4
\$\begingroup\$

I've run across strange opamp behaviour I can't understand. Voltages on TP1 and TP2 change drastically when I connect them to opamp inputs. Even more - when I swipe R2, voltage on TP1 changes too. Nevertheless, the circuit works ok - when voltage on R2 drops below voltage set on R1, opamp output goes high and then voltages on both pots seem to be right.

Opamp is LM6152. I thought that maybe i killed the opamp, but another brand new device works exactly the same.

If I change opamp to, say, LT1638 everything works as in ideal opamp tutorial.

Could someone explain me what's going on? Is it current feedback opamp? It's nowhere mentioned in the datasheet. Input bias current is 500nA. One strange thing is it lists maximum input current +-10mA in maximum ratings.

circuitlab schematic of circuit in question

Edit

MOSFET really is FDD6692.

\$\endgroup\$
6
  • \$\begingroup\$ Your circuit uses positive feedback, is that what you want? \$\endgroup\$
    – The Photon
    Jun 13, 2012 at 20:29
  • \$\begingroup\$ Yes, this was ment to be overcurrent protection in bigger circuit. I isolated this particular circuit on separate board to test for strange stuff I ask in question about. \$\endgroup\$
    – miceuz
    Jun 13, 2012 at 20:39
  • 1
    \$\begingroup\$ @miceuz, can you make it so that your image of the circuit in circuitlab clicks for me to look at the circuit itself? That way user can click and simulate and play with themselves. If you are not sure how to do this and give me a link to your circuit I can update it and you can look at what I did for future reference. \$\endgroup\$
    – Kortuk
    Jun 14, 2012 at 13:35
  • \$\begingroup\$ Kortuk is right: why don't you simulate it, since CircuitLab has a model? \$\endgroup\$
    – clabacchio
    Jun 14, 2012 at 13:44
  • \$\begingroup\$ CircuitLab does not have a model for it, you can just change name of a part in CircuitLab. \$\endgroup\$
    – miceuz
    Jun 14, 2012 at 13:50

3 Answers 3

3
\$\begingroup\$

That's an extremely nice opamp!
Also rather expensive.
But nice !! :-)

  • Something is fundamentally wrong with the real-world circuit that is not shown on the diagram.

The described behaviour makes no sense at all in one area so the rest is suspect.
That is, once one thing is completely un-understandable it can mean there is some major factor that has been missed.

LM6152 data sheet here.
The non inverting input is high impedance and NOTHING that the opamp does in normal use will affect it's voltage - so if something does then something is very wrong outside normal opamp behaviour.

The LT1638 data sheet here that you say worked is also a nice op amp, but in almost exactly the opposite way.
It couldn't pull the skin off a rice pudding downhill on a good day with the wind behind it.
Its forte is super low power and it has super low bandwidth and slew rate to go with it.
Whereas, the LM6152 is a 75 MHz bandwidth !!!! 45 V/uS slew rate !!!!!!!!!! stunner more at home in ski-jumping or single digit standing 1/4 miles.

SO I'd guess that the LM6152 is having major instability and oscillation problems which the LT1638 avoids by virtue of just being so darned slow that it is not pushed into oscillation by whatever is troubling the LOM6152. It may be that better power supply decoupling really close to the IC, or a whiff of low pass filtering in the feedback path (maybe a resistor << R3 in M1 source and a NF or less on inverting input to ground?, ...?) may help.

If you have no decoupling at all on the IC power supply near the IC, write it in the lessons learned book - even if it doesn't fix the problem.


OpAmp overloaded:

It does not seem to be what is causing the problem, but you opamp is sinking more current than allowed for a formally correct design when the LED is on.
With opamp output high Vout is 15V (rail-rail amp)
If VLED = 2.5V then Iout = (Vout-VLed)/ R4 = (15-2.5)/1k = 12.5 mA

Isource _max_typical = 6.2 mA, minimum = 3 mA and max = 17 mA.
So your attempted output is double typical available and 4 x minimum and less than best case.
This could do no worse tha pull Vout well below rail - and worst case could cause device malfunction.
However, it does not seem to be relevant to the problem that you describe.

Note: Formal professional design requires you to design for worst case parameters in the worst case situation that you require the design to work in.

This often leads to performance far below what it appears a device may usually achieve and often worst-case design is extremely conservative and functionally unnecessary.

However, if you want the circuit to always work, this is the correct way to design it.

\$\endgroup\$
3
\$\begingroup\$

Something about the output. The LED has a maximum forward voltage of 1.7 V, no minimum specified. The IRF530 has a \$V_{GS(TH)}\$ of maximum 4 V, so your FET may never switch on. You need a separate resistor from the opamp's output to LED and FET resp. The one for the FET is required because the reset switch would otherwise short the opamp's output.

Nice, a DC application for a 75 MHz GBW opamp. :-)

\$\endgroup\$
4
  • \$\begingroup\$ well, opamp was random ;) i was using half of it for constant current regulation, but even there it was way too fast as current sense amp has a bandwidth in kHz. I forgot to change MOSFET marking on schematics, sorry about that. I'm really using FDD6692 (html.alldatasheet.com/html-pdf/51327/FAIRCHILD/FDD6692/406/1/…) it has Vth of 1.6V and it does turn on and schematic works as expected, just voltage reading on TP1 was weird - it changed depending on voltage on TP2 and sometimes was more than 5V. \$\endgroup\$
    – miceuz
    Jun 14, 2012 at 13:44
  • \$\begingroup\$ @miceuz - Sorry, I have no answer for that, except Russell's: "Something is fundamentally wrong with the real-world circuit that is not shown on the diagram". Inputs of opamps do not influence each other. There's the law! :-) \$\endgroup\$
    – stevenvh
    Jun 14, 2012 at 13:47
  • \$\begingroup\$ I know about this law, hence, the question ;) now the circuit is EXACTLY as on my breadboard. I suspect @RocketSurgeon is right. \$\endgroup\$
    – miceuz
    Jun 14, 2012 at 13:54
  • 1
    \$\begingroup\$ [service message] @miceuz - you don't have to use the "@" in front of a user's name, unless you want him to be notified of your comment. And he will only be notified if it's his answer, or if he commented on it. \$\endgroup\$
    – stevenvh
    Jun 14, 2012 at 14:07
1
\$\begingroup\$

There are few differences to normal opamp.

  1. Datasheet says that opamp has a proprietary dynamic slew rate booster. So slew rate is not constant and goes higher with higher input value. What it means for DC, is possibly affecting differential input impedance for high diff voltages. The boost circuit is certainly some extra group of transistors wired closer to both inputs in some proprietary way (some TI trade secret).
  2. Datasheet does not specify differential input impedance, but only common mode one. So the differential input impedance can be few orders less. Say 5K\$\Omega\$ when common mode is 50M\$\Omega\$.
  3. The use looks like a comparator/latch thing, when opamp is typically expecting difference between inputs in range of few millivolts at most. If difference is more than 100mV then many specs are not valid. Right. For most amplifiers it is not noticable, but the device in question is a video speed amp, so it is specced for typical use case with very small diff voltage. Designers did not expect a comparator/latch mode use.

It does not mean that use of device is a mistake. The low impedance is simply a side effect with no catastrophic breakdown or errors.

To check the theory, you can make both voltages very close and see if common mode impedance is good when voltages are close and bad when voltages are far apart.

You are possibly close to reverse engineering some TI trade secret. By accident.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.