TL071 hand-made simulation using discretes

[steph_tsf]

hello,

can somebody say to me what's wrong in this Tina 7 T.I. simulation file ? I have replicated the TL071 schematic given in the datasheet. It delivers good figures regarding power supply drain, gain-bandwith product and slew-rates for all non-inverting gain configurations between 10 and 1000. I have not yet tested the inverting gain configuration.

For smaller gains like x2 or x1 unity gain buffer, the results are highly variable, depending on details like :
- load resistance
- load capacitance
- serial resistance of the generator
- input capacitance
If you simulate with a 22K load resistor and 33pF load capacitor, 100 ohm serial resistance of the generator, and 10 pF input capacitance, you will get alarming simulation results.

Set the load to 100K, decease the input capacitance to 5pF, and you'll get a unity buffer gain plot showing a MHz oscillator.

What will thus happen if one puts a TL072 wired as unity gain buffer after a volume potentiometer ? This is a case where the serial resistance of the generator can go from a few ohms (zero volume or full volume, or channel short-circuit by the balance pot), to more than 50K (a 100K pot at mid position).

Don't you think there is an issue, here ? Where am I wrong ?

I'm attaching a few files.
The .zip contains the Tina 7 T.I. schematic ready for simulation.

Cheers,
Steph

[wahab]

Your simulated schematic has too much bandwith,
as the GBW of the TL071 is no more than 3 mhz.

You should reduce the current in the input stage or in the vas
to reach this value;

Anyway, the jfet you use have not the required specs.
Try using toshiba devices..

[steph_tsf]

Originally Posted by wahab

Your simulated schematic has too much bandwith, as the GBW of the TL071 is no more than 3 mhz.

Definitely wrong answer. The simulated GBW is 3 MHz. See attached pictures, when the amp is configured for a x1000 gain.

[wahab]

Originally Posted by steph_tsf

Definitely wrong answer. The simulated GBW is 3 MHz. See attached pictures, when the amp is configured for a x1000 gain.

Yes, for a 60db gain , i m wrong, but for a 0db gain , i m right and
you are wrong....

Before jumping to this conclusion, make a test using toshiba jfets...

[steph_tsf]

hello wahab, I think we are half way the truth. Using a x1000 gain, the GBW of the simulated TL071 is very close to 3 MHz. Using a gain of x1 (unity gain buffer), the GBW of the simulated TL071 gets a little bit inflated, but not so much. Basically, I think the model behaves okay. I agree that the transistors used in the model surely need better parameters, to cope with IC technology. I don't think NPNs and PNPs in integrated circuits like the TL072 behave like NPNs and PNPs that are available as individual components. If you could be more specific in your answer, like suggesting some parameters tweak in the NPN and PNP models, that would help. Same for the PFETs : what model would you advise ?

[wahab]

Perhaps this models has better specs than the j175.

[kenpeter]

I had trouble with discrete transistors when I modeled a JLH
headphone amp based upon TI 47LS04 logic gates. Currents
passed by the datasheet specified internal resistors proved
insufficient to bias even the smallest LTSpice discrete model.

I remember ended up paralleling the resistance values of the six
gates instead of scaling down the transistors. So, I never really
solved the problem the correct way. But I seen the same what
you are fighting with for sure...

[satoru]

It's not clear to me what is the purpose of this simulation, but if you want to reproduce the behavior of TL071, then you have to pick an input JFET with lower gm, e.g. ~0.3 mS at 0.12 mA. J175 is not too far but still a little a bit too high, I think. Using 2SJ74 goes to the opposite direction from J175 (higher gm). Also, the Ft of PNP transistors in TL071 is merely a fraction of discrete PNPs (~5MHz). Maybe inserting resistors in base of output emitter follower Tr, or inserting a cap to connect the two bases can tame the oscillation?

Regards,
Satoru

[steph_tsf]

Originally Posted by satoru

It's not clear to me what is the purpose of this simulation

Hello, thank you for the valuable info you provided. Actually, I'm trying to see what kind of compromizes are done inside audio opamps for getting them (more or less) stable if used as unity gain buffers, with uncertain source impedances and with uncertain load impedances.
On one hand, I have the impression that internally compensated audio opamps like TL071 are building blocks for dummies, vastly over-compensated for allowing dummies using and abusing them, and never getting unstability.
On the other hand, I don't understand how faster audio opamps like OPA134 are stable in a huge domain. What kind of tricks do they use, for getting their behaviour nearly independent of loading, and nearly independent of source impedance ? If I'm designing such fast opamp using discretes, I can't get a stable unity gain follower. What do they put inside for getting speed, slew-rate, and stability, all at the same time ?

Those last weeks I'm concentrating in modelling power amplifiers, using an audio opamp as front-end. Using an audio opamp as front-end seems an interesting path for DIYers :

- they can select the audio opamp of their choice (if detailed instructions provided about stability for each supported model)
- they don't need to buy unobtainable dual matched JFETs
- they don't need to buy quantities of single JFETs, and match them
- they don't need to wire the associated current source, cascodes, and so

There are some interesting features, using an opamp as front-end :

- using a quality JFET audio opamp at the input, one may supppress the input capacitor, and one may apply DC amplification
- one can apply a conventional topology, using the opamp output, and boosting it a little bit by an intermediate gain stage using discretes
- one can apply a "current feedback" mode amp, using the current that is taken by the opamp supplies, and feeding the driver with dynamic currents that are more than twice the driver quiescent current
- one can apply weird topologies like improved BUZamp, revisited QUAD405, or nanoQSC, at minimal cost and minimal complexity

Last time I have used a pre-programmed model (TL072), I went fooled by the power supplies of the pre-programmed model not following the current that's delivered at the output. This ruins simulation for "current mode" topologies, where it is the supply taken by the audio opamp that's activating the driver.

Regards,
Steph

[steph_tsf]

hello dear friends,

here you can see that the built-in Tina 7 T.I. TL071 model is wrong.

Please have a look to the schematic. You can see both circuits being wired as x10 amplifiers, loaded by 330 ohm, with a 1V 1kHz sinus at the input. That's 10V on the output, hence about 30mA current into the 330 ohm load. This current is supposed to come from the power supplies.

Please have a look to the system waveforms. The built-in Tina 7 T.I. TL071 model is wrong regarding the current taken by the power supply pins.

The hand-made model using discrete is better. It shows the current being taken by the supply pin, and it shows the limited current that's available at the output of the TL071 (clipping).

For some audio applications, like current feedback topologies using a TL071 as front-end, the supplies are the signal.

Hope this explaination clarifies why this hand-made TL071 simulation has been undertaken.

Cheers,
Steph