OPA140: Gain/Phase datasheet plot not 100% matching simulation

Hi Andrew,

this need not necessarily to be an error. The phase response at the highest frequencies heavily depends on the capacitive load of OPAmp:

andrew_opa140.TSC

Kai

Hi Kai,

Very cool to know about this, appreciate your feedback!

Andrew

Hi Raymond,

I tried to simulate with different C_load but none of them could be 100% fit to datasheet.

Customer is building their database and needs raw data of Gain/Phase in datasheet. 

Customer is expecting we could provide the datasheet original data. Is it possible?

Thx!

Andrew

Hi Andrew,

The OPA140's Aol and phase plots are typical open loop bode plot. I do not understand why you need to fit both plots exactly. 

If you want to fit these both curve exactly, you may digitize the figure using available online software. Once you have the digitized points, you can perform a curve fit to smooth it out. 

Customer is expecting we could provide the datasheet original data. Is it possible?

The original Aol curve is likely simulated by IC designer as well (using IC design tool to obtain the Aol). I sent you the text file a few days ago. What are the issues? It is very close to the Figure 6-15. Aol bode plot among actual ICs may vary up to 20%.

Please let me know what your customer is trying to fit "100%" bode plot curve to a database. . 

Best,

Raymond 

Hi Andrew,

a 100% fit would mean that we are living in a perfect world

The model is very precise as it matches the phase curve up to more than 20MHz very accurately. 20MHz is one octave above the unity gain bandwidth and in this frequency range board parasitics become more and more relevant anyway.

It's a wrong attitude to think that the Spice model is the holy grail, always telling the truth, even if it is as accurate as for the OPA140. And it's wrong to rely on "the last place behind the point" of a simulation. The proper way is to design a circuit with enough headroom and to add components which ensure a proper phase margin. Then you will have so much headroom that you can even allow errors of the simulation.

An example: Assume you know your capacitive load at the output of OPAmp will be no more than 30pF. Then design your circuit by choosing the phase lead capacitance in such a way that it is stable even with 100pF, or even more. The more headroom you include in your design the less accurate the Spice model has to be. (Think about that designers some decades ago (like me ) had no Spice simulation at all. How could we design well working circuits??)

Another issue is that a simulation is only precise, if you are able to accurately add all the parasitics the OPAmp will see on the printed circuit board in the final application. As this can never be the case in a real world, you will need to carry out thorough experimenting and measuring with the real circuit anyway. And again you can check whether your circuit has enough headroom. Following the above example add 100pF load capacitance (or even more) to the output of OPAmp. Is it still running stable?

This is the proper way of using a Spice simulation. Consider it as a helpful tool but not as the holy grail. And don't think a Spice simulation is an excuse for omitting thorough measurings and experimentings with the real circuit. There's no shortcut

Kai

Thanks for comments from both gentlemen.

From a customer's point of view, it's nothing wrong to trust TI datasheet's especially most important one - Gain/Phase plot. I believe, different level of design experience and project budget will determine different design margin. +20% is a good start point for a robust design as Raymond mentioned and educated, thanks for sharing. However, I personally don't know how to link up phase margin of +20% margin, since it's not a straight line, so I think simulation were playing an important role when running over a complex load for a basic design skill person like me. 

I were double checking different part like OPA192, I found both OPA140 and OPA192 phase margin at unity gain seems too good.

OPA140 datasheet PM is 50+ but simulation is about 80. (unity gain, load of 2k ohms, no Cload specifying in datasheet, using 10pF as PCB parasitic load)

OPA192 datasheet PM is 35,  but simulation is about 67. (unity gain, load of 10k ohms, Cload 15pF)

Do you think these 2 famous parts simulation model a bit off than datasheet? 

Thanks.

Andrew

Hi Andrew,

can you show a schematic of the circuit you plan to build?

Kai

Hi Andrew,

It will be good to provide us your schematic as Kai suggested. 

Based on the Open-Loop Gain and phase plots published in the datasheet, phase margins in both parts are very different from the figures you stated. With Gain of 1V/V, the phase margin should not very close to the figure below, assumed the output has the minimum capacitive load.   

Best,

Raymond

Hi Andrew,

[FAE] Andrew Wang said:

OPA140 datasheet PM is 50+ but simulation is about 80. (unity gain, load of 2k ohms, no Cload specifying in datasheet, using 10pF as PCB parasitic load)

OPA192 datasheet PM is 35,  but simulation is about 67. (unity gain, load of 10k ohms, Cload 15pF)

Have you noticed that the scale of the phase is on the right side of the figures? This will give totally different results.

Kai

A good weekend to you too