This thread has been locked.
If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.
Hello David,
With regards to the choice of amplifier, both the OPA656 and OPA657 should meet your application requirements. The OPA656 is unity gain stable which sometimes makes it easier to stabilize. What value of transimpedance resistance are you planning to use?
With regards to oscillations, the high speed amplifiers are extremely sensitive to extra parasitics on the inverting and non-inverting pins so moving to a PCB should make those go away. Also, please ensure you have sufficient bypassing caps close to the device power pins. I believe a single board should suffice, however please ensure there is a solid plane for GND and power supplies. Also, please remove the planes from underneath the device to minimize any parasitic capacitance on the input pins.Please follow the layout guidelines in the OPA656 and OPA657 datasheet. Also, I would put bypass caps close to the Photo-diode cathode pins.
I dont quite understand some details of the ac coupling scheme. In my opinion all you need is a series cap for ac coupling. Given the low bias currents of these amplifiers, the transimpedance resistance will provide a path for the bias current of the amplifier without causing excess offset. Does this make sense or am I overlooking something?
Samir
Samir,
That was helpful, I really appreciate the insights. For this application the transimpedance resistance would range from as low as 25K to as high as 500K.
Since I'm currently operating dual supply, does that imply that I need a four layer board with a dedicated layer for signal, pos supply, neg supply, and ground? Or are these all just different sections of the same layer?
Also, I'm thinking of using through hole components initially for ease of prototyping until I become more proficient. If I indeed need dedicated solid planes for GND and power, does using through hole interfere with this, as the components will cross through the entire board?
Lastly, with regards to the AC coupling, my understanding was that the DC bias current from the photodiode required a return path, hence the need for a resistor or inductor to pass the DC in addition to the capacitor to pass the AC.
Best,
Dave
Hello Dave,
The GND plane should be a dedicated plane. For the power supplies, you could use a split plane. The power planes are more a precaution, you could get by without them. I would discourage the use of through-hole components for proto-typing. Generally we like to use 0805 and smaller components in order to extract maximum performance. I understand this does make it harder for quick changes, but sometimes failure to understand all the extra parasitics can cause longer debugging delays.
I understand now the AC coupling scheme. I would start off with a resistor as a safer bet and experiment with the inductor if necessary to see any change in performance but they should both work.
Samir
Samir,
Again, I really appreciate your help. There are a couple of small points I'd like to clarify. So ultimately I will have two identical circuits, with one representing a reference signal and the other representing a measurement signal. Each circuit will consist of two TIAs feeding a difference amp such as the INA 157. It is your opinion that these two circuits can coexist on the same board with little interference? The reason I ask was because when I had both circuits sharing a common breadboard and power supplies, the signal from the higher output circuit would appear at the other circuit., but at about 1/10 the voltage levell When I moved the two circuits to different breadboards with separate power supplies most of the cross coupling went away. Then again, this coupling was probably more a function of the breadboard than the circuits.
You also mention placing a bypass on the cathode. I hadn't thought of this. But now that I think about the circuit it actually makes a lot of sense. All four of my photodiodes share a common ground(cathode), so this probably serves to clean up some of the noise. Lastly, I think I'll give the surface mount components a try. My initial concern was that because I needed SMA connectors and header blocks that I'd have to go with through hole. But apparently those devices come in SMT as well. I don't have experience with soldering SMT, but I see a number of videos online that cover the topic.
Best,
Dave
Hi Dave,
I stronly believe it was the non-optimum bread board layout which caused the strong coupling between channels. That being said, if I had only a single shot for success I would go for a 2-board design.
With regard to the SMA we tend to use SMTs more often compared to the through hole.
I do have some questions with regards to your project. I assume you have 4 photo-diodes in your system - 2per channel. What signals are connected to these photodiodes? Are the PDs going to be producing a full-differential output? Also, how are you driving the PDs. I assume with a laser diode? If so which laser driver are you using. I too am working on a TIA project and I was having some trouble getting clean pulses from the laser. It looks to me like the laser driver is causing the issue. I had to use a snubber circuit to smoothen out the output from the laser driver.
Samir
Samir,
You are correct, there are 4 detectors arranged in pairs. Each detector has a separate anode, but are all connected at the same cathode. The detectors are Germanium integrated on a SOI platform and are about 2um x 12um each. We probe the detectors with a 9 pin Cascade Microtech DCQ probe and this brings the signal off the chip onto the breadboard/circuit. The bias is also provided through the probes. With regards to the laser, we couple to the chip using a single mode fiber from an Agilent 86100B tunable laser. The laser is split into two arms and each arm is modulated to a new frequency, f1 and f2 respectively. Part of each signal is incident on the reference detector pair, producing a reference beat at f1-f2, the remaining light couples off chip, reflects from an object, and is measured by the measurement detector. The measurement signal is compared to the reference to determine displacement through phase shift, or velocity through Doppler.
Dave
Thanks for the detailed information David. Few questions:
1. When you mention frequencies f1 and f2 are the sine waves, square pulse or some other data fomat?
Is this some kind of Time of Flight sensor application? It sounds to me like if you had a fully-differential amplifier front end (one that takes in a differntial signal and puts out a differential signal) which can then be fed to a differential input ADC, it would help you to reduce BOM cost and also reduce distortion by cancelling 2nd order harmonic components. Would this type of part be attractive for cost/performance reasons?
Samir
Samir,
1) f1 and f2 are sine waves.
2) The device is an interferometer/Doppler vibrometer. When you say a differential input, do you mean using something equivalent to a differential input TIA where we act on both currents simultaneously with the same amplifier, rather than using a separate TIA for each current? If so, this was something that I had an interest in doing when I started working on this design, but I wasn't quite sure how to do it. That's probably because I started with the OPA2380 and the non-inverting input was used to set the bias on the photodiode. However, with AC coupling this wouldn't matter, as the bias would be provided across either the inductor or resistor.
Are there any configurations or amplifiers that you would recommend to differentially convert the currents from a pair of photodiodes? We're currently sampling the data with an Agilent oscilloscope using Matlab and then performing signal processing on that data in Matlab. My original goal was to use ADCs rather than the oscilloscope, but I must admit that my ADC knowledge is a bit lacking. Are there ADCs that you would recommend for this type of application, where the frequency of interest is 500kHz or less and we require about 90dB SNR?
3) I placed inductors in series with the power supply where it comes onto the board and this seemed to help the noise performance. The inductors are fairly large, 47mH. The LC filter is formed with a 100uF decoupling capacitor. These are fairly large and produce a fairly low resonance around 75Hz, so I probably can get away with smaller values for Land C. When I start designing the PCB I think it would be useful to place inductors at the power supply inputs at the board, but I've also seen TI literature where ferrite beads are placed in series with the power supply pin at the actual IC. Is this recommended or necessary if there are inductors at the entry to the board?
4) Although the shunt capacitance Cf is often provided by the parasitics of the shunt resistor, I assume it makes sense when laying out the PCB to create a trace and pad for a separate Cf, given that the SMT Rf may not provide enough parasitic capacitance. Would you recommend using a tunable varactor to fine tune the Cf?
Best,
Dave