20W Halogen bulb viewed by a photodiode

[StillTrying]

I tried to capture the light flashes from a lightening storm, - and failed every time.

Changes in the brightness of the sky were enough to move the waiting level up and down about 1 Div. so the lightening flashes themselves should be peaks right up the screen. In among all the electrical noise I can't see any increase in light level at all. All I can think of is that the main bright flash doesn't occur until 2-3ms after the start of the electrical noise and I missed all of them.

[StillTrying]

1us wide 25mA pulse through a super bright blue 5mm LED, plenty of light, about 10cm between LED and PD.

Although these are the fastest light shape rise and fall times I've managed to get so far, I spent hours trying to improve the leading edge/rise time and failed.
Is this the best light shape I should expect from a SFH213 PD with -17V and 2 transistors, because further improvement of the leading edge is proving difficult.

[StillTrying]

4us 12A 60V through a 100W 10x10 daylight white LED COB, 15cm and 2 sheets of thin paper between COB and PD to get the light level down.

[soldar]

Years ago I designed a way to detect if a shaft had stopped rotating and it was a cross of Malta type thing rotating with the shaft and the blades interrupted the light from a tiny lightbulb to an LDR. The circuit detected the on/off light at the LDR and if it stopped it would activate a circuit. Except that it would just not work. After much investigation I found out the LDR response when receiving light from the light was not flat but was very visibly modulated at 100 Hz. The filament had low thermal inertia and light emitted varied over the cycle. Once I powered the light with DC the problem disappeared.

I also remember a circuit which would transmit voice over the light beam of a car's headlight. Obviously the light emitted by the filament was not modulated to 100% but there was enough modulation, maybe 5 or 10%, to transmit voice over the beam. I thought it was neat and always thought it would be a cool thing to build.

It was Elektor in 1979, before LEDs. See attachment.

[Karel]

I did some similar experiments using this sensor:

https://www.ti.com/product/OPT101

The LED lightbulbs in my house don't produce any flicker.
However, we have one designer lamp that contains a LED strip and a dimmer.
This lamp is really terrible when it is not set to the maximum intensity because it uses PWM at 100Hz 
The light produced is almost a perfect square wave between zero and maximum.

The picture is from the LED bulbs that don't produce any flicker and are "safe" to use 

[Zero999]

4us 12A 60V through a 100W 10x10 daylight white LED COB, 15cm and 2 sheets of thin paper between COB and PD to get the light level down.

For very bright light sources, I've found it's better to skip the amplifier and connect the cathode of the photodiode, to as higher voltage it can take and the anode to the oscilloscope set to 50Ohm impedance via a piece of 50Ohm co-axial cable.

Here's an oscillogram of a high power Cree® XLamp® CMA3090, driven with 60A pulses.


A schematic of the driver circuit used and the photodiode with no amplifier, directly driving the oscillscope's 50Ohm input.


It was discussed in another thread.
https://www.eevblog.com/forum/projects/ultra-short-ultra-fast-led-flash/msg2252829/#msg2252829

[soldar]

I worked for a company that sold and installed simultaneous translation equipment for congresses. In the large venue room they would install large panels with IR LEDs which were modulated with several carriers to carry the different channels with different languages. Users would use headphones with IR receivers and could select the language/channel.

In fact, I still have a couple of these IR panels with IR LEDs, waiting for a project where I could use them.

It makes me think that with overhead LED lighting at home it would be possible to modulate it with audio (from TV or music) and have headphones receiving the signal. Not the most practical idea but possible. IR is better suited for this purpose.

[StillTrying]

Here's an oscillogram of a high power Cree® XLamp® CMA3090, driven with 60A pulses.

Yeah, I studied it in the other thread! Real scope shots of linear a few 1us light pulses for comparison are very rare, I only know of 1 other, other than yours and mine, some others are current waveforms or just sims.
I can't identify which PD your SFH2201 is, 4mA seems like a massive (not very linear) PD current, I'm usually only around 20-200uA.

Here's one of my basic improvements to the "just PD and 50R" mostly to improve sensitivity.

* The impedance seen by the PD is <23R.
* The 510R gives X10 more signal than the PD + 50R.
* No feedback cap needed.
* The 15R makes it a 50R output. (I've not tried a 50R connection)
* Bandwidth -1@ ~50MHz and -3@ ~100MHz.
* The 0.65V emitter-collector is a bit low, and is a DC offset, but seems to work in practice.
* = in theory, and/or measured in sim, results in practice may do vary.

I've made the latest 2 TR version too small to add another stage to the 7660's -16V, I'll try inserting another -9V which would total -26V on the PD. The PD and 5k1 feedback R are already air connected to the RF TR's base, to save every 0.25pF.
Most of the time I don't know if the speed/edge limits are the PD, or the best that that LED can do.

[Zero999]

I agree 4mA is a lot, but I can assure you it's still well within the linear region of the SFH2201. I tested it by varying the distance from the emitter and it follows the square rule up to around 15mA, at which point it became non-linear. The tell tale signs of this is are a delayed turn off time and no further increase in current with intensity, similar to the storage time and emitter saturation of a BJT. Increasing the bias voltage also increased the maximum reverse current it could take before becoming non-linear. I upped the voltage to 30V, which is above the maximum rating and it would work at higher currents, but at increased dark current and reduced reliability, so isn't worth it.

