Create controlled 50v pulse

[(*steve*)]

The Vce of the 2N3904 is too low.

Not for a gate driver (it probably only needs to be powered from 12V). But, in any case, the 2N3904 can't switch anything like the current you need.

I'm testing the circuit in post #15 with a spare BC547C, a 910 ohm load and a USB oscilloscope and I can see how the turn off time increases with a higher resistance to base which explains the slow turn off time with saturation switching as Bob said.

If you place a diode from base to collector (preferably a schottky diode) you'll see a marked increase in speed.

But Steve, your MOSFET driver circuit uses a similar resistor>base configuration to switch the MOSFET off which you say is slower than a BJT?

The mosfet will switch much faster if supplied with a high gate current. This circuit supplies that higher gate current. A BJT has a gain that will fall at high currents. Effectively this means that switching high currents requires a higher base current. Try this with your test setup. Have a look at how the switching speed changes as you vary the collector load.

While BJT's are capable of operating at higher frequencies than MOSFETs, they have a broader linear region. Most MOSFETS are made for switching and have a far more narrow linear region.

If I dump the Picaxe could I create a purely 50v logic circuit that creates pulses of a precise length or would I need to ask that in a different part of the forum?

It probably doesn't matter what you use. And there's not a logic family that supports a 50V supply rail anyway. You could use one of the CMOS families that can operate from a 15V rail.

I guess we should ask what you're actually trying to do. We've made assumptions that may not be valid.

How fast does the load need to actually be switched? Do you need rise and fall times in single digit ns, or will a couple of hundred ns be sufficient? Can you use a low side switch? or is high side switching required? What is the nature of the load?

 

[CommanderLake]

I made a light with a CMA3090-0000-000Q0U0A40G (~172w LED) and a 50cm (25cm focal length) Fresnel lens that can light up the clouds but I figured that because LED's turn on and off super fast I could use it as a camera flash in a dark room to capture super fast moving objects with amazing clarity, FAR clearer than my 6D's fastest 1/4000th shutter speed and without any rolling shutter distortion.
I got an amazing result by discharging a 1uF PET through it but because it discharges logarithmically the duration of the flash is not well defined so in the name of science I'd like to experiment further with this and see how fast I can strobe the LED and get a clear image.
If I'm going to make this pulse generator I might as well make it as good as it can be, cost isn't much of an issue but I'd rather just stick to using a breadboard for now.
The LED starts to turn on at about 38v and the max voltage is 50v.

 

[CommanderLake]

I figured out how to get accurate sub microsecond pulses from my picaxe by using its PWM output with a controllable pulse length and a low frequency such as 1ms then turn it on and off with another of its pins via a MOSFET to isolate a single pulse from its PWM generator.

Do I use an n or p channel MOSFET to switch the load because neither turns on fully in the simulator I use.

 

[(*steve*)]

Do I use an n or p channel MOSFET because neither turns on fully in the simulator I use.

Neither. Turn the PWM on and off from the microcontroller.

The LED starts to turn on at about 38v and the max voltage is 50v.

Have a quick look at the resource on driving LEDs. LEDs should not be driven from voltage sources. Since you're driving it from a power supply, I recommend you set your current limit to a safe value for the LED. Note that if your power supply has any significant output capacitance it may well send a higher current pulse than you expect.

could use it as a camera flash in a dark room to capture super fast moving objects

OK, so you want to generate a single pulse. I was thinking you were generating a pulse train. Also, the rise and fall time for this are not critical as long as the total length of the pulse is not extended too far.

Can the picaxe create fairly accurate a pulse down to 1 microsecond, I've not tried.

You seem to have a handle on how to do this now. I would recommend that you code it like this:

code to start the pulse train (1us every 1ms)
delay 200us
code to stop the pulse train.

Assuming the pulse is emitted at the beginning of the period, that will work. Otherwise you may have to fiddle with the delay to ensure you get one and only one pulse.

I note that without a very high clock speed the PICAXE is not capable of such short pulses.

An alternative is to create a circuit which will shorten the pulse length, that way you can emit a single pulse (of whatever length) and it will be shortened to a constant value. With this you may be able to produce sub 1us pulses if you desire.

Here is a circuit (it will operate from 5V)



U1 to U6 are a 40106 -- this chip contains 6 schmitt triggers. R1 and C1 are selected to give the required pulse length of about RC. For 1us, you could use 1nF and 1kΩ. You could make R1 variable to create shorter pulses. The input pulse must be longer than the desired output pulse. The diode should be something like a 1N4148 or 1N914 (common small signal diodes).

