Where can I get a decent sensitive high-speed photodiode/phototransistor?

[Ben321]

I'm trying to send an optical signal across a room with an LED. A photodiode or phototransistor is needed on the receiving end. I'm using a Picoscope to digitize the signal, and will use my own software to decode it.

Problem is the bandwidth of the PD or PT I'm trying to use now is too narrow. Even just a signal with a constant frequency, of about 15khz, is about the fastest signal I can get through reliably. On the Picoscope software's display, there's an obvious exponential decay on the falling edge of a square wave. If the frequency is greater than about 15khz, the decay of the last cycle merges into the next cycle. I know very high speed data transfer should be possible with an LED/PD or LED/PT pair, but some capacitive effect is coming into play here, slowing the decay speed of the signal to an unacceptable level.

Problem is, I got the PD or PT out of another device that I took apart, and repurposed its components. I'm not sure even if it's a PD or a PT. The only thing I know it's not is a CdS cell (which has a distinctive look to it with a wavy line down the middle). The component I have looks like an LED, except I know its use in the device I took it out of was as a light sensor, not a light emitter. So I know it's either a PD or PT. I'm not sure if the component is a variant of PD or PT which is designed to have a slow response (as opposed to variants of PD or PT used in optical communications, which are built to have fast response times), or if the problem is with the circuit I'm using.

Of course, if I bought a PD or PT with good specs, sensitive and high-speed, then I could eliminate the PD or PT as the source of my problem, and work on debugging other parts of my circuit.
The diagram for the circuit I'm using (in both the PD and PT configurations) is attached to this post. The battery is 9V and resistor is 47kOhms.


If anybody knows of a good place to buy PDs or PTs that are both fast and highly sensitive, please let me know. It must be both fast and highly sensitive (not just fast), because my experiment is to send an optical signal across the room through the air, and not via fiber-optic cable, so there is a significant decrease in intensity of the light by the time it reaches the light sensor.

[james_s]

Photodiodes in general tend to be very fast, much faster than phototransistors. You need to use them with a transimpedance amplifier though and if you're using a high sensitivity photodiode they tend to have large areas and correspondingly high capacitance so you normally run them with a negative bias which effectively reduces the capacitance.

You might find this useful. https://www.analog.com/designtools/en/photodiode/

[tggzzz]

Firstly, post all the information not just some of it, e.g. voltages, resistors, waveforms. Why? See https://entertaininghacks.wordpress.com/library-2/good-questions-pique-our-interest-and-dont-waste-our-time-2/

Secondly in a free-space application consider that the output you are measuring will be the sum of the signal and ambient light. A more "sensitive" circuit's output will be saturated by a lower ambient light level.

Thirdly the ambient light level will result in a current that has an associated irreducable "shot noise". That may lead to signal-to-noise problems, depending on your link budget.

Hence, do the link budget calculations for your specific design and location, including ambient light and noise.

[Marco]

With an avalanche photodiode you won't necessarily need a transimpedance amplifier, though they aren't cheap they do often come with a bandpass filter you probably need any way.

Fast can mean a lot of things, what kind of bandwidth do you want and at what cost?

[Ben321]

Photodiodes in general tend to be very fast, much faster than phototransistors. You need to use them with a transimpedance amplifier though and if you're using a high sensitivity photodiode they tend to have large areas and correspondingly high capacitance so you normally run them with a negative bias which effectively reduces the capacitance.

You might find this useful. https://www.analog.com/designtools/en/photodiode/

Photodiodes are NORMALLY run with negative bias (it's not special to high sensitivity ones as your comment implied). In positive bias, they just conduct like a wire. A diode's special ability comes into play, ONLY when in reverse bias. Normal diodes block in the reverse direction. Photodiodes become photosensitive in the reverse direction (more current flows for more light hitting it).

As for transimpedance amplifier (current to voltage converter), you'll see my circuit does just that. The light sensor is wired in series with a resistor, and the voltage is measured across the resistor. As for the "amplifier" aspect of "transimpedance amplifier", my voltage measuring device (my Picoscope 2204A) has one builtin. The device itself is basically just a voltage amplifier who's output goes to the input of an analog to digital converter, and the output of the ADC then goes to a USB interface to send the data to the computer.

If you have any questions about my circuit, check the circuit diagram in my opening post in this thread. For the values of the components, read the last sentence of paragraph 4 in the opening post.

[Ben321]

With an avalanche photodiode you won't necessarily need a transimpedance amplifier, though they aren't cheap they do often come with a bandpass filter you probably need any way.

