Opamp instability and oscillation at the negative side of the waveform

[Tadaspadas]

Hello there,

For the last couple of days I have been working on OPA548 power opamp and could not tackle the instability issue. I am using the opamp to push current across the electrodes dipped into conductive water. Since the conductivity is high, we are talking about resistance levels below 100 Ohms. On the breadboard I build inverting amplifier with the gain of 3 (refer to figure 1). I have all decoupling connected as suggested in the datasheet. When the load is connected (30 ohm) I do get ringing on the negative side of the waveform. It's oscillation frequency is around 1MHz. I have tried snubber method - does not reduce the ringing, since I use low resistor. My opamp is powered by +-5V power supply. The input waveform is 1Vp-p at 10kHz and output is 3Vp-p. The lowest load is designed to be 15 ohm reference resistor in series with electrodes (thus it never gets short), max current = (3V/1.41)/ 15Ohm = less than 150 mA. Figures 2 (42kHz) and 3 (10kHz) shows the instability at different frequencies having the same 30 ohm load.

I am aware that open loop GBW is 1Mhz, thus I selected this opamp knowing that there will be plenty of space for minimal gain of 3 and operation at 10kHz.  Interestingly enough. when the load is much higher, say 10 kOhm, opamp performance is brilliant even at 100kHz. However I need to drive small loads. Figure 4 shows this case on the oscilloscope.

Could you guys please assist me on this issue? What are the options to stabilise the opamp even with low resistance loads? Is this opamp a good choice for such application? Do I need to add a current boost stage with a pair of pnp and npn transistors (although I don't think it is lack of current problem)? Looking forward to any advices and assistance.

Kind regards,
Tadas

[Doctorandus_P]

The trouble is likely to be in your breadboard wiring.

OPA548 is a fairly high powered opamp, with upto 60 between it's power pins and it can deliver upto 3A dc.
https://www.ti.com/lit/ds/symlink/opa548.pdf

The inputs of the opamp are also fairly responsive to very small voltage changes.

A 3A output current will generate voltage drop over breadboard contacts and even short pieces of wiring, and this means you have to be aware of the current paths through each part of your circuit.
How is the GND current routed after it goes through the 30 Ohm resistor back to the power supply circuit?

On the positive half of the sine wave the current on the "GND" side of the output resistor goes back to the power supply, and will generate a positive voltage over the wire (Use the output clamp of the power supply as reference), and during the negative half of the sinewave there will be a negative voltage between the power supply GND and the "GND" pin of the 30Ohm resistor. If a part of the voltage drop over the wiring is coupled back into the inputs of the opamp, then this will distort the signals.

Routing of the buffer and decoupling capacitors is also critical.
In a simplified world you want both very short wires between these capacitors and the opamp, but also a star GND to separate voltage drops over wiring between different parts of the circuit.

In a more complicated world, you want to keep the "star ground" clean by keeping high AC currents away from them, and when you go into higher frequencies then the influences caused by inductance of wiring and loop areas also increase.

Take a dive into the literature for designing audio power amplifiers. The constraints are nearly the same with both high currents in the output section and sensitive inputs and feedback circuits.

[Tadaspadas]

Hi there,

Thanks for assistance.

I do keep all capacitor and resistor legs as short as possible, and decoupling caps are closest to the voltage supply inputs to the opamp. Speaking of the grounding paths, since I built it on the solderless breadboard I have ground paths from two side alongside the power supply. This load ground is essentially the same as decoupling caps for the opamp. I just came across this datasheet of a audio amp which indeed has notes on stability and refers to your raised points.

Hovewer, after playing around this simple configuration I did get repetitive opamp response thus I doubt that the ground paths cause most of the problem.

Personally, I would love just to ditch this opamp and get another model. However in the literature it has been reported as general problem for most of the opamps that the behaviour with low loads and higher frequencies gets unstable.

Still seeking for answers and options! Thanks

[Cerebus]

The trouble is likely to be in your breadboard wiring.