It would actually be possible to replace the microcontroller completely and use a button to fire the "flash"



Although I show that using a 5V supply (which it will work with) a 40106 can also be run from 12V, so you don't need both 12V (for the driver) and 5V (for the logic).

This will directly drive a mosfet from 12V, but for shorter pulse durations I'd recommend you use the gate driver as well.

 

[CommanderLake]

When I asked what type of MOSFET I need I meant the one on the right thats connected to the LED, it wont switch on fully with your driver circuit with either n or p.
I use an appropriate current limit on the bench supply, if the current is slightly out of spec with a pulse it doesn't matter as thats a thermal/power dissipation limitation just like with transistors.
I tried just turning the PWM on and off quickly but that doesn't seem to restrict it to a single pulse, it creates a short pulse train in addition to the PWM pulse when turning off.

 

[(*steve*)]

N channel mosfet. And that driver is inverting, so without an input the mosfet will be ON.

What mosfet are you using, and are you testing in reality or using simulation software?

And are your grounds (-ve's) common (for 5, 12, and 50V power supply)?

 

[BobK]

Why are you messing with PWM to generate single pulses? All you need to do is turn the pin on, delay the required time, then turn it off. At 16 MHz, an instruction time is 250 ns, so, with no delay you can create a 250ns pulse, and with delays, any multiple of that.

Bob

 

[(*steve*)]

The picaxe runs interpreted code, it's a lot slower than native code. Even that doesn't explain how slow the pulseout command is, but other than changing to a different microcontroller (or just a hardware solution) you've got to work with what you've got.

 

[CommanderLake]

Pulsout 1 is 1.75us at 32MHz, I will experiment further with turning pwm on and off in code.
I only use an ideal simulator because its super quick and makes it immediately obvious in realtime if theres a problem with simulated scopes and current flow.
With an n-MOSFET the voltage from the driver is insufficient.

 

[(*steve*)]

With an n-MOSFET the voltage from the driver is insufficient

You should be seeing over 11 volts at the gate. What are you seeing?

 

[CommanderLake]

Its showing the full 12v at the gate with 14.33v across the load.

I just realised that because the driver is inverting, the way I cut off the PWM wont work.

 

[CommanderLake]

Actually I think saturation switching might be fast enough, here are the results from further testing with my test circuit with...
500 ohm load
500 ohm Picaxe to base
130 ohm base to emitter
1.74us pulse from Picaxe

Pulse width from/to half supply voltage
12v to load: 1.78us
15v to load: 1.74us

Also I noticed it gets faster with a lower load resistance and I have to tune the Picaxe to base resistance to eliminate the turn off delay without losing too much of the supply voltage.

 

[hevans1944]

All of the drivers I can find with the required voltage are SMT, I'd prefer through hole for prototyping.

Yeah, me too, but all the latest and greatest stuff seems to be readily available only in SMT packages. It's a worldwide conspiracy that we cannot do anything about. One possible solution is to purchase SMT-to-DIP adapters and solder the SMT devices to the headers. Now you have a package good for bread-boarding that you can actually pick up with your fingers and see with the Mark I eyeball, unassisted by microscope magnification.

Since we discover, finally, late in this thread that you want to use the LED for stroboscopic photography of fast-moving objects, have you not considered using a xenon flash lamp for this? Super simple, cheap, fast, bright and the flash can be relatively easily triggered (synchronized) with whatever event you want to capture photographically. Of course, you may be able to control the exposure more easily using the LED, but as @(*steve*) mentioned in post #24 you need to do this with a constant-current driver, not by using a MOSFET to switch power supply voltage to the LED. There are commercial constant-current laser diode pulsers available with the voltage and switching time characteristics you need. Perhaps you can find one, download their schematics, and build your own.

 

[CommanderLake]

I'm sorry I didn't tell you more sooner but often if I give too much info too fast it adds confusion so I feel I owe you some pictures:


2:40PM, not the brightest day but day nonetheless, on the right is a very bright 35w Xenon torch, on the left... well you can guess:


Here it is shining on the clouds with an exposure that represents approximately what the Mark I eyeball can see if your monitor has a gamma of 2.2:


And last but not least here it is in the fog:


I could get a Xenon flash like everyone else but I specifically want to see how this LED does as I already have it.

I'm going to try a TPH3206PSB GaNFET with a UCC27524P driver and a ZVN3310A to switch the PWM on and off as I cant quite get the Picaxe to turn it off fast enough, it sends out 2 pulses most of the time.
If the ZVN3310A doesn't work well directly off the Picaxe I will try it on the second output of the driver if the GaNFET works at 5v, else I will get another driver.
If none of it works I will try something else.