Fast can mean a lot of things, what kind of bandwidth do you want and at what cost?

The bandwidth of my PC-based ocilloscope (the Picoscope 2204A) is about 3MHz when using my own custom software written with official Picoscope SDK, or 500KHz when using the official Picoscope 6 software.
I'd like to use the full bandwidth of my scope (a little over 3MHz) if possible. That's how fast the entire signal path needs to be, all the way from the photodiode, up to the Picoscope itself.

I wonder if any companies make photodiode modules, instead of discrete components, where all the amplifier circuits etc are already in the module. Just apply power to the module from an external power supply or battery via its DC-power port, and a signal voltage is output from a BNC connector (or other similar connector) to whatever signal monitoring device (in my case the Picoscope) is being used.

Alternatively, maybe a photovoltaic cell could work as an optical signal receiver. How fast are photovoltaic cells? I mean like single cells, not complete solar-panels. And where can you even buy single cells? Most places that sell them, even small ones, are a complete panel, several inches wide (not a single cell about 0.5in on any side).

[tggzzz]

Photodiodes in general tend to be very fast, much faster than phototransistors. You need to use them with a transimpedance amplifier though and if you're using a high sensitivity photodiode they tend to have large areas and correspondingly high capacitance so you normally run them with a negative bias which effectively reduces the capacitance.

You might find this useful. https://www.analog.com/designtools/en/photodiode/

Photodiodes are NORMALLY run with negative bias (it's not special to high sensitivity ones as your comment implied). In positive bias, they just conduct like a wire. A diode's special ability comes into play, ONLY when in reverse bias. Normal diodes block in the reverse direction. Photodiodes become photosensitive in the reverse direction (more current flows for more light hitting it).

As for transimpedance amplifier (current to voltage converter), you'll see my circuit does just that. The light sensor is wired in series with a resistor, and the voltage is measured across the resistor. As for the "amplifier" aspect of "transimpedance amplifier", my voltage measuring device (my Picoscope 2204A) has one builtin. The device itself is basically just a voltage amplifier who's output goes to the input of an analog to digital converter, and the output of the ADC then goes to a USB interface to send the data to the computer.

That is wrong in all respects, too many for me to bother to correct.

You need to do some basic research on the topic.

[nfmax]

Try https://www.newport.com/c/high-speed-photoreceivers. More than enough bandwidth for your application, it would seem.

[Ben321]

Try https://www.newport.com/c/high-speed-photoreceivers. More than enough bandwidth for your application, it would seem.

At prices over $1000, I'm gonna have to say no to that option. Any alternatives in a hobby price range?

[Kleinstein]

Phototransistors tend to be slow. Even normal photodiodes are already relatively fast (e.g. good for 100 MHz) - the speed is usually more set by the amplifier circuit.

To reduce noise one should limit the wavelenghts and if possible the angle of acceptance. Chances are IR in the 950 nm range like used with remote control could be a good solution, as there is not that much background at the wavelength.

There are ready made chips for data transfer. It now is phased out for some time, but one may still be lucky to get some IrDA parts.
For low speed there are modules / chip for the IR remote controls (some 38 kHz carrier and modulation up to maybe some 2000 baud).

[nfmax]

Photodiodes in general tend to be very fast, much faster than phototransistors. You need to use them with a transimpedance amplifier though and if you're using a high sensitivity photodiode they tend to have large areas and correspondingly high capacitance so you normally run them with a negative bias which effectively reduces the capacitance.

You might find this useful. https://www.analog.com/designtools/en/photodiode/

Photodiodes are NORMALLY run with negative bias (it's not special to high sensitivity ones as your comment implied). In positive bias, they just conduct like a wire. A diode's special ability comes into play, ONLY when in reverse bias. Normal diodes block in the reverse direction. Photodiodes become photosensitive in the reverse direction (more current flows for more light hitting it).

As for transimpedance amplifier (current to voltage converter), you'll see my circuit does just that. The light sensor is wired in series with a resistor, and the voltage is measured across the resistor. As for the "amplifier" aspect of "transimpedance amplifier", my voltage measuring device (my Picoscope 2204A) has one builtin. The device itself is basically just a voltage amplifier who's output goes to the input of an analog to digital converter, and the output of the ADC then goes to a USB interface to send the data to the computer.

If you have any questions about my circuit, check the circuit diagram in my opening post in this thread. For the values of the components, read the last sentence of paragraph 4 in the opening post.

