Oscilloscope for Transient Measurements

[swami.me]

Hi,
    I am material scientist working on solar cells. I am rather a noob at oscilloscopes.  We are trying to build a transient photocurrent setup. In this setup we use a pulsed laser (1 ns pulse width) and measure the transient phocurrent out (time scale is less 10 ns). Repetition rate will be 1 Hz to 20 Hz but we need data points for every nanosecond. We are trying to buy a used oscilloscope for this purpose. Can you suggest me which would be a better choice?

1.Tektronix  TDS694C (They advertise for transient event capture)
2.Agilent AT-DSO7104A
3.Tektronix TDS-7254B

Please help me on this issue.

Swami

[free_electron]

A 1 ns pulse cannot be captured with a 1ghz scope ! You need a scope with at least 4ghz analog bandwidth to get anything slightly meaningful out of it. You will need to sample in the 16gs/s range... None of those scopes will work. The 694 comes close. The 7254 is $$$$$ . For both of them you will need probes... And that is going to be more $$$$, especially at those speeds.

With a 4 ghz scope you can still only see a risetime of 250ps... Time to switch technologies:

if your pulse is repetitive you can get away with a sampling scope. Look for a used 11801 from tek. They go cheap , but make sure it comes with the sampling head ! A lot of them are sold as mainframes only ( without the samplers and probes ) and that is useless.

Keep in mind that these sampling scopes ONLY work on repetitive signals ! You cannot do one shot events with them !

[AndyC_772]

When it comes to probes, it's highly unlikely that a used scope will come with anything suitable (if, indeed, it comes with probes at all, which is quite rare).

A standard (passive) oscilloscope probe just won't work because of the bandwidth - even the best top out at around 500 MHz - so you'll need to invest in an active probe which is compatible with your scope.

Most important of all is, when you do have suitable equipment, learning to use it correctly. Probing technique is of critical importance at high speeds, so you'll need to find someone who is familiar with the issues involved and who can show you how to get accurate results. Without this expertise, the measurements you'll make will be unpredictable and/or unrepresentative.

[EEVblog]

The others are correct.
Accurately measuring such a signal is very tricky and expensive business.
The $140K scope with it's $12K probe I showed in some previous videos would not be overkill for this task.
If you just want to "see" that a pulse is there, it's not that bad. But if you really want to accurately measure a 1ns pulse (with maybe 100-200ps rise times?), you need to spend the big bucks.
Thankfully, these scopes can be hired short term, so you don't have to buy them outright. The big companies will give a loaner for a while to see if it's suitable too.





Dave.

[AndyC_772]

Actually I'm quite interested in how you drive a laser to achieve a 1ns pulse in the first place. Are you able to offer any hints on the basic technique?

[IanB]

If I read the OP correctly, he wants to measure a pulse up to 10 ns wide with samples every 1 ns. Does this reduce the bandwidth requirements to something more accessible?

[AndyC_772]

Depends on how accurate the measurement needs to be, I guess. If the signal height each and every nanosecond is required, useful information, then the scope needs a rise time which is much less than 1ns.

For what it's worth, I'm not sure I could recommend the Tek 694C anyway; that generation of scopes is full of electrolytic caps which leak and corrode through PCB traces. They're a swine to fix when they do break, and Tek quote silly money for replacement boards. (My slightly later TDS754D doesn't have these, so should hopefully last a bit longer. Perhaps see if you can get a 794D instead if you think it's suitable).

[Balaur]

Actually, lasers with pulse duration widths in the fs/ps are quite usual for some applications.

One of recurring subjects in my line of work consists in the analysis of very short pulses (tens-hundreds of picoseconds) in rather deep logic (i.e. combinatorial gates in < 90nm microelectronic devices). Every new PhD student in this field immediately thinks about how wonderful an oscilloscope would be and why the previous researchers never considered this. It's quite a rite of passage, really.

