Nanosecond rise time for a plus of LED light

[Nima]

Hi all,

For an experiment I need a very fast single pulse of light. The turn-on time must be less than 10 nanosecond. I am using LEDs as the source of light. The shortest rise time by using batteries as the power supply would be 100 nanosecond. I am wondering if a LED driver can reduce the turn on time? Can you recommend one?

With thanks
Nima

[themadhippy]

alternatively try a large capacitor to store the energy

[ejeffrey]

Have you measured that?  Or you just guess?  The turn on time doesn't have much to do with the power source, it's going to be about the switch and the wiring.  10 ns transition time isn't terribly fast, you can do this with standard logic gates if the current needed is not too great.

[JustMeHere]

If you want 1 ns rise time, you will need a circuit that's capable of 1 GHz signals.  Keep your ground loop at a minimum.  4 layer board will help.  Then I'd use a series of different caps to help get the signal up.

Your Micro is going to need to be able switch fast enough.  You're probably going to need a MOSFET to handle the LED power.  It will need to be able to switch fast enough.  Don't forget you need to turn off the pulse quickly too. 

I'd use a very high frequency cap to help with switching speed.  Then probably a few bigger ceramic caps in parallel to provide the "bulk" charge.  You might need a third size between the high frequency and bulk caps to bridge the frequency response of the other two caps.

[richnormand]

here is something to look at for design, 1ns rise and fall time for an LED source:
https://arxiv.org/ftp/arxiv/papers/1011/1011.1954.pdf
Other possibilities are super luminescent sources and diode lasers that can do much more given the appropriated drive circuits.
Cheers.

[ejeffrey]

If you want 1 ns rise time, you will need a circuit that's capable of 1 GHz signals. 

A 1 ns rise time is only ~ 350 MHz.  The requested 10 ns rise time only about 35 MHz.

[Kleinstein]

10 ns is still in a range of standard logic (e.g. 74AC...). If it is just about a short pulse, one may use a series capacitor instead of a series resistor. When the gate switches the charge from the capacitor is transferred to the LED. One would get a relatively fast turn on and than a decay.
As other already noted, use low inductance, short leads and proper decoupling of the supply.

The light intensity of a LED is not directly following the current, but there can be additional delay for a build-up / decay of excess minority carriers. Especially some IR and red LEDs can be quite slow.  One could get an idea on how much delay by looking at the reverse recovery of the LED.

[T3sl4co1l]

The light intensity of a LED is not directly following the current, but there can be additional delay for a build-up / decay of excess minority carriers. Especially some IR and red LEDs can be quite slow.  One could get an idea on how much delay by looking at the reverse recovery of the LED.

It's my understanding that LEDs peter out above a few 10s MHz, for this reason.

There have been proposals for highly-asymmetric "LiFi" using illumination LEDs (basically bathe the room in data, let the clients sort it out; upload has to be handled separately, perhaps with a very low BW IR link), which would be "competitive" with WiFi, so, about this bandwidth (that is, the 2.45GHz WiFi, with many users and mediocre signal quality, giving 10s Mbps per client).

Incidentally, notice the yellow phosphors have longer delay times, so this would require blue filtered photodiodes to receive such data.

Far higher bandwidths are possible, optically, of course -- but they require laser diodes and/or acousto-optical modulators.  The power levels are quite modest, suitable for fiber use, nothing you'd be able to broadcast (well, "broad"cast; with good optics, a cableless point-to-point link might do).

Tim

[radiolistener]

so fast pulse edge needs very fast mosfet, also it needs to use RF feeder line (coax cable for example) from source to the load, because if you use usual wires they will attenuate high frequency components and you will get much slower edge on the load. Also, so fast pulse edge leads to a very wide bandwidth interference which may break almost all kind of communications in your area (up to 1 GHz).

If you still need it you can get 1 ns rise time with a cheap si5351 Chinese module, it will cost you about 2-3 USD, but it's output has too small power to drive LED, so it will require RF power amplifier with 1 GHz bandwidth, such device will cost much more...

For more fast rise time you can use ADF4351 module, it's output power even smaller that si5351, but it has rise time about 70 ps.

