What is photodetector bandwidth?

[faryafada]

Hi,
I would like to know how a detectors bandwidth can be measured? for example, if there is any relation with the produced current or what? I can't understand bandwidth concept.
Thank you.

 

[skeptic2]

Before you can measure the bandwidth, you must understand what you are measuring. First you need to decide which specific detector you are going to use.

What is it you don't understand? How were you thinking of measuring the bandwidth?

 

[davenn]

you would need to look at the manufacturers data sheet for a specific photodetector

Dave

 

[berkeman]

Hi,
I would like to know how a detectors bandwidth can be measured? for example, if there is any relation with the produced current or what? I can't understand bandwidth concept.
Thank you.

As already asked, what specific detectors are you looking to use? Are you going to use a pre-packaged module, or will you build it yourself with a PIN photodiode and amplifier? If you are building your own, be sure to use a good reverse bias voltage on the PIN photodiode to help get the best bandwidth out of the circuit.

 

[faryafada]

As already asked, what specific detectors are you looking to use? Are you going to use a pre-packaged module, or will you build it yourself with a PIN photodiode and amplifier? If you are building your own, be sure to use a good reverse bias voltage on the PIN photodiode to help get the best bandwidth out of the circuit.

It is a PIN photodiode and I am in simulation level. not any experimental work yet. Is it possible to measure bandwidth in simulation level or you need to build and measure it?

 

[skeptic2]

Does this help?

http://www.vishay.com/docs/81502/bpv10.pdf

 

[berkeman]

Does this help?

http://www.vishay.com/docs/81502/bpv10.pdf

Holy cow! 250MHz at Vr = 12V. That's impressive!

 

[dlgoff]

Holy cow! 250MHz at Vr = 12V. That's impressive!

I'm wondering if that's at an operating temperature of -40ºC.

 

[sophiecentaur]

Hi,
I would like to know how a detectors bandwidth can be measured? for example, if there is any relation with the produced current or what? I can't understand bandwidth concept.
Thank you.

In principle, it's no different from measuring the bandwidth of any receiver. You need to produce a modulated signal from a test source and then see what your detector (demodulator) gives you. The wider the bandwidth of the receiver, the more data or the faster the analogue signal it can detect. (Is that what you had in mind for your question - sorry if it's talking down but it's hard to assess your initial level of knowledge.
An obvious (?) way would be to amplitude modulate light of a suitable wavelength and see how the shape of the modulating signal waveform is affected - e.g. look for the effect on a very fast pulse. Alternatively, use a swept frequency for modulation and see the amplitude of the signal that your detector demodulates at different modulating frequencies.

Btw, thw signal bandwidth is distinct from the "spectral bandwidth" which is the range of wavelengths that the detector can actually detect - see the spec sheet in the link.

 

[faryafada]

In principle, it's no different from measuring the bandwidth of any receiver. You need to produce a modulated signal from a test source and then see what your detector (demodulator) gives you. The wider the bandwidth of the receiver, the more data or the faster the analogue signal it can detect. (Is that what you had in mind for your question - sorry if it's talking down but it's hard to assess your initial level of knowledge.
An obvious (?) way would be to amplitude modulate light of a suitable wavelength and see how the shape of the modulating signal waveform is affected - e.g. look for the effect on a very fast pulse. Alternatively, use a swept frequency for modulation and see the amplitude of the signal that your detector demodulates at different modulating frequencies.

Btw, thw signal bandwidth is distinct from the "spectral bandwidth" which is the range of wavelengths that the detector can actually detect - see the spec sheet in the link.

Thanks for your explanation, Sophie. I know something about bandwidth but I am concerned about any theoretical solution for measuring bandwidth, if any. i.e. how is it possible to measure the bandwidth if I have the current (I output) results for a photodetector?

 

[sophiecentaur]

Thanks for your explanation, Sophie. I know something about bandwidth but I am concerned about any theoretical solution for measuring bandwidth, if any. i.e. how is it possible to measure the bandwidth if I have the current (I output) results for a photodetector?

Like I said before, you need to provide the detector with a modulated input signal, swept or impulse. You then look at the I_output waveform and see how the system has modified it. This is what you'd do to measure the bandwidth of any channel. Of course, a single figure for 'bandwidth' would refer to the 3dB bandwidth (or whatever else you care to specify) but that, in itself, is not enough to char
acterise the system fully. How do you want to specify the performance? What parameters are relevant to your purpose?
Actually, I am not sure what you actually mean by your question. What is your level of knowledge of signalling channels in general? What do you understand the term 'bandwidth' to mean, in the context of this application? It's a word that has at least three connotations.

