THS4631: What is the capacitance to "ground" on the Vp and Vn pins of the opamp from the datasheet spec of 3.9pF

Frankly, I would suspect the THS4631 input C spec is incorrect on the Cdiff = 3.9pF - I would suspect it should be those 2-2pF on each input to ground - usually the Cdiff is much lower than Ccm

yes, looking inside the model briefly shows them right at the top with no Cdiff

Thanks for the information, less input cap on the THS4631 than datasheet would be good for my application, but the question on what is the correct interpretation of the specification is still open, I'm inclined to go with the application note; SLOA011A as this actually published by TI to interpret the datasheet.

well what I put in my article is correct, there is quite a bit of variability in specifications. That article kind of goes through fixing the model errors - you should also look at the OPA810. If you don't need high voltage JFET input, there are also plenty of lower voltage higher speed ones like the OPA656. 

For VFA, there are two larger Ccm on each input pin, and a normally smaller Cdiff - 

Here is what Karki put into SLOA011A circa 2001

The drawing is correct, his text on saying the input Ccm is the parallel combination of the two input common mode terms is nonsense - yes you could think of it that way if you run the op amp open loop and drive both inputs together, but no one does - you just need a Ccm on each input to ground as they get into the performance depending on external configuration. For instance just running non-inverting with a high source R will see just the Ccm on the V+ input (the Cdiff is bootstrapped out by the loop gain) and the Ccm on the V- input gets into the response as a feedback pole with the inverting side R's. 

This came up in a Sergio Franco article also, If you are driving the two input together you will see the sum of the two input common mode caps - but what you need for correct modelling is each one seperately on each input. So apparently following this odd little thing Karki was doing, the THS4631 reported the sum of the two Ccm values - that is a bit of a specification error - any of the VFA introductions I have done (in over 150 introductions) has the seperate Ccm specified that needs to be placed in each node of the model. 

If you go to section 16 of Bruce Trumps excellent "The Signal" you will see he is doing it how I do it as well, Yes, there is a lot of confusion on these specs in the datasheets, 

http://www.ti.com/lit/ml/slyt701/slyt701.pdf?&ts=1589912692790

It is also a little odd to see that 2018 date on SLOA011 as I know it was from the very early TI HSP days, yes, a broader search shows it was done in 1998 - prior to the BurrBrown acquisition. Bruce or I would have straightened Karki out had we been involved at the time, 

And to continue this sort of unnecessary confusion, I took a look at the much more recent OPA810 spec line here, 

Those comments are deceiving - these specs are aimed at what should be showing up in the TINA model to get good closed loop simulation matching across a wide range of external configurations, 

So, unless you are modeling an internally fixed gain device like the OPA832, these input C terms are by definition open loop - but what matters is what they do in the closed loop conditions and that depends on your external circuit. For instance, running inverting or transimpedance with the V+ input grounded, you see a single Ccm + Cdiff on the inverting summing junction for analysis purposes - the Ccm on the V+ input is shorted out by ground. So these comments in these datasheet spec lines are unnecessary and misleading, 

Thanks for all the points you have given. There is a lot of confusion in the datasheets.

in this application I need a high voltage opamp so the THS4631 seems a good choice

For many lower voltage applications I use the OPA659, has similar spec format to the THS4631

impedance, differential 10^12 Ω ∥ 1 pF

Input impedance, common-mode  10^12 Ω ∥ 2.5pF

Also used the OPA695, with its equally difficult to interpret input specs.

Noninverting input impedance  280 kΩ || 1.2 pF

Inverting input resistance (Rl ) Open-loop 32 Ω 

The article you mention " section 16 of Bruce Trumps excellent "The Signal" "  this is the same as reference 2  in my original submission, 

Based on all you have said it would appear that the App note SLOA011A 

is likely to be incorrect for common mode capacitance in the data sheet,

and the the Ccm should be interpreted as in your or Bruce Trump  articles.

I think you have it worked out, the OPA695 is a current feedback so it only has a non-inverting input RC then a low output impedance on the inverting input to sense the error current - the input stage is a unity gain buffer from the V+ to V- side. 

Incidentally Ciaran, 

I put together this discussion yesterday as well to canvass this issue more broadly, 

https://e2e.ti.com/support/amplifiers/f/14/t/907063