Amplifier Distortion when Combining Stages in Series

[ajderoui]

I am having some issues designing a simple amplifier circuit and was wondering if anyone could help me. I feel like I am missing something fundamental, but can't seem to grasp it.

I'm using an AD826 dual operational amplifier using ±15 V supplies. I've breadboarded some very simple noninverting amplifiers with gain. Analog Devices recommends not using feedback resistors > 1 kΩ, so all of my feedback resistors are at that value. Lowering the value to 860 Ω allows me to add one stage before my problem occurs again but otherwise has little effect.

Individually each stage yields a clean output. The feedback resistance is 1 kΩ and the sense resistor is 330 Ω. I also have a 220 Ω input resistor. I'm using a 1 μF cap and a 0.01 μF cap for decoupling at each power supply rail of each op amp. There is no non-inverting amplifier (with gain) configuration shown in the datasheet: http://www.digikey.com/product-detail/en/AD826AR/AD826AR-ND/611660 however, the inverting amplifier configuration shown does the same thing (I am assuming the 50 Ω resistor to ground at the input is the equivalent circuit of the function generator and not a component of the circuit).

The problem is that once I connect enough stages in series, I see distortion in the output waveform (voltage output was less than half rail voltage) with a 150 kHz input sine wave (50 Ω generator output). A similar effect happens if I load down the output of a single stage - but I'm not sure how my chained circuits are loading each other down and how to remedy this problem.

Any advice would be much appreciated.

 

[psparky]

Have your second amplifier in series peaked?

In other words it can't deliver more than 15 volts according to your rail voltages.

So your first one can handle it fine, but once you multiply it by the second stage you may be over the limit.

Just my two cents.

 

[ajderoui]

Thanks for the reply, but I've tried adjusting the input to -20 dB attenuation. The output doesn't saturate, it only distorts. The signal I'm starting with is about 0.1 mVp-p. I have successfully set up several stages in series with small gains (about 4 each). I should, to my knowledge, be able to add as many additional stages with the same setup as I want (although obviously this is a poor design and I'd like to have higher gain at each stage) assuming my decoupling is adequate. The problem is with 3 stages each at a gain of say 8, I start to see distortion, and with 5 stages each with a gain of 4 I start to see distortion. I'm not sure if somehow I'm loading down the chips but I just don't see how

 

[Averagesupernova]

I think a schematic would be helpful.

 

[ajderoui]

amplifier circuit

This is really all I'm building using 5% tolerance 1/4 W resistors. I have more of these stages in series but that should give you an idea. I was originally trying to do a little filtering as well but was experiencing problems so I dialed it back to the most basic case I could think of.

 

[Averagesupernova]

Do you have the node where your 15 volt sources connect together connected to ground?

 

[ajderoui]

I'm assuming you are referring to a connection between two 15V supplies, where the negative of one would be connected to the positive of the other. In this case, I'm using a ±15V supply rated to 3A on either rail, so that connection is provided internally to the supply. I need only connect +15, 0, and -15V pins to my breadboard.

Like I mentioned before, each stage works perfectly individually, its only when I connect them in series that I have issues (there is no problem with the supply. it pulls a maximum of 200 mA when I connect three chips implementing the same circuit). The pins 4 and 8 shown in the diagram are the power pins for the op amp and are decoupled using 0.01 and 1 μF capacitors similar to the recommendation in the datasheet (although they show 3.3 instead of 1 μF this should not make a significant difference)

 

[Averagesupernova]

.2 amps seems REALLY high. Something is wrong.

 

[ajderoui]

The chips quiescent current draw is 15 mA each (typical). I have 4 hooked up. That's 60 mA just to turn them on. 200 mA is the max I ever see, and that's when I up the input voltage to nearly saturate the 6th stage output (25 Vp-p approx). 5 Vp-p at the input of the final stage across the 220 Ω resistor is 8 mA-rms. Each stage before that should be about 1/4 the previous, so we can say about 11 mA. Total that should be more like 100 mA max. I agree it is high but I can assure you the circuit is connected correctly and the above stated conditions are true. I have also replaced the chips with new ones in case I had some damaged ICs. The question is why

 

[Averagesupernova]

You have 4 chips hooked up or 4 individual amplifiers? This makes little sense. I still think you have something wrong.

 

[ajderoui]

The highest number of chips (2 amps each) was 4, which is when I saw the 200 mA. What doesn't make sense?

