Maker Pro
Maker Pro

Create controlled 50v pulse

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
Getting it perfectly in focus is tricky as I have an extremely shallow depth of field to work with.


A solution to this is to manually focus when the blade is not moving, then arrange a trigger to fire the flash when the blade is in that position.

Typically you'll use an optical trigger that uses an infrared beam to detect the blade (positioned well out of shot).

In a darkened room you set the fan spinning, open the shutter, arm (and then automatically fire) the flash, before finally closing the shutter.
 
I do all of the above but with a full frame Canon EOS 6D and a fast lens up close the plane of focus is probably a millimeter or 2 deep and the blades aren't all in the same plane.
Anyway with continued development with the picaxe firmware I made an adjustable PWM output with a pulse width/frequency ratio much greater than the built in PWM so I can freeze the fan to the naked eye.
I was even able to take a video without strobing or banding with magic lantern on my 6D by fine tuning the framerate and recording RAW video.
 
When I say I do all of the above that's excluding the optical trigger.
Nice pictures, I should get a macro lens.

On the subject of microcontrollers and programming, I started experimenting with making a boost converter regulator out of the pulse generator circuit and in turn optimizing the code(which I'm pretty good at) of the 8x2 dot matrix OLED display I'm using as it uses a Picaxe which receives commands over RS-232 from another Pixaxe.

The display usually works up to 4800 baud which is rather slow so by sacrificing some functionality of the display which I dont need I was able to get the RS-232 interface up to the fastest a Picaxe can do, 38400 baud and at that speed the transfer of one 8 byte line takes a relatively miniscule 3ms!

Oh and the voltage regulation is good but a bit bouncy and rather coarse at low load but I'm sure I can improve it further.
 
I prefer Sigma lenses of which I have 3 because one gets more for ones money and the sharpness and chromatic aberration performance are amazing from comparisons I've seen.
 
I got some fast photo diodes to accurately measure the rise and fall times of the led and finally got around to getting a reading and its faster than I thought:
Image1.png
That's a 3.7μs 10-90% rise time and a 1.5μs 90-10% fall time.
 

hevans1944

Hop - AC8NS
That's a 3.7μs 10-90% rise time and a 1.5μs 90-10% fall time.
That looks like a nice, fast, PIN photo-diode you found. Specs say it is capable of nanosecond response times, which are about three orders of magnitude faster than the pulse rise and fall times you measured with it.

It is possible that the rise time was increased by parasitic inductance between LED and driver. Note the exponential rise in light output. The fall time may be also have been increased by parasitic capacitance between LED and driver after an initial very fast decay followed by an exponential decrease in light output. But this looks fairly typical for an LED directly illuminating a photo-diode. To observe a faster response (if it existed) would probably require either a photo-multiplier tube or a PIN photo-diode detector operating in avalanche mode. For your stop-motion photography, the response you posted above appears to be more than adequate.

I have noticed a delay in the light decay output from some so-called "white light" LEDs that I believe is caused either by slow power supply discharge rates, or by delayed fluorescence decay in the phosphor used to convert short-wave blue or ultra-violet LED excitation to visible "white" light. I have noticed this only in commercial high-brightness LED lamps operating from 120 VAC line voltage. The delayed "off" phenomenon is easily observed when it occurs, but it may not occur with the high-wattage LEDs you are using. Clearly it didn't happen when you made the oscilloscope image shown in post #47. Good job!
 
Last edited:
Thanks, I had to get another breadboard because the MOSFET switching causes so much kickback I couldn't get a clean measurement.
You can still see a small spike on the graph there with a completely separate breadboard and bench supply for the photo-diode, I increased the resolution enhancement from 8 to 10 bit to clean it up a bit(or 2).

The wiring is far from optimal, relying on the output capacitance of the bench supply, I could try running the positive wire to the breadboard where I can add a capacitor before the LED and make the wiring shorter.

I read that using a high impedance circuit with an op-amp is best for a fast response but then there is no low impedance path for the photo-diode's capacitance to charge and discharge and its extremely slow and is limited by the bandwidth of the op-amp, so I use a resistor to balance sensitivity and speed.

