse normalization? This 5nA limit really worries me since I fear my signal will be buried in noise. In there any inexpensive improvement that could be implemented? Would it be required at all?
Thank you for your support!
Mat
omes
rld
e .be
ats!
Yeah you may want pulse to pulse normalization. How stable is the OPA? and how big is the signal you are looking for? I wouldn't worry too much about the 5nA. Isn't there a 100 Meg ohm resistor as part of the detection circuit. (Or some value like that.) Unless I'm making a math mistake that will give you a 500mV pulse height. piece of cake! (Though you will have be careful to keep stray capacitance down to a minimum.)
George H.
From the 10 µJ/pulse and 0.48 µA/W values given above, the pyroelectric detector is generating 4.8 pC per pulse or 3E7 electrons per pulse. Charge sensitive pre-amps used for x-ray and gamma ray spectroscopy have equivalent noise charge of a few hundred electrons. In fact Amptek is advertising silicon drift x-ray detectors (essentially a silicon diode with a charge sensitive pre-amp) with a full width at half maximum resolution of 136 eV in the curious units used in that community. Since Si requires an average of about 3.7 eV per e-h pair, this corresponds to a full width at half maximum of charge collected of 37 electrons or a noise equivalent charge of less than 16 electrons. Look for application notes from Amptek and Ortec. Amptek used to and may still sell charge sensitive pre-amps with external input FETs so you could mount them close to your pyroelectric to minimize capacitance.
If you have a nuclear physics group handy, they might have suitable charge sensitive pre-amps for measuring a few million electrons that they no longer use because they are too noisy.
Bret Cannon
comes
sensor
hold
value
!...
to=20
say, a
pulse=20
my=20
that=20
pyroelectric=20
Charge=20
equivalent=20
=20
eV in=20
of=20
maximum=20
=20
external=20
minimize=20
charge=20
longer=20
There is also the possibility of getting charge amps for small piezoelectric accelerometers from the vibration community. Scales like 3 pC/V were/are readily available.
There's a teensy capacitance problem with pyros, though. Charge dispensing is really the right answer IME.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
Hello,
I actually wonders: under the action of the laser beam one gets a charge output. These charges can then be used in an electronic circuit: how comes then that when the sensor is "relaxing" one does not see a negative signal?
Ok, I have a question which remain here: a MOSFET off-resistance should be of the order of a few Ohms, say about 50 Ohms. That means two things:
- the electrical pulse width will be very short.
- with my 5nA signal, I will only get a 50*4E-9=2.5E-7V, which is very small! And I have the feeling that amplifying the signal thereafter at 1kHz by a factor ~1E7 is not that trivial with very short pulses. Am I wrong in my reasoning?
Synchronize the data acq system so that
This is otherwise feasible in my setup! :-D
This is another extremely useful information! I could apply it somewhere else successfully! Thank you very much!
Best regards, Mat'
You do. If the temperature returns to its original value, the net integrated charge is zero. That's one of the inconvenient things about pyros. That's why you want to turn on the integrator just before the pulse arrives and turn it off just afterwards--it avoids error from the cool-down slope and gives you the results from a single pulse.
The pulse from the pyro isn't that short. The point is to integrate the charge on the feedback capacitor of an op amp, and digitize that. You can reset it whenever you like, but it's important to measure the voltage before and after reset and subtract.
Cheers. Use it in good health.
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.