From your original post, it sounds like the device you have has around 200pF effective capacitance, so that it achieves around 15kHz bandwidth with a load resistance of 47 kohm. To get a faster response, you could use a photodiode with a smaller capacitance - but to achieve this, it would have a smaller active area, and so give a lower photocurrent for the same luminous flux. In some cases, such as your optical data link, this can be mitigated by focussing the received light into a smaller area. Think of a telescope pointed at your LED, focussed to generate a real image in front of the eyepiece: place your photodiode here. It needs to be only just big enough so that the whole of the image falls on its active area. Higher magnification gives a smaller image (until diffraction becomes significant) but makes alignment more critical.

Using a focussing telescope has the additional advantage that (if correctly-designed baffles are used) the amount of stray light reaching the photodiode can be greatly reduced. As @tggzz points out, this gives excess photocurrent, which even if it doesn't saturate your amplifier, will give increased shot noise.

Also, of course, the telescope has to be correctly aligned & focussed, and kept that way. This may not be trivial.

As most other posters have suggested, you can make things a lot easier for yourself by using a transimpedance amplifier, with a low input impedance. This means that instead of being loaded by 47 kohm your photodiode may see only an ohm or so. This raises the bandwidth dramatically. You do this by moving your load resistor to the feedback path of an OPAMP, with the photodiode connected directly to the non-inverting input. The Analog Devices tool posted by @james_s shows this configuration. Using this tool, and assuming a 200 pF photodiode capacitance, I was able to achieve a 3MHz bandwidth and a nicely-damped pulse rise time of 170ns, using the two stage configuration. (You should AC-couple between the two stages to remove the effect of background light)

The AD tool allows you to set the photodiode bias voltage, but this makes little difference. Often, you use a DC 'bias' voltage of 0V. This gives the maximum photodiode capacitance, but also gives zero leakage current - important if you want to measure absolute light levels. Your application may benefit from a reverse bias of a few volts (check the photodiode datasheet) but beware - noise on the bias supply plays directly into the amplifier input through the 200pF photodiode capacitance & gets the full gain. As you appreciate, you don't want to operate in photovoltaic mode, where the photodiode is forward-biased - stored charge will massively slow its response.

Almost none of the above depends on whether the device is a photodiode or a phototransistor, except for the speed of response. For that, you want a photodiode and preferably a p-i-n photodiode. But at your modest bandwidth I don't think that is essential. The actual photoelectric effect within the device is very fast (picoseconds). What slows it down is the transit time of the photocarriers to the device terminals, and the self-capacitance of the device, if working into a finite load.

Bob Pease wrote some about this here: https://www.electronicdesign.com/technologies/analog/article/21801223/whats-all-this-transimpedance-amplifier-stuff-anyhow-part-1, and the book by Jerald Graeme he mentions is also a good read.
Graeme, Jerald G. Photodiode Amplifiers: Op Amp Solutions. New York: McGraw Hill, 1996, ISBN 978-0-07-024247-0
Phil Hobbs also covers the development of transimpedance amplifiers for photodiodes, for example in:
Hobbs, Philip C. D. Photodiode Front Ends: The Real Story. Optics and Photonics News 12, no. 4 (1 April 2001): 44–47. https://doi.org/10.1364/OPN.12.4.000044.
And of course, the 'Bible":
Hobbs, Philip C. D. Building Electro-Optical Systems: Making It All Work. Wiley Series in Pure and Applied Optics. New York: Wiley, 2nd Ed. 2009.

I believe there is a PDF of the Graeme book lurking on the web somewhere

[Marco]

Mouser has some evaluation modules for LT1328 in stock ... I'd get those.

Simple On-Off-Keying will be very sensitive to interference and doing more complex modulation yourself will be a bit of a pain, the LT1328 solves a couple of problems for you.

[james_s]

Photodiodes in general tend to be very fast, much faster than phototransistors. You need to use them with a transimpedance amplifier though and if you're using a high sensitivity photodiode they tend to have large areas and correspondingly high capacitance so you normally run them with a negative bias which effectively reduces the capacitance.

You might find this useful. https://www.analog.com/designtools/en/photodiode/

Photodiodes are NORMALLY run with negative bias (it's not special to high sensitivity ones as your comment implied). In positive bias, they just conduct like a wire. A diode's special ability comes into play, ONLY when in reverse bias. Normal diodes block in the reverse direction. Photodiodes become photosensitive in the reverse direction (more current flows for more light hitting it).

As for transimpedance amplifier (current to voltage converter), you'll see my circuit does just that. The light sensor is wired in series with a resistor, and the voltage is measured across the resistor. As for the "amplifier" aspect of "transimpedance amplifier", my voltage measuring device (my Picoscope 2204A) has one builtin. The device itself is basically just a voltage amplifier who's output goes to the input of an analog to digital converter, and the output of the ADC then goes to a USB interface to send the data to the computer.