Here is the test with a dirty fan spinning at full speed.
First is the maximum 1/4000th shutter speed of my 6D:


And here it is with a 1uF capacitor at 50v discharged through the light:

I swear its on!
Actually that particular edge of the fan blade is a tiny fraction out of focus, its sharper where its in focus.

 

[CommanderLake]

but as @(*steve*) mentioned in post #24 you need to do this with a constant-current driver, not by using a MOSFET to switch power supply voltage to the LED.

Have you used one of these high power LED's, with small low voltage LED's that have a very narrow margin between vf and the current limit yes it does help to use current regulation but with this high power LED the margin between vf and full power is 11-12v that's plenty of room to find the right voltage.
But I do use current regulation at 3.5A of its 3.6A rated maximum and limited to 50v for constant operation.

An example of LED power limits, I just tested a 3mm green LED on my bench supply for a few minutes at 3v 70mA of its rated 20mA and while it got a touch warm it didn't explode or burn out but if operated constantly at 70mA yes it would eventually burn out, and you seem to be worried about a high power LED with overkill cooling maybe a few % over its rated current for LESS THAN 10 MICROSECONDS?!

 

[CommanderLake]

OMG I just realised I dont need the small MOSFET to switch the PWM signal on and off I can just use the enable pin on the UCC27524!

I did have it working but accidentally connected the output of the 5v regulator to the output of the UCC27524 therefore burning it out, I was tired and its an easy mistake to make on a breadboard anyway.

So I've ordered a replacement yesterday along with suitable components for the RC snubber as specified in the GaNFET datasheet and a ferrite ring because there was a lot of ringing and overshoot on the output when I tested with 12v.

 

[(*steve*)]

Sorry about not getting back to you earlier.

An example of LED power limits, I just tested a 3mm green LED on my bench supply for a few minutes at 3v 70mA of its rated 20mA and while it got a touch warm it didn't explode or burn out but if operated constantly at 70mA yes it would eventually burn out, and you seem to be worried about a high power LED with overkill cooling maybe a few % over its rated current for LESS THAN 10 MICROSECONDS?!

As long as you keep the current to a reasonable value then you'll be fine.

However, as a counter example... I accidentally connected a 10W LED to a capacitor that had a voltage across it significantly above the typical Vf of the device. I don't remember seeing a flash, but the LED never worked again.

 

[WHONOES]

Try the attached circuit. It just uses a monostable HC 123. It will produce a pulse of almost any width. R1 may be made variable so that your pulse width may be set accurately. The output stage is triggered from the inverting output of the monostable. This means that Q2 is normally on which keeps Q1 turned off. U1 is triggered via a capacitor coupled pulse. This is necessary otherwise the pulse from U1 will be lengthened by the width of the trigger pulse.

 

[CommanderLake]

I got a 1GS/s Picoscope 2208B to analyse the final working prototype and the rise and fall times are <30ns.
I just did a test at 1us of a 23cm desk fan rotating at 1960 RPM, thats a velocity of 23.6 M/s.
Using a Canon EOS 6D and a Sigma 85mm f1.4, processed with DxO PhotoLab.
The exposure is f1.8, ISO 6400, 0.5s with manual flash triggering.
Getting it perfectly in focus is tricky as I have an extremely shallow depth of field to work with.
The motion blur at 23.6M/s at 1us should be 23.6 microns but the LED doesn't respond fast enough so the light pulse is somewhat triangular.

Thats a 1024x1024 crop btw.

 

[Ground_Loop]

I would like to create a controlled 50v pulse of a configurable duration for a high power LED with less than 4 amps possibly using a Picaxe as I've used one before, so I have 2 questions...

1. Can the picaxe create fairly accurate a pulse down to 1 microsecond, I've not tried.

2. What circuit would I need to switch a 50v power supply on and off very fast with a 5v signal from the picaxe?

I have a 60v 5A bench supply.

Thanks.

Edit: I realise this might not even be possible because of the slew rate required but it only needs to be fast in the 40-50v range which is where the LED turns on, not all the way from 0v.

You could use a 74LS121 Monostable MV which can reliably deliver pulse widths down to less than 100nS and it can be triggered by the picaxe 5V output. Pulse timing can be made adjustable with a switched timing capacitor and/or variable timing resistor. From there I would drive a MOSFET with the 74LS121 'Q' output and use the power supply to control voltage and available current.