Having some problems with overshoot

[smk037]

having some reflection issues (looks like a reflection at least)

circuit is real simple,

I thought it was reflecting at the input at first, but we matched the input impedance to the function generator output impedance and it didn't change a thing

we tried adding some capacitors to lower the bandwidth too but that was also unsuccessful

weird thing is that it is oscillating at the input too, so I can't think of anything else that would be doing this unless its reflecting from the op amp

http://img23.imageshack.us/img23/5967/waveform.jpg

Uploaded with ImageShack.us

http://img526.imageshack.us/img526/1272/circuitl.th.jpg

Uploaded with ImageShack.us

 

[smk037]

http://img526.imageshack.us/img526/1272/circuitl.jpg

Uploaded with ImageShack.us

 

[smk037]

actually, that waveform picture is a bit off
it is overshooting on both the high to low and the low to high parts

 

[berkeman]

What kind of oscilloscope probe are you using? Have you seen this recent thread here in EE about probe ringing?

https://www.physicsforums.com/showthread.php?t=458430

.

 

[berkeman]

And what is the frequency of the drive waveform?

 

[smk037]

The overshoot is apparent at frequencies from 10Hz-15MHz (the range of the function generator)

as far as the oscilloscope goes, the output is actually connected to the laser (the laser diode is coupled to a fiber optic cable) using an optical to electrical converter.
The input is attached with a BNC/banana plug, with the banana plugged attached to the input of the op amp using a wire and alligator clip

 

[smk037]

that thread about oscilloscope probes does provide some insight

the thing is, the overshoot is almost double the value of the waveform value
I wonder if the oscilloscope probe could be responsible for that kind of overshoot?

I should mention, the overshoot is apparent at the output even when only the optical output is connected to the oscilloscope (there is no electrical connection to the oscilloscope)

 

[berkeman]

I'm not understanding what you are measuring. Are you measuring the output of the opamp, or the input to the opamp, or the output after the optical conversion?

You won't get probe ringing at low frequencies, unless you haven't calibrated your probe. Have you done the compensation adjustment on the probe?

I have to bail for now. Will try to log back in after a couple hours.

 

[smk037]

I am measuring the output after the optical conversion

but I also measured the input to the opamp in order to check for oscillation there. At the input of the op amp, I found that it was oscillating, but overshooting to a much lower percentage

I have done calibrated the probe

thanks for the help

 

[Bob S]

One thing you did not mention is whether the overshoot is visible primarily at high frequency, e.g., 10 MHz function generator frequency. If so, the width of the overshoot is less than 10 ns. Such an overshoot could be the result of the propagation time from the op-amp output through the laser diode and optical link back to the op-amp input.

Light travels at a foot per ns in vacuum, and 8" per ns in optical fiber. Could you trigger the 'scope on the function generator (or + input of op-amp) and look at propagation delays at various points in the system; op amp output, the 130-ohm resistor on the laser diode driver transistor emitter, at the photodiode cathode (what is the resistor to ground?), and at the inverting input of the op-amp.

What is the op-amp? How long is the cable from the photodiode to the inverting input, and does it need to be terminated at both ends?

Bob S

 

[Adjuster]

The first thing that occurs to me is that physical layout is very important for this kind of set-up. For instance, if you are using such things as banana plugs and crocodile clips for your input connections, you won't get such a good line termination at high frequencies.

Generally the circuit should be physically compact, and the connections to the laser diode and the photodiode should be as short as practicable. You should also make sure that your power supply impedances are good and low (don't forget decoupling capacitors to common from the amplifier and laser supplies).

As referred to by the last poster, the total feedback loop time delay is an important feature. Is your feedback photodiode inside the laser package, or is it linked up using fibre, perhaps with a tap coupler? If the latter, the fibre time delay may be very significant.

Finally, it may be helpful to apply only as much input amplitude as is necessary for adequate modulation. Driving the laser on from below threshold may cause overshoot in itself, at least if the circuit can supply a sufficiently fast current rise. Additionally, there may be worse transient effects if the circuit is driven into clipping.

 

[Adjuster]

Looking at this again, you seem to be using a rather large DC input bias (4V). Does this imply that you need about 8Vp-p input for full modulation?

If so, are you sure that this will not exceed the amplifier input common-mode range? Also, what will be the minimum photo-diode bias voltage when the laser is fully on? (Do the amplifier and the photo-diode also operate from an 8V supply?)

 

[smk037]

One thing you did not mention is whether the overshoot is visible primarily at high frequency, e.g., 10 MHz function generator frequency. If so, the width of the overshoot is less than 10 ns. Such an overshoot could be the result of the propagation time from the op-amp output through the laser diode and optical link back to the op-amp input.

Light travels at a foot per ns in vacuum, and 8" per ns in optical fiber. Could you trigger the 'scope on the function generator (or + input of op-amp) and look at propagation delays at various points in the system; op amp output, the 130-ohm resistor on the laser diode driver transistor emitter, at the photodiode cathode (what is the resistor to ground?), and at the inverting input of the op-amp.