If you have any questions about my circuit, check the circuit diagram in my opening post in this thread. For the values of the components, read the last sentence of paragraph 4 in the opening post.

Well, you can lead a horse to water, but you can't make it drink...

[Ben321]

From your original post, it sounds like the device you have has around 200pF effective capacitance, so that it achieves around 15kHz bandwidth with a load resistance of 47 kohm. To get a faster response, you could use a photodiode with a smaller capacitance - but to achieve this, it would have a smaller active area, and so give a lower photocurrent for the same luminous flux. In some cases, such as your optical data link, this can be mitigated by focussing the received light into a smaller area. Think of a telescope pointed at your LED, focussed to generate a real image in front of the eyepiece: place your photodiode here. It needs to be only just big enough so that the whole of the image falls on its active area. Higher magnification gives a smaller image (until diffraction becomes significant) but makes alignment more critical.

Thanks for that info. Telescope will be impossible though. My photosensor is a bare component. I have nothing to mount it to a telescope's eyepiece.
Also, in my case though, an LED isn't actually what's being used. It could be used for sending digital data, but in my case, I'm actually trying to read an image off of a normal CRT (the old green monocrhome type). Each pixel on the screen is illuminated only at the moment the electron beam hits it (unlike with modern LCD and similar screens where all the pixels are illuminated simultaneously), so as a whole, the CRT acts like an analog optical image sender.

Using a focussing telescope has the additional advantage that (if correctly-designed baffles are used) the amount of stray light reaching the photodiode can be greatly reduced. As @tggzz points out, this gives excess photocurrent, which even if it doesn't saturate your amplifier, will give increased shot noise.

Ambient light won't be a problem, as my light source (the CRT) is far brighter than ambient light, and I am turning off the ceiling light in my room when conducting this experiment to reduce the ambient light.

As most other posters have suggested, you can make things a lot easier for yourself by using a transimpedance amplifier, with a low input impedance. This means that instead of being loaded by 47 kohm your photodiode may see only an ohm or so. This raises the bandwidth dramatically. You do this by moving your load resistor to the feedback path of an OPAMP, with the photodiode connected directly to the non-inverting input. The Analog Devices tool posted by @james_s shows this configuration. Using this tool, and assuming a 200 pF photodiode capacitance, I was able to achieve a 3MHz bandwidth and a nicely-damped pulse rise time of 170ns, using the two stage configuration. (You should AC-couple between the two stages to remove the effect of background light)

Thanks for that information. However, ambient light won't be a problem, as I don't keep my room light on when doing this experiment, so AC coupling isn't needed. In fact, AC coupling is absolutely not going to work in this purpose. It must be DC coupled, or I will end up creating additional artifacts in the picture being received.
Also, decreasing that resistance to 1ohm like you suggest will decrease the amplifier's gain dramatically, to the point where my analog-to-digital converter (the Picoscope 2204A) won't even be able to see the signal, even on its smallest range setting. Even when I reduced it from 47 kOhms to 5 kOhms (not all the way to One ohm), the resulting signal was so weak that the signal to noise ratio was horrible. This noise isn't from ambient light either. Rather it is RF interference, from the AM broadcast stations in my city. The AM broadcast band is from 550kHz to 1800kHz, which is within the passband of my Picoscope, and low-pass filtering isn't going to work, as that would also block the analog optical image signal I'm trying to capture. Maybe I'll need to use some kind of faraday cage around my entire apparatus.

The AD tool allows you to set the photodiode bias voltage, but this makes little difference. Often, you use a DC 'bias' voltage of 0V. This gives the maximum photodiode capacitance, but also gives zero leakage current - important if you want to measure absolute light levels. Your application may benefit from a reverse bias of a few volts (check the photodiode datasheet) but beware - noise on the bias supply plays directly into the amplifier input through the 200pF photodiode capacitance & gets the full gain. As you appreciate, you don't want to operate in photovoltaic mode, where the photodiode is forward-biased - stored charge will massively slow its response.

I'm using a bias of 9 volts from a 9 volt battery (not just a "few volts"). I thought a photodiode or phototransistor acts like a light-dependant-resistor, so the more voltage across it, the more current will flow through it for a given light level, so I thought that the more voltage I use the better the performance of this circuit. Is using a lower voltage actually better? Does it decrease the response time?

