OPA847 oscillation

Hello Liu,

One possibility is that since OPA847 is not unity gain stable, using too large a CF could be the cause for instability. Here is an E2E post with an application note and some notes within which may help you:

http://e2e.ti.com/support/amplifiers/high_speed_amplifiers/f/10/p/293655/1025819.aspx#1025819

In addition, here are some things which in the past have helped me stabilize a fast TIA, specifically when dealing with the LMH6629 in the TIA configuration which had the photodiode placed at a ~3" distant from the inverting input, as shown in the schematic below:

1. A resistor (R6= 100ohm) in series with the inverting input to see if the instability is related to loop phase delay of R1 looking into the "Flex Board" (which is essentially a PC board with two traces on it with the net effect of moving the photodiode (not shown, on the left) further from the amplifier + its parasitics)

2. An RC across the device inputs (I've used 55ohm in series with 1.2pF for R5 and C1)

3. Placing R4 in series with the 1/2 supply decoupling cap (C3) so that the non-inverting input AC resistance is not zero (will be 1kohm instead in this case). I think of this fix as adding base resistance to high frequency BJT for stability

Note that in my particular case, I have no cap across R1 (feedback resistor) as the parasitic capacitance of the 25k resitor (+ board) was enough to stabilize the loop. Your case may be different

Hope these hints help you out.

Regards,

Hooman

Hi ,  Mr Hooman Hashemi

Thanks for you quick answering.

I put my  Tina simulation file here, are you saying that, though simulation shows that OPA847 can get a 3-db bandwiddh @ 100 MHz with 50 p photodiode capacitance and 2.5 p feedback capacitor, in reality, due to the unit-gain unstable, it will make system starts oscillation once the CF larger than 1 p?

7762.847.TSC

Hello again Liu,

I was only suggesting that if CF is raised too high, the following expression may cause the Open loop gain plot intercept with the Noise Gain plot to occur at below the minimum stable gain of the device (12V/V for the OPA847):

From: http://www.ti.com/lit/an/sboa122/sboa122.pdf

With the values you've shown in your TINA-TI file (Cs= 50pF, CF= 2.5pF), your setup shows a value of 21V/V which is above the minimum stable gain of the OPA847:

20.log (1+ Cs/CF)= 20.log (1+ 50pF / 2.5pF)= 26.4dB = 21V/V

You may run into this issue if CF approaches 4.5pF such that the expression above equals 21.6dB (12V/V). So, if your layout / parasitics adds much capacitance to CF, it is conceivable that you reach this critical value (4.6pF) with even a smaller CF (discrete component).

So, with your present conditions (assuming a low parasitic setup / layout) your instability is not caused by the phenomenon I've tried to describe.

Regards,

Hooman

Hello Liu,

Regarding my post on October 3 about trying the non-zero AC resistance to ground to see if stability is achieved:

I've modified your TINA-TI file (with R5 and C4 on the non-inverting input) by adding the resistance R1 (100ohm in this case) to clarify what I mean (your original circuit showed R5=1.2k R1=0ohm which does not serve the purpose of raising the AC resistance):

Regards,

Hooman

thanks, Hooman, I will try to add R1, and give you feedback this afternoon.

Hi,  Hashemi ,

shortly update 

Good thing is adding a R1 in in the non-inverting node do helps.

Bad thing is it does not help a lot.  

Without R1,  i can increase cf to 1p unitil system unstable. with R1,  this value rises to 1.4p.   Still,  I can not reach the calculated unstable value (4p).