WEBENCH® Tools/THS4631: Trans Impedance Amplifier noise

Hello Herve,

Sorry for the delay in reply. Would it be possible for you to attach the THS4631 schematic for which you are trying to calculate the total output noise? Also, what is the photo-diode model/part number you are using or it's photo-diode capacitance?

As you may have noted from the book references, the total output noise is primarily dependent upon the noise gain of the op-amp. At higher frequencies, the noise gain is given by NG = 1+(Cg/Cf) which determines bulk of the integrated total output noise. The Cg in the noise gain term consists of the photo diode capacitance as well as the input capacitance (both differential and single-ended cap) of the THS4631; whereas Cf is the feedback capacitance. I believe the WEBENCH calculates the total output noise based on the above noise gain calculation. It would be helpful if you can attach the WEBENCH schematic for further understanding of your problem.

Best Regards,

Rohit

Dear Rohit,

Thanks for your reply. I have attached both design reports (WEBENCH and mine). I hope I made  mistake in my calculations that will explain the discrepancies in noise, specially with laser light on photodiodes.

0412.TI Forum - Help with TIA.pdf2112.WebenchReportsServlet.pdf

Kind regards,

Herve

Hi Herve,

Sorry for missing this post.

I don't think you have made a mistake in your calculations for the total output referred noise. For some reason, the WEBENCH noise numbers seem to be too high for the circuit you are designing. I have asked the WEBENCH team to look into this and they are yet to get back.

I have simulated your circuit in TINA-TI and able to correlate with your calculation. The only difference between my simulation and your calculation is the gain bandwidth product (GBP) where I am simulating ~150MHz of GBP based on Aol curve, whereas in the calculation you have considered 325MHz (fc). Simulating a GBP of 150MHz gets me closer to the total output referred noise of 977uVrms. In the device datasheet, the GBP is mentioned to be around 210MHz which is the fc number I would recommend to use in your calculation.

Now, regarding the issue of measuring 10 times the output referred noise than calculated, are you making there is no other source of parasitic capacitance at the inverting input of THS4631 other than the ones considered in your calculation?

Best Regards,

Rohit

THS4631_TIA_noise.TSC

Figure 1: Aol and Noise Gain (1/Vfb) curves

Figure 2: Total Output Referred Noise

Dear Rohit,

Many Thanks for the reply. The length of the trace connecting the photodiode to the THS4631 inverting input is 10.387mm or 0.409in. Assuming a pcb trace capacitance of 3pF/in, we get a worst case scenario of 1.22pF. I used 3pFparasitic capacitance in my calculations.
I did make sure the area around the inverting input was void of copper pour, on all 4 layers of the pcb.

I spoke to more experienced colleagues and they too could not find anything wrong with the design. They did suggest I measure the spectral noise density of the circuit which I plan to do at some point.

I will let you know the outcome.

By the way I have installed TINA at work. How do you rate the pcb engine of TINA's paid version from Designsoft? is it better than ultiboard (we have to use it at work just because we have a license agreement with NI and the tool is available. It is an average package, not so good when you compare it to the like of Cadence Allegro and Mentor Graphics PADS)

Thanks again Rohit and regards,
Herve

Hi Herve,

I have never used the pcb engine of TINA's paid version from Designsoft myself. So, I would not be the right person to comment on this.
You might find ADS to be a good tool for pcb simulations, although I have used it limited as well.

Best Regards,
Rohit