 

[meBigGuy]

Given a spice model for a photodetector, you can connect it to an amplifier model, include estimates of parasitic (pcb?) capacitance and measure the system response (or just measure the photodetector current to begin). The model will have a calibrated way to provide stimulus. You sweep the frequency of the input, and see how the amplifier or current output varies as the frequency increases. The 3dB bandwidth is simply the frequency at which the output drops 3dB.

 

[sophiecentaur]

Isn't spice a simulation? How does that help with actually measuring the characteristics of a real detector? Have I missed something about Spice? Is there a hardware version?
3dB bandwidth is a partial solution to finding the 'usefulness' of the device (for instance, the noise bandwidth). It's fine if you can treat it as a single pole low pass filter but, more detail would be needed if you want to include it in a channel with shared Tx Rx filtering (a common requirement).

 

[the_emi_guy]

It is a PIN photodiode and I am in simulation level. not any experimental work yet. Is it possible to measure bandwidth in simulation level or you need to build and measure it?

Interesting question. Bandwidth of photodiodes depends on things like material, doping levels, and physical size (physically larger = lower bandwidth). You would need to extract equivalent circuit from these physical characteristics.

You may want to do a literature search of this. Are you in a university environment where you have access to journals?

I have attached an example, Let us know what you find.

 

[meBigGuy]

The OP said "It is a PIN photodiode and I am in simulation level. not any experimental work yet. "

One of the first steps to understanding the actual bandwidth would be a simple frequency sweep in spice. No argument that more analysis is appropriate, as are more complete models. 1 step at a time.

 

[the_emi_guy]

The OP said "It is a PIN photodiode and I am in simulation level. not any experimental work yet. "

One of the first steps to understanding the actual bandwidth would be a simple frequency sweep in spice. No argument that more analysis is appropriate, as are more complete models. 1 step at a time.

Yea, we are all working on very little information from the OP. I am assuming he knows how to drive his simulator (whatever it is) and needs to know how to develop a model for his photodiode.

Maybe I am wrong and OP needs tutorial on spice.

@OP: more info please. What exactly do you mean by "I am in simulation level".

 

[carlgrace]

If you want a rough idea of the bandwidth, the easiest way (and it works great in simulation) is to put a sharp pulse into the detector and measure the rise time of the response. If it is close to first-order then the bandwidth is about 0.35 over the rise time of the detector. (so if the 10%-90% rise time of the detector is 1 ms, the BW is about 0.35/1ms or about 350 Hz.)

 

[Njorl]

I'll assume your simulator is a pencil and paper. Calculate the capacitance of your detector. Multiply by the load resistance. Take the inverse. That's a decent approximation of your bandwidth. You could use the intrinsic resistance built into your detector instead of load resistance, but nobody ever does that.

The capacitance calculation is the only difficult part. You can treat a mesa detector as a parallel plate capacitor - area is the size of the mesa, separation is the thickness of the undoped region, don't forget the dielectric isn't a vacuum. Interdigital finger detectors are a more complicated calculation.

 

[sophiecentaur]

What I don't understand is how, if you are doing a simulation, someone hasn't already defined the frequency response of the photodetector.? Else, how can it have been simulated?

 

[berkeman]

For the OP -- I did a google search on Modeling A PIN Photodiode In Spice, and got a lot of good hits. Try the same search if you'd like to read through the list.

 

[meBigGuy]

What I don't understand is how, if you are doing a simulation, someone hasn't already defined the frequency response of the photodetector.? Else, how can it have been simulated?

We are talking spice models. The bandwidth is not a parameter for a typical spice model. Models are made up of current sources, voltage sources, transistors, resistors, capacitors, etc. When you simulate in a proper testbench, with proper stimulous, loads, amplifier models, and parasitics, you can determine the bandwidth response of the device under test.

But, we are just talking to ourselves as the OP has gone away.

 

[sophiecentaur]

Presumably, the simulated components are based on measurements of a real detector? A light / current transducer may not have a simple equivalent circuit model but surely the process is to start with the measured performance and then work out the list you quoted, using a minimum of parameters to give a 'good enough' match to reality.
I have lost track of this a bit but isn't the detector in question a 'real' one, somewhere down the line? I suppose my main problem was with the word "measurement", which made me think of wires and dials. haha

The use of simulations is an interesting phenomenon. Is there not a danger of over-reliance on the modelling process, for students, particularly? There must be a tendency to assume that parasitics have all been included, followed by total surprise when the system does something strange, when it's actually built. The real-life experience must come a lot later in the learning process and I realize that simulation is the only way into the production process for complicated systems. I seem to remember things, like which way round to connect a device and how to solder reliably, came in the first few months of my Electronics education. I still assume that it's down to my building skills as much as anything else, when something doesn't do what was expected.

