PD Front-End Design Question

[Cigarsnob]

Hi Everyone!

I'm having some issues with understanding the bandwidth reduction of a transimpedance amplifier when adding a second stage non-inverting amplifier. I original posted my question on Reddit: http://www.reddit.com/r/AskElectronics/comments/35rnz4/pd_frontend_design_question/

If any of you can take a look and maybe provide any sight into what's going on.

Thanks for the help

[T3sl4co1l]

Does the text say anything about the GBW or fT of A1 and A2?  Are they assumed equal?

Tim

[Cigarsnob]

Does the text say anything about the GBW or fT of A1 and A2?  Are they assumed equal?

Tim

Hi Tim,

At the moment, I don't see any references to GBW or fT. I would have thought that as long as fT and the gain you're working with at A2 (second stage) provides the proper bandwidth for that stage then the overall bandwidth would be unaffected. I would understand seeing rolloff if the second stage bandwidth was lower than your signal bandwidth.

In the text, all I see is:

"...a second limit to signal bounds the attainable improvement due to the finite bandwidth of the added amplifier...The signal bandwidth limitations of both amplifiers now bound the upper end of the attainable circuit bandwidth...the bandwidth limit of A2 (the second stage amplifier) makes the signal bandwidth and noise curves peak at the same optimum value of Av_0o (optimal gain of stage two). Above Av_0o, this bandwidth limit rolls off both signal and noise, equalizing their bandwidths..."

They also make a reference in chapter 6, so I'll read though that to see if there's any details there.

For our prototype that we're doing in the lab, we're trying to get 50kHz BW out, but if we use a non-inverting amplifier with a gain of 3 for the second stage then the signal bandwidth gets reduced to 15kHz or so. But if the second stage is an inverting amplifier (same gain) then the signal bandwidth is untouched and we get our 50kHz signal bandwidth. Both op amps has an fT in the 65MHz+ range.

[Marco]

They assume a constant gain, not a constant feedback resistor ... with the second amplifier they reduce Rf, the 0.64 factor is an artefact of their obfuscated math (I assume it's 2/pi, but I'm not interested in doing the math).

As for why your bandwidth changes purely based on the load you hang on the transimpedance amplifier, there is no simplified circuit based explanation for it ... either you're doing something wrong or your opamp is behaving strangely (the transimpedance amplifier is loaded differently with the inverting amplifier, it shouldn't make a difference but you could try adding a load resistor to ground with the non inverting amplifier).

[LaserSteve]

You might want to take a look at Phil Hobb's work...

http://www.electrooptical.net/www/frontends/frontends.pdf

Then call or email Dr.  Hobbs..  His website is at  www.electooptical.net
He takes the odd call, rather enjoys it, actually... My bet is that he would settle this for you.

Steve

[Cigarsnob]

You might want to take a look at Phil Hobb's work...

http://www.electrooptical.net/www/frontends/frontends.pdf

Then call or email Dr.  Hobbs..  His website is at  www.electooptical.net
He takes the odd call, rather enjoys it, actually... My bet is that he would settle this for you.

Steve

Hi Steve,

Thank you for your reply. I'm a fan of Dr. Hobbs and own his Building Electro-Optical book. Prior to prototyping, I read the paper you referenced along with his other paper: http://electrooptical.net/www/canceller/shotnoise_opn.pdf and none of them addressed this 0.64 BW reduction. That's why I was surprised when I saw the reduction. At that point, my boss pointed out the 0.64 BW reduction addressed blindly in the Photodiode Amplifiers Op Amp Solution book gives.

I didn't think about contacting Dr. Hobbs, but I will certainly do so after this post. 

[Cigarsnob]

They assume a constant gain, not a constant feedback resistor ... with the second amplifier they reduce Rf, the 0.64 factor is an artefact of their obfuscated math (I assume it's 2/pi, but I'm not interested in doing the math).

As for why your bandwidth changes purely based on the load you hang on the transimpedance amplifier, there is no simplified circuit based explanation for it ... either you're doing something wrong or your opamp is behaving strangely (the transimpedance amplifier is loaded differently with the inverting amplifier, it shouldn't make a difference but you could try adding a load resistor to ground with the non inverting amplifier).

We tested the opamp by its self prior to connecting it to the transimpedance amplifier and the gain and bandwidth is all good. It's not until it's added to the transimpedance amplifier that it gets reduced. Mind you, we also scaled the transimpedance gain resistor down by the gain of the second stage, so the overall transimpedance gain remains the same as if it was just the transimpedance amplifier.

[T3sl4co1l]

"...a second limit to signal bounds the attainable improvement due to the finite bandwidth of the added amplifier...The signal bandwidth limitations of both amplifiers now bound the upper end of the attainable circuit bandwidth...the bandwidth limit of A2 (the second stage amplifier) makes the signal bandwidth and noise curves peak at the same optimum value of Av_0o (optimal gain of stage two). Above Av_0o, this bandwidth limit rolls off both signal and noise, equalizing their bandwidths..."

Holy crap, what an unreadable book.  I'm impressed you've made it this far, I would've given up and Googled for my problem rather than try to find a solution in that morass!

Tim

[Cigarsnob]

"...a second limit to signal bounds the attainable improvement due to the finite bandwidth of the added amplifier...The signal bandwidth limitations of both amplifiers now bound the upper end of the attainable circuit bandwidth...the bandwidth limit of A2 (the second stage amplifier) makes the signal bandwidth and noise curves peak at the same optimum value of Av_0o (optimal gain of stage two). Above Av_0o, this bandwidth limit rolls off both signal and noise, equalizing their bandwidths..."

Holy crap, what an unreadable book.  I'm impressed you've made it this far, I would've given up and Googled for my problem rather than try to find a solution in that morass!

Tim

I tried googling it, but found nothing. Contacted Phil Hobb's on a Google Groups and he couldn't get any reasoning for it. Unless he's assuming something that isn't clear, this shouldn't be an issue. However, we are experiencing it in the lab. I just need to prototype it myself and see what's going on.

[Marco]

I think you're too easily discounting the possibility that you are making a mistake. In my opinion it's the only logical answer for why putting a couple pF instead of a couple kOhm load on the output of the transimpedance amplifier makes the slighest bit of difference ...

Are you driving these opamps too close to the rails maybe?

[Cigarsnob]

I think you're too easily discounting the possibility that you are making a mistake. In my opinion it's the only logical answer for why putting a couple pF instead of a couple kOhm load on the output of the transimpedance amplifier makes the slighest bit of difference ...

Are you driving these opamps too close to the rails maybe?

I'm sure you're right as I'm not doing anything new/special. I no longer have access to the prototype, so I plan on building one up myself to see what's going on.