OPA858: input capacitance calculation

Hi Sonja,

do you want to run a phase stability analysis?

Why do you want to use a T-network in the feedback loop of TIA?

Kai

Hi Kai,

thank you for replying!

Yes, I want to do that.

I am using this T-network to get the really small capacity of 0.492 pF in the feedback loop which I need.

Sonja

Morning Sonja, 

Your capacitance you need for the transimpedance design is just the photodiode + layout+ Ccm +Cdiff on the OPA858, 

Your feedback capacitor is what you solve to get a particular Q in the closed loop response and is not part of your input capacitance as the feedback loop servo's the inverting input to have only error signal on it. 

Typical thick film 0602 and smaller resistors are about 0.18pF parasitic, you can get the rest of that (without the T network) with interlayer overlapping metal under the resistor - or for test with a 2 wire ribbon cabled clipped down to the right length and tacked across the R. Somewhere I have the measured C/inch but something like 0.5pF/inch. 

Transimpedance design is described (with useful simplifications) here. 

0537.Transimpedance design flow using high speed op amps.pptx

Hello Michael,

thank you for you answer!

Are you saying the input capacitance is "photodiode + layout+ Ccm +Cdiff on the OPA858" for the T-network or only for the case without?

I am asking specifically for the calculation of input capacitance when using the T-network.

When running the simulation with only 1 feedback capacitor I am getting the results you are referring to in your work.

However, when using the T-network the results are different. ... So there must be an impact of the grounded part of the T-network (Cf1 and Cf3 in series) or not?

Kind regards,

Sonja

Hi Sonja,

can you post your simulation file?

Kai

Hi Kai,

here are my simulation files.

I plotted the AOL (VF3/VF1) and the Noise_Gain (VF2/VF1) using the post processor.

Tina TI OPA858 dc trans T ss.TSC

Tina TI OPA858 stability T ss.TSC

Kind regards,

Sonja

Thanks Sonja, 

I ran the AC, this looks pretty good with the T network, what is the issue. The available F-3dB for a butterworth should be about 286MHz - getting 200Mhz below, not bad 

Hey Michael,

thank you!

My issue is that I want to write down the calculation of my stability analysis.

The noise gain is calculated (1+Cin/Cf)... for that I need to know  the impact of the T-network on the input capacitance.

If there is only one Cf --> Cin = Cj + Ccm + Cdiff 

How is it calculated when using the T-network?

Kind regards,

Sonja

Well, have fun with that, I am betting Gerry Graeme has done that in his book. 

Calculations are one thing, but here it is easy to just run the LG sim to extract phase margin - looks pretty good at 56deg. 

Here is the file, coming from this article, #6 continues this one

https://www.planetanalog.com/stability-issues-for-high-speed-amplifiers-introductory-background-and-improved-analysis-insight-5/

OPA858 Zt LG.TSC

Hey Michael,

you're right. The simulation should be fine.

Thank you very much,

Sonja

Hi Sonja,

unfortunately, the T-networking will increase the noise:

sonja_opa858.TSC

Kai

Hi Kai,

thank you for all the work you spend looking into my circuit. I really appreciat it!

I already knew that the noise is higher when using the T, thanks.

Sonja

But going back to the calculation of the input capacitance..

Without the T, it is just Cin = Cj + Cdiff + Ccm

With the T, isn't it: Cin = Cj + Cdiff + Ccm + Cx

And Cx is the series connection from Cf1 and Cf3?

When I run the simulation, one time with the T and one time without the T but using 2 series capacities at the inverting input, I get almost exactly the same result.

So, do you think, the calculation can be done like that?

(Tina TI OPA858 stability.TSC)

Thanks,

Sonja

Hi Sonja,

I applied the Kirchhoff's laws and did some math to derive the differential equations. Does this make sense to you?

Kai

You are welcome :-)

Kai

Hi Kai,

looks like I was too hastily..

Can you explain to me, how you calculated the second part of the second equation --> Cf1*((Cf2+Cf3)*Ux' - Cf2*Uout')/(Cf1+Cf2+Cf3)

Thank you in advance,

Sonja

Hi Sonja,

hope you can decipher it :-)

As you can see I throw out UCF3.

Kai

Thank you so much Kai!

Looks like I had the same equations as you, but could't arrange them in the right order to get the result.

Thank you,

Sonja

You are welcome :-)

Kai

Hey Kai, I have tnot been following this - but early on, I think i noted we don't normally think of the feedback C as part of the input capacitance - the op amp loop gain makes the V- nodes look like a virtual ground, 

The better way to think about this is the noise gain equations - the inverse of the voltage divider from the output back to the V- node, that is where the feedback C's get into the response. 

Hey Sonja, 

I had another question on using a T feedback C to get really low values, I had not really looked at that much before, but pretty simple to describe - this one was 50Mohm design needing a 0.04pF feedback C, 

So the 1pF from the output to the T center sees its signal divided by the two paths, most going into the grounded cap. I just chose 1pF values for that one and the one to the inverting input. Then the effective feedback C is 1pF/(1+24pF/1pF) here, or .04pF. You can tune that value with the cap to ground. The response here is identical to what you get with just a 0.04pF feedback C. 

And again, the feedback is not part of the input impedance, the source drives into a virtual ground due to the op amp loop gain. The necessary feedback C is solved using the total C on the inverting node including diode, Ccm + Cdiff and the feedback R and target response peaking. 

Hi Michael,

thank you for having a closer look!

I run the simulation for getting the noise gain (Noise gain is calculated: NG = (1 + Cin/Cf))

1. my circuit with the T network

2. my network without the T but with the "added" capacity of the feedback network to the input

3. the normal circuit with only the feedback capacitor

When calculating the noise gain, I get the same results for 1. and 2., also the simulation is the same. Only for 3. it is different.

So to me it seems that there is an impact of the T-network on the input capacitance.

Or am I completely wrong?

Noise Gain.TSC

The blue line is the noise gain of 3., the green line is the noise gain of 1. and 2. ..completely identical.

Kind regards,

Sonja

Well Sonja, 

As I said earlier, I have not been following this too closely, but I can go back to your earliest attached TINA file to see what you appear to be trying to do and make some suggestions, those are in this attached Word file

Review of Zt design flow for OPA858.docx

And also a word file stepping through some OPA858 model testing I had done some time back, 

6087.OPA858 model testing.docx

Then, the Zt design flow I developed quite some time ago now is in this ppt file, the useful thing here is the really simple resulting equations with some simplificataions in the full equations. 

1602.Transimpedance design flow using high speed op amps.pptx

And then the TINA file for the modified original ckt 

OPA858 Zt with T network.TSC