 

[meBigGuy]

The IC industry totally relies on spice models. Every thing is simulated over voltage, temperature and process and back-annotated for best-case and worst-case parasitics. Board designers also rely on models, especially for tricky circuits. Manufacturer's component models are extracted or developed from the circuit design and then validated through testing and characterization. They are not developed from test data.

Modern designers have to rely on models, and have to know how to rely on them or else they are going to take multiple turns to get designs into production, and then redesign them again as they come across production volume component variations. With good modeling and good design skills you can get it all correct with one turn. Our company created a mixed signal chip containing a CMOS 5th order sigma delta class D pwm power amplifier with dynamic feedback, and it worked 1st turn and was shipped in volume. Can only be done with simulation-simulation-simulation.

Designers always need to get burned a few times before they understand how important the details around the models are.

The way school should approach this (and maybe they do) is to provide the student with a board containing basic circuits, and component models for the componets used. The students should annotate parasitics, simulate and then characterize the board in the lab, comparing it to simulation.

Prototyping and breadboarding still have a place, but it is becoming less practical as components get smaller and designs get denser and models become more accurate. The soldering skills now are how well you can change an 0402 resistor under the microscope, or add an 0603 bypass capacitor.

 

[sophiecentaur]

I would be extremely surprised if the equivalent circuit to a length of conductor, a transducer or a solid state device was not derived from measurements somewhere. Predicting the EM behaviour of a bit of copper on a substrate theoretically would be pretty fraught, for instance. That was my original point. Of course, Spice has to use the parameters it 'knows about' but you can bet your life they came from some original measurements of physical quantities.

I do take on board the absolute necessity for simulation for large and complex circuits, though. There would be no point in breadboarding an integrated circuit just to see if it worked. The behaviour of the breadboard realisation would be too far from the target device.

BTW Very impressed by the story about your company's successful, first time design. Some clever people at work there.

 

[meBigGuy]

I love my work. Surrounded by people smarter than me that are highly professional and self motivated (like this forum --- lol). A premium "startup".

I'm sure you are correct in that measurement might sometimes be used to set parameters. But the models are generally pretty sophisticated and the parameters are often 1 removed from the way we intuitively think of the functional characteristics of the component.

Consider this model for a UMX-5601 PIN diode.
http://www.microsemi.com/index.php?option=com_docman&task=doc_download&gid=13985
How would you measure the parameter "I region stored charge vs. frequency"

Based on measurements you might tweak parameters (Better know what you are doing), but I call that characterization or validation of the model. The model is theoretically based and the parameters are generally determined by other simulations (say, at the semiconductor physics level) or parameter extraction processes.

Even traces on a circuit board are modeled and simulated by 3D solvers based on materials and thicknesses, including corner effects and all. PADS PCB software includes that capability. Of course you can start with simple lumped elements, advance to simple lossy transmission lines based on PCB and metal geometry/materials or move on to the 3D solutions.

 

[sophiecentaur]

I love my work. Surrounded by people smarter than me that are highly professional and self motivated (like this forum --- lol). A premium "startup".

Been in that position. You need native cunning to survive with people that smart. Great environment.

How would you measure the parameter "I region stored charge vs. frequency"

By looking at the very effects it causes - namely integration and phase change. I could ask how you could arrive at it theoretically and be sure of your result.

Even traces on a circuit board are modeled and simulated by 3D solvers based on materials and thicknesses, including corner effects and all. PADS PCB software includes that capability. Of course you can start with simple lumped elements, advance to simple lossy transmission lines based on PCB and metal geometry/materials or move on to the 3D solutions.

A similar situation exists with antenna design- there are programs for calculating the way a metal shape will behave as a radiator but, in the end, you have to measure it and modify the predictions before you use them in a (simulation) link budget or propagation program.

I think this conversation is more a matter of terminology than anything. I think we could just agree to differ in where we are putting the relevance of measurements. There are many good reasons (you have mentioned some of them) why simulation is the backbone of your field. That is pragmatic. I am just making the point that any mathermatical characterisation of a component must be based on measurement, somewhere along the line - either initially or as part of the characterisation procedure.

 

[meBigGuy]

I am just making the point that any mathermatical characterisation of a component must be based on measurement, somewhere along the line - either initially or as part of the characterisation procedure.

Can't argue with that.