Not it's not, it's more fundamental than that, or at least I suspect it is.

A 3A output current will generate voltage drop over breadboard contacts and even short pieces of wiring, and this means you have to be aware of the current paths through each part of your circuit.

A red herring. It's a 30 load with a 3V peak signal => 100mA current, not enough to cause a significant problem in a circuit with a gain of 3. If the ground resistance was 100m that would only produce a ~40mV sinusoidal shift in output and it would evidence itself in distortion of the whole wave, not a kink as seen here

I the likely cause of the problem problem can be found from the data sheet.

From the data sheet:



The OP is using a ±5V supply, and is trying to produce a ±3V signal. Absolute minimum supply for this part is ±4V. Combine that with the rather ungenerous output margins (check the graph later in the datasheet for output swing @ the actual output current) and that's where I'd investigate first. Try ramping the supply up to, say, ±10V and see what you get. I know it's not obviously clipping, but I've known op amps exhibit 'odd' behaviour when pushed into the corners of their performance envelopes before.

You'll get a much better feel for the stability of the amp if you feed it some square waves and see what they look like.

[E-Design]

But the kind of instability you have found in the literature isnt whats going on with your setup / amplifier. What you have isnt oscillating, it has a non symmetrical distortion.

Does this distortion become less as you make the 30 ohm load higher - try 300, but keep same frequency and amplitude. Is it improved? Whats the supply rails? (Edit Cerebus has it, min supply issue)

[Someone]

But the kind of instability you have found in the literature isnt whats going on with your setup / amplifier. What you have isnt oscillating, it has a non symmetrical distortion.

Yes, looks a bit like crossover distortion.

[jonpaul]

Suspect UHF oscillation that is on the distorted part of wave and invisible as above max freq capability the scope and probe. The oscillation changes the bias conditions.

Check power bypass caps at pins of opamp, perhaps add a damping RC at output.

Jon

[T3sl4co1l]

It's not clipping, the +/-5V supplies will deliver 6Vpp just fine, twice what is shown.

No reason to believe it's oscillation, the scope doesn't show anything obvious (but, it's in averaging mode, how many averages? and why?), and there aren't other errors like DC shift or distortion.

But do show layout, just in case.

If a bias resistor, say 220 ohms from +5V, or -5V, to output, changes the position of the hiccup, it's definitely crossover distortion.

It might also be characteristic of the load, but that's a pretty fast transient to blame on some cyclic voltammetry phenomenon.

Tim

[Vovk_Z]

I bet it is crossover distortion, because of several reasons:
1) a quite high signal frequency - 10-40 kHz;
2) an opamp is of 1 Mhz bandwidth only,
3) TS says it is everything ok with decoupling and layout.
4) As everybody mentioned it doesn't seem like oscillation.
A higher power supply voltage (or absence of a load) may possibly decrease or vanish this phenomenon. But in general, my opinion is that OPA548 is not fast enough for 10 kHz signal and 15-32 Ohm load.

[Vovk_Z]

To amplify low voltage high-frequency signal with a load of 32 Ohm I used LM6172 (two opamps in one case in parallel). Actually, it was my headphone amplifier. If TS has 15 Ohm load then he may possibly need four such opamps in parallel (two cases). They are relatively cheap and really fast (or they were relatively cheap 10 years ago).
Here is a spectrum of that my LM6172 amplifier working onto a 33 Ohm load:

[StillTrying]

I was going to suggest a 220r to the -5V as T3sl4co1l did.

The problem is described here, but I've not read the 5 replies to it.

https://e2e.ti.com/support/amplifiers/f/14/t/539524?OPA548-ringing-right-after-cross-over-but-to-negative-voltage-output-only

[Tadaspadas]

Thanks for involvement. I will make a list of suggestions and comments from my side what is the effect, whether it improved operation or not.

The trouble is likely to be in your breadboard wiring.

