Fast LED Strobe for Macro

[mawyatt]

These Cree CXB3070 LEDs look very promising, with ~ 20 times the output of the CXB1304. ....
The controller design now supports 4 higher voltage LEDs such as the Cree CXB3070 which provides ~10,000lm at ~2A 36V, and separate LED current control under continuous and pulsed conditions. The number of LEDs can be easily expanded beyond 4 by paralleling LEDs with a small ballast resistors (~1/2 ohm) for conditions where more LEDs are required.

Mike, great idea. But what about LED cooling ?

If you are mostly interested in the pulse performance and can operate with a much lower continuous "modeling" light then a small LED heatsink may be sufficient, however if you need considerable continuous "modeling" light or a very high pulsed output with wide pulses at a fast repetition rate then a larger heatsink will be required. Using a tiny LED fan under the modeling output conditions, but stoping it for pulse use, is also a possibility.

These are some things we'll have to experiment with I suspect.

Best,

[JKT]

What cooling? 1ms pulse every 3 s adds up to 36 mW continuous.

I wonder how well the modeling light would work, though. I had target value of 200...300 lm, which worked pretty well. Even that shows (just) in pictures, if the flashes don't fire. Is it possible to get such little output reliably from those powerLEDs without PWM? If not, high CRI LED strips are very easy to arrange with 12V constant source.

My target value was 400 lms and with these latest LEDs the required number is 32 for 1 ms pulse. So this is beginning to sound feasible. And I quite agree that LEDs would be just about every way superior compared to traditional flash tubes.

Any chance of getting the time control to work more or less logarithmically? Or with fixed length steps corresponding to constant stop change? With enough LEDs one stop difference can also be had by turning off half the LEDs.

[Macro_Cosmos]

This might actually be useful for the current (job) project I'm working on. We are hitting a bottleneck where the constant LED is too bright, however lowering the brightness will result in bad images. Pulsed LED would give a strong burst of light that's preferable.

I'll take a detailed read later.

[Chris S.]

Frans (Fotoopa) made very nice system, which produced great results:
https://www.flickr.com/photos/fotoopa_h ... 4329376535

Frans must have a lot of time on his hands. Incredible link, that's for sure. Thanks for sharing.

-JW:

PS: I wonder if he has a website.

Frans is, to our delight and honor, a member of our forum, posting as “fotoopa”. (I think this means “photo-grandpa” in German.) To my mind, Frans' postings have immensely enriched our forum. The work Frans does is astounding

Though Frans seems not to have been active on our forum for a couple of years, perhaps a few polite requests might welcome him back? I for one would greatly value having Frans active again in our community.

--Chris S.

[mawyatt]

What cooling? 1ms pulse every 3 s adds up to 36 mW continuous.

I wonder how well the modeling light would work, though. I had target value of 200...300 lm, which worked pretty well. Even that shows (just) in pictures, if the flashes don't fire. Is it possible to get such little output reliably from those powerLEDs without PWM? If not, high CRI LED strips are very easy to arrange with 12V constant source.

My target value was 400 lms and with these latest LEDs the required number is 32 for 1 ms pulse. So this is beginning to sound feasible. And I quite agree that LEDs would be just about every way superior compared to traditional flash tubes.

Any chance of getting the time control to work more or less logarithmically? Or with fixed length steps corresponding to constant stop change? With enough LEDs one stop difference can also be had by turning off half the LEDs.

The controller works by having completely separate LED current for the modeling and pulsed outputs, so no chance of continuous illumination bleeding into the pulsed illumination even with modeling at a very high level.

The Cree LEDs seem to have an output of 100~150lm per watt, so 200~300lm modeling would equate to 2~3 watts total dissipation and not require much of a heat sink.

Log control is certainly possible, as are multiple discrete steps. Computer control vis a RPi or Arduino is also possible, but all these would complicate the design and cost The design essentially has 2 steps to cover ~100us to 11ms, each step covering ~10X range with linear coverage within each range. That seems adequate, however we've supplied the schematic and one could adapt to specific needs if desired to DIY

Also you could use precision multi-turn pots (Bourns 3590) with dials off the PCB for greater setting precision and repeatability. The PCB supports directly supports on the PCB the Bourns 3296 multi-turn and 3386 single turn pots to allow user preferences for setting controls.