There are some tricks that we've used for better results:
- test structures with various configurations but easy to capture logic output (1/0). A mathematical analysis of the captured events provides insightful data about the physical phenomenon
- test structures able to measure the duration of the logic event. Captured data can allow the computation of some parameters of the mathematical event model
- a fast analog memory that captures a few thousand points of the event with a fast (ps/tens of ps) sampling rate.

Usually, all these structures are on the chip itself, added during the design phase. This is not at all helpful for your stated application.

We have also experimented with building sensor chips that can be connected to the Device Under Test. While still useful, a lot of post-processing is required for getting back to the original physical event.

I realize that this discussion is not directly useful to you, but I wanted to make you aware about the inherent difficulty of directly observing the analog event. You may be more productive using some low-cost hardware tricks and software/mathematical analysis.
Maybe you can consider some signal conditioning/digitization circuitry or a series of voltage comparators to transform the pulse into logical events. Possibly a Logic Analyzer with adjustable-threshold inputs or a Pulse height analyzer or a Pulse Counter can be relevant to your application?

Cheers,
Dan

[free_electron]

Your best option is that 11801 or a csa8000 with the appropriate head.
An 11801 will do 50GHz signals with the right head. And you can have these machines for a few thousand dollar on ebay. 3k will get you one.

These machines sample in the hundred kilosample per second. They have a very precise trigger detector and then inject a random ( but known) offset to this trigger and take a sample ( actually a burst of samples ) Every triger generates a few datapoints. By overlaying all the triggers you get to see a restituted signal. This technique works only on repetitive signals as you need hundereds of periods of the signal to build the image on the screen.

[swami.me]

Thanks a lot guys for your input.

Andy - We are using a dye laser which is pumped by a Nitrogen laser. As Balaur said there a variety of lasers in the market with femtoseconds and pico seconds resolution.

I just wanted to make clear that we are not going to analyze the pulse rise and fall times. We pulse a photodetector and see the quality of photodetector by measuring the transient current decay after the pulse is turned off. The photodetector has decay time from 10 - 100 nanoseconds which is of interest to us. If I see the literature and journal publications, researchers have used

1. DSO90254A - http://arxiv.org/pdf/1207.3341v2.pdf (See Experimental)
2.Tektronix 1 Ghz Scope - http://arxiv.org/pdf/0907.1401.pdf (First Page 3rd Paragraph)
3. http://pubs.acs.org/doi/suppl/10.1021/jp109697w/suppl_file/jp109697w_si_001.pdf - This is the best one where they explain the technique.


I am really sorry to bother you guys with too much information. I find these discussions very interesting. Again thank you all for the advise.

Swami

[swami.me]

Also if you see the third one I posted (i.e http://pubs.acs.org/doi/suppl/10.1021/jp109697w/suppl_file/jp109697w_si_001.pdf) they have used a cheap scope (TDS 3032). Our experimental setup is almost identical to theirs.

Please let me know. Also do I need to get active or passive probes. I have no idea

Swami

[IanB]

If your pulse width is 10 ns, that would give you a basic frequency of 100 MHz. If you want to see any detail in that pulse, for example to measure its amplitude and duration with any precision, then you will need a scope input bandwidth somewhat higher than 100 MHz, for example 1 GHz.

While I am no expert, my reading suggests that 1 GHz is getting beyond the capabilities of a standard passive probe design and therefore an active probe may be indicated. However, I have read an article somewhere that describes how to make a low cost passive probe with a much higher frequency response. Such a thing may meet your needs. If I remember the link I will post it.

[IanB]

Here is a Tektronix tutorial document on probes:

http://circuitslab.case.edu/manuals/Probe_Fundamentals-_Tektronix.pdf

[AndyC_772]

As Balaur said there a variety of lasers in the market with femtoseconds and pico seconds resolution.

I know  but there are still gaps in my knowledge about how such devices are driven with such short pulse durations. Do they use an avalanche device? High speed logic? An inductive spike? Alien technology?

I just wanted to make clear that we are not going to analyze the pulse rise and fall times. We pulse a photodetector and see the quality of photodetector by measuring the transient current decay after the pulse is turned off. The photodetector has decay time from 10 - 100 nanoseconds which is of interest to us.