OPA548 is a fairly high powered opamp, with upto 60 between it's power pins and it can deliver upto 3A dc.
https://www.ti.com/lit/ds/symlink/opa548.pdf

How is the GND current routed after it goes through the 30 Ohm resistor back to the power supply circuit?

To answer this question I am adding the view of the breadboard and my connections:

Figure 1: breadboard from the top

Figure 2: breadboard connections to the signal gen and oscilloscope

It's not clipping, the +/-5V supplies will deliver 6Vpp just fine, twice what is shown.

But do show layout, just in case.

Here above is the breadboard view as asked.

Try ramping the supply up to, say, ±10V and see what you get. I know it's not obviously clipping, but I've known op amps exhibit 'odd' behaviour when pushed into the corners of their performance envelopes before.

Tried this, changing voltage to +-12V does not solve the problem. All the other uploaded pictures are at the operation with +-12V supply.

You'll get a much better feel for the stability of the amp if you feed it some square waves and see what they look like.

You are very right here. When I switched to a square wave gen, the problem became more than obvious. Let's have a look at the figure 3.


Figure 3: square wave at 10kHz frequency with 30 ohm load.

But the kind of instability you have found in the literature isnt whats going on with your setup / amplifier. What you have isnt oscillating, it has a non symmetrical distortion.

Does this distortion become less as you make the 30 ohm load higher - try 300, but keep same frequency and amplitude. Is it improved? Whats the supply rails? (Edit Cerebus has it, min supply issue)

You are right, now I see it can be distortion problem.
I have made a typo on my first post saying load of 10 kOhm (not 10 kHz). So, yeah, when the load increases, the distortion effect significantly reduces (but don't completely vanish with 10 kOhm). If you go to open output, then there is not distortion. The problem is that in my application small loads need to be driven (15 Ohm to 1 kOhm).

Yes, looks a bit like crossover distortion.

As I found out it very much is a crossover distortion. Check the video attached below.

Suspect UHF oscillation that is on the distorted part of wave and invisible as above max freq capability the scope and probe. The oscillation changes the bias conditions.

Check power bypass caps at pins of opamp, perhaps add a damping RC at output.

Jon

Scope is capable of 200MHz and it is nowhere near the limits. Tried RC snubber, does not help.
_________________________

It's not clipping, the +/-5V supplies will deliver 6Vpp just fine, twice what is shown.

No reason to believe it's oscillation, the scope doesn't show anything obvious (but, it's in averaging mode, how many averages? and why?), and there aren't other errors like DC shift or distortion.

But do show layout, just in case.

If a bias resistor, say 220 ohms from +5V, or -5V, to output, changes the position of the hiccup, it's definitely crossover distortion.

Hey Tim, your answer helped to diagnose the problem. Check the attached video for the crossover effect.
Averaging on the scope is turned on for 2 cycles. it helps to reduce white noise.
There is no DC shift and I have enough supply for the opamp to not start clipping the waveform (+-12V supply voltage).
When I add a resistor as you say between the output and -5V or +5V supply rail, the hiccup position moves and it once again proves the crossover distortion problem.

The link to the video: https://photos.app.goo.gl/zQ89u8tnV5SJdncV9. In the attached video I change the input signal offset. The input peak to peak is 1V. Th offset move between -0.6 to +0.6 and it shows that the hiccup disappears when the waveform leaves the negative voltage area or goes below the hiccup level.

Now we identified the issue - it is CROSSOVER DISTORTION. What are the ways to solve it? I assume it some sort of biasing problem inside the opamp. Is the only way to change the opamp to a better one? I am powering a reference resistor and electrodes, so I have mainly resistive loading. A simple example schematic of the load is below.

Figure 4: the load.

Any help and suggestions would be appreciated! and thank you guys who already helped me a lot!

[T3sl4co1l]

Excellent, comprehensive response, thank you!

Yes, the best you can do is either: get a different amp (higher bias, lower distortion, faster), or bias the output for class A operation.

