Strange behavior OPA336 as integrator

Hi Marek,

I think you have misunderstood my discription of the cicruit. The way you have drawn it obviously won't work since there is no bias for the inverting input.

My application is a straight forward classic integrator circuit where the input signal is fed via a resistor to the inverting input. The non-inverting input is connected to ground which should be no problem for the OPA336 as its common mode input voltage range extends to 200mV below the negative supply.

The integrator is part of a larger circuit. In my application the output DC voltage of the integrator is kept steady at about 1.8V, so well within the linear region of the output.

As I mentioned I have tried some other opamps which don't show the oscillation problem and function as expected. The reason I choose the OPA336N is because it has a reasonably low input offset voltage and low supply current but then I ran into this unexpected oscillation problem.

Regards,

Paul

--

Paul,

It would be very helpful for you to provide at least the part of your schematic directly controlling the output of the integrator as it is critical in determining the cause of the problem you see-without it, depending on the polarity of the input offset voltage, the integrator output would end up at one of the rails (see below).

Please make sure that the loading of the output does NOT exceed +/-4mA at 25C since this would result in the trioded output more than 1.5V from the rail (3.3V-1.8V) and thus non-linear operation of the OPA336 output stage - see graph below.

If you need additional assistance, please upload complete schematic by clicking on the Options bottom above.

Hi Marek,

Here's the cicuit including the part that controlls the output of the integrator. I used a different type of MOSFET in the example because I don't have a model for the one I'm actually using. Therefore the gate voltage will settle at a higher value but still within the output capability of the OPA336.

As you will see when running a transient simulation, the output settles at about 2.92V. On the bench the OPA336 will oscillate in this configuration where other similar opamps don't.

Paul

--

Paul,

There is a problem with the MOSFET control circuit below - you picked NMOS type device to close the loop but you use it as if it were PMOS - in NMOS current should flow from Drain to Source but in your application current flow is reversed.  By doing so you seem to turn-on parasitic devices that operate in a very narrow current-bias window - see second schematic with V4=1V).  Even in a symmetrical NMOS device you must pay attention how you connect the body of the MOS transistor - the body must connect to Source and since current in NMOS flows from Drain to Source, Drain must be at the higher potential than Source.

In order to correct the problem, flip the NMOS (or if possible, swap the body connection) so its Source is at lower potential allowing current flow from Drain to Source.  This will allow the circuit to work properly irrespective of the input voltage level or direction of the output current - see below.

Hi Marek,

The MOSFET being reversed is actually intentional. The MOSFET is supposed to work in the third quadrant in it's linear region in this application. The input voltage of the circuit will never be higher then the 100mV depicted in the simulation. In fact in reality it will be even lower, so the parasitic body diode will not become forward biased and the MOSFET will act as a controllable resistor. The 1Volt input signals you used in the simulation will never occur in my application. Also negative input voltages can not occur.

The use of MOSFETs in the third quadrant is not uncommon as it is also frequently done  in synchronouos rectifier topologies.

I have used this circuit many times in earlier prototypes with different opamps without any problems. It's just that the OPA336 seems to behave differently and becomes unstable. When loading its output with a 10k resistor the problem is gone but I have no explanation for that.

EDIT: A 10kohm resistor from the output to Vcc seems to work even better in order to kill the oscillation. It looks like the OPA336 needs some DC load in order to work properly in this configuration.

Regards,

Paul

--

Paul,

I am not familiar with operating MOSFET 'in the third quadrant' and believe that with the very low voltage across parasitic Body to Drain diode such condition may results in a very high loop impedance leading to oscillation you experience; with parasitic diode hardly forward biased, and VDS voltage below 100mV, I am skeptical that MOSFET will act as a true controllable resistor. 

All in all, I would suggest that you try to set up the NMOS in a conventional way, with its body connected to a lower potential (Source), and by keeping VDS<100mV operate it in Ohmic region; this would create a true current controlled resistor by simply changing the magnitude of IS1 current source while keeping MOS in the Linear Region  - see below.

Marek,

Ok, I just reversed the MOSFET in my circuit. So drain and source are now swapped the way you suggested. The circuit still works as its supposed to do,  but without the DC load at the output of the opamp the oscillation is still there. So no difference in that respect.

I also tried one other type of opamp that I had handy; an OPA170. This one also has no problems with oscillation and works just fine.

EDIT:  Third quadrant operation is actually not that much different from 1st quadrant operation (see attached pic). The characteristics in the third quadrant are merely extensions of the first quadrant graphs for small voltages. Only as the voltage reaches values where the body diode becomes conductive things change. 

Regards,

Paul

--

Paul,

I thought you bias your circuit with 12mA load?  Anyway, your circuit will NOT work without any DC load at the output because in order to close the loop around OPA336 IS1 must sink enough current to reliably bias external MOS transistor in addition to 1uA required to satisfy loop equation (100mV/100k) - see below.

According to simulations below, it would be enough to load the output with 1.2uA current to assure proper operation of the circuit BUT the MOSFET model is not what you use and OPA336 macro-model is more than 15 years old and thus may not exactly match the OPA336 performance. For that reason, you must provide at least 2uA load- 1uA for the MOS transistor and 1uA for feedback network - see below.

Attached please find the Tina-TI circuit you may use for your simulations.

Marek,

Sorry if my description of the problem gives rise to confusion.

The 12mA is indeed the bias of the MOSFET drain current. 12mA is just an example. In my application this current can range from a few mA to 100mA. But this is not the problem. With many other types of opamps the cicruit does what it's supposed to do without any oscillation and behaves exactly like the simulation predicts.

But in the case of OPA336 things behave differently. It starts oscillating at approx 77kHz. And this appears to be related to the DC output current of the opamp itself because if I connect a 10k resistor directly to the output of the OPA336 this mitigates the oscillation problem. But this resistor shouldn't be needed because other opamps work fine without it in the exact same circuit.

I just would like to know why this particular opamp (OPA336) shows this oscillation problem where others don't.

Regards,

Paul

--

Paul,

Adding a resistor in series with the output increases capacitive drive of any op amp as it tend to cancel a pole created by the output impedance and load capacitance. Since OPA336 is designed to drive around 300pF in a unity gain (see below), driving a large MOSFET might be just too much without an additional loop compensation. 

For that reason adding a resistor in series with output or CFET across the MOS gain (Gate to Drain) improves the capacitive load drive of OPA336 (see below).

Marek,

As I mentioned in my initial post I tried a 100ohm resistor in series with the gate of the MOSFET exactly as you suggest in your previous post. The capacitive nature of the load was also what I first suspected to cause the problem. But the series resitor didn't make any difference for the oscillation.

The only thing that did make a difference is a 10k resistor parallel at the OPA336 output to provide some DC load (see attachment).

Regards,

Paul

--

Paul,

Your method of using 10k resistor to ground loads the output stage with extra 284uA and works by improving the phase margin of the close-loop at the cost of higher current - higher current in output stage increases bandwidth of the last stage and thus pushes out a high frequency pole to a higher level thus improving the phase margin.  The Riso resistor method (in series with the MOS Gate) does NOT increase the output current but should work by creating a canceling Zero in a close-loop transfer function.  Perhaps you should try to stabilize the circuit by using the 100 ohm gate series resistor BUT the Body connected to Source instead of Drain as you originally have had configured the circuit - see T2 transistor below (second schematic).  You may also try to increase the close-loop gain to 100 (R7=1k instead of 100k), which improves the phase margin at the cost of lower bandwidth - see last schematic below.

The difference between some op amp like TLV2241 or OPA170 being able to drive the MOS transistor directly may simply come down to the phase margin difference between various op amps.