[Siwastaja]

Depending on how you exactly define rise/fall time, 10ns is nearly reachable with a simple voltage source + series resistor + low side MOSFET solution. For example, I have had acceptable results with BSS316N modulating at 20MHz in a time-of-flight application. LEDs are marginal at 10ns rise/fall, anyway, so a laser diode should be considered instead.

[Kleinstein]

The speed really depends on the LED type. Slow ones may have a time constant in the low µs range. I remember an old article driving blue LEDs for ps range pulses. So there is a quite large range.

Even with a relatively slow LED one can speed up the turn off by shorting the LED, or applying a slight reverse votlage so that a reverse current flows. The LEDs still emitts some light until reverse recovery is over. It may take relatively high current to get significant intensity from short pulses.

[T3sl4co1l]

Um, who are you replying to? You can condense all those into one post / edit...

Tim

[BillyO]

Sub 2ns times are available with simple cheap 74AC logic without really trying hard.  And it can sink or source up to 75ma for short pulses (<2ms).  If you parallel all inverters in a 74AC04 you can sink/source nearly half an amp for very short pulses (<100us, 10% duty cycle).

I think the real problem will be finding LEDs that will turn on/off that fast.

[rstofer]

Here's a simple circuit with fast rise time:
https://www.eevblog.com/forum/projects/super-simple-_fast-edge_-pulse-generator-for-scope-based-basic-tdr/msg221694/#msg221694

[tggzzz]

Here's a simple circuit with fast rise time:
https://www.eevblog.com/forum/projects/super-simple-_fast-edge_-pulse-generator-for-scope-based-basic-tdr/msg221694/#msg221694

That mentions 2ns transition time.

This jellybean circuit is almost an order or magnitude faster: 250-300ps 2.5V into 50ohm. You do need to be careful with decoupling and layout at those speeds.
https://www.eevblog.com/forum/testgear/show-us-your-square-wave/msg1902941/#msg1902941

[Siwastaja]

I think normal LEDs accessible to the OP would be in the 10ns or more range. There are mentions of faster LEDs but when I looked into it, it seems no one are using such LEDs for example in time-of-flight applications, but go to lasers instead beyond 20-30MHz modulation frequencies (with 3D ToF, the transition time could be literally the only reason to use a laser diode, because the coherency does not matter and all - all you need is light. The narrow wavelength is arguably a bonus if you can manage to design a matching filter over the sensor, for even better ambient light rejection.)

If the OP is limited to the usual 10-20ns LEDs, then driver circuit with 2ns transition time is fine as it's not the limiting factor.

[Nima]

Thanks

[MrAl]

Hi all,

For an experiment I need a very fast single pulse of light. The turn-on time must be less than 10 nanosecond. I am using LEDs as the source of light. The shortest rise time by using batteries as the power supply would be 100 nanosecond. I am wondering if a LED driver can reduce the turn on time? Can you recommend one?

With thanks
Nima

Hi,

This might have been said already but you need a mosfet with a mosfet driver and low ESR caps and short short lead lengths.
Your LEDs are important too as some are not made for pulsing at all.

[Terry Bites]

This should be of interest.

[David Hess]

here is something to look at for design, 1ns rise and fall time for an LED source:
https://arxiv.org/ftp/arxiv/papers/1011/1011.1954.pdf

I wonder why they did not place the LED as the feedback element of the operational amplifier.  Compensation for the current feedback AD8009 is then implemented with the normal feedback resistance in series with the inverting input to compensate for the LED capacitance, which is how current feedback integrators are made.  Select the resistance for good response.

Using the amplifier has the advantage of sweeping charge out of the LED when turning off for better response, but the same thing can be done with emitter/source transistor switching by adding a pull-down resistance or current source.

[richnormand]

I just posted that to provide Nima some info about that it was possible to modulate an LED even beyond his requirements.
I did something similar years ago in the "almost" GHz range. Sometimes analog/digital gets fuzzy for high speed optoelectronics but we went through several types/brands to find the best candidates.
I was using direct LED output, no idea what happens once you put in phosphor response, like in "white" LEDs or multi-chips ones.
Cheers David