1
\$\begingroup\$

In my signal processing chain, the first stage is an inverting gain stage using Op AMP LT6200CS8-10.

My output DC operating point is 1.24 V. I am providing an exponential signal (Amplitude 1mV) as the input. This is my schematic: Schematic

C43 is not populated. The layout is shown below:

Layout

This stage is driving another op amp gain stage using AC coupling cap C38 shown in layout above.

When I don't apply a signal, this is how the output (net op_amp_1) looks:

Waveform 1 - no signal at input

When I apply an exponential signal with 1 mV amplitude high and 0 mV amplitude low, the output (net op_amp_1) looks like this Waveform 2- signal applied at input of op amp

The flat portion shows oscillation:

Waveform 3-flat portion zoomed in

How I can reduce this oscillatory behavior of the output?

\$\endgroup\$
  • \$\begingroup\$ What do you mean "when I apply a signal"? How do you apply or not apply, by disconnecting a cable, or how? \$\endgroup\$ – Ale..chenski Aug 10 '17 at 18:11
  • \$\begingroup\$ I have the coax cable from signal generator to my board. When there is no signal, I mean the signal generator's output is off. \$\endgroup\$ – Ash Aug 10 '17 at 19:27
  • \$\begingroup\$ What do you mean "Off"? Is the 50-Ohm source disconnected? Then you effectively change the feedback network parameters (R23-R24-siggen), which changes the gain, and cable delay also changes the feedback transfer function. The gain is about 50X-100X, it is easy to cross the stability threshold. \$\endgroup\$ – Ale..chenski Aug 10 '17 at 20:23
  • \$\begingroup\$ I mean output off, the cable is still connected. So the source impedance should be still there. \$\endgroup\$ – Ash Aug 10 '17 at 20:42
  • \$\begingroup\$ It is not certain that the impedance is still there. If yes, then did you consider that the oscillations could be coming from the source? \$\endgroup\$ – Ale..chenski Aug 10 '17 at 20:45
1
\$\begingroup\$

1) Your most likely culprit is the ground connection on pin 4 of the LT6200. That skinny trace is just asking for trouble. Make a via to ground right at the pin.

2) Your decoupling caps likewise. A via for each. And R29/C40, too.

3) Your R23/R24 combination will produce an effective input impedance of 20 ohms, assuming the LT6200 is working correctly. This will not cause oscillation, but it does represent a gross impedance mismatch to your presumed 50 ohm signal.

ETA - Point 3 applies to AC components only. DC doesn't count since it is blocked by C35.

\$\endgroup\$
  • \$\begingroup\$ Thank you for the answer. Could you suggest any rework which I can do to improve this on the current revision of board? \$\endgroup\$ – Ash Aug 10 '17 at 1:59
  • \$\begingroup\$ R24 is connect to the (-) pin which is a high impedance path. The low impedance path is at the opamp output. 50 ohm is actually parallel to (33 + 2k7) ohm. \$\endgroup\$ – Jason Han Aug 10 '17 at 2:58
  • 1
    \$\begingroup\$ @Ash - Sure. Drill a small (.050 or less) hole just below pin 4. Insert a fine wire and solder on both sides. Do the same for R29 and C40. \$\endgroup\$ – WhatRoughBeast Aug 10 '17 at 21:24
  • \$\begingroup\$ @JasonHan - Sorry, no. Negative feedback will keep the - input at 1.24 volts regardless of signal - it's called a virtual ground. All AC components of the signal will therefor see a 33 ohm to ground. And there is no way 33 adds to 2.7k. \$\endgroup\$ – WhatRoughBeast Aug 10 '17 at 21:27
  • 1
    \$\begingroup\$ @JasonHan - exactly. No (appreciable) current flows into the op amp. However, the pin is forced to remain at a DC level of 1.24, which is indistinguishable (in AC terms) from ground. So AC flows into the node and is compensated for by the output through the feedback network. The node is kept at a DC level and is indistinguishable from ground from an AC point of view, so the 33 ohms is effectively feeding ground from the point of view of the input signal. This is in parallel with the 50 ohm terminator. \$\endgroup\$ – WhatRoughBeast Aug 11 '17 at 0:17
0
\$\begingroup\$

Use a star ground topology at C40. Use a trace from J3 Gnd and connect directly to C40, then punch a via to the ground plane. Pin 4 of U10 should also have it's own via to the ground plane.

You may also consider the new value for R15, R29 and C1 (new) electrolytic cap. Which gives better low pass filter from the 5V supply. Also consider using a new C40 value that can lower down the ripple in your scope. It's in the capacitor datasheet that shows ceramic cap impedance vs frequency curve.

schematic

simulate this circuit – Schematic created using CircuitLab

enter image description here

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.