Best,

[Macro_Cosmos]

I follow Frans on both Flickr and YouTube.

He's also been building a massively overkill marble machine. He made his own controller for that, it's just wow.

The high speed in-flight insect capturing device he makes, which utilises two Nikon D7100 cameras is amazing as well, I think the one offered by Cognisys is based on his.

https://live.staticflickr.com/65535/489 ... b79f_h.jpg
This looks like a nuclear reactor

Bit off-topic, but I just love his work.

[mawyatt]

Agree, Frans work is indeed amazing

Especially like the high power LED mounting he developed.

The Pulsed LED controller Fran developed evidently works well, but the design doesn't use differential negative feedback techniques, is discrete device based and depends on a transistor Vbe base emitter junction potential as a "reference". This creates a fundamental temperature dependency on LED current, and each "driver section" requires individual calibration for discrete device variations, due to the nature of these Vbe junctions this requires a higher voltage drop across the current sense resistor and the current control range is limited and can't go below ~0.8V, so lower tightly controlled current range is ~Vbe/Rsense. With the approach presented here, utilizing high gain negative feedback with differential sensing, current can be sensed and controlled to ~0, and no temperature dependance nor calibration is required as each "driver section" is an independent precision current sink. All sections can be accurately controlled from one potentiometer rather many, although one could claim having individual driver calibration is good since you can calibrate to each LED. The large capacitors are not required either because the current sensing voltage is lower which lowers sense resistor power (V*V/R) and less voltage headroom is required.

Anyway, a traditional discrete design approach vs. a precision analog integrated design approach. Each should work well, so hat's off to Fran for his pioneering work on these pulsed LEDs for macro use

Best,

[mawyatt]

Haven't been able to work this lately because of available time.

JKT made some good points about the required illumination being ~400lm for proper exposure, so working with that as a reference.

400lm of continuous light using LEDs with 100~150lm/watt requires 3 to 4 watts. Assuming 4 LED modules this equates to only 1 watt per LED Module, heat sinking should be relatively easy at this level.

For flash use, a 1ms flash burst would require 400lm/T, where T is the flash period. This is 400,000lm, likely way too much for practical consideration. At 10ms then 40,000lm are required and this can be achieved with 4 or 8 LED modules of 10,000 ~ 5,000lm per LED module.

A typical speedlight has ~60Ws of available energy (1/2 CV^2). Xeon tubes are ~50% efficient, with another ~10% lost to I^2R heating in the capacitors and wiring & tube resistance. So a reasonable estimate of optical output is ~25Ws.

High power LED modules usually require 38~43V and with 100~150lm/watt would require 2~3 amps for the 10,000lm modules and 1~1.5amps for the 5,000lm models. This is well within the capability of the controller previously mentioned.

4 LED modules with 10,000lm per module would produce a total 40,000lm, and require under 2 watts average power per module to produce 100lm per module continuous or with flash. The peak optical power is limited to 10,000lm per module at ~100watts.

So an optical LED flash could produce ~40,000lm and ~400 watts peak power, whereas the speedlight would produce ~25Ws. The time cross over point where the LED modules equal the output of the speedlight is 25Ws/400W or 0.0625 seconds. At this time the LED module would dissipate an average power of 16watts. It would take about six times more LED output to equal a speedlight output at 10ms, or ~ 250,000lm, which would require ~25 10,000lm modules!!

This may be the reason there are no high power LED based speed lights or strobes available

As JKT mentioned earlier, may require too many LED modules to get a good LED flash output power in millisecond ranges, and the high power LED modules aren't cheap either

Need to think about this more, maybe others will comment.

Best,

[Saul]

... the required illumination being ~400lm for proper exposure, so working with that as a reference...

It depends - if used on the microscope with DF, DIC etc at high magnification - it will be not enough. Same problem with monitoring - more Lm, more heat, bigger heatsink, bigger assembly, etc & etc ...

[mawyatt]

... the required illumination being ~400lm for proper exposure, so working with that as a reference...

It depends - if used on the microscope with DF, DIC etc at high magnification - it will be not enough. Same problem with monitoring - more Lm, more heat, bigger heatsink, bigger assembly, etc & etc ...

