4
\$\begingroup\$

I am attempting to build the following circuit using an LM6171 op-amp. I know that this op-amp is unity-gain stable. I previously had this circuit working with an LT1637 op-amp, but I had to switch parts for a higher GBWP.

schematic

simulate this circuit – Schematic created using CircuitLab

I want to put a low amplitude sine wave (10 mV, peak-peak) on the input. The output is connected to a low impedance load (< 10 ohms) through a relay, which is normally off (so normally open circuit).

When I switch on the power, the output shows a ~20 V, 20 MHz sine wave (without me applying any voltage to the input). Why is this op-amp oscillating? I am not sure what part of the datasheet would give me a clue here - I figured that the main issue was the phase margin.

\$\endgroup\$
7
  • \$\begingroup\$ The load is reducing the phase margin by adding at least one pole in the feedback. At 20MHz it doesn't take much shunt capacitance or series inductance to do that, and that's a pretty 'hot' op-amp (100MHz GBW) so it would be quite happy to oscillate at 20MHz. \$\endgroup\$ – Spehro Pefhany Mar 30 '15 at 13:54
  • \$\begingroup\$ @Spehro Would that still cause an issue with no load? (I guess it would - stray capacitance doesn't take days off.) More importantly - how do I lower those? \$\endgroup\$ – Greg d'Eon Mar 30 '15 at 13:57
  • \$\begingroup\$ See answer below- of course you can try to reduce the stray capacitance as a first choice, but that doesn't appear on the schematic. \$\endgroup\$ – Spehro Pefhany Mar 30 '15 at 14:36
  • 1
    \$\begingroup\$ Does it oscillate with the load attached? The ten ohm load is mostly resistive? (what is it?) Do you have a long (few foot) cable going to the load? (That could add some capacitance.) \$\endgroup\$ – George Herold Mar 30 '15 at 15:49
  • 1
    \$\begingroup\$ Why is the feedback connection to the - terminal after the resistor? If you put it before the resistor, R1 would form part of an RC filter with any load capacitance, reducing any phase issues experienced by the opamp. \$\endgroup\$ – Nick Johnson Mar 30 '15 at 16:06
4
\$\begingroup\$

Try something more like this- you may have to fiddle with the value of the capacitor to optimize it if you need really high frequency response.

schematic

simulate this circuit – Schematic created using CircuitLab

\$\endgroup\$
3
  • 1
    \$\begingroup\$ Can I ask where you're getting the component values from? \$\endgroup\$ – Greg d'Eon Mar 30 '15 at 14:37
  • \$\begingroup\$ It's a very rough guess that would cause it to roll off .. st like 8MHz with a 10 ohm load. \$\endgroup\$ – Spehro Pefhany Mar 30 '15 at 14:45
  • 1
    \$\begingroup\$ Also make sure you have good decoupling of the supply rails. Maybe good high quality ceramic 0.1uF and .01uF caps on each supply pin of the IC with a good low inductance connection to ground. \$\endgroup\$ – John D Mar 30 '15 at 14:47
4
\$\begingroup\$

In my experience, what you're seeing is a result of various bad practices on your part - that is, you're doing (or not doing) things in ways that worked just fine the first time, but are now biting you on the butt.

First, note that you've made an enormous jump in the performance of your op amp. You've increased GBW from 1.1 MHz to 100 MHz, and slew rate from 0.4 v/usec to 3600 v/usec. This increase in performance comes at a price. You must start paying attention to layout in new ways. First, you MUST have a ground plane. Second, you MUST properly decouple your power supplies. See Figure 58 of the data sheet, and read the text. Third, you need good high-frequency layout, keeping traces as short as possible, particularly the traces connecting to the inputs of the op amp.

If you're building this on a solderless breadboard, you are probably doomed.

With that said, one other possibility suggests itself. You say that you turned on power with no input. Does that mean the input was floating? That's a major no-no. Short the input to ground when it's not connected to the nominal source.

Finally, are you SURE you need this fast an op amp? You haven't specified the frequency of the 10 mV sine wave you're designing for, and if it's less than about 10 MHz you should be looking at a slower amp.

People who ask for more bandwidth than they need deserve what they get.

\$\endgroup\$
2
  • \$\begingroup\$ This is exactly the kind of answer I was hoping for. I'll respond to the points separately when I have a chance (ie when I'm not on my phone) \$\endgroup\$ – Greg d'Eon Mar 30 '15 at 21:14
  • \$\begingroup\$ Let's see. I have a ground plane. I have decoupling capacitors, although they are not what the datasheet recommends (0.1 uF != 0.01 uF) - this could be an issue. My signal traces are not particularly short or overly long. The input is being driven by another op amp, so it is not floating. Please notice that the output will be larger amplitude than 10 mV. The maximum frequency is 100 kHz. I am not totally happy with the (very high) slew rate, but I thought that too high > too low... maybe I was wrong. \$\endgroup\$ – Greg d'Eon Mar 31 '15 at 2:21
1
\$\begingroup\$

The LM6171 has 40 degrees of phase margin, which will make it ring to a step load, even under good conditions. Any loading or circuit layout that degrades phase margin, and it will quickly become an oscillator. Layout will be really important, short runs and small loop areas. If the amplifier is not on top of the load, you may need to build it into a probe head to make sure it is close.

In this case, circuit as drawn and switching the load, will present the amplifier with not only a load step but also a gain step, from unity gain to gain of > 10 (for a 10 Ohm load). Some how the 100 Ohm output resistor has to be outside the amplifier loop, or an inner loop needs to be added.

You may be asking too much of this circuit. An additional amplifier may be needed to complete sensing and loop feedback.

\$\endgroup\$
1
  • \$\begingroup\$ +1 - this is helpful information. Neat idea on the probe head: I hadn't thought of literally moving the amplifier closer. It looks like I've just been naive (which usually works for slow circuits) and I've finally stepped into the world of analog where some real expertise and caution is required. \$\endgroup\$ – Greg d'Eon Apr 1 '15 at 11:01

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.