Hello dear AAC forum!

I am attempting to build a fast amplifier, but have stumbled upon a bit of a wall. It's behaving quite strange!
First of all, here is relevant schematic:


It consists of a wideband current feedback amplifier (OPA695) set to a gain of G=+4, AC-coupled to the input signal. It is operated with V-=0V, V+=9V, giving it a midpoint voltage of 4,5V. A 50k resistor on the non-inverting input opens a path to ground for the input bias current. A slow voltage-follower op-amp (µA741C) sets a bias voltage on the inverting input, enabling me to raise and lower the base output level as I wish.

I have designed and simulated this circuit in TINA-TI, a spice-based circuit simulator from Texas Instruments, and it works as expected. The OPA695 amplifies input signals by 4, and the bias circuit allows me to set the output base voltage by adjusting the P3 potentimeter. However, after assembling the circuit, it is behaving unexpectedly:

1. The bias applied to the inverting input is also seen on the non-inverting input
2. The signal applied to the non-inverting input is also seen on the inverting input, BUT:
3. The amplified signal is miniscule (~10% of input signal) AND negatively amplified!
4. The base output voltage is always 880mV (negatively saturated) regardless of bias voltage

At first, I thought I had fried the OPA695 during soldering and thus a low-impedance path had been created internally between the two inputs. I assumed this because when probing the two inputs, they showed nearly equal base voltage, and equal signal shape and amplitude riding this base voltage. Meanwhile, the output was constantly negatively saturated, with a tiny, negatively amplified signal riding the DC value of 880mV. Especially the negative amplification puzzled me, as the circuit is designed for positive gain.
After scrutinizing my PCB to make sure it was just not simply due to a misshap when laying out the components or from a short between the two inputs, I replaced the OPA695, taking care to not fry the tiny SOT23-6 package. However, all of the same unexpected properties remained.

I am at a loss here. The circuit is showing such wildly differing behaviour from the simulated circuit that I am wondering if I have missed something fundamental. Any help or tips are appreciated.

PCB layout:

The 450ohm feedback resistor is located directly below the OPA695, as recommended by the datasheet.

OPA695 datasheet: http://www.ti.com/lit/ds/symlink/opa695.pdf

 

Especially the negative amplification puzzled me, as the circuit is designed for positive gain.

No, it isn't. R9 sets the positive input at DC ground. The positive voltage on the inverting input makes the op-amp try to null that input to zero volts, but it doesn't have a negative voltage power supply, so it can not do as you told it to do. Bad design.

 

To correct things, disconnect the 741 from C10/C11 and connect it to the GND end of R9. With C10/C11 in place, the gain of U1 reduces to unity at DC. So if the + input voltage reference is 4.5 V coming from the 741, the output will sit at 4.5 V. Note that the 741 inputs and output do not function near its rails. Adjusting its input below about 2 V will not work.

#12's analysis leads to a question: how did this circuit pass simulation? This is not a subtle or device-specific error, and the sim should have shown the saturated output.

ak

 

how did this circuit pass simulation?

I expect thousands of simulations work correctly every day, but the sims that don't work arrive here. That gives me an average input of, "Simulators can't be trusted". Some of it is operator error, like when Martinez tried to use a 10Hz triangle wave to simulate an audio signal. (The input capacitor simply wasn't the right size for that frequency.) This one is more like, "How could a simulator miss that?"

I don't Sim, but other people here might examine the sim files if the Thread Starter posts them.
Then again, maybe nobody here runs TINA.
Smily shrugs.

 

Or, TI - ever the late-comer and analog wanna-be - has whipped up a sim program to compete with LT spice, and it isn't as good.

Which means it is an excellent simulation of TI parts vs. LT parts.

ak

 

No, it isn't. R9 sets the positive input at DC ground. The positive voltage on the inverting input makes the op-amp try to null that input to zero volts, but it doesn't have a negative voltage power supply, so it can not do as you told it to do. Bad design.

That is what I assumed as well, before simulating. But it turns out that the value of R9 decides what voltage the positive input will sit at. I think this is due to a constant bias current flowing out of the positive input, creating a voltage drop over R9. Setting R9 to 50k gives me a positive input base voltage of ~2,5V. Removing R9 causes the positive input to charge the capacitors C9 and C8 and saturate. I could choose a lower value for R9 to lower the base voltage, but this would decrease my input signal.

To correct things, disconnect the 741 from C10/C11 and connect it to the GND end of R9. With C10/C11 in place, the gain of U1 reduces to unity at DC. So if the + input voltage reference is 4.5 V coming from the 741, the output will sit at 4.5 V. Note that the 741 inputs and output do not function near its rails. Adjusting its input below about 2 V will not work.

#12's analysis leads to a question: how did this circuit pass simulation? This is not a subtle or device-specific error, and the sim should have shown the saturated output.

ak

Thank you both for the responses. Yet, I am not sure if this explains the strange behavior of the inverting and non-inverting input showing practically equal values, regardless of what the output is doing. Particularly, how would the DC bias applied to the inverting input be able to show up on the noninverting input?

I would attach the sim file, but AAC does not allow the file extension, unfortunately.

 

Relying on a very broadly specified error term to set a critical output parameter is a horribly bad design practice. And no one is suggesting removing R9. Ever. The non-inverting input needs a DC reference potential.

When the negative feedback loop is closed, the inverting and non-inverting inputs will be at the same potential. That's what negative feedback does.

ak

 

Relying on a very broadly specified error term to set a critical output parameter is a horribly bad design practice. And no one is suggesting removing R9. Ever. The non-inverting input needs a DC reference potential.

When the negative feedback loop is closed, the inverting and non-inverting inputs will be at the same potential. That's what negative feedback does.

ak

Well, it turns out you were right in attaching the bias to R9. Thank you kindly for your help! You too, #12.

 

#12 doesn't know how right he is! I have had to troubleshoot thousands of analog designs, and usually I am assured that the simulation works, but the breadboard doesn't work. When I hear this I know that the designer knows next to nothing about circuits. I wrote a magazine article that pointed out that very, very few op amp models match the real item for ac performance. Stray capacitance etc. is not taken into account by the simulator. Wiring mistakes, wrong pinouts, exceeding voltage capability, etc. are part of another batch of errors that the simulator never sees. If I was once again put in the penalty box as a design manager (please no, no) I would require that all designs be built and tested prior to simulation. A simulator is not a circuit designer; it is a very powerful slide rule and nothing more; it analyzes data (sometimes poorly) and nothing more.

 

...................
If I was once again put in the penalty box as a design manager (please no, no) I would require that all designs be built and tested prior to simulation. .................

Then what's the purpose of the simulation?
I would think you would want to build and test the design after the simulation.
Or is that what you meant?