Struggling w. a Single Supply Integrator

[eev_carl]

Hi,

I simulated 2 op amp integrator circuits from a Mims notebook.  The dual supply version works fine and I built it on my breadboard.  A square wave is converted to a triangle wave.  This is the DualSupplyIntegrator attachment.


I'm having trouble with the second circuit which is a single-supply integrator.  I'm not sure if this is an LT Spice problem in the way I set up the supply or something else I'm missing.  This version of the circuit AC couples the input, sets V- to ground, and replaces the voltage at the + terminal with V+/2.  SingleSupplyIntegrator

Thanks in advance,
Carl

[ogden]

I'm not sure if this is an LT Spice problem

Rest assured this is not LTSpice problem. Make both circuits as one simulation so you can compare waveforms 1:1. Beauty of simulation - you can measure voltages/waveforms in any part of the circuit. So do it - compare two circuits, go through them starting from input. Find differences and think cause of them. Maybe input capacitor is too small? If you increase input capacitor - it most likely will not have time to charge during quite short simulation time? ... and so on...

[eev_carl]

Hi,

I didn't mean a problem with the software itself but how I arranged the components in LT Spice.

My question is about this specific circuit.  The following attachment is the pair of circuits from the book.

Thanks,
Carl

[ogden]

First test you could make - shift not only opamp supplies but your signal source as well. Make +/- 2.5V pulse going around vref (4.5V) which is 2V to 7V and see what's coming out of the circuit. ANALYZE, do not ask me to do it for you

Also name signal lines in your 2nd circuit. How would I know where vn002 and vn004 are?

[eev_carl]

Is this what you mean by making the pulse go around Vref?

[chris192]

I think your circuit should work.
The bias resistor is a good addition.

I don't use LTspice but I would check the netlist to make sure everything is connected how you would expect it to be.

That's a very neat notebook.

[eev_carl]

Thanks for confirming the circuit.  I think it was an LT Spice connection issue because this is working for me after I recreated the schematic.  I pulled out the bias R in accordance with the original schematic. 

[eev_carl]

And to close the loop, my problem was as Chris said, connections.  The feedback capacitor wasn't hooked in. Additionally, there was a problem referencing V2 as the power source for V+ on the op amp.  The resultant triangle wave was misshapen so I created a V+ label between the voltage + terminal and the 10k resistor.

Here's the original LT Spice simulation that I reworked.

[rstofer]

It seems to me that your AC coupling is a form of differentiator.  So, you have a signal which you differentiate and then you integrate.  I would suspect that the result would be the same shape as the original signal with perhaps some amplitude difference.

[eev_carl]

It seems to me that your AC coupling is a form of differentiator.  So, you have a signal which you differentiate and then you integrate.  I would suspect that the result would be the same shape as the original signal with perhaps some amplitude difference.

How would you implement a single-supply integrator without the AC coupling?  I breadboarded this and got the same triangle wave as in LT Spice, not a square wave (though with a big overshoot).

[rstofer]

I am just guessing but in a single supply system, there are no negative voltages.  So, why does your square wave not center around 2.5V and swing from 0V to 5V.  The absence of input, what would be 0V in a dual rail configuration is 2.5V in the single supply design.  You would expect that with the input held to 2.5V, the output would eventually become 2.5V

You could post your .asc file so others could play with your circuit without having to create it from scratch.

The latest simulation (just above) doesn't show a triangle wave at all.  There is too much exponential stuff going on.

[ogden]

I am just guessing but in a single supply system, there are no negative voltages.  So, why does your square wave not center around 2.5V and swing from 0V to 5V.

Many lo-end audio circuits are exactly like that - single supply, input capacitor, signal centered around ground ( 0V ).

[Zero999]

Try increasing C2 to 10μF

Increase R2 to 470k or even 1M. It only has to allow enough current through to bias the op-amp. If the op-amp isn't a bias compensated, J-FET or CMOS input, then there should be a resistor between Vref and the non-inverting input, with the same value as R2. I think the LT1001 is bias current compensated, so the resistor in series with the non-inverting input would make it worse, but if you're breadboarding it with the uA741 which isn't bias current compensated, then include it.

[rstofer]

I am just guessing but in a single supply system, there are no negative voltages.  So, why does your square wave not center around 2.5V and swing from 0V to 5V.

Many lo-end audio circuits are exactly like that - single supply, input capacitor, signal centered around ground ( 0V ).

True, but they usually bias the input pin at Vcc/2.  The problem with the circuit above isn't so much a matter of bias as the fact that the signal can go negative, below ground, and the integrator can't follow along.

I don't see how adding bias to the input pin will work with the signal going below ground and I don't see how the bias will work at all without affecting the input - the reason the capacitor was put there in the first place, I suppose.

I would like to see what happens when the input signal is changed to 0-5V and the input capacitor removed.  But not enough to lay out the circuit from scratch.

Analog computing, my main interest in op amps, is usually done with integrators using dual supply op amps but I don't see why it couldn't be done with single supply op amps just as well.  It's just a case of Vcc/2 is the new 0V.

I also don't see how the single supply design can get to the rails because rail-to-rail is a marketing term, not an engineering fact.  In an analog computer scenario, we would probably leave a volt or so off of each rail so, for a Vcc of 5V, the signals would vary between 1V..4V and the center would still be at 2.5V

ETA: I would adjust the signal source to center on 2.5 and have a 3V peak to peak amplitude.

For the old 741 integrators, the supply was +-15V and the signals were constrained to +-10V.

[eev_carl]

if you're breadboarding it with the uA741 which isn't bias current compensated, then include it.

It's an LT1001. 

[rstofer]

Maybe it's complicated like Maxim says in this white paper:

https://www.maximintegrated.com/en/app-notes/index.mvp/id/1155

I was thinking about using modern single rail op amps for another analog computer.  I think I'll give that a pass.

[bson]

You're using the V2 node for a supply voltage while V2 is actually a voltage source, not a node name...  Not sure what Spice will do in this case, but quite probably not anything reasonable.  (The reasonable behavior IMO would be to produce an error.)

[rstofer]

Right, use an OUTPUT port where you generate Vref and an INPUT port at the op amp.  The port icon is on the toolbar between Gnd and Resistor.

[Wimberleytech]

Right, use an OUTPUT port where you generate Vref and an INPUT port at the op amp.  The port icon is on the toolbar between Gnd and Resistor.

Ditto...I simulated it with ports configured properly and it works fine.