Let me give a very broad overview of my problem: I have a circuit that works fine with one stage. Ideally I should be able to add duplicates of these stages in series as many times as I want and continue to get the same effect, assuming I'm not amplifying all kinds of power supply noise into the output and assuming that the input noise of the amplifier is much lower than the signal of interest (which it is). However, when I connect multiple stages, the output is severely distorted (the signal I'm looking at should be about 100 mV but is more or less clipped at the negative end and highly distorted at the positive). I can combine more stages if each successive stage has less gain (higher input resistor but constant feedback resistor of 1k). This is an extremely simple circuit and I've changed everything out ensuring that my power supply, oscilloscope, amplifier, etc. are not the problem. The question is, what characteristic of the amplifier using this circuit causes my output to distort. I notice during debugging that loading down the op amp (i.e. small value resistor at output) has a similar effect of distorting the originally sinusoidal waveform

 

[Averagesupernova]

Are you sure you don't have a factor of 10 or 100 confusion going on here? Like a x10 scope probe mix-up? If you adjust the input signal way down does there get to be a point where the distortion goes away?

 

[ajderoui]

I've tried the probe at x10 and x1 settings see if the measurement was playing a role (and set the oscilloscope accordingly). I've also tried dialing the input signal way down to see if there was some limit in the output voltage swing. Neither change affected the shape of the output

 

[Carl Pugh]

My guess is that there is a high frequency oscillation.
The oscillation may be at a high enough frequency that you can't see it on an oscilloscope.

This may or may not work, place a 0.004 microfarad capacitor across each 1,000 ohm resistor. If this works, use smaller capacitors or a better layout.

There is a dual LM741 operation amplifier. If may be possible to unplug the operational amplifier you are using and just plug in a dual LM741.

Good Luck
Carl

 

[Averagesupernova]

Carl has a good point about the 741. When we run out of ideas it is time to start swapping out things to get something to change or to change it to something closer to what we are confident in and have more experience in. Then maybe something new will occur to us that hadn't before. The 741 is a very tried and tested part. Not very fancy, but pretty stable.

 

[ajderoui]

Thanks Carl, I tried some 3 pF capacitors in parallel with the feedback resistors. Filtering down high frequency oscillations did not seem to cause any change.

I didn't have any LM741's laying around but I do have LF353 which are JFET input general purpose amplifiers which I am quite familiar with using. After switching them out the circuit works fine, but the bandwidth of the AD826 is what I was really interested in, as I need about 100 dB amplification in the 100 kHz to 1 MHz range.

These are some oscilloscope readouts using the AD826 and the circuit I showed previously (plus the 3 pF feedback resistors). The first is with 4 stages. The second is with 5 stages. All are exact copies of one another. Adding the 5th stage causes some major issues as you can see. This problem did not occur with the LF353s.

I should also mention that the problem occurs with less stages if the gain on each stage is increased (regardless of input voltage). For example, if I were to change the sense resistor value to 100 Ω instead of 220, I would expect to see similar distortion appear at an earlier stage.

 

[Carl Pugh]

Some general suggestions:
AC couple each stage.
Put first stage(s) in a metal container with good ground to metal container.
Use batteries for first stage(s) until you get everything working.

 

[ajderoui]

As an update, while I said my LF353's work, they actually worked a bit better than I expected. Their GBP is almost good enough for this application - so I decided to try adding a few more stages to compensate for the attenuation at high frequency. The same type of distortion appeared in my output waveform after adding one more stage. Whatever is happening is being made worse by the inclusion of more stages (I did not add a chip, just connected the second op amp of another I already was using).

 

[Averagesupernova]

I am guessing you have power supply decoupling issues.

 

[Carl Pugh]

The input operational amplifier's (OA) offset voltage is driving the final OA stages into saturation.
AC couple the OA's.

 

[turbo]

In tube amps, the outputs of initial stages are often (softly) clipped and the output to the next stages can be attenuated. I tried tinkering with solid-state amps and found myself frustrated. I realize that this reply may not be helpful, but after ~50 years tinkering with multi-stage audio, I have found that some things do not work too well. Good luck!

 

[Averagesupernova]

If it is an offset issue, this would be a good reason to use inverting stages. If the OP is using 4 chips (8 stages) with each stage having a gain of 4, it would be understandable why this could be troublesome. My quick math puts the voltage gain of this chain of amplifiers at 65536. My rule of thumb with op-amps is to always try to use inverting stages. Sometimes a super high input impedance is required and of course this can be accomplished with a non-inverting buffer stage on the input with a gain of one.