What is the op-amp? How long is the cable from the photodiode to the inverting input, and does it need to be terminated at both ends?

Bob S

The cable to the photodiode is just a wire about 8 inches long
I am not sure what you mean by it needing to be terminated at both ends?
The op amp is a TL071

The first thing that occurs to me is that physical layout is very important for this kind of set-up. For instance, if you are using such things as banana plugs and crocodile clips for your input connections, you won't get such a good line termination at high frequencies.

Generally the circuit should be physically compact, and the connections to the laser diode and the photodiode should be as short as practicable. You should also make sure that your power supply impedances are good and low (don't forget decoupling capacitors to common from the amplifier and laser supplies).

As referred to by the last poster, the total feedback loop time delay is an important feature. Is your feedback photodiode inside the laser package, or is it linked up using fibre, perhaps with a tap coupler? If the latter, the fibre time delay may be very significant.

Finally, it may be helpful to apply only as much input amplitude as is necessary for adequate modulation. Driving the laser on from below threshold may cause overshoot in itself, at least if the circuit can supply a sufficiently fast current rise. Additionally, there may be worse transient effects if the circuit is driven into clipping.

The photodiode is packaged in with the laser diode
Hmm, we do have coupling caps for the op amp but I did not think about coupling caps for the laser and photodiode supplies
I will definitely try this first
As far as the physical layout of the board goes, this is the same suggestion one fiber optics professor had for me, and I do plan on soldering down the circuit and making everything as tight as possible. I don't really see why this would make a huge difference though, especially at low frequencies. Could you shed some light on this?

And the voltage is as low as I can get it for the required modulation. It needs to modulate between 20-40mA, which is above threshold. But I suppose I can try to turn the gain up and the input voltage down and see if it improves anything.

Looking at this again, you seem to be using a rather large DC input bias (4V). Does this imply that you need about 8Vp-p input for full modulation?

If so, are you sure that this will not exceed the amplifier input common-mode range? Also, what will be the minimum photo-diode bias voltage when the laser is fully on? (Do the amplifier and the photo-diode also operate from an 8V supply?)

We are only using a 2Vp-p signal, the DC bias is required so that the laser is operating above threshold.

I will have to check the CM range.

The amplifier, the photodiode, and the laser diode are all operating from the 8V supply.

 

[smk037]

I mean 1Vp-p

Basically our input voltage should be going from 3-5 volts

 

[berkeman]

Have you characterized the delay from the output of the opamp, through the optics portion, and back to the - input? It may be that the delay is causing stability issues, and hence the overshoot and undershoot...

 

[Adjuster]

The cable to the photodiode is just a wire about 8 inches long
I am not sure what you mean by it needing to be terminated at both ends?
The op amp is a TL071

The photodiode is packaged in with the laser diode
Hmm, we do have coupling caps for the op amp but I did not think about coupling caps for the laser and photodiode supplies
I will definitely try this first
As far as the physical layout of the board goes, this is the same suggestion one fiber optics professor had for me, and I do plan on soldering down the circuit and making everything as tight as possible. I don't really see why this would make a huge difference though, especially at low frequencies. Could you shed some light on this?

And the voltage is as low as I can get it for the required modulation. It needs to modulate between 20-40mA, which is above threshold. But I suppose I can try to turn the gain up and the input voltage down and see if it improves anything.

We are only using a 2Vp-p signal, the DC bias is required so that the laser is operating above threshold.

I will have to check the CM range.

The amplifier, the photodiode, and the laser diode are all operating from the 8V supply.

If your system can operate at 10MHz then it is not really "low frequency" (except perhaps as viewed by a microwave engineer). It is quite fast enough for parasitic capacitances and inductances to have serious effects.

For instance, that 8" long connection to the rear monitor diode could be detrimental. As you say that this is "just a wire", I guess you are not trying to make it a properly terminated transmission line. What is the photodiode load resistance? Could there be enough capacitance to common from this point to add significant delay to the loop?

Finally, I would echo Berkeman's advice to characterise your system for delay. One word of warning though: be very careful when attaching probes - don't do it with power on. The laser could be ruined if you were to create a momentary short-circuit, e.g. grounding the monitor diode load.

 

[Bob S]

The two attached thumbnails simulate the overshoot problem caused by excessive delay in a closed loop feedback situation.

The first thumbnail shows a differential amplifier (Q1 and Q2) followed by an emitter follower (Q3) driving a 50 ohm load. A 2000-ohm feedback resistor R10 is coupled back to the inverting input (base of Q2). In the thumbnail, a 10 mV, 200-ns pulse is applied to the non-inverting input (base of Q1). The output waveform at R9 is shown in the thumbnail.

In the second thumbnail, everything is the same except that a 10-ns, 50-ohm transmission line delay is inserted in the amplifier output just before R9. The transmission line is terminated in its characteristic impedance (R9, 50 ohms) to minimize reflections. This delays the application of the output waveform back to the inverting input of the amplifier. The resultant output waveform, shown in the thumbnail, depicts the overshoot caused by the extra 10-ns delay. In this simulation, the extra delay is only about 5% of the input pulse width.

Bob S