Current Pulser Flash

Discussion in 'The Projects Forum' started by brianvg, May 29, 2014.

  1. brianvg

    Thread Starter New Member

    May 28, 2014
    6
    0
    Hello! This is my first post here...

    I am working on a project where we are trying to photograph droplets spraying at a fixed frequency. This frequency is in the 10 to 100 kHz range. Presently we have been using an averaging approach to photographing, where multiple underexposed images can be analyzed.

    We have calculated that to optimally photograph the droplets we need a flash width which is 100 ns max, so that the acquired image does not have any blur with our optical set up.

    Up till now we have been using a mosfet driven high power led to provide lighting, which works admirably, but seems to hit a wall around 500ns. When we attempt to go lower than that we are spending so much of the pulse in the rise and fall time of the mosfets, as well as in the charging time of parasitic capacitance in the system and LED, that the LED barely has time to begin generating light. Recently, we tried a well laid out design, with a gate driver for the power mosfet. This improved things somewhat, but was still not fast enough. Additionally, we have been considering the use of a BJT(s) operating in avalanche mode, but this has proven more difficult than expected to realize, especially for the current and time involved.

    The goal is to find a way to supply the high power LED with a 30+ amp pulse, with a duration of less than 100 ns. Does anyone here have experience with a circuit which might be usable?

    Thanks for any and all suggestions!!
     
  2. AnalogKid

    Distinguished Member

    Aug 1, 2013
    4,542
    1,251
    How often is the flash fired? That is, are you taking individual photos a few seconds apart so the circuit has time to recover, or are you firing off a rapid series of flashes? If the former, than you might need a more complex driver. Above you mention the MOSFET - singular. You probably need two or four MOSFETs in a half-bridge or H-bridge configuration to suck out the electrons for faster turn-off.

    Have you got a photo so we can see the wiring, and a schematic of your circuit?

    ak
     
    brianvg likes this.
  3. bertus

    Administrator

    Apr 5, 2008
    15,648
    2,347
    brianvg likes this.
  4. DickCappels

    Moderator

    Aug 21, 2008
    2,656
    632
    That's a great link that Bertus found. In the circuit below, you can stretch the discharge time by connecting a length of coax with the shield to ground and the center conductor connected to the collector of the transistor.

    [​IMG]

    Here is a link that leads to avalanche generator with a transmission line to stretch the pulse. http://www.aholme.co.uk/Avalanche/Avalanche.htm (Schematic below -note that circuit below is free-fun but can be triggered with a lower power supply voltage).

    [​IMG]

    By the way, I would be interested in seeing what your solution turns out to be.
     
    Last edited: May 29, 2014
    brianvg likes this.
  5. brianvg

    Thread Starter New Member

    May 28, 2014
    6
    0
    Thanks for the suggestions!

    Analog kid:

    At the moment we are using a single mosfet, but one with a very small Rds_on (something like 0.02 ohm), as well as a driver which can deliver on the order of 5 amps to charge the gate capacitance.

    I can post the schematics for the current pulser, but they are nothing fancy... just a 24 V powered mosfet with driver, with a smattering of fast 200nF tantalum caps, and a monster 1000uF to keep the supply stable during the pulses. The layout is poor, and it rings a lot at pulses. I will get them on monday and post them here.

    The duty cycle is very low. The absolute max we would want to repeat pulses is a couple of thousand times a second. We would like to stay flexible as regards the momentary repeat rate... to do things like trigger the pulse 20 times, with a 100ns pulse length, and pauses of 300ns, once every 500ms. The triggering electronics are already finished, and we need to be able to trigger the pulses using TTL signals.


    Bertus and Dick Cappels:

    I have already had a look at the site that Bertus linked to, it is what sent me down the avalanched transistor route... I think that may be a good way to solve the problem, but every circuit I have breadboarded I have not been able to get to work, even with a tiny LED. I am not sure what I am doing wrong. I am also not entirely sure that even at 200ns any of the small signal transistors that are used in those avalanche designs would be able to hold up to a 30 amp pulse. I guess the magic smoke would escape.

