Intermittent Oscillation on Transimpedance Amplifier Output

I am debugging a colleagues OPA381 based transimpedance amplifier and am getting this weird 300khz oscillation riding on top of the output. The amp is connected to a photodiode as part of a particle detection sizing instrument. I get decent scatter signals but I also have this

300 khz that I can't nail down to any causes:

  1. Power Supplies look good, ripple ~2mv

  1. Tried adding up to 100 pf in parallel with feedback resistor - I can still make it "oscillate" by properly positioning something static (clear acrylic with a scribe line) in the beam path.
  2. If I increase laser power it may go away or it may stay the same.
  3. Does not appear to be a microphonic effect.

Questions:

  1. Are there optical scattering phenomena in in the 300 khz range off of stationary (to the naked macroscopic eye at any rate) that I don't know about (there is much I don't know)
  2. Would more power supply bypassing help.

Any and all suggestions welcome as always.

Thanks much, Ed V

Reply to
EdV
Loading thread data ...

What's the load on the opamp like?

John

Reply to
John Larkin

If that also shows up on the photodiode bias, not so good.

What's the light source look like? Could there be spectral stuff riding on it?

Might want to try 30-40 ohms in series with the output pin. Also, there might be noise peaking up in the hundreds of kHz area depending on the photodiode capacitance.

Ed, I won't be able to see your response until someone else answers (because you are using Google).

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

The OPA381 feeds directly into the non inverting input of an AD829.

Reply to
EdV

s
  1. The input is a 650 nm Laser diode powered by a WK brand LD driver configured with the internal diode for constant power. The power to the laser look clean and the amp behaves the same way with a known good lab laser diode supply.
  2. No PD bias we are running in photovoltaic mode.

Another interesting aspect to this is there are three PDs on a concentric detector. The one that oscillates to the largest amplitude is the inner most one. The inner one also will more often show the oscillation after a scatter event than the other outer PDs. Is this a phase margin issue?

Thanks much for taking the time to consider my plight here.

Happy holidays.

EdV

Reply to
EdV

(echoed so Joerg can see it...)

Reply to
John Larkin

If you're adding additional gain downstream, maybe you're getting feedback, through the supply rails, stray capacitance, or ground loops.

The AD829 has a compensation pin. You might try hanging a cap on there and see what happens. It is undercompensated.

It's kinda hard to troubleshoot stuff like this remotely. A full schematic would help.

John

Reply to
John Larkin

Thanks, John.

Ok, that rules out the source as the culprit. Darn, would have been too easy :-)

Ok, so that also rules out a clandestine feedback path through there. I'd consider running them with bias though. Not so much for the improved frequency response but the capacitance drops one heck of a lot, meaning more stability for the TIA. I usually keep them reverse-biased between

30-50% of the abs max limit.

Possibly, but you wrote that hanging 100pF across the FB resistor didn't change a thing. So phase margin problems are unlikely unless the photodiodes are the size of a saucer.

Not knowing the photodiode type and amplitude of the oscillation: You could try disconnecting the whole amplifier and hanging a resistor across the most offending photodiode. Make the resistor small enough so

1/RC is around 300kHz or higher, but not much higher. C being the photodiode capacitance at zero bias. Then hang a FET-probe onto it and take a look with the scope. Is it still there?

As John wrote, a schematic (of the whole chebang) along with a couple photos or layout would help our crystal ball diagnostification devices :-)

Merry Christmas to you as well.

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

nd

t.

n

this

tic

.

ff

't

ng

re

e

de

a
)

Will do. Thanks!

Reply to
EdV

  1. Is the time constant of the RC network across the OPA381 suspiciously near 300kHz? Does the frequency of oscillation change if you change this time constant?
  2. The OPA381 is a 5V-only part. Most TIA circuits I've seen assume you have split rails. How is the photodiode tied to mid-rail?
  3. The OPA381 GBW is about 18MHz. What gain does this stage have?
  4. Are you sure the 300kHz is from the first stage? Let's say there's a
400kHz filter downstream. The output-to-input delay would be a bit slower, say 300kHz. If you have a high impedance high gain node, it can pick up feedback from the output cabling (basically, you may have a radio receiver).
  1. If you're using a switcher for the supply rail, try running this stage off a battery just to ensure it really is not coupling through the supply rails. If the oscillation disappears, try running the whole thing off batteries. If the oscillation reappears, you know there is coupling through the supply rails.
  2. If it's a particle detector... is the detector separated from the amplifier by a cable? If so, what cable type are you using?
Reply to
Nemo

Post a schematic. It's likely novice bungling ;-) ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
 Click to see the full signature
Reply to
Jim Thompson

s

=A0 =A0 ...Jim Thompson

=A0 =A0| =A0 =A0mens =A0 =A0 |

=A0 | =A0 =A0 et =A0 =A0 =A0|

=A0|

=A0 =A0 =A0 |

Joerg has had a look and made some recommendations. The primary step forward on this is the next go includes the development and manufacturing partner as opposed to a guy with too many other things going on ;-)

Thanks much to all.

EdV

Reply to
EdV

Speaking of that... how exactly DOES the noise on a PD bias couple into the output current?

It's not as simple as a proportional coupling to the bias level right? Because the output current does not actually depend on the bias, it depends on the incoming light.

I mean, 10V bias with 1 mV noise does not necessarily translate to an added noise of 1/10000 of a test output current.

Regards, /Bjorn

Reply to
BW

I think it couples directly. A mV on the bias gives you a mV on the output. I don't have an obvious way of seeing that. How about this. Imagine the photodiode is grounded but that the non-inverting input of the opamp is the bias point. It's raised to some voltage. Now any noise on the bias goes directly to the output.

George H.

Reply to
George Herold

Through the PD capacitance.

John

Reply to
John Larkin

As John said, it's through the capacitance. Look at it this way: Say the bias rattle tried to move the IN- of the opamp up or down by a millivolt. The opamp's "job description and incentive plan" contains a clause that at all times it shall maintain IN- at the same potential as IN+. So the opamp will try its darndest to regulate itself so that IN- remains at the same potential as IN+. The only way it can do that is by wiggling its output.

In a TIA the output voltage level depends on the current into or out of the IN- node. Whether that current comes from actual light or, via the photodiode capacitance, from a noisy photodiode bias doesn't matter. To a TIA a current is a current is a current ;-)

Not sure what you mean with test output current. If you respond I'll have to wait until someone else answers, since gmail is blocked here.

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

s.

Opps, well I opened my mouth and stuck my foot in. So it goes pretty much as the 'noise gain' from the opamp? Add the bias supply noise to the opamp voltage noise. (In quadrature)

George H.

Reply to
George Herold

Photodiode bias noise actually produce a noise current into IN-. The magnitude is frequency dependent because the "series resistance" is the capacitance of the diode, plus whatever source impedance the bias itself has.

[...]
--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

d
.

his

ic

f
t

ly

is

u
a
n

he

ng

g

=A0 =A0 ...Jim Thompson

=A0 =A0| =A0 =A0mens =A0 =A0 |

=A0 | =A0 =A0 et =A0 =A0 =A0|

=A0|

=A0 =A0 =A0 =A0 |

|

bit hard to spot what you've missed with no circuit diag

NT

Reply to
Tabby

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.