Sup y'all.
A while back I was asking for some examples of what to look for when an op amp oscillates so I figured I would post this in hopes that it will help someone.
I did a sloppy breadboard mockup of this schematic and tested the noise from 3 dual chips: the OPA2134, NE5532 and the LM4562.
The noise floor of the soundcard was -92.3dB. The OPA2134 was -95.5dB. The NE5532 was 95.1dB. Then going to the LM4562 I noticed that it was -67dB and while sitting at idle I noticed that it was hot to the touch. Here is the noise on an FFT:
Here was the circuit at idle and playing a 30Hz sine wave:
Here is the oscillation of the chip zoomed in showing the frequency of the oscillation:
When I added the 2- 2200uF caps to the rails of the LM4562 the noise floor went down to -95dB and the oscillation stopped.
The LM4562 is the most sensitive chip that I've used so far. It also gave me problems when I replaced my soundcard's JRC2114s in the I/V section with them. I guess it's the same with all chips that have such a high gain bandwidth product (55MHz)?
A while back I was asking for some examples of what to look for when an op amp oscillates so I figured I would post this in hopes that it will help someone.
I did a sloppy breadboard mockup of this schematic and tested the noise from 3 dual chips: the OPA2134, NE5532 and the LM4562.
The noise floor of the soundcard was -92.3dB. The OPA2134 was -95.5dB. The NE5532 was 95.1dB. Then going to the LM4562 I noticed that it was -67dB and while sitting at idle I noticed that it was hot to the touch. Here is the noise on an FFT:
Here was the circuit at idle and playing a 30Hz sine wave:
Here is the oscillation of the chip zoomed in showing the frequency of the oscillation:
When I added the 2- 2200uF caps to the rails of the LM4562 the noise floor went down to -95dB and the oscillation stopped.
The LM4562 is the most sensitive chip that I've used so far. It also gave me problems when I replaced my soundcard's JRC2114s in the I/V section with them. I guess it's the same with all chips that have such a high gain bandwidth product (55MHz)?
Tell your lab assistant to install various different capacitors AND various different (1.0 ohm in series with different capacitors). Then find the hairy edge, the cliff of death. What capacitance is the max capacitance which still oscillates? What capacitance is the min capacitance that doesn't oscillate? What are the numbers with 1.0 ohms of series damping resistance?
You can repeat these experiments with diyAudio's favorite very high bandwidth, very quick-to-oscillate opamp, the venerable AD797. Then you will learn you some data and some facts, that most people don't know. And you will feel right good about yourself. Yessir.
You can repeat these experiments with diyAudio's favorite very high bandwidth, very quick-to-oscillate opamp, the venerable AD797. Then you will learn you some data and some facts, that most people don't know. And you will feel right good about yourself. Yessir.
Thanks Bonsai.
I followed Mark's advice and experimented with different types/value of capacitors while monitoring whether the LM4562 oscillates. Thanks to this experiment I am not afraid of this chip anymore.
With no bypass capacitors vs. 1- 100nF ceramic across power pins 1 and 4, is a difference of 15.4dB. The chip still oscillates under both conditions, it just does it not as bad with the ceramic.
The bare minimum where this chip stops singing is a 100nF (or higher) cap across negative rail and ground. This stops the oscillation even under load while the chip is running a 10k sine wave at about 1VRMS. It does not matter at all whether that cap is ceramic, film, or electrolytic. I think it is interesting that a cap on the negative rail will stop the oscillation but not for the positive, does anyone know the reason this might be?
The next to bare minimum is while there is a 100nF across power pins 1 and 4, 1 cap of 100nF or higher across positive rail and ground OR negative rail and ground will stop oscillation. These caps were placed over 2" from the chip's power pins and oscillation was tested with a load of about 4k at idle and while running a 10k sine wave. I used various values from 100nF to 2200uF with film, ceramic, and electros and I could not measure a difference between any of them.