Almost none of the above depends on whether the device is a photodiode or a phototransistor, except for the speed of response. For that, you want a photodiode and preferably a p-i-n photodiode. But at your modest bandwidth I don't think that is essential. The actual photoelectric effect within the device is very fast (picoseconds). What slows it down is the transit time of the photocarriers to the device terminals, and the self-capacitance of the device, if working into a finite load.

Bob Pease wrote some about this here: https://www.electronicdesign.com/technologies/analog/article/21801223/whats-all-this-transimpedance-amplifier-stuff-anyhow-part-1, and the book by Jerald Graeme he mentions is also a good read.
Graeme, Jerald G. Photodiode Amplifiers: Op Amp Solutions. New York: McGraw Hill, 1996, ISBN 978-0-07-024247-0
Phil Hobbs also covers the development of transimpedance amplifiers for photodiodes, for example in:
Hobbs, Philip C. D. Photodiode Front Ends: The Real Story. Optics and Photonics News 12, no. 4 (1 April 2001): 44–47. https://doi.org/10.1364/OPN.12.4.000044.
And of course, the 'Bible":
Hobbs, Philip C. D. Building Electro-Optical Systems: Making It All Work. Wiley Series in Pure and Applied Optics. New York: Wiley, 2nd Ed. 2009.

I believe there is a PDF of the Graeme book lurking on the web somewhere

Thanks for the info. I think this will be very helpful.

[nfmax]

From your original post, it sounds like the device you have has around 200pF effective capacitance, so that it achieves around 15kHz bandwidth with a load resistance of 47 kohm. To get a faster response, you could use a photodiode with a smaller capacitance - but to achieve this, it would have a smaller active area, and so give a lower photocurrent for the same luminous flux. In some cases, such as your optical data link, this can be mitigated by focussing the received light into a smaller area. Think of a telescope pointed at your LED, focussed to generate a real image in front of the eyepiece: place your photodiode here. It needs to be only just big enough so that the whole of the image falls on its active area. Higher magnification gives a smaller image (until diffraction becomes significant) but makes alignment more critical.

Thanks for that info. Telescope will be impossible though. My photosensor is a bare component. I have nothing to mount it to a telescope's eyepiece.
Also, in my case though, an LED isn't actually what's being used. It could be used for sending digital data, but in my case, I'm actually trying to read an image off of a normal CRT (the old green monocrhome type). Each pixel on the screen is illuminated only at the moment the electron beam hits it (unlike with modern LCD and similar screens where all the pixels are illuminated simultaneously), so as a whole, the CRT acts like an analog optical image sender.

Using a focussing telescope has the additional advantage that (if correctly-designed baffles are used) the amount of stray light reaching the photodiode can be greatly reduced. As @tggzz points out, this gives excess photocurrent, which even if it doesn't saturate your amplifier, will give increased shot noise.

Ambient light won't be a problem, as my light source (the CRT) is far brighter than ambient light, and I am turning off the ceiling light in my room when conducting this experiment to reduce the ambient light.

As most other posters have suggested, you can make things a lot easier for yourself by using a transimpedance amplifier, with a low input impedance. This means that instead of being loaded by 47 kohm your photodiode may see only an ohm or so. This raises the bandwidth dramatically. You do this by moving your load resistor to the feedback path of an OPAMP, with the photodiode connected directly to the non-inverting input. The Analog Devices tool posted by @james_s shows this configuration. Using this tool, and assuming a 200 pF photodiode capacitance, I was able to achieve a 3MHz bandwidth and a nicely-damped pulse rise time of 170ns, using the two stage configuration. (You should AC-couple between the two stages to remove the effect of background light)

Thanks for that information. However, ambient light won't be a problem, as I don't keep my room light on when doing this experiment, so AC coupling isn't needed. In fact, AC coupling is absolutely not going to work in this purpose. It must be DC coupled, or I will end up creating additional artifacts in the picture being received.
Also, decreasing that resistance to 1ohm like you suggest will decrease the amplifier's gain dramatically, to the point where my analog-to-digital converter (the Picoscope 2204A) won't even be able to see the signal, even on its smallest range setting. Even when I reduced it from 47 kOhms to 5 kOhms (not all the way to One ohm), the resulting signal was so weak that the signal to noise ratio was horrible. This noise isn't from ambient light either. Rather it is RF interference, from the AM broadcast stations in my city. The AM broadcast band is from 550kHz to 1800kHz, which is within the passband of my Picoscope, and low-pass filtering isn't going to work, as that would also block the analog optical image signal I'm trying to capture. Maybe I'll need to use some kind of faraday cage around my entire apparatus.