Edit: Using shorter wiring and a 100μF 100V polyester I was able to get about 2μS rise and 1.4μS fall, I also tried reducing the resistor in series with the photo-diode from 1K to 250Ω but that made no difference to the rise and fall times.
It looks like the inductance of my wiring is slowing it down more than anything else.
Image2.png
 
Last edited:

hevans1944

Hop - AC8NS
It looks like the inductance of my wiring is slowing it down more than anything else
I agree 100%.

It appears to me, from viewing your oscilloscope waveforms, that there are possibly two things going on here: (1) a delayed "turn on" caused by slow-rising current to the LED; and (2) a delayed "turn off" after an initial rapid decay in light output, perhaps caused by improper PIN diode signal conditioning. Please upload a schematic of how you are connecting the PIN photo-diode to the Pico Technology oscilloscope. I can also show you how we did it at UDRI back in the 1970s.

Your high-brightness LED is providing more than enough illumination to "turn on" the PIN diode, but the PIN diode needs to be reverse biased, almost but not quite to the point of avalanching, to achieve its fastest turn-off, measured in nanoseconds instead of microseconds. I would recommend about 30 to 40 volts of reverse bias for the PIN diode you are using. You will need a series resistor between the PIN photo diode and the reverse bias power supply to protect the photo diode from damage if exposed to continuous bright light. A small capacitor connected to ground on one end and to the junction of the photo diode and current-limiting resistor on the other end will allow the PIN photo-diode to deliver enough charge during pulse-mode operation without affecting the protection function.

The same is true for achieving a fast turn-on, but the PIN diode's fast response for increasing light intensity is being overwhelmed by the parasitic inductance in series with the LED that limits how fast current can increase in that device. No amount of capacitance added to the source terminal of the MOSFET is going to improve this, and may in fact increase the turn-on time because of capacitor ESR and its own parasitic inductance.

You need to minimize the series resistance and the series inductive reactance to the LED, hopefully without compromising its safe operating area. I know I stated in an earlier post that you needed a constant-current source for your LED, but for short duration pulses (microseconds) with low repetition rates (a few hertz to perhaps a few kilohertz), inductive and resistive parasitics are doing the current-limiting job. Note that the LED has reached "steady state" conditions after about four microseconds.

If we assume that reaching steady state required about five time-constants, then the time constant for "charging" the parasitic inductance is about 4/5 microseconds which is a result of the inductive reactance in series with the parasitic resistance of the wiring and the resistance of the forward conducting LED. You want to minimize the effect of that time constant, without destroying the LED, by hitting the LED as hard as possible when you turn the MOSFET on. You should use the Pico Scope to monitor the LED forward voltage as close as possible to the LED connections. Make another oscilloscope image of this configuration and post it here.

You are going to see the normal LED forward-bias conduction voltage drop superimposed on the IZ voltage drop of the connections, but there is no way that I know of that you can easily separate the two with one measurement. A vector voltage analyzer, perhaps?

What you can do is take steps to minimize the IR and IX drop in the connections to the LED by making those connections as short as possible and using wire gauge that is as large as possible. Then measure the remaining voltage drop by temporarily replacing the LED with a short-circuit. Do this and make another oscilloscope image with the "shorting wire" temporarily replacing the LED. The resulting waveforms should be essentially identical in terms of rise time, but the voltages will be different with and without the LED in the circuit. The point here is to identify and minimize the parasitic inductive reactance and parasitic resistance that is slowing down the turn-on time of the LED.

Another approach to this problem, if permanent fixed-width pulses are acceptable, is to use a delay line for energy storage. After charging the delay line, it is shorted with a fast-acting switch (MOSFET, SCR, or other) causing a square pulse to propagate down the delay line to the LED load. Very simple, very efficient. Depending on pulse width, the delay line can be a spool of coaxial cable or a few dozen lumped-component sections consisting of capacitors and inductors. The more sections there are, the better defined the "square" pulse will be, but the design of lumped-element delay lines is another topic entirely. Just "food for thought" offered here.
 
I replaced all the wiring with 22 AWG solid core and made it as short as physically possible, this is the photodiode circuit:
PDcircuit.png
thats a polymer capacitor
I fashioned a ground pin for the scope probe out of a piece of thick bare copper wire wrapped around the tip and crammed into the breadboard so the entire circuit is smaller than my thumbnail.

