measuring the long tail of a transient decay

[eigenroot]

I am trying to measure the long tail of a transient current decay of a photodiode which is excited by a pulsed laser. The transient from the photodiode is fed into a current amplifier and the amplified current is measured by an oscilloscope with 50ohm input impedance. See the circuit here View attachment circuit.pdf

My goal is to amplify the long tail of the photocurrent as much as possible, and I am not very interested in the initial part of the decay (say, from the peak to its 1%, spanning about 10us from the falling edge ). But this initial part limits the amplifer gain I can use because if I make the gain too large it either saturates the amplifier or overloads the oscilloscope. Right now I insert an analog switch between the photodiode and the amplifier, and turn on the analog swith with certain amount of delay (e.g., 10us) from the pulse falling edge, so that only the current 10us after the falling edge is passed to the amplifier. The problem is there is always charge injection effect in analog switch in real practice. Although the charge injection of the switch I use (ADG701) is only a few pC, I can still see a huge peak induced by the switch turn-on and turn-off in oscilloscope.

How can I get rid of the charge injection in this case? It is important since any injected charge will affect the weak and long-tail current. Or, is there a better way to circumvent the gain limit caused by the inital part of transient decay? Any thought is greatly appreciated!

 

[zorro]

Hi eigenroot,

How if you use a logarithmic amplifier?
There are monolythic log amps with excellent log characteristic over several decades.
Additionally, if you are interested in the measurement of the time constant of an exponential decay (really I don´t know if this is your interest), that shape becomes a linear decay whose slope can be readily determined.
Regards

Z

 

[FvM]

Below 1 % of initial photo diode signal, it may be tricky to determine, if the observed "long tail" is originated from the photodiode itself or just parasitic behaviour of your circuit. I wonder if you're sure which effects are brought up when you "insert an analog switch".

As long as you can't verify correct circuit operation by injecting a clean test pulse without long tail, I won't believe any measurement result.

 

[chuckey]

have you tried to balance the "charge injection" transient with an equal and opposite one generated by you?. Use an anti-phase square wave an a suitable adjustable capacitor (such as a pair of 2cm long twisted wires).
Frank

 

[sky_123]

You mention measuring the long tail, but (to me) it wasn't clear what parameter you were trying to measure.
If you already know the shape of the decay, then are you just trying to measure when it is approximately down to 1% of the initial value?
If so, then you don't need to measure it at 1%. You could measure the amplitude earlier (when, say it is down to
30%) and then make a computation.

 

[eigenroot]

You mention measuring the long tail, but (to me) it wasn't clear what parameter you were trying to measure.
If you already know the shape of the decay, then are you just trying to measure when it is approximately down to 1% of the initial value?
If so, then you don't need to measure it at 1%. You could measure the amplitude earlier (when, say it is down to
30%) and then make a computation.

Sorry for the confusion. I want to measure the photocurrent in the long tail part, maybe starting from 10us after the falling edge and the end point is preferrable hundreds of ms (or the point when the signal is overwhelmed by the noise). I know it may sound funny to get photocurrent over such a large time window, but I would like to resolve the information in the long tail which is related to the material properties of the diode.

- - - Updated - - -

Below 1 % of initial photo diode signal, it may be tricky to determine, if the observed "long tail" is originated from the photodiode itself or just parasitic behaviour of your circuit. I wonder if you're sure which effects are brought up when you "insert an analog switch".

As long as you can't verify correct circuit operation by injecting a clean test pulse without long tail, I won't believe any measurement result.

Thanks for your comment. If you mean the RC constant is a concern here, it is below 1us. And I am measuring the current starting from 10us and beyond (the current at t=10us is at least less than 5% of its peak value), which shows non-exponential decay shape, and apparently larger value than expoential decay. So I think the long tail is mainly not contributed by the parasitics.

But I agree that inserting an analog switch between the diode and the amplifier introduces an amount of injected charge that is not so easy to control. Maybe it is not the best way to do. Do you have alternative ways in mind? Thanks.

- - - Updated - - -

have you tried to balance the "charge injection" transient with an equal and opposite one generated by you?. Use an anti-phase square wave an a suitable adjustable capacitor (such as a pair of 2cm long twisted wires).
Frank

From the documentation from Analog Devices, they balance the charge injection by inserting a dummy capacitor between the gate and drain of NMOS to compensate for the gate-to-drain compacitance differnece between NMOS and PMOS. However, the charge injection is still a problem in my measurement (see the circuit View attachment circuit.pdf).

Can you elaborate on your "anti-phase square wave" method?

 

[FvM]

I didn't expect this long time range. I agree that a limiting amplifier ór similar should work in this case. If the photo diode shows the said effects, it seems necessary to create clear operation conditions for it. You didn't mention however the intended photo diode circuit.

I apreciate the log amplifier sugestion. Feeding the photo diode current directly to a transistor log circuit would allow to record the diode current over a maximum dynamic range without even limiting it.

 

[chuckey]

Switch your photo diode off, increase the CRO gain to max to inspect the transients (blow through from the analogue switching clock). Find or generate an edge that is anti-phase with these (clock inverter may be needed). generate a transient by coupling this edge into the amplifiers input via a very small capacitor, which you adjust to reduce the unwanted signal.
Frank