I did try using a high speed op-amp based transimpedance amplifier, in the hope that lowering the impedance seen by the photodiode would make it more linear, but found it wouldn't work at such high currents and didn't offer any real benefit, at such high intensities. I agree, for much lower intensities, an amplifier is necessary, but past 1mA it's generally pointless, since the oscilloscope has adequate sensitivity and 50Ohm is still low, compared to the high impedance of the photodiode.

[StillTrying]

I'm surprised the SFH2201's typical rise and fall times are 40ns, I should be able to get 5ns PDs to be faster.

Changing the reverse bias on the SHF213 PD between -9V and -27V made absolutely no difference at all to the rise and fall times or light shape.

Perhaps it's the way I'm driving the blue (or green) LED.

The green trace is the 3.7V directly across the legs of the blue LED, I'm not sure if the slight curve at the beginning of the 3.7V is real or not.

There's lots of claims out there that LEDs can switch from fully off to fully on in 20ns, 2ns, or even instantaneous, I don't believe it myself.   Should I switch to flashing an old red laser pointer.

[Zero999]

Yes, no doubt there are faster photodiodes, than the one I used, but I knew it wouldn't be the weakest link, so deemed it to be reasonable.

Upping the forward voltage helps to increase the speed and I agree, the LED is probably the limiting factor. LEDs do have some capacitance, but interestingly I found it doesn't keep the LED on, probably because it discharges quickly enough. Here's an oscillogram showing the forward voltage of the  LXM2-PH01-0070, a plain red LED, with a more modest current than the one I posted above. When the MOSFET turns off, the voltage quickly drops, but then it takes awhile to decay. Your circuit is push-pull, which will be faster than a single ended MOSFET drive. You could try bypassing the current limiting resistor with a small, 10pF to 100pF capacitor, to speed things up.


And here's an oscillogram of the photodiode saturating, with it placed close to the LED. I didn't record which LED I used.

[StillTrying]

"You could try bypassing the current limiting resistor with a small, 10pF to 100pF capacitor, to speed things up."

I had to go up to 10n across the LED's 5R to get 10ns off the rising edge of the light shape, perhaps not too surprising with the low 5R and only a spare 0.2V above the LED voltage to play with. I'm not sure much is visible on the voltage directly across the blue LED, but I've seen before that a very hard to find change of 2 or 3mV on the voltage across the LED can make a very visible change to the received light shape. I'll leave the 10n on for now.
I don't want the change the 1us pulse applied to the LED too much because the original idea was to apply an as near perfect as possible 1us square pulse to the LEDs and view the differences in light shapes, especially between phosphor and non-phosphor colours.
I'll probably try flashing a red laser before attempting any more improvements to the PD side.

[Circlotron]

Would be interesting to *listen* to some of those waveforms while pointing at various light sources. I wonder what the sun sounds like?

[StillTrying]

"Would be interesting to *listen* to some of those waveforms while pointing at various light sources."

Many years ago I transfered audio through a torch bulb and slow LDR, similar to soldar's Reply#30, but just AM modulating the bulb. It would only transmit over about 2 feet, speech sounded muffled but perfectly understandable.

From experiments in this thread
https://www.eevblog.com/forum/projects/infrasound-sensor-what-makes-a-narrowband-2-7-hz-signal/msg1485156/#msg1485156

I can convert most of what the scope can record into an audible .wav. I can't think of a very interesting light shape to listen to. Maybe a TV remote slowed down by 4 times so the 38kHz becomes 9.5kHz.

[StillTrying]

The red laser out of an old red line maker couldn't be fast flashed, there's a good regulator inside it. Green is the current, a 400mA spike dropping to about 35mA @3V3.

[StillTrying]

200mA 1us through a 5mm superbright green LED gives a much faster light turn on and rise time ~19ns, I think that's the best I'm going to get, @ 200mA and 2.5% duty the green colour turns slightly bluey.
It's more proof that when used normally @ 5-20mA all LEDs turn on quite slowly, if 30-150ns counts as quite slowly!

A superbright medium-warm white 5mm @25mA had the slowest turn on and off light curves I've seen @ ~5us long, had to increase the 1us pulse width to 10us to fit them in. Green trace is the 3V directly across the white LED.

Testing complete!

[Zero999]

Yes, green LEDs shift towards blue, at higher currents.

The delay in the white LED could be due to the phosphor.

Here's an oscilloscope showing the phosphor lag of the LXM2-PL01-0000, a phosphor converted amber LED. I tested other LEDs, in the same range, but without the phosphor and they were much faster.
https://www.lumileds.com/uploads/265/DS68-pdf

[StillTrying]

After getting the light rise and fall times down to around 18ns by over driving a superbright green LED as above, I decided that would do, because that's 10 times faster rise times than any real life LED lamp or torch I've seen.

After another not very successful attempt to capture lightning flashes.

After some analogue/linear stuff.
www.eevblog.com/forum/beginners/floating-probe!-for-$2-50/msg1849505/#msg1849505

After viewing some neons and IR remote controls.

Playing with the 100W white COB including taking some 8 us @ 15 Amp photos.

I've run out of LED/light experiments to do.