I'm using a TI CSD19531KCS MOSFET, the last 2 in stock, driven by a TI UCC27524 driver and controlled by a Picaxe with an OLED display connected.
I wrote a program for the picaxe to generate single or continuous pulses with an adjustable width and interval.

I was able to get a 10-90% rise time of about 1.3μS while the fall time remains at about 1.4μS and now the shape of the rising edge looks like the inverse of the falling edge:
Image4.png
NOTE: The measurements are not aligned as I could not take them at the same time with reasonable ease.
Green is the photodiode and red is the inverse voltage at the led, the 50v supply is isolated so I connected the ground wire to positive and the probe tip to the switching negative side, the ringing is just the ground clip wire.
The MOSFET takes about 70ms to discharge after the LED turns off, the pulse interval is 65.6ms, width 10.5μS.

This PDF on "Pulsed Over-Current Driving of Cree XLamp LEDs" states on page 8 that at 1KHz, "For duty cycles less than 10%, do not exceed more than 300% of the maximum rated current."
I'm using a far lower duty cycle than that and 3 x 3.6 = 10.8 amps.
 
Last edited:

hevans1944

Hop - AC8NS
It does appear that you are doing almost everything right, except I would opt for 10, 12, or 14 AWG bare copper wire for any circuit conducting LED current. 22 AWG is hook-up wire, not wire useful for large currents.

And I am confused by the slow rise and fall times of the green oscilloscope traces for PIN photo diode output when the LED excitation in red shows negligible rise and fall times! The PIN photo diode appears to be correctly reverse biased from the +20 VDC supply, although the 100 μF polymer electrolytic capacitor with the low ESR rating is doing nothing where it is connected, except smoothing the +20 V DC supply. A smaller-value capacitor needs to be connected from PIN diode cathode to ground, as its only function is to supply current for fast changes of the illumination. I presume the 250 Ω resistor provides sufficient current-limiting to protect the PIN photo diode from steady-state room illumination. You should be able to increase its value (for satety) without affecting the PIN photo diode response to pulsed illumination conditions.

Back in the day, we would connect the anode of the PIN photo diode directly to the inverting input of a fast op-amp and the cathode directly to a reverse-bias power supply output, usually the +15 VDC op-amp positive rail. A small, low-esr capacitor was connected across the diode. The non-inverting op-amp input was connected to ground.

A feedback resistor from op-amp output to its inverting input set the "gain," or conversion factor of illumination current to output voltage. For really low-light, or other sensitive situations, the feedback resistor could be several hundred megohms. Such large feedback resistors could really slow down circuit response caused by stray capacitance to ground, but you could probably get away with a few thousand ohms of feedback resistance and still have a few dozen millivolts of output.

If there were a large amount of ambient light, that has to be offset to allow the op-amp to remain within its output voltage operating range. This we did with either another PIN diode, shielded from ambient light, connected to the inverting input, but with a negative polarity bias supply, or with a potentiometer and voltage divider network. The dual PIN diode method has the advantage of providing temperature compensation for dark current variations, but the op-amp approach provides a linear output versus illumination response over at least six decades of input current... if you need that.

Your measurement method should be okay for building a strobe light based on a Cree power LED.

Any idea why there is such a huge difference is rise and fall times for your measurements?
 
I doubt the static resistance of the wire would make much difference to the rise time and just testing 2 bits of wire thick and thin with my Keysight U1733C LCR meter, the thicker wire of precisely the same length has a much higher inductance and impedance at 100KHz, also thicker wire would risk ripping the solder pads off the led.

I know the phosphor in white LEDs that turns UV light to visible wavelengths makes them slower as it acts as a sort of optical capacitor so maybe thats what we're seeing.

The datasheet for the photodiode shows 5ns with 50Ω at 20v so I assume its fine with 250Ω.

After trying IR remote controls to find a test light source thats as fast as possible I tried a laser I have and it outputs a high frequency where the rising edge goes from 500 to 800mV on the scope and the 10-90% rise time is, wait for it... 5ns!
So there's no doubt how fast this photodiode is.

Surely adding a capacitor in parallel with the photodiode in the op-amp configuration would act as a low-pass filter and slow down the response time?
 
