Differential opamp troubles

[SethGI]

Hi guys,

I'm building an amplifier for a signal processing board with this chip as the main opamp: http://www.ti.com/general/docs/lit/getliterature.tsp?genericPartNumber=ths4521&fileType=pdf
It's a fully differential opamp, and I'm having some troubles with it. It's almost working, but it's being a bit weird. I give it a Vocm input of 2.5v (well within recommended range). The positive output is correct - it is centered around 2.5v and gives pretty much exclusively the positive side (a bit negative, but I'm not too concerned).

The negative output, however, is by default at 5ish volts, which is the supply voltage. When the positive input rises above 0v, it goes negative (which is right). I just don't know why it starts at 5ish volts and not 2.5 like it should.

Here's the schematic:



THANKS!!!

[dmills]

Vocm = +2.5V, non inverting input is AC coupled via C12, but inverting input is connected to Agnd via R11.....

Voltage gain is -4 by a guick look, so I would be expecting the negative output to be trying to peg at 10V or so.

I think another cap in series with R11 would probably help matters.

Incidentally, you might want to look at what sort of -3dB corner 100n + 1k gives you, if you are trying to work anywhere near the audio band?

Regards, Dan.

[SethGI]

Hi,

I added a cap and that helped with the bias offset issue. I don't quite get what you meant by " you might want to look at what sort of -3dB corner 100n + 1k gives you, if you are trying to work anywhere near the audio band?"

Can you explain a bit more please?

Also, the cap I added (.1uF ceramic) introduced a ton of noise. Like a real ton of noise. Am i using the wrong capacitance/type of cap?

Thanks,

Seth

[whollender]

The combination of a 100nF cap and a 1k resistor results in a high pass filter with a 3dB corner of ~1.6kHz.  If you are trying to do audio frequencies (20Hz - 20kHz), you'll need a much bigger cap.

The cap type also makes a big difference.  If you use a ceramic cap that isn't a C0G/NP0 type, it will introduce a lot of distortion due to the capacitance changing with voltage.  The may also introduce more noise due to microphonics.  Note that ceramic caps with the C0G/NP0 dielectric aren't available in large capacitances, so if you want a good DC block for low frequencies, you'll probably want to go for an electrolytic.

I used the same amp you're looking at (the THS4521) in my SuperAudioBoard project (https://hackaday.io/project/5912-teensy-super-audio-board), but used 100uF electrolytic caps on the inputs to maintain a very low corner frequency on the inputs.  The schematic and a design guide where I walk through the design decisions are in my github repo at https://github.com/whollender/SuperAudioBoard

[dmills]

You did add that cap in series with R11, not across it right?

Also, that input stage looks like something for an electret mic to me, are you sure it will be ok into 1k ohm, seems a magnitude low to me?
If that is the source, resistor noise in this amp is a non issue because the mic itself will have huge self noise.
73 Dan.

[SethGI]

I added a 100uF in series and it helped a bunch. There's still a lot of noise going on, but it seems to go away when I remove C33 (on the output). On a scope it is clearly a cap charging and discharging. Why would this happen? Is it a problem if I just ditch that cap (the noise goes away without it)?

Thanks,

Seth

[whollender]

It depends on the purpose of your amplifier circuit.  I've usually seen that circuit topology used to drive differential input ADCs.  If that is the case, then C33 is required for good ADC performance (it acts as a charge buffer so that the sample and hold cap in the ADC doesn't significantly change the input voltage).

Just to double check, you added 100uF in series with both the input, and R11, right?

Otherwise, the only other thing that looks weird to me is that your feedback capacitors (C31/C32) seem much larger than necessary. As a quick rule of thumb, the feedback resistor and capacitor will tend to start rolling off the gain around 1/(2*pi*R*C), which results in a corner frequency of only 7kHz (this isn't exactly correct because the feedback R and C aren't directly in parallel, but it will give us a rough idea).

I would try reducing the feedback caps to around 1nF, or a bit lower.  If you go too low, the load capacitor will make the circuit unstable, but 1nF shouldn't be any problem.

What kind of signal are you using to test the circuit?

[SethGI]

You're right that it's for an ADC. I do have it in series with both the input and R11.

I'll try lower caps. I was using the recommended values from the ADC datasheet on page 39.

I'm using this for underwater sonar processing. I have a hydrophone (underwater microphone) hooked up to the circuit to test and I've just been banging on it.

If I do need that C33 cap, how can I keep it from causing so much noise?

[dmills]

Hydraphones are usually a piezo ceramic transducer and really need a jfet input stage not a low noise resistance bipolar opamp, but that aside, your issues with noise may be thing thing bursting into oscillation, that opamp is good for 150MHz GBP, so good layout is going to matter, stick a scope on the output and have a look.

Also, those pole splitting caps in the feedback network are at least an order of magnitude too large 4k+ 5.6nf? Try more like 220pF as a first cut...
The value of the cap should go down as the feedback resistor goes up if you want to maintain constant bandwidth... 

I am not at all sure I would have started a sonar preamp from here, it might be a sane opamp to use after a couple of cascoded jfets, but not IMHO directly off most sonar transducers.

Regards, Dan.

[whollender]

What ADC are you using?

As Dan mentioned, the circuit you show is great for driving the ADC inputs, but it won't work well as a first stage after your transducer.  The diff amp has a very low input impedance (only 1k ohm), so you'll want some sort of buffer in between it and the transducer itself.  A simple voltage follower would probably work very well because it will present a high impedance to the transducer, and a low impedance to the diff amp.  As Dan mentioned, a JFET input OP will have the best characteristics (noise, distortion) for this kind of application.

If you are biasing the transducer via R13 from the +9V supply, any noise on the +9V supply will be directly added to your signal.  You may want to look at alternate biasing arrangements, or make sure that the +9 volts is very low noise.

If you want to determine if the 'noise' is coming from the diff amp circuit, disconnect the input and ground it (also disconnect R13 if possible).  This will allow you to look at the noise of the diff amp circuit by itself.

The output cap (C33) is required for the ADC to work properly, so you do need it.  The only way that it can create noise is if it's causing the circuit to oscillate, or if it's a ceramic cap that isn't a C0G/NP0 type.  Ceramic caps with lower grade dielectrics exhibit a piezoelectric effect, so they can pick up vibrations in the environment and translate them to electrical signals.

All of the caps in the signal path (DC block, feedback, and output cap) should be of C0G/NP0 type if they are ceramic caps.  In addition to exhibiting the piezoelectric effect I mention above, the lower grade dielectrics also change capacitance significantly with varying voltages, which causes distortion.

What frequency range are you looking for?  This will help us understand what is required from the circuit.