OPA858: DC Offset Errors ???

Hi Stewart,

have you checked the circuit with a fresh OPAmp? And have you installed supply voltage decoupling caps close to the OPAmp?

You are right, you should not observe such a high offset voltage at the output of OPA858. Your circuit is stable, I have confirmed this with a phase stability analysis. But: When you connect a multimeter directly to any pin of the OPAmp, so much unwanted stray capacitance and load capacitance may be added that the OPAmp immediately starts to oscillate. With a multimeter you will see an averaged (!) DC value of this oscillation looking like a huge and unexplainable offset voltage. So, do not directly touch any pin of OPAmp with a multimeter or scope probe. Insert an isolation resistor of 50...100R first.

Keep in mind that a digital multimeter (DVM) can have an input capacitance of several nF.

By the way, what is connected to the output of OPAmp? Any capacitive load?

Kai

Hi Kai,

many thanks for the quick reply and for the thoughts.

Just to clarify ( which I probably should have done in my original post... ), I saw the DC offsets using a 500MHz Rigol oscilloscope with 10x probes. The good news is that the opamp was stable throughout ( no oscilation ). So, I'm still a bit puzzled as to why there is such a large offset.

Thanks again for the ideas.

Stewart

Hi Kai,

sorry, I hit the wrong key before completing... The opamp connects directly to the input of a comparator ( no other capacitors ).

I'll give your idea of adding some 'decoupling resistance' to the measurment a go and see if the reduction in load capacitance has any effect.

Again, many thanks for taking the time.

Stewart

Hi Stewart,

another good idea is to replace the OPAmp by a fresh one and to check again.

Kai

Hi Kai,

again, many thanks for the thought and one which I had myself early on in trying to figure out what was happening.

I've experiecnced the issue on several boards with this circuit arrangment and all show the same offset, to the same or lesser degree.

I am still unsure why this is happening.

Thanks again,

Stewart

Hi Stewart,

do you see any oscillation wih the scope at the output of OPAmp? Or is there only this nasty DC offset?

Kai

Hi Kai,

no, just the DC offset - The amplifier is stable and is not oscillating.

Thanks,

Stewart

Hello Stewart,

  Your calculation looks like a good estimate on the expected output offset voltage. What is your input signal, and is this coming from a photodiode? The reason I ask is because the AC coupling capacitor probably will not have a DC path, which might be causing the amplifier to act in a strange manner. For DC blocking TIAs, we usually recommend a DC servo loop/bootstrap or using an integrated TIA with an ambient light cancellation functionality.

Thank you,
Sima 

Hi Sima,

many thanks for the note and for your suggestions.

Yes, there is a photodiode ( TIA ) also using an OPA858 in front of the above stage ( LIDAR application ).

This is a fast application ( ~100MHz ) so the 1nF is not coupling any signal in quiecent conditions and I was expecting R3 ( 50R ) in the above to be providing a DC path to GND. Maybe I was mistaken or have missed something...

I was also coming round to the idea of adding a slow DC servo loop to correct the DC offset, but I'd still really like to understand where the error is coming from.

As to your kind suggestion about an integrated TIA, do you have any recommendations.

As with everyone who's tried to help with this, I really appreciate the time you've given to answering my enquiry.

Thanks again,

Stewart

Hi Stewart,

another idea: Have you soldered the thermal pad of OPA858 to VS- ?

Kai

Hi Kai,

again, many thanks for the thought.

I had to go back and check the PCB layout and yes, I've got a soldered thermal pad under the chip and it's soldered and is connected to pin 5 ( -Ve ).

I forgot to mention in one of my earlier replies, but yes I have decoupling caps ( 100nF // 6u8  ) within about 10mm of the chip on both the +Ve and -Ve supplies ( all of which are connected to power and ground planes under the chip ).

Good thought though...

Thanks again,

Stewart

Hi Stewart,

Stewart Forbes said:

I forgot to mention in one of my earlier replies, but yes I have decoupling caps ( 100nF // 6u8  ) within about 10mm of the chip

I don't want to sound nitpicky or pedantic , but 10mm might be to far away from the OPAmp. 10mm copper trace is about 10nH of unwanted inductance which gives an impedance of about 350 Ohm at 5.5GHz ! This renders the decoupling caps to be nearly useless. The decoupling caps have to be mounted closest to the OPAmp. Every millimeter counts.

What could happen is, that the input stage of OPA858 may oscillate. This might not be visible at the output of OPAmp but can result in an increased supply current and an unwanted and unexplainable DC offset voltage which appears as a huge output offset voltage because of the high gain of your circuit.

A decompensated 5.5GHz OPAmp is ultra critical and does not forgive even the least mistake. All involved copper traces at the output, the inputs and supply voltage pins must be as short as ever possible. Even the least inductance or stray capacitance caused by a too long copper trace can result in weird behaviour.

You might find this thread helpful:

https://e2e.ti.com/support/amplifiers-group/amplifiers/f/amplifiers-forum/1089964/opa855-q1-output-layout-instruction

Kai

Hi Kai,

thanks for the thought ( and as you say, don't worry about being "nit-picky or pedantic", as you're quite right that these devices are 'highly strung' and need treating with care ).

After I saw your note, I went back to the PCB and checked the dimensions and the decoupling capacitors are actually considerably closer than I remembered ( the closest is approx. 2.5mm from the device ).

I’ve got one board to repair today, so will do a little more investigation in the process.

Again, good thought on the decoupling arrangement and certainly worth checking.

Kind Regards,

Stewart

NB. Thanks for the referenced thread as well !

Hello Stewart,

  Nope, you are right! I was looking at this as a TIA rather than a standard voltage feedback amplifier. I switched the inputs of the amplifier, thanks for pointing out the input resistor to ground. 