Last edited:

hevans1944

Hop - AC8NS
I doubt the static resistance of the wire would make much difference to the rise time and just testing 2 bits of wire thick and thin with my Keysight U1733C LCR meter, the thicker wire of precisely the same length has a much higher inductance and impedance at 100KHz, also thicker wire would risk ripping the solder pads off the led. ...
Doubts should be based on good measurements and calculations, not speculation. Resistance and inductance have a huge effect on fast pulses, degrading both rise time and fall time. Do the math. I would also take a careful look at the LCR meter to make sure you are using it correctly because wire self-inductance will decrease with increasing wire diameter. Use this online calculator to get a reality check on what your LCR meter should measure using various wire diameters.

All connections to pads on the Cree lamps should be made with attention to mechanically minimizing or eliminating the tension in the connections. Pads are fragile, but with good assembly practices it should be possible to use wire sizes up to the diameter of the pad. This will require some creative thinking on your part.

I have successfully designed and manufactured device holders for a gallium arsenide photo-conductive semiconductor switch (PCSS) that carries thousands of amperes for a few nanoseconds with picosecond rise times. The PCSS was sliced from a 100mm (4-inch) GaAs wafer into devices typically 10mm wide by 25mm long, capable of holding off 50 kV or more when immersed in Fluorinert dielectric fluid. See image below of IBIS Photoconductive Semiconductor Switch developed by me and Rabi Bhattacharya under government contract with the Defense Threat Reduction Agency, the military arm of Homeland Security in the United States of America.

upload_2019-8-4_7-8-48.jpeg

Just to give you a clue as to what is involved, I used gold-plated phosphor-bronze finger contacts to make a spring-loaded pressure contact with a gold top layer on a thin multi-layer contact we vacuum deposited on each end of the device. The finger contacts were soldered to thick copper strips which were clamped on the other end to massive copper blocks. All this was necessary to minimize series resistance and inductance.

Further processing in our lab was required to control avalanche conduction in parallel conductive tracks between the contacts on each end. Once triggered by a near-infrared pulse of light from a laser diode, the PCSS continued conduction in avalanche (electron impact ionization) mode until the pulse-forming network it was switching was discharged.

I know the phosphor in white LEDs that turns UV light to visible wavelengths makes them slower as it acts as a sort of optical capacitor so maybe thats what we're seeing. ...
I believe, without proof, that this is what is going on.

... The datasheet for the photodiode shows 5ns with 50Ω at 20v so I assume its fine with 250Ω. ...
That will be the impedance the PIN photo diode photo-current "sees" and it appears to be okay for measurement purposes.

... After trying IR remote controls to find a test light source thats as fast as possible I tried a laser I have and it outputs a high frequency where the rising edge goes from 500 to 800mV on the scope and the 10-90% rise time is, wait for it... 5ns!
So there's no doubt how fast this photodiode is. ...
After reading the datasheet I had no doubts about your PIN photo diode speed. Actually, because the intrinsic layer (the "I" in PIN) is so thin, it is difficult to NOT make a fast PIN photo diode... assuming the semiconductor fab can make one in the first place. IIRC, this is done with an epitaxial deposition process, similar to that used to make heterojunction bi-polar transistors (HBTs), but with a very thin layer deposition of intrinsic semiconductor material. This used to be a tricky process and early PIN diodes were very expensive because of it.

... Surely adding a capacitor in parallel with the photodiode in the op-amp configuration would act as a low-pass filter and slow down the response time?
Not really because the PIN photo diode is operating as a current source and the capacitor provides charge for that purpose when the light intensity changes quickly. In the op-amp transimpedance mode, a small-valued capacitor is often placed across the feedback resistor to control bandwidth and reduce noise. In your circuit the reverse bias supply, current-limiting resistor, and PIN photo diode are all in series, connected to ground. Placing the capacitor across the PIN photo-diode provides an alternative source of current for the PIN photo diode when a fast change in light intensity occurs. See the schematic below from this application note:
upload_2019-8-4_7-50-14.png

Please download and read the application note.

Is this discussion helping you? Have you got your "cloud burning" Cree LED working to your satisfaction yet? Can you post some more pictures like you did before?

Hop - AC8NS
 
Last edited:
This is more about the science than the photography, I love learning these things and experimenting.

That's one mighty PCSS you have there, do they give out free samples?:p
I got 5 of the SFH 203 P so it doesn't matter much if I discover a limit.
I misunderstood your explanation of the circuit and where the capacitor goes.