    Cappels, I dont think I have seen the site that you link... It is interesting, in so far as it gives values for components, which a lot of the scholarly stuff written on the subject neglects. I guess the physics community uses these kinds of circuits to produce really short pulses of light for calibrating detectors and such... Unfortunately they do not divulge too many of the details. I would love to be able to scale up that design...

    Unfortunately the avalanche mode stuff is made more difficult because there are not many spice models for BJTs that include avalanche behavior. I found one for a specialized BJT from the FMMT series, but they are expensive, and evidently not too different from a small signal BJT.

    I also think if I go the route of the avalanched BJTs, then I will probably have to do an array of them so they don't go pop... and I am not sure how to get them to all trigger at the same instant.

    Thanks for all of the suggestions!
     
  6. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,291
    1,255
    If you are using white leds the problem may be the leds and not so much the driver. Most white leds are made with phosphor and are kind of slow to respond. So a check of the data sheet (if they spec it) might be in order. You can mix colors to get white, but it will be tricky. Anyway: here is a fast FET (low gate charge) that might help.
    http://www.infineon.com/dgdl/IPP147...90004&fileId=db3a304412b407950112b4277a093c43
     
    brianvg likes this.
  7. brianvg

    Thread Starter New Member

    May 28, 2014
    6
    0
    Ronv:

    The LED is a very high end power LED, it is blue and uses I think a matrix of multiple LED cells. We also have them in red and green. I can recall the manufacturers name. Its about a 20W LED as I recall (dont have the data sheet in front of me)

    Coincidentally, the mosfet you mentioned is very similar to the one we are using now, if not identical. I know if is from the same infineon series...

    I wonder how these guys do it?:

    http://www.picolas.de/index.php?opt...&catid=8:diodenlasertreiber&Itemid=19&lang=en
     
  8. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,291
    1,255
    Last edited: May 29, 2014
    RichardO and brianvg like this.
  9. brianvg

    Thread Starter New Member

    May 28, 2014
    6
    0
    Excellent link Ron. Unfortunately I do not have access to the data sheets or the equipment, as it is a long weekend here. I won't be back to the lab until monday. I will keep you all posted though!
     
  10. shortbus

    AAC Fanatic!

    Sep 30, 2009
    4,015
    1,531
    Wouldn't it be better to use something made for the application instead of a LED? Xenon flash tubes are what I'm suggesting. If the flash needs to be a small diameter/area, the tube could be masked by a metal housing with an aperture.

    Could be done on the cheap too. Used disposable camera flash units are free at most photo shops. To get fast repeat flashes more than one could be used, triggered separately. As one is flashing the rest are charging.
     
  11. RichardO

    Well-Known Member

    May 4, 2013
    1,235
    384
    I have been playing with using an MMBT2369 as an avalanche pulser for amplifier testing. I am getting a risetime of less than 2 nS on a breadboard driving 50 ohms. The pulse width is on the order of 10 nS.

    I have attached the simulation done in LTspice. Note that I simulate the avalanche transistor using an SCR-like circuit of a PNP and NPN feeding each other.

    Tens of amperes at the speeds you want are going to require a lot of attention to circuit layout -- especially ground planes and the high frequency quality of the capacitors.

    This application note shows a simple example of how to parallel avalanche transistors for higher current. They claim that currents of over one hundred amperes can be produced:

    http://diodes.com/download/4539
     
  12. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,291
    1,255
    I knew I had seen this somewhere:
     
  13. brianvg

    Thread Starter New Member

    May 28, 2014
    6
    0
    Shortbus:

    If you find an application note that shows xenon tubes flashing in the tens of nanosecond range, do let me know. As far as I know though, those tend to operate in the microsecond range, at best. I would guess a camera flash would struggle to get into the single digit microseconds. I guess that if you want a standard flash to flash at that speed you are looking at high voltages and probably some sort of spark gap triggered device?

