This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LM324: Surprising behavior of a simple amplifier

Part Number: LM324

I have years of analog designs, normally doing complicated, or fast or high precision designs, however this time I was requested to design a simple photocell data logger with only 8 bits of precision therefore I used a simple LM324 device as a transconductance amplifier with 32.4Kohm gain with the +Input grounded and the -Input to the diode and the feedback resistor (+a 1nF high frequency cut of capacitor also in parallel to the 32.4Kohm feedback resistor) Power supply is only positive 15.5V.

I was surprised it didn't work , until I touched the output of the amplifier with the 10Mohm scope probe. 

I notice that by loading the OpAmp output by a 4.7Kohm to GND it worked OK. 

I assumed the device was faulty and I replaced with an other from an OTHER batch, with the same results.

By the specs the input common mode includes GND , therefore I have NO explanation on the behavior.

I delivery the circuit with the 4.7Kohm load... but why?

Does someone has a clue?

It troubles me that I have no problems making conditioners with 24bits precision, acquisition systems with 64GBPS throughput... but this simplest project didn't work as expected, it shows that we can never be sure.

Eli Jacob

  • Eliezer,

    Seeing that it didn't work until it was touch makes me suspect a bad connecting in the circuit.

    Does the output connect to anything other then the feedback resistor and cap?  I ask because LM324 can only have an output near ground when sourcing current or sink only a small current <30uA.

    Can you provide a couple input current and output voltage points for not working and working (with 4.7k)

    Does it not work for a certain input / output range or is the gain incorrect for the whole range?

    This is a simulation

  • Thanks,
    In fact I had 4 cards with 6 x channels each, all working the same, therefore is neither a failure in contact .
    Also the problem is not that the output did not go near GND because of lack of drive, the strange thing is that the output behaved noisy (but not oscillating) at a low voltage , fluctuating between 0.2 to 0.5V .
    Only when loaded with the 4.7K the output got to the right voltage (between 1.5 to 2.5V) ==> The voltage RISED when loaded with the resistor.
    It behaves like the common mode barely includes GND and the device is unstable on those conditions WITHOUT a slight load.
    A load as light as 1Mohm suffice. With the 10Mohms of the probe it still fluctuating but between 0.8 to 1.5V (closer to the expected but not enough)
    The fluctuations may be triggered by the noise of the PS, since because I doesn't need precision I didn't add a second LDO for the analog (as I commonly do)
    Is it possible that with the high PS value 15.5V with noise of ~0.1V impair the near GND common mode of the device?
  • Eliezer,

    I have two competing ideas. Before I can chose the right one, I need to know the voltage on the inverting inputs.
    Ideally they should be very close to 0V. If they are more negative than -400mV then the output may have a phase reversal which will cause an incorrect output voltage.
  • Right!
    With the load on the output, the output voltage goes about to 1.5V cancelling the photodiode that should be in that case around 50uA (1.5 V/32K) , and the I- is effectively virtual ground.
    Therefore without the cancelling current the I- may go NEGATIVE
    That may explain the unstable operation when the output was near zero.

    However that doesn't explain why the 4.7K load fixed it. Unless the drive requirement provides a way (in the internal circuit) to bias the input stage to tolerate some more negative millivolts and that is what the LM324 needs to get back to correct operation , making toe output go more positive further makes the I- closer to 0 and the LM324 jumps to normal operation.

    If that assumption is true I would recommend to make a note that for a photodiode to be correctly operated with LM324 family and any other amplifier that can be unstable if less than -500mV is present on their input (Open voltage of a photodiode) should be biased accordingly .

    Also it should be mention that the LM324 family is unstable and may produce a phase reversal if the input go under -300mV.
    On the contrary now looking in detail on the spec I found there is even an example of the same connection I did with 1Mohm feedback resistor! That won't work, as I experimented.

    Can you explain why the 4.7K load fix the operation? I would expect that the additional load drive even more close to 0 the output getting more negative the I- and therefore aggravating the results, but on the contrary ... the load to GND make the output get more positive fixing it...
  • Eliezer,

    I tested this setup in the lab today. I was able to get the output voltage move around for a fixed input (while VCC was varied) and other poor output voltage issues.

    I was able to get a good transfer function if I removed the 1nf feedback capacitor or keeping the capacitance on pin 2 to ground low. I believe the photodiode capacitance plus the feedback capacitor is causing an oscillation that creates the moving DC output you saw.

    Can you try removing the 1nf capacitor to see if that stops the issues you are seeing?
  • Hi

    I did that, in fact I tested removing the cap completely and also changing it to 100nF (so see if I can worsen the behavior, also gives info)

    I also connected 2.2nF and 100nF between the inputs.

    I so than 100nF between the inputs without feedback capacitor helps, but not so much than the 4.7K load.

    The point that 4.7K load fix the problem makes me think that is something related to DC biasing.

    I don't know the detailed internal diagram of the LM324, but from the simplified diagram there is a 6uA current source to the input stage and also other current source of 100uA for the output transistors drive.

    Can be that lack of load on those transistors affect the current source , and that the implementation of the current sources causes a reduction of the magnitude of the input stage current source, leading to the problem?

  • Eliezer,

    I ran a few more tests today and I problem saw yesterday was caused by the source meter (current source) and the op amp being an unstable condition if connected directly (with 2 foot cable). At least some of the time, I saw it was the source meter the caused the oscillation. Adding some resistance between the op amp and my cable to source meter fixed this for me.

    I'd like to simulate and lab test closer to your board's conditions. Can you describe the impedance of your current source (photo LED) and describe the impedance of any loads on the output? Is your input signal a static DC current or some time varying signal?

    Adding a 4.7k from output to ground changes the DC output current may making it a higher sourcing current. For the entire input range from zero current to maximum current will engage the Darlington NPN source current driver in the LM324. This driver has a low output impedance.
  • Hi Ronald
    My source is between 1m to 10 cm away from the input in my tests, I did not see any difference in behavior with the distance, however a capacitor of 100nF between the inputs also made some improvement.
    But the Resistor made a better job.
    In the meantime I found a better (more detailed) implementation description showing the collectors of the current sources.
    Therefore I think the increase in the current source of the input stage (caused by the current demand of the load on the output stage) made the difference.
    Don't you think?
  • Eliezer,

    Input common mode at ground is fine. Usage with or without with the feedback capacitor is also fine. I even tried to raise GND to +200mV thus making the noninverting inputs 200mV below GND and the LM324 worked well (for VOUT >200mV because output  VOL is relative to GND pin).

    The absolute maximum input voltage is -0.3V per data sheet page 4. If the input current precedes LM324 VCC then input voltage spec can be violated.  A schottky clamp diode should be sufficient to keep input voltage below 300mV. This would require adding a component which requires a new circuit board layout. A quicker fix is to make LM324 VCC power up first or try the LM324A which still has a -0.3V maximum input however it is resists phase inversion at the inverting input better.