Nanosecond pulse stretching 1

[matchhead]

I work for a small fiber-optics company. We build nanosecond pulsed lasers. The problem that I have been presented with is: if you have a nanosecond pulse from a photodiode, how can you determine its power, energy or even amplitude with a 10MSPS ADC? More specifically, how can I stretch and amplify a 1ns pulse to something greater than 1us while still preserving information about the original pulse like amplitude and duration?
-Paul

 

[itsmoked]

Don't use an A/D.

Buffer and amplify the signal then feed it to a magnitude detector and a timer or counter.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[matchhead]

So what you're saying is there is no reasonable way to do what I said?

 

[IRstuff]

You can, but it'll be cheaper to do it the conventional way.

How fast is the rep rate. If it's fast enough, you can still do sampling, but I would guess that a 10MSPS ADC has too low a bandwidth. I would think that you'd need something more like a 250MSPS ADC, but it would have to have at least 5 GHz bandwidth to even come close.

Why don't you just rent a scope?

TTFN

FAQ731-376

 

[matchhead]

The rep rate is less than 1 khz, but I'm passing the laser through a computer controlled monochrometer to gather the spectrum information of the laser, so I could watch an oscope screen and jot down a thousand samples of data, but I'd rather go for something controlled by a PIC.
Our company needs an instrument that we can keep for the express purpose of pulsed laser spectroscopy, and we can only afford to spend a thousand dollars. My specialty is embedded systems, so my experience with high frequency stuff is very limited.

 

[IRstuff]

Well, you may be SOL:

http://www.national.com/pf/AD/ADC08B3000.html

Cited part is a 3GSPS, 3GHz full power bandwidth ADC. You can get by, probably, with something like:

http://www.national.com/pf/AD/ADC08D500.html,

which has a 1.7GHz bandwidth at 500 MSPS.

As for power/energy, your existing meters should already be able to do one or the other.

TTFN

FAQ731-376

 

[matchhead]

The ADC09B3000 is listed as $4900... ouch, but thanks for the suggestions.

 

[richs]

Hiya-

I wonder if you could get one of the high speed comparator chips to have sufficient rise time to construct an analog sample and hold circuit?

I've been fighting firmware fires (not real fires Keith) all day and my head hurts, so I'm not going to think though this.

But, maybe some of the aforementioned high speed comparators, surface mount, with a well laid out/shielded PCB *MIGHT* be in order.

Since the rep rate is low (1KHz mentioned by the OP) A PIC can certianly be used for the control of the circuit, but I's sure watch my grounding, etc.

Right now, I wouldn't even hazard a guess as to a suitable dialectric for the cap at this point.

Maybe a chat with an RF or microwave buddy for the front end of this critter might be in order. Maybe going as far as having him design the front end out of microwave capable chip transistors.

It would sure eat up that $1000.00 budget pretty quick though. Best of luck and keep us posted!

Cheers,

Rich S.

 

[jimkirk]

If you need to get measurement information from a single pulse, then I suggest a comparator & integrator to determine pulse width, and an integrator on the output of your photodiode amplifier to get some energy measurement.

If you can use a series of pulses, you can reconstruct a pulse with a high speed sample and hold and vary the timing of each sample, (trigger on start of pulse, trigger + 10 pS, trigger + 20 pS, up to the width of your pulse...) digitize each sample fairly slowly (1 kHz rate), and analyze the results in software.

The timing circuit and sample and hold that are going to be operating at gigahertz frequencies are the difficult, expensive bits, and you don't want to cut corners there.

Another realtime approach is to have a small bank of comparators with various thresholds, each with an integrator to get a pulse width measurement.

Basically, you're making an AD converter with good amplitude accuracy but lousy resolution, and pretty good time resolution.

Develop some a priori information about the pulses, rent a high speed DSO for a month and make some measurements. Then you should be able to correlate the results from your ADC to some decent results.

 

[itsmoked]

Digitizing a 1ns pulse is non-trivial.

You essentially need to do a bunch of analog domain work before hitting the digital realm.

At these speeds you need excellent layout quality. I'm hard pressed to see you doing this for under $1k.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[Comcokid]

A few days after reading this post, something occurred to me. You may not need to stretch the pulse if you can read the energy in it - depending on your application. Go to

http://www.analog.com,

look under their "Log Amps/Detectors" catagory, and look at the various detector chips and the application notes. These go to as high as 10 GHz. Some of these are charcterized for their response to pulses as well (e.g. radar use). They have development boards for a reasonable price, and one might serve fine for your finished detector. Linear technology makes some log detectors as well. These could work if you PRF rate is sufficient.

Just a thought.

Otherwise, to measure the pulse width and amplitude in a manner sufficient for a processor without extreme cost, you will need to convert it using an equivalent time detector method. Most TDR-type position and fluid level sensors use equivalent time detection means that take pico-second pulses and equivalent-time detect them over many samples such that the standard timer-counter input of a small micro can measure the time-converted pulse width. Tektronix does this with many of their super-fast scopes. The 11801 series could display 20 GHz signals, yet the equivalent-sample time base in the scope is 250 kHz (basically the 20 GHz signal is sampled over many repeated periods, and is reproduced exactly in shape as a 250 kHz waveform on the screen).

 

[bythomas]

Normally, the energy of a single laser pulse is proportional to the charge of the photodiode pulse. We usually use a charge amplifier to make a pulse with an amplitude proportional to the charge of the original pulse but decays at microseconds time constant. Digitize the output of the charge amplifier with your slow ADC and process the waveform in your software. One kHz should be very easy. A low-cost charge amplifier can be found here:

http://www.cremat.com/charge_amplifier.htm.

 

[DHambley]

Isn't the output light energy what you eventually want to measure and control? This is accomplished by light power detectors. When I designed pulsed-laser drivers, this was the power of concern. The power dissipated in the driver circuitry was only needed to size the heatsinks and such.