Transimpedance Amp Oscillation Troubleshooting Help Needed

[Ravaner]

Hello. I have developped a transimpedance amp using LF356, unfortunately I get an oscillation at 500 kHz of low level (10 mV) but not acceptable. I've tried all classic tricks, decoupling power supplies ... With no real improvement and so I'm going to developped a similar amp using discrete components. Has someone experience in that area (BW > 4 MHz) ? Which kind of problem I'm going to encounter ?

 

[Studiot]

If you are seeing oscillation, there must be a positive feedback path at that frequency.

Have you tried techniques such as a guard ring. Have you properly investigated the input and ouput paths for unwanted coupling? Screening?

 

[berkeman]

Hello. I have developped a transimpedance amp using LF356, unfortunately I get an oscillation at 500 kHz of low level (10 mV) but not acceptable. I've tried all classic tricks, decoupling power supplies ... With no real improvement and so I'm going to developped a similar amp using discrete components. Has someone experience in that area (BW > 4 MHz) ? Which kind of problem I'm going to encounter ?

There are two main problems/challenges with building an opamp up out of discretes. First, you don't have a good way to match transistor pairs. Even if you buy transistor pairs in the same package, their matching characteristics usually are not very good.

Second, you will have at least an order of magnitude higher capacitances in a discrete circuit, compared to the capacitances on an IC. This limiits the bandwidth of what you can do with discretes.

 

[waht]

The feedback loops in general are very complicated to model as they depend heavily on the most minute details of the op-amp and the feedback path.

A simple or solution or a suggestion would be to skip that part and replace the op-amp with many op-amps of different specs, quality and manufacturer just to see which one works better.

Another solution might be to add a second negative feedback path to null out the effects of positive feedback. But that can also be really tricky to do, and may not ever work if you don't know the exact behavior of the feedback path.

 

[Studiot]

Here is an AD516 kitted out to avoid a similar situation.

Note the items of good practice.

Diecast box
ground plane
guard ring on track side of pcb
large Earth shield on track side of pcb
note in particular the feedback track run between two large earthed lands in the second pic. (sorry I forgot to highlight it)

Other tricks to consider.

Ferrite beads in the supply leads
Zobel network on the amp - it may just be too lightly loaded

 

[berkeman]

Here is an AD516 kitted out to avoid a similar situation.

Note the items of good practice.

Very sweet. You forgot to mention the Windows Server textbook though...

 

[Ravaner]

Many, many thanks for all that suggestions, I will redesign my board ...