Thanks for the input.

For continuous use the LEDs seem fine and a good option. A single LED module can produce 10,000 or more lumens , however for flash use they are inadequate & limited compared to a traditional Xeon flash tube.

Kind of like trying to force a nail into a board, rather than hitting it with a hammer

Seems like a better use of my time would be to try and modify a manual old school type strobe, applying an IGBT for a fast pulsed output.

Best,

[JKT]

Seems like a better use of my time would be to try and modify a manual old school type strobe, applying an IGBT for a fast pulsed output.

That controller could still provide adjustable voltage and/or current for modeling LEDS...

[dickb]

Frans is, to our delight and honor, a member of our forum, posting as “fotoopa”. (I think this means “photo-grandpa” in German.)

You could probably say so in German as well, but in Frans's case it means photo-grandpa in Dutch. The word probably makes more sense when you realise that it consists of "foto" and "opa", the double o makes it easy to misinterpret.

[mawyatt]

Seems like a better use of my time would be to try and modify a manual old school type strobe, applying an IGBT for a fast pulsed output.

That controller could still provide adjustable voltage and/or current for modeling LEDS...

Yes, an adjustable current is the way it works, current mode operation. However without the effective use for flash, this just becomes a video LED light, and many of these are available at reasonable costs.

Without a much higher output LED model at reasonable cost, the LED flash doesn't seem worthwhile.

It would work well, and have a nice quick output flash, but just not enough optical power to be useful for serious macro use I believe. If someone wishes to give this a try I'll be happy to help, but for now I'm going to pass on getting the PCBs fabricated and ordering the components. The schematic and PCB layout are already provided above.

BTW, a RPi controlled version has also been developed and I'll show it below. It features 12 bit amplitude and 12 bit timing control, remote LED temperature sensing, readout and over-temperature protection (hardware). This was designed to mount to the back of a heatsink which had the 4 LED modules and a Bowen's compatible mount for various reflectors. Kind of a LED Studio Strobe equivalent, but at lower optical output.


Maybe someone can change my mind, I'm open to good reasons to continue with this pursuit.

Best,

[Chris S.]

Mike,

This question is moot now, but I've been wondering: Do you think latency would be a significant issue in using LED COBs for motion-stopping flash--say 1/20,000 second or faster? (By "latency"--not sure this term is exactly right, here--I mean, "Would it take overmuch time for the LED array to come up to full brightness when energy is applied, and to drop to full darkness when energy is removed?")

With some LED COBs I've been working with for continuous light, it appears that when I turn them off, they go dark not immediately, but fall off over a brief but observable moment of time. As I haven't been attempting flash, this hasn't bothered me--but it does make me wonder about the practicality of high-speed LED flash. This said, the fall-off time I perceive may be a function of my power supply, my visual perception, the phosphors that give these LEDs a spectral curve closer to black-body radiation than most LEDs have, or something else.

Your thoughts?

--Chris S.

[mawyatt]

Chris,

I think what you are "seeing" is the result of the LED power source capacitance slowly discharging and the result is a slow "tail" on the LED going off. LEDs have an extremely wide range of output illumination, so even small currents still produce an output. As for turn-on latency, this is highly circuit dependent, since the LED should be very quick to respond to an applied current.

This is exactly why we chose the particular circuit topology shown, it's pure current mode without any capacitance capable of delivering LED current other than small parasitic capacitances when commanded off, and the LEDs for use have no additional capacitance on the LED module. So the only residual output would be from the LED self & parasitic capacitances and the phosphors. Being current mode, the LED current is "forced" thru the LED, so turn on latency is very quick. Without the phosphor induced "tail" this current mode circuit topology is fundamentally capable of doing 1/100,000 (10us), even 1/1,000,000 (1us), or faster with some different component values.

To evaluate the LED response one could construct a relatively simple test circuit to measure this. I know way back in 1980s when we were working with the fiber-optic gyros, LEDs were utilized as the optical source and they were capable of being modulated at fractional gigahertz rates.

So it seems the phosphor would be the limiting LED speed factor, and the circuit the final limiting factor which shouldn't be an issue for the speeds you indicated if current mode is utilized.

Best,