I took another reading of the voltage rise at the LED but this time with 1 probe at the LED and the other at the breadboard end of the wires going to the LED, RED is at the LED and GREEN is at the breadboard:
Image5.png

I also took the same measurement without the big 100μF film capacitor:
Image6.png
 

hevans1944

Hop - AC8NS
That's one mighty PCSS you have there, do they give out free samples?:p
I wish we had. Maybe then someone besides DTRA would have bought one.
This device was discovered in the 1990s by scientists at Sandia National Laboratories (SNL), Albuquerque, New Mexico. SNL is co-located on Kirtland Air Force Base, home of the Air Force Weapons Laboratory. I worked at the Weapons Lab as a contractor for UDRI in the 1970s, as an technician supporting high-energy laser development for the Airborne Laser Laboratory airplane, which is now a moth-balled exhibit at the Air Force Museum in Dayton, Ohio. Last time I looked, the USAF is STILL trying to develop a weapons-grade laser to fit onto airplanes. But I didn't know diddly about SNL at the time, and there was no Internet World Wide Web to educate and inform me.

Turns out that SNL is into all sorts of stuff involving high energy. At one time they built and operated an nuclear EMP simulator constructed on a wooden trestle large enough to drive a B-52 bomber onto... which they did:

upload_2019-8-4_20-36-48.jpeg

SNL is also involved as stewards of the USA nuclear weapon stockpile, along with Laurence Livermore National Laboratory (LLNL). LLNL played a pivotal role in our development of a working PCSS whereas SNL spent about ten years in the 1990s trying to get it to "fire" for more than a few shots before self-destructing. Problem was, once triggered into avalanche conduction, the conductive path was random between the two contacts, very much resembling lightning. The conductive streamers had various widths and current densities and that is what led to the self-destruction. What we did, at the suggestion of DTRA, was to implant oxygen ions in parallel tracks between the contacts. The oxygen ions "poisoned" the GaAs and prevented it from behaving as a photo-conductor. Areas that did not receive oxygen ion implants carried the current streamers within narrow confines between the contacts. We could "see" that this was happening because the GaAs emitted infrared light, created by the current streamers, that could be photographed using a gated CCD camera.

Our partner in all of this was a sub-contractor named L3 Pulse Sciences in San Leandro, California. This was a case of the tail wagging the dog, as L3 is a humongous defense contractor heavily embedded with the Pentagon and Congress. How I was able to get them to sub with UES is another fascinating but longish story, for which I received no credit and very little appreciation. A few years and a couple of DTRA contracts later L3 took a CCD image and e-mailed a copy to me. It showed sixteen (IIRC) parallel conducting tracks on one of the GaAs switches we had fabricated and sent to them for testing. Next thing I know my boss and I are on an airplane to Albuquerque to talk with a bunch of engineers and scientists at SNL. Most of them didn't believe we had done what we said we had done, mainly because we didn't have "theory" for what was going on that agreed with what SNL thought was going on. UES had just been awarded a million dollar "fast track" contract to provide some of these PCSS devices to L3 so they could build a portable, modular, expandable, battery-operated nuclear EMP simulator. Which they did.

L3 finished the simulator about a year later, and that is the last I heard of it, except for signing away to UES any patent rights I might have had. Rabi and I were named as co-inventors on the UES patent application, which was rejected twice by the United States Patent and Trademark Office with the "explanation" that it was based on "prior art." No mention was made that the so-called prior art didn't work until we came along and fixed it.

I tried for two years after that to find another sponsor willing to fund development of our PCSS prototype into a real product, with a specification datasheet, that could be manufactured by or for UES and distributed by the likes of Farnell or DigiKey or even AliBaba. Nada. Dead ends everywhere I looked. UES carried me on overhead at half-time pay during this period before finally "letting me go" with a nice letter saying it wasn't my fault, but if I wanted to seek another job, at age 70, I would be eligible to collect Ohio unemployment compensation while doing so. I finally saw the writing on the wall and accepted retirement to Florida in 2016. Well, except for playing around in this and other forums, just to keep active whatever remaining brain cells I have left.

It is nice finding another person who seeks knowledge "just for the fun of it" as you appear to do.
 
Just an interesting fact (if my calculation is correct), the same 3μs pulse of light could freeze something moving at the speed of sound (343M/s) to within about 1mm.
 
Top