# Transimpedance Amplifier (TIA) - How to shunt high current spikes?

I am working on the design of a high-bandwidth transimpedance amplifier (TIA) to detect a relatively weak optical signal (100μW) but also be able to withstand short pulse, high power optical signals (>10W peak, 100ns long). The aim is to achieve a 50MHz bandwidth with a transimpedance gain of 5kΩ. I wish to use the FGA21 photodiode and the OPA657 Op-Amp to achieve this, key specifications include:

• Gain Bandwidth Product: GBW = 1.6GHz
• Photodiode Capacitance: C_d > 100pF

To limit the effect of input capacitance and push the bandwidth out I am using the bootstrapped architecture described by Hobbs. The SPICE model of this circuit is illustrated in the figure below:

Simulations suggest that this can achieve the required gain and bandwidth. However, I need to extend this design further so that it can shunt the ~10A pulse generated in an efficient manner. My thought is to use a Zener diode between the inverting and non-inverting op-amp inputs to regulate the differential voltage and provide a low impedance path for the current spike to ground. For this I require a Zener capable of handling high currents and that has low capacitance (as any capacitance >5pF will limit my TIA bandwidth), which is a big ask.

So I am wondering if there alternative methods to shunt high currents with minimal capacitance or is there a way to increase the maximum power dissipation of Zener diodes?

• Hobbs also wrote a nice book. Do you have it? – jonk Apr 24 '18 at 1:51
• Also, are you using a beam-splitter situation and a partial pass mirror (dichroic or otherwise) using an emitter that overwhelms the amplifier? Perhaps a fluorescence or phosphorescence application? – jonk Apr 24 '18 at 1:55
• @jonk Sadly I do not have a copy of the book, but based off of the bits I've read about bootstrapping it would be very useful to have a copy! – J-Pease Apr 24 '18 at 1:58
• I have a copy on the shelf. And you didn't answer me about the splitter situation. Is it? – jonk Apr 24 '18 at 2:02
• @jonk In regards to the optics I have two sources, one is a high power pulsed laser (100ns pulses, 10kW peak power) the other is a low power CW laser (~10mW). The issue is that for my experiment these two sources are matched in wavelength and polarisation so there is no easy way to seperate the two using optics such as dichroic mirrors or polarisers. I can attenuate to reduce the incident power, but then I'll need higher gain to see the CW source. – J-Pease Apr 24 '18 at 2:02

Since the current will be converted into voltage at the op-amp input I would suggest a pair of anti-parallel 1N4148 diodes which will clamp at ~.7 volts.

Their capacitance is 4 pF and recovery time is 4 ns. Maximum clocking rate is 100 MHZ @ 2 volts 50% duty cycle. Their continuous limit is 100 volt @ 200 mA but it is wise to keep volts and current to 50% of maximum for long life.

Their are tiny SMD transzorbs with a capacitance of just 1.5 pF, but I do not know for sure how many joules they can handle in 100 ns. Their KA rating is usually in the uS range.

The problem with tranzsorbs is they do not clamp less than 5 volts. They are more like back-to-back zener diodes.

• Thanks for the very helpful response! I haven't come across the 1N4148 diodes before, having looked at the spec sheet the peak current they can handle is very impressive (2A in 1us). In general do you believe that this maximum peak current will improve as the pulse time is reduced? – J-Pease Apr 24 '18 at 3:49
• Well, if your talking of current over time then that is joules. However a joule is one watt/second so for 100ns that is ten million times smaller. By the way, you may have to buy the 1N4148 in a package of 200 for about \$5 USD. They are handy for voltage clamps and fast low-current rectifiers. – Sparky256 Apr 24 '18 at 3:56
• Sorry for the confusion, when I say the maximum peak current I am referring to the maximum current the 1N4148 can tolerate. I guess what I am trying to ask is that if you know whether the power dissipation of diodes improves as the impulse time decreases (i.e. if the diode is rated to 2A peak in 1us in 100ns it could handle >2A peak)? – J-Pease Apr 24 '18 at 4:17
• Yes of course. Temperature of the diode junction is what is important. At 100ns the diode may tolerate 10 amps, but the datasheets do not clarify that. – Sparky256 Apr 24 '18 at 4:46
• Thanks again for all the help, I think there is no better test than putting them in and seeing if they survive! – J-Pease Apr 24 '18 at 4:49

I think you've read the data sheet and seen the responsivity figure: -

$$\Re (λp) = 1.04 A/W (typ)$$

And concluded that if the incident power is 10 watts, then you'll get 10 amps from the device. This won't happen by a long way for one of these: -

You need to speak with ThorLabs to find out what it could be but I think it will be a handful of mA maximum. After all, if you shone the laser at it and it was unconnected to a circuit would it produce an arc across its terminals of 10 amps? You also need to speak with ThorLabs to see that it can handle the incident power of 10 watts in 3.1 sq mm of optical window (3.2 MW per sq metre).