The AD tool allows you to set the photodiode bias voltage, but this makes little difference. Often, you use a DC 'bias' voltage of 0V. This gives the maximum photodiode capacitance, but also gives zero leakage current - important if you want to measure absolute light levels. Your application may benefit from a reverse bias of a few volts (check the photodiode datasheet) but beware - noise on the bias supply plays directly into the amplifier input through the 200pF photodiode capacitance & gets the full gain. As you appreciate, you don't want to operate in photovoltaic mode, where the photodiode is forward-biased - stored charge will massively slow its response.

I'm using a bias of 9 volts from a 9 volt battery (not just a "few volts"). I thought a photodiode or phototransistor acts like a light-dependant-resistor, so the more voltage across it, the more current will flow through it for a given light level, so I thought that the more voltage I use the better the performance of this circuit. Is using a lower voltage actually better? Does it decrease the response time?

Almost none of the above depends on whether the device is a photodiode or a phototransistor, except for the speed of response. For that, you want a photodiode and preferably a p-i-n photodiode. But at your modest bandwidth I don't think that is essential. The actual photoelectric effect within the device is very fast (picoseconds). What slows it down is the transit time of the photocarriers to the device terminals, and the self-capacitance of the device, if working into a finite load.

Bob Pease wrote some about this here: [url]https://www.electronicdesign.com/technologies/analog/article/21801223/whats-all-this-transimpedance-amplifier-stuff-anyhow-part-1[/url], and the book by Jerald Graeme he mentions is also a good read.
Graeme, Jerald G. Photodiode Amplifiers: Op Amp Solutions. New York: McGraw Hill, 1996, ISBN 978-0-07-024247-0
Phil Hobbs also covers the development of transimpedance amplifiers for photodiodes, for example in:
Hobbs, Philip C. D. Photodiode Front Ends: The Real Story. Optics and Photonics News 12, no. 4 (1 April 2001): 44–47. [url]https://doi.org/10.1364/OPN.12.4.000044.[/url]
And of course, the 'Bible":
Hobbs, Philip C. D. Building Electro-Optical Systems: Making It All Work. Wiley Series in Pure and Applied Optics. New York: Wiley, 2nd Ed. 2009.

I believe there is a PDF of the Graeme book lurking on the web somewhere

Thanks for the info. I think this will be very helpful.

A photodiode acts as a light-dependent current source, not as a light-dependent resistor. You don't get any more signal out of it by increasing the bias voltage (unless and until you reach the avalanche gain region - but that is whole different type of device). Your amplifier needs to measure this photocurrent.

It sounds like you stil haven't grasped what a transimpedance amplifier, using an OPAMP with resistive feedback, actually does. Suppose we keep your 47 kohm resistor. We know this generates enough signal voltage for your Picoscope to measure. All we do now is to take that same 47 kohm resistor and make it the feedback resistor around an OPAMP, between the output and the inverting input. The non-inverting input is connected to ground: the photodiode is connected to the inverting input.

Since the OPAMP input terminals take essentially zero current (OPAMP rule 1), all the photocurrent from the photodiode must flow through the 47 kohm feedback resistor. Since this is the same resistor you have now, the voltage across it will be the same. However, by OPAMP rule 2, the OPAMP will adjust its output voltage to keep the two inputs at the same voltage - zero. This means the OPAMP output is an inverted copy of the output voltage you have now.

BUT - the OPAMP input voltage remains constant at zero volts regardless of the current from the photodiode. Specifically, this means the voltage across the photodiode capacitance does not change with light level, so no current is diverted to charge & discharge this capacitance as the light level changes. This makes the response time, from the same photodiode with the same 47 kohm resistor, much faster (roughly by a factor of the open loop gain of the OPAMP, at the overall 3dB bandwidth of the system).

It is essential to understand the operation of the transimpedance amplifier/photodiode combination - all high speed photoreceivers use some variation on this circuit as it is far better than the other alternatives - except at bandwidths above 10 GHz or so. The book by Graeme explains this well - he also wrote some articles for EDN on similar topics which may be still around somewhere.

[Ben321]

From your original post, it sounds like the device you have has around 200pF effective capacitance, so that it achieves around 15kHz bandwidth with a load resistance of 47 kohm. To get a faster response, you could use a photodiode with a smaller capacitance - but to achieve this, it would have a smaller active area, and so give a lower photocurrent for the same luminous flux. In some cases, such as your optical data link, this can be mitigated by focussing the received light into a smaller area. Think of a telescope pointed at your LED, focussed to generate a real image in front of the eyepiece: place your photodiode here. It needs to be only just big enough so that the whole of the image falls on its active area. Higher magnification gives a smaller image (until diffraction becomes significant) but makes alignment more critical.

