OPA2156: OPAMP Differential Amplifier with output swing

Hi Martin, 

Martin Rupp49 said:

The PWM frequency of the current controller is between 250kHz and 300kHz. The OPAMP has a fairly high slew rate and GBW product of 25MHz, which is results in around 2MHz Bandwith with Gain 12.41

If I understand the above statement, you want to capture the PWM current between 200-300kHz, then you will need to have at least 2.5MHz-3MHz BW to capture the switching signal. 

Please try the following, it may be good enough for the application. 49.9kohm||1pF is approx. 3.2MHz at -3dB, and it seems to be ok to measure up to 300kHz signal (with minimum attenuation and adequate phase margin). In theory, you would need slight higher op amp BW or reduce the gain slightly, say down to 10V/V.  

An op amp can not drive capacitive load directly and it needs to have a resistor in front of it, say 100Ω to 1kΩ, but you should open up the pole in RC filter up to 2MHz range (the simulation is seen the pole for RC filter is up to 1.6MHz). 

8510.OPA2156 Welding 07182025.TSC

Please let me see if the above suggestion will work out. 

Best,

Raymond

Hi Martin, 

Ok, I opened up the thread. You can simply overwrite the thread status and open it up again, since you are the customer. 

Best,

Raymond

Hello Raymond,

I am back with a few more questions regarding OPA2156 and the circuit above.

We use the OPA2156 with supply voltage from 10V to 26V (single supply). Do you see any performance (like bias voltage, offset voltage..) which can impact the output result with a signifcant output error (let's say 10% to 20% output voltage deviation with the voltage range of 10V to 26V )? the datasheet has some indicators, that above 18V or so, some parameters are completely off the charts, but you cannot tell how much off the charts.

Also I notice, that in our application the OPA2156 becomes significantly warm and are also depdendent on the supply voltage of the OPAMP. With 12V I can measure around 34°C, with +24V i can measure 39°C. 
This comes as  a suprise to me, as I would not expect the OPAMP to dissipate that much power. Also the Supply voltage impact is significant to me (5°C from +12V Supply to +24V Supply seems to me like a lot).

We are seeing this issue only with U1000A circuit (U1000B circuit seems not to be affected) which means that it might be related to the fact, that U1000A is dealing with a gain of 12.4 while U1000B deals with a attenuation of 2.5. Either way, it is one single chip, so supply voltage must affect both OPAMP's which it clearly does not.

If you feel like this OPAMP is the wrong choice, feel free to suggest a different product.

Thanks
Martin

Hi Martin, 

Martin Rupp49 said:

(let's say 10% to 20% output voltage deviation with the voltage range of 10V to 26V )?

An op amp's output errors can be from various sources, such as ripple voltage of the supply rail (e.g. switching power supply), poorly matched resistor values from the difference amplifier, common mode noise, op amp's gain etc.. 

Martin Rupp49 said:

Also I notice, that in our application the OPA2156 becomes significantly warm and are also depdendent on the supply voltage of the OPAMP. With 12V I can measure around 34°C, with +24V i can measure 39°C. 

I do not know what is OPA2156's load. When Vcc rail is increased, the power dissipation across the op amp's output stage will increase. The power dissipation in power is approx. (Vcc - Vout)*I_out + Vcc* Iq current. So increasing in op amp's temperature is not surprising. 

For instance, the quiescent power will increase 24V*5.2mA = 0.125W, which will double vs. 12Vdc. Operating at 39C is a normal and it does not consider high temperature. 

Martin Rupp49 said:

We are seeing this issue only with U1000A circuit (U1000B circuit seems not to be affected) which means that it might be related to the fact, that U1000A is dealing with a gain of 12.4 while U1000B deals with a attenuation of 2.5.

The gain is one of the factor, if you are talking about the higher noise at the op amp's output. What was the resistor matching? You should consider 0.1% or better resistor tolerance for the application, if output noise is a concern.  My guess is that op amp's choice is ok, but you may have certain issues with the difference amplifier's configuration. 

If you have other questions, please let me know. 

Best,

Raymond

Hello Raymond,

Thanks for your reply. The Resistor must all be 0.1%, that is set by design and the vendor has verified that.

The Supply rail indeed has ripple (because each time the MOSFET turns on, the supply voltage for the OPAMP drops).. I tried for ruling that out supplying the OPAMP with an External power supply. I could not see much difference in output variation against the supply voltage.

So that is the main issue we are still having, the output result (the amplfied analog Voltage from the current measurement varies by 10 - 20% from 10V to 26V supply voltage with the same set current demanded from the current controller). It is actually on some boards we see that the current measurement output from the OPAMP increases by 20%, sometimes it decreases by 20%. It is, if component tolerances decided if the OPAMP output voltage goes up or down with the same set current.

The load for the OPA2156 is the ADC from a microcontroller. So what you see in the schematic above is mainly what there is, except for the Microcontroller ADC input pin.

As for the Output load: Since R1022 (10k) is not assembled, there is no real load for the OPAMP, except there is on the feedback circuit. Considering the installed resistors (which we reduced by factor 10 with respect to the above Schematic), according to my calculation there should be power dissipation in the range of 10mW which should not lead to 39°C case temperature. But ok, i agree, 39°C is still ok, but the thermal imager shows this area as one of the hotspots, which is still somewhat unclear to me.

To be clear, noise is not the issue here. Noise is in fact quite ok. It is as mentioned above, that we see a significant change in the OPMAP's output voltage, when we sweep the power supply from 10V to 26V, while the current controller is active. We do know, that the current controller also has issues in maintaining a constant current throughout the whole supply range and that is ok. The issue really is, that we sometimes see a decrease in the OPAMP's output voltage, sometimes an increase in the OPAMP's output voltage. If it were always an increase, I would not complain, but there must be more, than the eye meets.

As for GND, the GND has been carefully isolated from the current controller power.

R1001 has been changed to 1k, still the same behaviour.

Input CM voltage is ok in my opinion. You can verify yourself by assuming 10V to 26V for VSYS on the above schematic with the voltage dividers before the OPAMP + and -. I think we have enough room to the rail and CM voltage is overall within range for every voltage from 10V to 26V. Do you agree?

Here is the current schematic.

Any suggestions are highly appreciated.

Thanks

Hi Martin, 

I noticed that there is ringing on the waveform. Is it generated by your LC filters?

Let us try a few things. 

1. Place 100 ohm at VSYS_HS rail and form a LPF

2. Add 2.2pF to the marked locations. The input difference amplifiers need to match the input impedance very precisely. Your difference amplifiers' CMMR is no better than 60dB currently. You may consider to consider to limit the op amp's BW by increasing 2.2pF. 499kohm and 2.2pF are approx. 145kHz. so the lower difference amplifier should match the top one, or 159kohm with 7.2pF/6.8pF or 7.3pF capacitor in parallel. 

3. The Rsense may have some parasitic inductance in them. Although parallel these helps a lot, please make sure that individual one has parasitic inductance < 5nH. Say at 300kHz, the impedance is approx. 0.94 mohm,  4X of 50mOhm is approx. 12.5mohm. Even with 1nH parasitic capacitance (sum of Ls), you are talking about 1.9mohm, which is approx. 1.9/(12.5+1.9) = 13% error in the best scenario. So I can see that 20% of Isense variation just from the parasitic inductance. 

4. with fx = 300kHz, your LC filter pole should be at least 1/10 to 1/20 of 300kHz. I do not know what the system is configured for.

Please improve some of circuit above, please let me know, we can fine tune the Isense measurement. It seems that you are doing DCM in current mode. Did you compensate all the pole and zero in the system and the load. 

Best,

Raymond