Which, if you're going to do the latter, the trouble is, you need to bias it heavily enough that the crossover is never...crossed.  You need to draw a lot of current from the thing.  It would be nice if you could just turn up the internal bias (transistor to transistor), but that's buried deep inside, no such luck.  So you're stuck pulling current through just one transistor (to -V or +V).  Messy.

And so, if you were going to go that way, you might as well go dumb and simple -- a regular op-amp (TL071 say) paired with an emitter follower (TIP31 or better?) and a big fat pull-down resistor.  Grossly inefficient, simple as hell, low distortion; fairly fast, too.

Probably, there are better power amps out there, just search around and select a few.  Try them out and see which works best.

Tim

[Cerebus]

It's not clipping, the +/-5V supplies will deliver 6Vpp just fine, twice what is shown.

Ha! My bad, I mis-read the oscillogram^* as showing 1V/div, so as already showing 6V_ptp.

^*Does anyone other than me still use this word?

[T3sl4co1l]

It's not clipping, the +/-5V supplies will deliver 6Vpp just fine, twice what is shown.

Ha! My bad, I mis-read the oscillogram^* as showing 1V/div, so as already showing 6V_ptp.

^*Does anyone other than me still use this word?

Heh, it's correct, but "look at the 'scope" rolls off the tongue a little easier I guess.

(Also, edited above response)

Tim

[Tadaspadas]

Thanks for suggestions again.

A higher power supply voltage (or absence of a load) may possibly decrease or vanish this phenomenon.

Increasing the supply voltage does not decrease the problem. Absence of lead does! BUT, I still need to drive 15 ohms in my application, so not very practical to keep testing circuit unloaded.

To amplify low voltage high-frequency signal with a load of 32 Ohm I used LM6172 (two opamps in one case in parallel). Actually, it was my headphone amplifier. If TS has 15 Ohm load then he may possibly need four such opamps in parallel (two cases).

I checked the datasheet and it looks like it is not possible to supply enough current (50mA per channel) whereas at highest setup I need to draw 300mA easily.

The problem is described here, but I've not read the 5 replies to it.

https://e2e.ti.com/support/amplifiers/f/14/t/539524?OPA548-ringing-right-after-cross-over-but-to-negative-voltage-output-only

INDEED the guy Thomas explains that this is the problem of the opamp architecture. If you guys are interested in this more, definitely read the link. I can copy the main bit of the response here as well.

""" Crossover distortion becomes more evident with a low-level vs a high-level output swing because at low levels the crossover region is a larger percentage of the total waveform. It occurs in a region near the output swing midpoint where both output transistors are be conducting minimum operating current. Crossover distortion does benefit from the application off negative feedback and the more loop gain available the more it will be reduced.

The OPA548 uses an all NPN output stage, which is quite different than the more conventional complementary NPN/PNP output stage. The upper NPN is an emiiter follower, and the lower NPN is a common-emitter amplifier. Looking into the output pin a low-impedance is presented when the output is in the positive portion of the swing, and a high impedance is presented when the output is in the negative portion of the swing. The 2-ohm external load is driven by two very different sources and in those cases and electrical characteristics will be different depending on which transistor is more "on" or "off" than the other.

The vertical spike in the crossover region occurs when the amplifier stage output slews between one transistor and the other. There is some overshoot on the ends which produces the ringing. """

I think this suggests that this particular opamp will only work like this with low loads and another opamp should be selected.

Any suggestions for a good power opamp with a heatsink able to work with +-5V, drive 15 ohm load with at least 0.5A current, operation frequency at 10kHz without issues (open loop GBW should be more than 1Mhz then in thus case)?

Thanks for you help and looking forward for more comments

[Vovk_Z]

Probably, there are better power amps out there, just search around and select a few.  Try them out and see which works best.

there are several quite cheap, popular (easy to buy) high output current ones, for example: AD8397 (24 VDC or +-12VDC Vs max, 300 mA max output current), AD811 (current feedback, +-15 Vs, 100 mA max output current).
Here you can find more..  I used "High Output Current Op Amps ≥ 100mA" menu section.