Thanks for that info. Telescope will be impossible though. My photosensor is a bare component. I have nothing to mount it to a telescope's eyepiece.
Also, in my case though, an LED isn't actually what's being used. It could be used for sending digital data, but in my case, I'm actually trying to read an image off of a normal CRT (the old green monocrhome type). Each pixel on the screen is illuminated only at the moment the electron beam hits it (unlike with modern LCD and similar screens where all the pixels are illuminated simultaneously), so as a whole, the CRT acts like an analog optical image sender.

Using a focussing telescope has the additional advantage that (if correctly-designed baffles are used) the amount of stray light reaching the photodiode can be greatly reduced. As @tggzz points out, this gives excess photocurrent, which even if it doesn't saturate your amplifier, will give increased shot noise.

Ambient light won't be a problem, as my light source (the CRT) is far brighter than ambient light, and I am turning off the ceiling light in my room when conducting this experiment to reduce the ambient light.

As most other posters have suggested, you can make things a lot easier for yourself by using a transimpedance amplifier, with a low input impedance. This means that instead of being loaded by 47 kohm your photodiode may see only an ohm or so. This raises the bandwidth dramatically. You do this by moving your load resistor to the feedback path of an OPAMP, with the photodiode connected directly to the non-inverting input. The Analog Devices tool posted by @james_s shows this configuration. Using this tool, and assuming a 200 pF photodiode capacitance, I was able to achieve a 3MHz bandwidth and a nicely-damped pulse rise time of 170ns, using the two stage configuration. (You should AC-couple between the two stages to remove the effect of background light)

Thanks for that information. However, ambient light won't be a problem, as I don't keep my room light on when doing this experiment, so AC coupling isn't needed. In fact, AC coupling is absolutely not going to work in this purpose. It must be DC coupled, or I will end up creating additional artifacts in the picture being received.
Also, decreasing that resistance to 1ohm like you suggest will decrease the amplifier's gain dramatically, to the point where my analog-to-digital converter (the Picoscope 2204A) won't even be able to see the signal, even on its smallest range setting. Even when I reduced it from 47 kOhms to 5 kOhms (not all the way to One ohm), the resulting signal was so weak that the signal to noise ratio was horrible. This noise isn't from ambient light either. Rather it is RF interference, from the AM broadcast stations in my city. The AM broadcast band is from 550kHz to 1800kHz, which is within the passband of my Picoscope, and low-pass filtering isn't going to work, as that would also block the analog optical image signal I'm trying to capture. Maybe I'll need to use some kind of faraday cage around my entire apparatus.

The AD tool allows you to set the photodiode bias voltage, but this makes little difference. Often, you use a DC 'bias' voltage of 0V. This gives the maximum photodiode capacitance, but also gives zero leakage current - important if you want to measure absolute light levels. Your application may benefit from a reverse bias of a few volts (check the photodiode datasheet) but beware - noise on the bias supply plays directly into the amplifier input through the 200pF photodiode capacitance & gets the full gain. As you appreciate, you don't want to operate in photovoltaic mode, where the photodiode is forward-biased - stored charge will massively slow its response.

I'm using a bias of 9 volts from a 9 volt battery (not just a "few volts"). I thought a photodiode or phototransistor acts like a light-dependant-resistor, so the more voltage across it, the more current will flow through it for a given light level, so I thought that the more voltage I use the better the performance of this circuit. Is using a lower voltage actually better? Does it decrease the response time?

Almost none of the above depends on whether the device is a photodiode or a phototransistor, except for the speed of response. For that, you want a photodiode and preferably a p-i-n photodiode. But at your modest bandwidth I don't think that is essential. The actual photoelectric effect within the device is very fast (picoseconds). What slows it down is the transit time of the photocarriers to the device terminals, and the self-capacitance of the device, if working into a finite load.

Bob Pease wrote some about this here: [url]https://www.electronicdesign.com/technologies/analog/article/21801223/whats-all-this-transimpedance-amplifier-stuff-anyhow-part-1[/url], and the book by Jerald Graeme he mentions is also a good read.
Graeme, Jerald G. Photodiode Amplifiers: Op Amp Solutions. New York: McGraw Hill, 1996, ISBN 978-0-07-024247-0
Phil Hobbs also covers the development of transimpedance amplifiers for photodiodes, for example in:
Hobbs, Philip C. D. Photodiode Front Ends: The Real Story. Optics and Photonics News 12, no. 4 (1 April 2001): 44–47. [url]https://doi.org/10.1364/OPN.12.4.000044.[/url]
And of course, the 'Bible":
Hobbs, Philip C. D. Building Electro-Optical Systems: Making It All Work. Wiley Series in Pure and Applied Optics. New York: Wiley, 2nd Ed. 2009.

