Another (ultra)fast photodiode design circuit

[stcoso]

Hi everyone...
One year ago i was designing a TIA around a BPV10 pin photodiode ( https://www.eevblog.com/forum/beginners/photodiode-circuit-design/ ) to monitor the light output from a LEDs driver that i've built. Since then i've learned a lot and designed other (faster) TIA with better opamps around the same photodiode (bpv10)to accomodate for my necessities (faster LEDs driver).

Now i'm a bit stuck .

Let's say i need to measure light waveforms in the uv region with pulse width of 1ns and peak irradiance at about 2(kW)/((m^2)(s)). How can i achieve this (other than buying a properly built product  )? Can a circuit like this be designed to operate linearly (inside a 5% range) from dark to peak irradiance?
What about linearity with respect to pulse width? Can be built to operate (linearly) even with continuos light waveforms?

[tggzzz]

You should invest in Phil Hobb's book https://www.electrooptical.net/Building_ElectroOptical_Systems/

Alternatively, if you can ask a good and interesting question which doesn't look like you are merely after free consultancy, he may respond on usenet group sci.electronics.design.

[stcoso]

Thanks! I'm already looking at it!

[jmelson]

pulse width of 1ns

YIKES!  I hope you have a VERY small PIN photodiode, and very excellent circuit construction.  This is going to be pretty tough!
I never work with anything that fast, and generally anything with a risetime shorter than 10 ns gives me fits.

Jon

[stcoso]

The photodiode is just the first (and probably the main) of a long list of problems  ... Especially for the UVA range.

The BPV10 that I'm using has (from datasheet) 2ns rise/fall when opportunily biased.
Looking at mouser/digikey/rs/farnell catalogues to find something suitable (rise/fall time and spectral range) led me to look at SiPM...

[jmelson]

Looking at mouser/digikey/rs/farnell catalogues to find something suitable (rise/fall time and spectral range) led me to look at SiPM...

SiPMs have a lot of small avalanche photodiodes paralleled with a series quench resistor.  I think the quench resistor will interfere with the fast response you need.  They have LOTS of gain for very low light levels, which doesn't sound like what you need.

Jon

[awallin]

For simulating bandwidth and noise-floor I wrote some python scripts over here: https://github.com/aewallin/TIASim they are based on the formulas in Hobbs and other references.

So far I only built a ca 60 MHz BW detector with S5973 photodiode, OPA657 opamp, and RF 10 kOhm.
I am building one withe HMC799 next, stay tuned... (https://www.analog.com/en/products/hmc799.html)
The new TI opamps like OPA859 also look interesting (but WSON footprint is harder to solder), because they have lower input-capacitance than the older OPA657.

For 1 GHz you probably need a photodiode with <1pF capacitance (at suitable bias), and an op-amp + layout with minimal input capacitance and parasitic capacitance over RF. Rumor has it that e.g. Thorlabs FDS015 and FDS025 are in fact Hamamatsu/other? photodiodes - so they might be available elsewhere also?

If a 'digital' photoreceiver with just 0 or 1 output is suitable, then 1 Gbit SFP optics for telecom are dirt-cheap, and probably available for ~800nm, 1300nm, 1500nm. As a bonus the SFP has a TX channel also

[stcoso]

I've a Si PIN photodiode candidate: https://www.hamamatsu.com/eu/en/product/type/S9055-01/index.html

Not easy to find though

[stcoso]

Looking at mouser/digikey/rs/farnell catalogues to find something suitable (rise/fall time and spectral range) led me to look at SiPM...

SiPMs have a lot of small avalanche photodiodes paralleled with a series quench resistor.  I think the quench resistor will interfere with the fast response you need.  They have LOTS of gain for very low light levels, which doesn't sound like what you need.

Jon

Yep... The fastest SIPMs have got a fast output with sub-ns edge. But the fact that they're sensitive even to single photons is a problem. I've got something like 10^18 photon/m^2s

[mcookieman]

The new TI opamps like OPA859 also look interesting

Beware that the input current noise is not white for FET input opamps...

[khs]

Yep... The fastest SIPMs have got a fast output with sub-ns edge. But the fact that they're sensitive even to single photons is a problem. I've got something like 10^18 photon/m^2s

I think there is no need for an amplfier.

Just connect some reverse voltage to the diode and connect it directy to the 50 Ohm input of your oscilloscope.

[Marco]

Why would you need a transimpedance amplifier with 0.5 pF of capacitance? Can't you just feed it directly to a high gain 50 Ohm matched MMIC?

50*0.5e-12 doesn't seem a relevant RC time for a couple GHz signal.

[stcoso]

I think there is no need for an amplfier.

Just connect some reverse voltage to the diode and connect it directy to the 50 Ohm input of your oscilloscope.

Why would you need a transimpedance amplifier with 0.5 pF of capacitance? Can't you just feed it directly to a high gain 50 Ohm matched MMIC?

50*0.5e-12 doesn't seem a relevant RC time for a couple GHz signal.

Inspired by both of this comments i tried with the bare BPV10 pin photodiode that i have at hand. When biased at 50V it as a 2pF Cj (with 50 Ohm is -3dB at 1.6GHz)... It actually works better than i remember but show it's limits with less than 0.05A/W at the wavelenght of interest.

I'll look into MMIC.

Thank you all.