EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: guymo on August 08, 2021, 07:14:41 am
-
I have an op amp that's oscillating when I probe it to view the signal on my scope. I would like to track down and better understand the cause. This is my first time working with a high-speed op amp. I've stripped the circuit back as far as I can.
I now have two stages of LT1819. The first is connected as a unity gain buffer, with a dead short from output to inverting input. The second is configured for non-inverting gain of 10, with resistors of 1.8k and 200R.
The first stage is very happy. When I probe the output of the second stage (passive 10x probe), the amplifier starts oscillating at around 15MHz. I am pretty sure this is caused by the probing, because all the other signals are affected by it, and so is my hi-fi.
My understanding of these things tells me the culprits are a combination of the probe capacitance and unwanted capacitance at the inverting input. But simulating in LTSpice it looks like there would need to be something around 100pF of capacitance at the inverting input to cause the kind of thing I'm seeing; even then the Spice plot shows about 15 degrees of phase margin.
I tried adding a 47 Ohm series resistor to the output but that did not suffice. (In this configuration, occasionally I could successfully probe the circuit without it oscillating.) I tried using a coaxial cable directly soldered after this resistor, instead of the scope probe; this made no difference. Increasing the output resistor to 1k does quell the oscillations. But I feel like I should not need such a large output resistor.
I fear my PCB layout must be to blame. An image is attached. The inverting node is second from the bottom on the right, and the associated trace is somewhat long because of the three passives it connects to. (Right now I only have two of those populated -- the compensation cap is not there while I experiment). There's a ground plane below and around those passives. But could that be adding so much capacitance at that node? How can I find out? Thanks for any pointers!
-
The layout does not seem to show a voltage follower driving the non-inverting amplifier.
What is the decoupling arrangement? The application notes recommend bulk decoupling if a heavy load is driven.
-
The layout does not seem to show a voltage follower driving the non-inverting amplifier.
The first stage has been subject to surgery for this investigation -- I've wired the first stage as a voltage follower by hand.
What is the decoupling arrangement? The application notes recommend bulk decoupling if a heavy load is driven.
Good point. I only have 100n ceramics. So far though I have not asked this circuit to drive much current, at least I don't think so -- just the scope probe / scope input which should be a light load, should't it?
Thanks as always!
-
What is the decoupling arrangement? The application notes recommend bulk decoupling if a heavy load is driven.
Good point. I only have 100n ceramics. So far though I have not asked this circuit to drive much current, at least I don't think so -- just the scope probe / scope input which should be a light load, should't it? [/quote]
It does not take much of a heavy load to cause a problem, and the feedback network counts toward the load on the output.
-
Ok, that's something to try -- thank you! I've tried to make some capacitance measurements on a bare copy of the PCB using a signal generator and scope to measure rise times etc and I can't see anything untoward from the layout so other than this idea I am drawing a blank.
-
I added 10uF of capacitance to each rail (just electrolytics that I had on hand) and the oscillation is gone.
Thank you so much! I've never seen this kind of impact from power supply decoupling before -- mainly have been doing audio frequency things and it hasn't come up at all. This has been a great practical learning experience and I know what I need to do on my next PCB revision now.
-
I added 10uF of capacitance to each rail (just electrolytics that I had on hand) and the oscillation is gone.
Thank you so much! I've never seen this kind of impact from power supply decoupling before -- mainly have been doing audio frequency things and it hasn't come up at all. This has been a great practical learning experience and I know what I need to do on my next PCB revision now.
"Just electrolytics" may be exactly what is needed. The ESR acts as an AC termination for the transmission line formed by the power bus. If a zero ESR capacitor is used, then the unterminated transmission like acts like a resonator.
-
The trick for probing sensitive analog circuits like this is to have a low loading probe.
It is worth having an active probe around for such cases (can sometimes be picked up cheep on ebay) as those have a amplifier in the very tip letting them have <1pF loading.
Otherwise there are some ways of DIYing it. Turns out a 1K resistor on the end of a coax cable going into a 50Ohm termination actually makes a pretty good low capacitance probe that works into GHz. If you don't need the bandwidth and just want to reduce the impact of loading on a regular probe you can also just add a resistor in series with that. This does create a RC low pass filter but the resistance also helps hide away some of the probe capacitance that might be upsetting a fast analog circuit.
-
The trick for probing sensitive analog circuits like this is to have a low loading probe.
It is worth having an active probe around for such cases (can sometimes be picked up cheep on ebay) as those have a amplifier in the very tip letting them have <1pF loading.
I often prefer the heavier loading of a common passive probe because it gives some idea of the operating margin. If the circuit is unstable with an additional 10 to 20 picofarads, then maybe it needs to be made more tolerant.