What you describe sounds exactly as though you're measuring the fall time of the detector's output pulse. You just need to ensure that the rise / fall time of your scope is faster than that of the signal you're measuring. Rise time = 0.35 * bandwidth, roughly.

You'll need active probes if the bandwidth you require is greater than a passive probe can do. Passive probes are available up to about 500 MHz bandwidth and are much cheaper than their active counterparts.

What's the output impedance of your photodetector? If it's very high, an active probe might present a lighter load and would give more representative results.
[/quote]

[IanB]

Here is a note with a description of how to make a high bandwidth resistive input passive probe:

http://www.signalintegrity.com/Pubs/straight/probes.htm

The downside is that it places a higher load on the circuit under test than other probe designs. This would be a problem if your photodetector cannot provide that amount of current.

[prenato]

There,s also Doug Smith's 1 GHz probe:

http://emcesd.com/1ghzprob.htm

It's a similar design to the one mentioned above , so only adequate if the circuit can tolerate the lowish impedance (these are sometimes called LoZ probes). Tektronix also sells some passive LoZ probes that are not too expensive (less than $100 on eBay) .

[kg4arn]

One of our forum members did some testing on the Howard Johnson 1K Zo probe.
I have built some myself and got similar results, but I have much less test capability than Janne does.

http://koti.mbnet.fi/jahonen/Electronics/DIY%201k%20probe/

[MDG]

Actually I'm quite interested in how you drive a laser to achieve a 1ns pulse in the first place. Are you able to offer any hints on the basic technique?

Regarding the question of how one achieves ns pulses, check out the discussion about the origin of the short pulses in the popular home-built nitrogen lasers  http://jossresearch.org/2011/05/16/your-diy-nitrogen-laser-is-not-a-blumlein/#Conclusion . The page also has links to the figures in the original 1974 Scientific American article, where you can see the clever design of the spark gap.

[Wim_L]

I just wanted to make clear that we are not going to analyze the pulse rise and fall times. We pulse a photodetector and see the quality of photodetector by measuring the transient current decay after the pulse is turned off. The photodetector has decay time from 10 - 100 nanoseconds which is of interest to us. If I see the literature and journal publications, researchers have used

For this kind of duration, and a smooth exponential decay, any of the scopes you were thinking of would be suitable. Risetime and peak amplitude will not be detected correctly, but measuring the decay of that long tail (>10 ns) is easily within reach of scopes with a few hundred MHz of bandwidth and 1GSa/s.

For connections, if the device is high impedance, a fast enough preamplifier (an active probe for example) connected right at the sample, driving a scope 50 ohm input (note, check that the scope has 50 ohm inputs or plugins). If the device itself can source lots of current without affecting its behaviour, 50 ohm coax connected directly to it will be cheaper (though one might legitimately ask how one you be sure this won't affect the device without also having tested with a preamplifier).

[ejeffrey]

I know  but there are still gaps in my knowledge about how such devices are driven with such short pulse durations. Do they use an avalanche device? High speed logic? An inductive spike? Alien technology?

Generally when you want pulses from 10 ns down to 30 fs you use the laser physics to get short pulses, very little to do with the driver circuit (although you may still have to pulse the drive, it isn't responsible for the pulse length).  Mode locked, Q-switched, and gain-switched lasers all take advantage of the physics of the laser cavity and gain medium to generate short pulses.  http://www.rp-photonics.com/index.html has lots of information about lasers.  Semiconductor lasers can be gain switched, or they can be driven directly by a fast pulse.  If you only want pulses, avalanche pulsers are a good way to go.  More often you want to modulate the waveform to carry data.  In that case, a high speed DAC or digital logic signal will do it.

