Short pulser.

s a

But

Well the 74HC series logic I have failed to give me anything less than

10ns. I've got to order something faster. (Except for John L's circuit, I got ~2ns rise times, but didn't play too much with C values. (10pF)

How about this, (perhaps someone mentioned it already.)

I take a step and apply it to both an inverter and OR gate. Take the output from one and send it through a piece of coax. (1 foot per ns?) and then send both steps into an AND gate. (Or is there going to be too much ringing from the coax?)

George H.

After all this, I still have not seen any indication of the objective. What the heck are you trying to accomplish with this pulse? Sometimes knowing the "what" helps in designing the "how". Make that always...

Oh, Sorry. I just want to check the minimum pulse width spec. of a PIC counter. (10ns). I've got occasional ~5ns pulses and I'm wondering if the pic can see them or if I need to stretch them a bit.

(The pulses come from a pmt, I was mucking about with gate widths and counting pulses on the 'scope and counter at the same time. With long trace widths I worry the DSO misses some, with short widths the two always seemed to agree, but... the apparatus works at very low light levels and it's a bit of work to get a lot of light into it.... and yet not too much!

Here's a link since you asked.

Two slit interferece one photon at a time. Low light level is made with an interference filter in front of a light bulb, and a detector slit moves across the face of the pmt to detect the interference pattern.)

There must be a lot of people that could use a short pulse.

George H.

Old 7474 type flipflops has such low gain that the RC oneshot thing could hang in a linear half-flopped state. An RLC would fix that. Or better, add a schmitt gate in the reset loop. I don't know of HC is safe in this circuit. Old ECL wasn't, but the newer ones, EclipsLite and such, seem to be. Still, it's tricky with just an RC.

I see from your "some time ago" schematic that you have an ample long-term supply of stained, grey paper and dull pencils.

John

On a sunny day (Tue, 07 Dec 2010 08:21:28 -0800) it happened John Larkin wrote in :

That was a 74LS74. Yes I did throw out an ordner with about 500 of those works of art some time ago. The luck finder will get millions for it in the next century, You can buy one from me for only 1000$, I will sign it for you too :-) Think of it as a gift to your grand-children.

Anyway, that circuit worked OK for 30 years in a floppy disk interface. Zero error super circuit.

Eh? So two monostables, one at 50 ns, one at 51 ns, and the NAND output is a single nanosecond. The idea is to use a variable capacitor (small trimcap) to make one of the monostables a 50 to 80 ns variable, and the other a 50 ns fixed value, and then the difference is knob- tunable to whatever value you want.

It doesn't matter that there's logic margin and time delays in the gate, only the slew rate limits the pulse width (because shortest pulses aren't full height).

I don't know if I can dig up any monostables or not. Perhaps some old 74LS123's

Do you have any suggestions for newer one-shots? (Might as well add some to a parts order.)

George H.

74HC123? 74LV123? ;-)

an

Heee hee hee ... well I would have searched there first.

George H.

Try it. The rule of thumb is that if the transmission line is longer than the rise time then you have to pay attention to reflections.

My straw man would be to series terminate the coax line.

--
These are my opinions, not necessarily my employer's.  I hate spam.

The 'LS123 is probably going to have some shot noise-related jitter; Digikey lists 74VHC123 parts that might work well, but not now in stock. TC7WH123 is in stock, but lacks the ".not.Q" output.

In any case, I'd advise to use two chips instead of the two elements of the '123 dual, so the power supply pins can be bypassed independently (might be important when the edges get close).

No sign of 74F or other more snazzy families with any such part numbers; looks like the low end is served by '555, and the high end has been largely replaced with synchronous logic.

It might pay to capacitor-couple to an ECL gate, to get fastest output pulses.

"My straw man would be to series terminate the coax line."

Excellent, I never thought of that. mostly I'm brute force, (ie stupid, but can get job done.) I was just going to try hanging smaller and smaller resistors at the end in parallel termination (R to ground). I figured I would gang up on it with a bunch of inverters in parallel.

George H.

e

?)

Coaxial cable mostly needs to be terminated with a 50R or a 75R resistor - RG62A/U is listed as 93R. TTL doesn't like driving impedances that low. A resistive divider can get you there, at a cost in signal amplitude.