    Richard O:
    I had read that application note, and I bread boarded some of their circuits out with some small signal transistors (from what I have read, specialty avalanche BJTs are just regular small signal transistors in better housings, and tested to avalanche at a known voltage). I was not able to get anything like the numbers used in the application note, but I guess that is probably down to the terrible parasitic inductances of a breadboard circuit. I don't have any experience laying out really high speed circuits and PCBs. I think I am going to price the pre-made pulsers linked earlier in this thread, then depending on the quote I get, decide if it is time to get my hands dirty and really dig into low inductance and impedance matched PCB design.

    I would definitely be interested in seeing your LTspice simulations! I don't seem to be able to see them at the moment though...
     
  14. RichardO

    Well-Known Member

    May 4, 2013
    1,235
    384
    If you were trying to use a solderless breadboard then I would expect that you would have problems. For something like this you need to use a perfboard with copper on one side and solder the parts together with the shortest possible lead lengths. The caps are especially critical. The ones discharging into the LED must be ceramic -- preferably surface mount to eliminate pesky lead inductances.


    Jim Williams wrote a number of application notes for Linear Technology. You may want to look at these if you haven't already.

    AN47: "High Speed Amplifier Techniques -- A Designer’s Companion for Wideband Circuitry"
    See figure 32 (and others) for how he built his circuits!

    AN95: "Simple Circuitry for Cellular Telephone/Camera Flash Illumination --
    A Practical Guide for Successfully Implementing Flashlamps"
    The name says it all.


    You need the LTspice circuit simulator to view my simulation. It is a free download from Linear Technology.

    (I can't seem to get a screen shot quite small enough to post to this forum).
     
  15. RichardO

    Well-Known Member

    May 4, 2013
    1,235
    384
    I have been having some fun simulating avalanche circuits driving LED's. What I am seeing makes me think that an avalanche driver is not the way to go. The avalanche waveform has a long tail caused by the exponential discharge of the capacitor. This tail might cause blurring in your image. In addition, the avalanche circuit has no way to disharge the capacitance of the LED's. AnalogKid alluded to this in an earlier post in his suggestion to use a push-pull (half-bridge driver).

    I am curious about the LED's you are using -- one's that can be driven with 30 amps. Can you supply data sheets?

    The one high power LED I have used is a white LED array that has 12 LED's in series so it needs about 44 volts at 1 amp maximum. The series LED arrangement might be easier to drive.

    LED Engin make a blue version like the one I used: LZ4-00B200. It needs 14 volts at 700 mA (which is about 10 watts) and supplies about 130 lumens at 460 nm.
     
  16. brianvg

    Thread Starter New Member

    May 28, 2014
    6
    0
    This the the LED series we are driving:

    http://www.luminus.com/products/CBT-120.html

    I am also beginning to think that avalanche is not the way to go... researching a good half bridge rectifier solution now. The commercial pulsers are very $$$ :eek:.
     
  17. RichardO

    Well-Known Member

    May 4, 2013
    1,235
    384
    Nice LED's. Thanks for the link.

    Wow, the PCO-7120 is a nice module but at $588 apiece it better be!


    I have tried some simulations of MOS-FET LED drivers and I don't think a push-pull driver is necessary. More info when I get a chance.
     
  18. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,291
    1,255
    What is the part number of the FET and driver you are using. You should be able to get there with the one I posted and a suitable driver. What are you using to limit the current to the LED?
     
  19. RichardO

    Well-Known Member

    May 4, 2013
    1,235
    384
    I am using parts from the standard libraries of LTspice. I chose the IRF2805S for the FET and the LT1693-1 for the FET driver. Nothing that special about either of them. They do seem to work -- at least in simulation.

    Your IPP147N03L looks great. Actually it looks incredible. I need to find a Spice model for it.

    I am doing the simplest thing you can imagine. I use a 0.1 ohm resistor to limit the current and an LM317 set to a few volts more than the LED forward drop. To reduce the fall time caused by the LED capacitance, I use a second FET which is turned on after the first FET is turned off.

    For this to work, the power is not allowed to sag during the pulse. This is the reason for all of the caps. I have not looked in detail at the minimum capacitor requirements so what I have simulated is probably overkill.
     
Loading...