  What is the input current range? The set non-inverting gain + shunt resistor conversation could be violating the input/output range of the amplifier.

   Our integrated TIAs with ambient light cancellation (DC blocking) would be our LMH34400 (1-channel), LMH32404 (4-channels), and LMH32401 (1-channel). These devices should fit your bandwidth depending on your gain requirements and your input capacitance (photodiode). 

Thank you,
Sima 

Hi Stewart,

if the problem persists, I would change the feedback resistances to values which from the datasheet should work. So I would choose R4 = 75R and R5 = 453R for a test. C3 should be removed for this test. Now the circuit should really work. If not, then there's something else going on. Don't forget the isolation resistance at the output of OPA858 when probing.

If the circuit works with the changed feedback resistors, you can move back to your originally wanted gain in a step-by-step fashion. Play a bit with the values of feedback resistors and feedback capacitance. If the circuit only works with a decreased gain, then you may need to use two OPAmps in a row and split the gain evenly onto these two OPAmps.

Another idea for a test is to leave the feedback resistors untouched and only remove C3, or to decrease C3.

You can additionally short the +input of OPA858 to signal ground (closest to the OPAmp) for the DC test.

By other words, play a bit with the circuit to get a feel for it and to find the cause of trouble.

By the way, is the OPA858 abnormally warming up? What about the supply current? Is is increased? The best way to measure the supply current is to mount a small series resistance in the supply voltage line (close to the decoupling cap) and to measure the voltage drop across it. This small resistance also serves as an enhanced supply voltage decoupling measure by forming a low pass filter. Something between 2R2 nd 10R should work.

Kai

Hi Kai / Sima,

thank you both for the above and sorry that it's taken me a while to get back to you both ( I've been diverted onto another project this past week ).

Taking inspiration from your collective thoughts, I think I now understand the problem.

It would appear that the tiny amount of noise being generated by the front end TIA is passing through the DC blocking capacitor ( C2 in the schematic snip ) and while this should present no DC bias to the non inverting amplifier in question, it does appear to have that effect. Maybe the noise is 'coloured' / asymetric in some way and/or is being 'rectified' in the input stage to the OPA858: It's difficult to know the exact mechnaisum at work.

I think my two possible courses of action are :- 1) look at an alternative to the OPA858 or 2) apply a slow DC correcting servo loop around this stage.

Again, thank both for taking the time to help with this issue: It is VERY much appreciated.

Hi Kai / Sima,

thank you both for the above and sorry that it's taken me a while to get back to you both ( I've been diverted onto another project this past week ).

Taking inspiration from your collective thoughts, I think I now understand the problem.

It would appear that the tiny amount of noise being generated by the front end TIA is passing through the DC blocking capacitor ( C2 in the schematic snip ) and while this should present no DC bias to the non inverting amplifier in question, it does appear to have that effect. Maybe the noise is 'coloured' / asymetric in some way and/or is being 'rectified' in the input stage to the OPA858: It's difficult to know the exact mechnaisum at work.

I think my two possible courses of action are :- 1) look at an alternative to the OPA858 or 2) apply a slow DC correcting servo loop around this stage.

Again, thank both for taking the time to help with this issue: It is VERY much appreciated.

Hi Kai / Sima,

thank you both for the above and sorry that it's taken me a while to get back to you both ( I've been diverted onto another project this past week ).

Taking inspiration from your collective thoughts, I think I now understand the problem.

It would appear that the tiny amount of noise being generated by the front end TIA is passing through the DC blocking capacitor ( C2 in the schematic snip ) and while this should present no DC bias to the non inverting amplifier in question, it does appear to have that effect. Maybe the noise is 'coloured' / asymetric in some way and/or is being 'rectified' in the input stage to the OPA858: It's difficult to know the exact mechnaisum at work.

I think my two possible courses of action are :- 1) look at an alternative to the OPA858 or 2) apply a slow DC correcting servo loop around this stage.

Again, thank both for taking the time to help with this issue: It is VERY much appreciated.

Hi Stewart,

Stewart Forbes said:

apply a slow DC correcting servo loop around this stage.

Hhm, if there's a deeper problem causing this huge DC offset voltage, then such a DC servo wouldn't cure much. You may run into trouble elsewhere.

Stewart Forbes said:

It would appear that the tiny amount of noise being generated by the front end TIA is passing through the DC blocking capacitor ( C2 in the schematic snip ) and while this should present no DC bias to the non inverting amplifier in question, it does appear to have that effect.

This can easily be checked: Disconnect the TIA from the left side of C2. Does this change anything?

In a second step, connect the left side of C2 to signal ground. Does this change anything?

In a third step, unpower the TIA. Does this change anything?

Keep in mind that the TIA may oscillate. Also keep in mind that the TIA and "U2" may see some unwanted coupling via the supply voltages. In this case the supply voltage decoupling measures should be enhanced (-> Pi filters at each OPAmp).

Two another tipps:

1. Replace C2 by a fresh one. It may have been damaged during the testings and may leak.

2. Insert an isolation resistor between the upper side (hot side) of R3 and the +input of "U2". Try different values between 22R and 220R. This isolation resistor not only works as a (more or less mild) low pass filter in combination with the common mode input capacitance of "U2" but also dampens any unwanted LC resonances caused copper track inductances, internal bond wires, etc. In my designs I very often have such an isolation resistor at the +input of HF-OPAmp 

A 50R isolation resistor directly at the output of TIA can also help to dampen unwanted resonances. In this case you may want to increase R3 a bit to not divide down the signal all too much.

Divide and conquer!

Kai