ECL in - all its flavours - uses NPN-emitter-follower-buffered outputs which can drive 50R. If you bias the ECL between 0V and -5.2V or -4.5V or whatever (see the data sheet of the stuff you are using) and return

50R the resistor to -2V, you can use this resistor to terminate the far end of your coaxial cable. and if you short the output to ground, it won't blow up.

PECL is just ECL biased between TTL power rails - +5V and 0V. It can still drive coax, but the 50R resistor has to be returned to +3V. An

82R + 120R resistive divider between 5V and 0V is close enough to 50R to 3V (48.7R to 2.97V). If you short the output to ground, it blows up. 100K ECL is specified for use with a 4.5V power rail, but - IIRR - only gets 11% warmer if you drive it from 5V, which isn't usually problem in the lab.

-- Bill Sloman, Nijmegen

Now, you've done it!!! I went up into the attic and brought down my metrology soapbox.

Don't take this personally, but I'm gonna talk about idiots like me. We think we're smart enough to end-run the laws of physics and do hard stuff the "easy way".

I do it all the time. And I've been burned most of the time. On this particular topic, I do have credentials. I spent most of the '70's designing pulse generators and counters and scope calibration stuff.

I can't tell which counter and pulse generator, but I can give some rules of thumb. If you buy equipment from the big guys, like Agilent, you can expect that there are specs, they have the devices required to determine if their products meet specs and a production line that can make it to spec. They have a reputation for making good stuff. And it worked correctly under the EXACT test configuration defined in the cal procedure when it left the factory...40 years ago. For the rest of the suppliers, you have no idea whether the stuff does what they say. With hobby stuff made in a garage and sold on ebay, you can be certain that the performance is iffy near the spec limits.

It's all about price/performance. You can bet that the system is optimized to give the best performance under the specified test condition, possibly at the expense of performance you'd expect in different conditions from the spec.

If you ignore all the significant second order effects, a counter is a threshold detector behind a band-limited nonlinear filter. There are frequency dependencies, slew rate issues, resonances...on and on. At the limit of performance, the ability of a counter to count is critically dependent on input frequency, pulse width, slew rate, dc level and on and on. And we haven't even begun to discuss all the issues with getting the signal from the source into the counter. If you test it under a limited number of amplitudes, frequencies, pulse widths...about all you'll do is get a false sense that the counter might work for what you're putting into it.

Bottom line is that people who need short pulses need to be able to vary all the parameters and count on the output being what they dialed in on the knobs. It's easy to cobble together something that puts out pulses. Pulses of known and variable parameters into unknown loads is a much harder problem. And you want the parameters known at the LOAD END of the connecting cable. The devil is in the details.

Metrology is HARD.

OK, I'll put the soapbox away...for now...

Getting back to your problem. If you're measuring narrow, low rep-rate pulses, you should probably put your effort into the counter, not the pulse generator. For a ripple counter, it's all about the input amplifier and first stage. Build yourself a blazing fast divide by two that can easily work with your inputs and feed that to your existing counter. Remember that the first stage has to count reliably at a frequency that's 1/the time between adjacent inputs. This time may be WAY shorter than implied by the average frequency.

Any decent oscilloscope will have a trigger output. You can use that to precondition signals into a counter.

As for the resolution/trace length issue, Many digital scopes have two useful modes. Glitch capture that stretches the input to make sure it shows up on the trace. High-res mode that samples at high rate and compresses that into a longer sample when you have a trace memory that's too short to hold all the fast samples.

I skimmed the site you linked. The frequency is the average rate of photon arrival. What's the probability that two photons arrive very close together? A plot of the distribution of time difference between adjacent (in arrival time) photons would be interesting. Depending on the distribution, maybe missing those doesn't matter...unless the experiment is making unwarranted assumptions... Maybe the spatial distribution (interference pattern) is caused by those photons that really did get thru at approximately the same time. Counting the AVERAGE frequency may be misleading. What if the deflection is not about the slit at all, but the interaction with all the other photons that got deflected off the slit and didn't make it thru? I'm basically uncomfortable with the explanations why there's only one photon at a time and if that's the relevant conclusion.

