Transimpedance amplifier

my approach would start with choosing a good 1st stage then followed by 3 or so second stages with different gains, in my opinion your dynamic range is too high to be handled by one stage, but maybe a PGA will work, have no expirience with that.

how can you be so sure about your currents if I may ask, I am working with TIAs and photodiodes as well, so feel free to share.

And for the rest of your criterias, cant you just find an OpAmp that reaches your specs, as you already know what you want.

Greetings,

Seb

For For your 1st stage, it is the bandwidth and the noise which need to be optimized.

Here is an expression for the traditional TIA stage (photodiode feeding the inverting input of an OPA directly) achievable bandwidth taken from the LMH6629 datasheet:

So, if you need a large bandwidth (f_3dB), you would choose a large GBWP device, low RF, for the total input capacitance (Cin) that you have.

With respect to noise, consider "ini" from the equation below. Figure 62 below shows which term is dominant for the chosen RF (x-axis).

For this particular device (LMH6629), Figure 62 shows the noise is dominated by "it" (thermal noise of RF), for RF < ~2.5kohm because the device noise is low enough to not have an impact. For RF > 2.5kohm, the dominant term for this device would be "in" (device input noise current). The device noise voltage term (en/RF) does not play a role at any x-axis value, which is what you want so that your noise limitation does not come from the device noise performance.

With multi-stage designs, like in your case, you want the best noise performance in the 1st stage. And, you like to have as high a gain as possible in this same 1st stage, that would not limit your bandwidth / speed.

Having said that, I'd chose the highest RF value possible and use a spreadsheet to analyze Equation (10) above for several devices (e.g. OPA657, LMH6629, etc.) to make sure you choose a device that optimizes "Ini" for the chosen RF value. A CMOS / JFET input device (e.g. OPA657) has lower input noise current "in" but may not have as low a noise voltage term (en/RF) as a BJT input device (e.g. LMH6629) may offer.

Here is a good TI tool that allows you to search our entire OPA database for the right GBWP, "in" and "en" noise terms:

Regards,

Hooman

Hello Ken and Seb,

I've put together a spreadsheet of the most promising devices for your review and consideration. This spreadsheet uses your conditions:

• Cin= 80pF
• Required Bandwidth > 10kHz (I've used 100kHz to allow margin)
• Min / Max input current: 1pA / 1uA (Hooman 7/9/15: Your minimum was 10pA and I mistakenly used 1pA instead But the procedure is the same. Sorry).

Here is the procedure I have used:

1.  To maximize SNR (dB) under Min input current, for each device, I had Excel compute the largest RF value that still allows 100kHz of bandwidth. I then eliminated devices with too large of RF (100's of Mohm) that were impractical.

2. Computed the SNR (dB) for each device and sorted the list from the highest to the lowest. I then eliminated devices with negative or low SNR.

3. Computed the TIA output voltage (Vout max) under your maximum input current (1uA). I then eliminated devices which had RF values which caused "Vout max" > 10V.

The final list of device that meet these set of requirements is highlighted below (e.g. OPA365 through LME49721) and in the Excel attachment:

/cfs-file/__key/communityserver-discussions-components-files/14/7607.TIA-solution-E2E-Hooman-7_5F00_9_5F00_15.xlsx

Regards,

Hooman