I believe there is a PDF of the Graeme book lurking on the web somewhere

Thanks for the info. I think this will be very helpful.

A photodiode acts as a light-dependent current source, not as a light-dependent resistor. You don't get any more signal out of it by increasing the bias voltage (unless and until you reach the avalanche gain region - but that is whole different type of device). Your amplifier needs to measure this photocurrent.

It sounds like you stil haven't grasped what a transimpedance amplifier, using an OPAMP with resistive feedback, actually does. Suppose we keep your 47 kohm resistor. We know this generates enough signal voltage for your Picoscope to measure. All we do now is to take that same 47 kohm resistor and make it the feedback resistor around an OPAMP, between the output and the inverting input. The non-inverting input is connected to ground: the photodiode is connected to the inverting input.

Since the OPAMP input terminals take essentially zero current (OPAMP rule 1), all the photocurrent from the photodiode must flow through the 47 kohm feedback resistor. Since this is the same resistor you have now, the voltage across it will be the same. However, by OPAMP rule 2, the OPAMP will adjust its output voltage to keep the two inputs at the same voltage - zero. This means the OPAMP output is an inverted copy of the output voltage you have now.

BUT - the OPAMP input voltage remains constant at zero volts regardless of the current from the photodiode. Specifically, this means the voltage across the photodiode capacitance does not change with light level, so no current is diverted to charge & discharge this capacitance as the light level changes. This makes the response time, from the same photodiode with the same 47 kohm resistor, much faster (roughly by a factor of the open loop gain of the OPAMP, at the overall 3dB bandwidth of the system).

It is essential to understand the operation of the transimpedance amplifier/photodiode combination - all high speed photoreceivers use some variation on this circuit as it is far better than the other alternatives - except at bandwidths above 10 GHz or so. The book by Graeme explains this well - he also wrote some articles for EDN on similar topics which may be still around somewhere.

Thanks for the info. Do you know any wide bandwidth opamps capable of amplifying signals who's frequency is at least as high as 3MHz, with a gain of at least 10x? Most common opamps seem to top out at about 20kHz for even a 1x gain (output voltage = input voltage), as they are intended to be used in audio circuits.

[tggzzz]

Thanks for the info. Do you know any wide bandwidth opamps capable of amplifying signals who's frequency is at least as high as 3MHz, with a gain of at least 10x? Most common opamps seem to top out at about 20kHz for even a 1x gain (output voltage = input voltage), as they are intended to be used in audio circuits.

It looks like Digikey has about 5000 different types in stock, with 3dB bandwidths between 20kHz and 12GHz.
https://www.digikey.co.uk/products/en/integrated-circuits-ics/linear-amplifiers-instrumentation-op-amps-buffer-amps/687?k=&pkeyword=&sv=0&sf=0&FV=-8%7C687%2Cmu20000Hz%7C2153%7C0&quantity=&ColumnSort=-2153&page=1&stock=1&pageSize=25

Being able to do your own basic research is a key attribute of an engineer (and most other jobs). Sooner or later people will become tired of spoonfeeding you.

Here's https://entertaininghacks.wordpress.com/library-2/good-questions-pique-our-interest-and-dont-waste-our-time-2/]more help for you[/url].

[StillTrying]

I'm actually trying to read an image off of a normal CRT (the old green monocrhome type). Each pixel on the screen is illuminated only at the moment the electron beam hits it (unlike with modern LCD and similar screens where all the pixels are illuminated simultaneously), so as a whole, the CRT acts like an analog optical image sender.

There's no chance you'll see the video signal in the phosphor's light, the phosphor's much to slow!

www.researchgate.net/publication/2540782_Characterising_Sources_of_Ghosting_in_Time-Sequential_Stereoscopic_Video_Displays#pf4

[james_s]

There was an interesting early color TV system I read about a few years ago. The "camera" was actually a flying spot scanner and used a special CRT with a very short persistence phosphor and then there were two pickup assemblies consisting of photomultiplier tubes with color filters mounted up in the corners of the room. Strobe lights that fired during the vertical blanking interval were utilized so that the people could see what they were doing. Apparently it worked fairly well for things like news broadcasts but I don't think it was ever widely used.

Persistence on standard CRT phosphors varies but I do agree that it's likely to be much too slow to get a usable image, however it might still be fun to try.