I've got two speed boats at relativistic velocities going under a bridge. They're bouncing off the shock waves of each other. One makes it, one hits the concrete pillar. I'm having a hard time believing that the shock wave from the crash doesn't affect the trajectory of the boat that made it thru. Or that the trajectory wasn't determined by interactions BEFORE the slot. The trajectory of the lucky boat is not the same as it would have been without the other boat. Still one photon at a time (at the detector), but possibly different experimental conclusions.

I shoulda paid more attention in physics class.

a

.)

n.

lly

Hmm that's a good idea. I should bring my Rigol 'scope in and see if it can see the pulses. It's got a lot longer memory depth than the TEK. I've neverheard of glitch cacpture, but I did have the 'scope set to peak detect.

"I'm basically uncomfortable

Yeah join the club! I think lots of physicists are uncomfortable with it... but that's the way nature seems to work. You do experiments and look at the results.

The photon arrival time is described by Poisson statistics. Random events. Poisson statistics are a bit weird. The most probable time between two events is zero, with a probability that decays as an exponential, the 1/e time is the average count rate.

Now in the experiment you can start with a high light level and then slowly reduce it. You see no change in the character of the two slit pattern. When you are done with the light bulb set very dim. You turn off the high voltage to the PMT, and open up the apparatus, You take out the detector slit, and the double slit. Now you close it back up and measure a new much higher count rate. Knowing the PMT=92s quantum efficiency you can calculate the number of photons hitting the pmt. And that is how many are in the apparatus.

When you do this you get count rates of ~100kHz, with ~10% efficiency that=92s about 1 million photons per second. The length of the apparatus is a bit more than one meter. What=92s the average time between photons? What=92s the average =91distance=92 between photons? (about 300 meters) Let=92s say the count rate is 300kHz so that the average distance between photons is ~100 meters. That means (knowing Poisson stats.) that in one out of every 100 counts there were two photons in the apparatus together. But if the interference pattern only comes when there are two photons, then it should change as the count rate is changed. (A weak two slit interference pattern on top of the one slit diffraction pattern.) But that is not observed. QM is weird.

Look up the Feynman lectures on Physics at your local library. The first volume has an very nice discussion of the two slit experiment. There is not any refelction or 'bouncing' going on. There is diffraction, from the size of the slits (and the wavelength of the light), and interference becasue there are 'two ways' for an event to happen. (The event being the counting of a photon.)

This is the fundamental Quantum paradox. I don't think it helps to try and understand it.

George H.

The wave behaviour is relatively intuitive, as long as you don't have a picture of particles as tiny billiard balls. The really deeply mysterious part is the identification of the squared modulus of the wave function with the probability density for finding the particle.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

Some famous person said that if quantum mechanics doesn't scare you, you haven't studied it enough.

One photon does interfere with itself in the double-slit experiment, even when it's provable that it goes through only one slit or the other.

John

On a sunny day (Thu, 09 Dec 2010 15:06:19 -0800) it happened John Larkin wrote in :

There has been some talks in sci.physics that in the 2 slit experiment, it was repeated with buckyballs, and the buckyballs showed an interference pattern, there could be a guide wave happening in the ether. That buckyball experiment is particularly mind challenging as it is a macro size object. So I now adhere to the fact that there is a wave structure running in front of the buckyballs that causes displacement of those AFTER they pass the slits. Else QED would be really scary, it is only scary because it is probabilistic nonsense, an approximation of reality, a model that replaces logic with probabilities, and it will have to be superseded by some ether theory,. You can think of those waves as the bow waves of a ship perhaps, and the buckyball as the ship. It seems De Broglie in his later years suggested something like this. Give it some more hundreds of years, flat earth will become round.

pattern,

size object.

the buckyballs

nonsense,

and it will have to

buckyball as the ship.

A long-baseline, image-splitter interferometer is even scarier. A photon hits a half-silvered mirror and obviously either passes through or is deflected by 90 degrees; you can measure that. It takes one path or the other. The two paths are later merged, and you have interferance... even when the paths are kilometers apart. So the photon took both paths.

If you treat the photon as a distributed-in-space wave bundle, it is somehow ripped into parts and later reassembled, even though it is always a single photon of fixed energy.

Adding fast optical electro-optical gates within the interferometer makes things even worse.

John