[Vovk_Z]

Increasing the supply voltage does not decrease the problem. Absence of lead does! BUT, I still need to drive 15 ohms in my application, so not very practical to keep testing circuit unloaded.

Yes, I see now your text about the supply voltage. Of cause, it is not practical to work w/o a load , it was suggested (to play with a load value) only to check if it is a crossover distortion.

To amplify low voltage high-frequency signal with a load of 32 Ohm I used LM6172 (two opamps in one case in parallel). Actually, it was my headphone amplifier. If TS has 15 Ohm load then he may possibly need four such opamps in parallel (two cases).

I checked the datasheet and it looks like it is not possible to supply enough current (50mA per channel) whereas at highest setup I need to draw 300mA easily.

That means you need 3-4 LM6172 cases in parallel to have 300 mA. But you can find another a bit more powerful opamp.

Any suggestions for a good power opamp with a heatsink able to work with +-5V, drive 15 ohm load with at least 0.5A current, operation frequency at 10kHz without issues (open loop GBW should be more than 1Mhz then in thus case)?

Yes, you need an opamp with GBW more then 1 MHz. For about and order more.
I have suggested a couple opamps above (look for AD8397, two in case in parallel) and there is always an option to order 'A king', or rather 'A queen' - LT1210CT7. this one is a bit costly but if you need only one then this is not a problem. It depends on the maximum needed output voltage too. AD8397 has rail-to-rail output but has a bit special case with a thermal pad. And LT1210CT7 is easy to mount - it has TO220 case type but it is not rail-to-rail output type.

[StillTrying]

"A laser-trimmed monolithic integrated circuit provides excellent low-level signal accuracy and high-output voltage and current."

I find it hard to be impressed with that OPA548 and its data sheet. I'd expect a +/- 3A output to still be good or even better at +/- 150mA.

3Vpp 0.5A 15R @ 10kHz sounds like an audio amplifier, especially if you've got a bit more than +/-5V for the supplies.

[floobydust]

Don't forget water is a massive capacitive load due to the double-layer.
It depends on the electrode area but I would expect many uF. This is what I would think is your problem with the "bottom-side fuzzies" as the oscillations are called, with the op-amp.

Also, you usually don't DC-couple to electrodes in solution because electrolysis will occur as the op-amp keeps it centered around 0V. Galvanic potentials will be on the electrodes and over-riding those causes corrosion.

[Vovk_Z]

3Vpp 0.5A 15R @ 10kHz sounds like an audio amplifier, especially if you've got a bit more than +/-5V for the supplies.

Yes, but audio-IC usually have a minimum gain of 20. So they are not convenient to use as usual opamp.

[vk6zgo]

It's not clipping, the +/-5V supplies will deliver 6Vpp just fine, twice what is shown.

Ha! My bad, I mis-read the oscillogram^* as showing 1V/div, so as already showing 6V_ptp.

^*Does anyone other than me still use this word?

Why yes, I was looking at the oscillogram on my oscillograph just the other day!

[exmadscientist]

I would recommend avoiding the AD8397. It has a rather unusual input stage, some consequences of which are mentioned on page 11 of the datasheet. In general this part seems much more prone to oscillation than usual, particularly under heavy load. The datasheet didn't mention anything about that, but instability is not uncommon in rail-to-rail output stages, so perhaps they felt they didn't have to. Someone on an audio forum described it as "high-strung", and that seems an adequate summary of the AD8397. It is difficult to achieve its signature performance (wide swing into heavy loads) and if you don't get that, I see little reason to bother with it over choosing something less temperamental.

If you end up spinning a PCB, it may be useful to provide footprints for a series RC network between inverting and non-inverting inputs ("lag-lead compensator"). This type of network is often useful with tricky compensation problems, like calming down the AD8397, and having the footprints available will only make life easier.

If you are not so interested in outputs swinging rail-to-rail, and they otherwise meet your needs, the "line driver" or "DSL driver" class amplifiers tend to be generally better behaved, in my experience.