@swami.me

So you only care about the fall time of the electrical signal, which is much longer than the optical pulse?  The minimum rise/fall time of a scope is approximately 0.35 / BW.  Thus a 100 MHz scope will have a minimum rise/fall time of about 3.5 nanoseconds.  That is already faster than the fastest signal (10 ns) you expect, but I would still go for a bit faster for two reasons.  First, the finite rise time of the scope will cause an error in your measurement of the decay time.  You can calculate the effect and compensate for it, but always start with the best data possible.  Second, the finite rise-time blurs out the leading edge also.  That means you can't start looking at your data for at least one rise time after the trigger.  For your shortest pulses (10 nanoseconds), you will be left with a very small amount of signal to fit.  Even a 200 MHz scope will be a big improvement, and any of the ones you suggested will be more than adequate in the bandwidth department.

It looks like your signal is only a few mV.  This presents its own set of problems.  You will want a scope with low input noise, good signal averaging, and a stable trigger.  As long as you have a good clean trigger from the laser that should be fine.  More troubling, a 10x probe will probably lose too much signal even with signal averaging.  You can see a 600 uV pulse on a good scope with averaging, but it is hard to get data from it.  You may want a FET probe or a dedicated high speed amplifier at the detector.

[PA4TIM]

One of our forum members did some testing on the Howard Johnson 1K Zo probe.
I have built some myself and got similar results, but I have much less test capability than Janne does.

http://koti.mbnet.fi/jahonen/Electronics/DIY%201k%20probe/

Nice test setup. You wrote the carbon resistor has 0.5 pF. That could be true. But carbon composite will have much more, some metal film could also have a little bit more ( distributed capacitance, and it can have inductance)

My pobetest  http://www.pa4tim.nl/?p=3059

But the capacitance between your resistor tip and the groundlink is much more as 0.5 pF
You can lower that by removing the shrinktube from the resistor ( that is a dielectrium)
You can calculate or measure it.

What I ask myself, If you use coax and no termination at the end of the probe. ( but i have not read all about this probe so maybe you used termination, it lookes like a normal high Z probe) The coax will be a huge capacitance. I work a lot with LCR bridges and if I use coax to extend the meaurement and not use a guarded 3 point construction the coax adds around 80 pF of capacitance ( depends on the cable lenght)

[ejeffrey]

Low-Z probes like this are designed to be used with scopes that have internal 50 ohm termination, or with an external terminator attached to the scope.

[LeoUCDavis]

I would assume that these sub-sampling scope wouldn't work very well on signals with high jitter?

Your best option is that 11801 or a csa8000 with the appropriate head.
An 11801 will do 50GHz signals with the right head. And you can have these machines for a few thousand dollar on ebay. 3k will get you one.

These machines sample in the hundred kilosample per second. They have a very precise trigger detector and then inject a random ( but known) offset to this trigger and take a sample ( actually a burst of samples ) Every triger generates a few datapoints. By overlaying all the triggers you get to see a restituted signal. This technique works only on repetitive signals as you need hundereds of periods of the signal to build the image on the screen.

[tggzzz]

A standard (passive) oscilloscope probe just won't work because of the bandwidth - even the best top out at around 500 MHz - so you'll need to invest in an active probe which is compatible with your scope.

Not quite true, though I know why you say that.

Consider "low impedance" or "Z0" passive probes, e.g the 6GHz Keysight 54006A.

Most important of all is, when you do have suitable equipment, learning to use it correctly. Probing technique is of critical importance at high speeds, so you'll need to find someone who is familiar with the issues involved and who can show you how to get accurate results. Without this expertise, the measurements you'll make will be unpredictable and/or unrepresentative.

Very true.

It always surprises me that people don't realise the probe and scope are integral parts of the circuit under test.

[Precipice]

If you're strapped for cash, a Tek 7104 / R7103 has the bandwidth it looks like you need (>1GHz) , and even at your low rep rate, you'll get a visible trace in a normally lit room (there's an image intensifier plate on the front of the tube).
Sampling scopes have a lot of advantages, but the 7104 has some real charm (and some annoying 1970s features)
http://w140.com/tekwiki/wiki/7104
A brutal, brutal 'scope - built for exactly this sort of thing, and now available ridiculously cheaply.