It turns out that this chip isn't so unstable and scary after all. You can't wire up a board more sloppy than the way the breadboard is in that picture after all. For the folks who are getting a lot of 50/60Hz noise for their designs and they don't have access to a scope, the problem might not be ground loops, it might be oscillation due to no local supply bypassing to ground. For all above cases of oscillation it was centered at 4.11MHz and it varied from 2-4 volts peak to peak.
I have not messed around with snubbers at all yet and I'm pretty ignorant when it comes to them still. Can someone advise the values/schematic to use to filter the power to the chip?
I followed Mark's advice and experimented with different types/value of capacitors while monitoring whether the LM4562 oscillates. Thanks to this experiment I am not afraid of this chip anymore.
With no bypass capacitors vs. 1- 100nF ceramic across power pins 1 and 4, is a difference of 15.4dB. The chip still oscillates under both conditions, it just does it not as bad with the ceramic.
The bare minimum where this chip stops singing is a 100nF (or higher) cap across negative rail and ground. This stops the oscillation even under load while the chip is running a 10k sine wave at about 1VRMS. It does not matter at all whether that cap is ceramic, film, or electrolytic. I think it is interesting that a cap on the negative rail will stop the oscillation but not for the positive, does anyone know the reason this might be?
The next to bare minimum is while there is a 100nF across power pins 1 and 4, 1 cap of 100nF or higher across positive rail and ground OR negative rail and ground will stop oscillation. These caps were placed over 2" from the chip's power pins and oscillation was tested with a load of about 4k at idle and while running a 10k sine wave. I used various values from 100nF to 2200uF with film, ceramic, and electros and I could not measure a difference between any of them.
It turns out that this chip isn't so unstable and scary after all. You can't wire up a board more sloppy than the way the breadboard is in that picture after all. For the folks who are getting a lot of 50/60Hz noise for their designs and they don't have access to a scope, the problem might not be ground loops, it might be oscillation due to no local supply bypassing to ground. For all above cases of oscillation it was centered at 4.11MHz and it varied from 2-4 volts peak to peak.
I have not messed around with snubbers at all yet and I'm pretty ignorant when it comes to them still. Can someone advise the values/schematic to use to filter the power to the chip?
There are also times when the op amp's feedback loop needs to be compensated to make the amp stable. This would be done by paralleling a capacitor across the feedback resistor. So, consider this as well as bypassing the power supplies to cure stability problems, such as those caused by the proto-board capacitance between pin 2 and pin 3 of the op amp. (I'm surprised that you can stabilize an amp worth listening to on a proto-board!)
It's a useful technique for altering any feedback loop. My point was that op amp stability also has to do with the feedback loop, not just supply bypassing. Honestly, I've never seen an amp fast enough to be worth listening to work in a proto-board without becoming unstable due to the 2-3pF of capacitance from pin 2 to pin 3. In that case, a lead network in the feedback loop is going to fix that problem, whereas extra PS bypassing will likely do nothing useful.
The 4562 was surprisingly stable on the proto board, it only took 1 100nF cap to prevent oscillation on it's negative Vcc in to ground. Why do you think that is?
So paralleling a capacitor across the feedback resistor can help stabilize a cranky op amp, but how do you know what value to pick for it?
Do you guys have any experience in experimenting with snubbers for filtering the power to a fast op amp to help it with stability?
So paralleling a capacitor across the feedback resistor can help stabilize a cranky op amp, but how do you know what value to pick for it?
Do you guys have any experience in experimenting with snubbers for filtering the power to a fast op amp to help it with stability?
Forge me if im misunderstanding since I'm new to opamps, but aren't pins 4 and 8 the power pins (V+ and V-)?
I would take a guess that isn't a coincidence since 0.1uF is such a common value folks use for bypassing. The chip's designers probably made sure the design was stable with at least 0.1uF. As for the negative rail probably something to do with that particular chip's design, or where the leakage capacitances are in your particular setup there.
For many if not most Opamp's the PSRR of the negative rail is significantly worse than the positive rail, and the 4562 is no exception. Look at the data sheet for the 4562 starting on page 14.
(Here) http://www.ti.com/lit/ds/symlink/lm4562.pdf
The PSRR graphs on the left side are the positive rail, the right side for the negative rail.
Mike
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.