• State Resolved
  • Locked Locked
  • Replies 7 replies
  • Subscribers 47 subscribers
  • Views 990 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

OPA858: Paralleling opamps

Emmanouil Tsachalidis
Intellectual 850 points

Part Number: OPA858
Other Parts Discussed in Thread: ADS4225, LMH6518, BUF602, OPA820, OPA656

Hi,

Paralleling opamps will decrease noise and will therefore prevent degradation of SNR. But, what is the right way to do it?

I mean, just paralleling pin by pin is one possible way but then i will have to think about the input offset voltage and the feedback and gain resistors.

Would it be a right approach to just use one feedback resistor and one gain resistor for a set of say 4 opamps? I read somewhere that the offset would then be the mean value of all opamps. Is that correct?

The way i see it, it would be more convenient if i just parallelized pin by pin all opamps and then use one Rf and one Rg. What about the decoupling caps? Would i have to use the same number of caps just like in a case of one opamp? Since decoupling caps must be extremely close to the pins there is no way to use one set of de-caps for 4 opamps unless i find a way to place them one on top of the other and so on. Still, it doesn't sound right. One negative off course is the fact that the quiescent current will be multiplied by the number of opamps that will be used and given 20mA for the OPA858 the total for 4 opamps will be 80mA.

I am thinking about using 4 OPA858 in parallel followed by a LMH6518 which will directly drive an ADS4225 through a low pass.

Could you pls comment on my thoughts regarding the paralleling of multiple OPA858?

Thanks in advance

Regards

Manos

  • 0
    Guru 48395 points

    Morning Manos, 

    The only pin you tie together is the V+ input, and I would ;come from a common source signal through 20ohm isolating resistors to V+. The outputs have to be summed together through "Ballasting" resistors to avoid high currents due to offset mismatch, something like 100ohm each to give a 25ohm source to the next stage. this was that scope application wasn't it. To 100MHz I think. 

  • 0 in reply to Michael Steffes
    Intellectual 850 points
    Hi Michael,

    Yes you're right, the 100MHz scope projejct.

    So if i understand well, each opamp will have its' own decoupling caps, Rf & Rg, right?

    What about the input voltage offset adjustment then?

    The Rg instead of ground would have to be driven by a really low impedance and high bandwidth opamp (probably by a OPA858).
    Should i connect one end of all Rg resistors (instead of the ground) to the same output of one single opamp that will set the offset?

    What about the common mode input capacitance? That will be summed as they will be in parallel, right?

    As far as the output resistors are concerned i get it, non-similarities in voltage offset will create different output voltages (especially if gain>0) at each opamp's output and while each opamp will struggle to compensate for its' inputs difference there will be some considerable high currents which are somewhat balanced across the ballast resistors.

    Regards
    Manos
  • 0
    Guru 48395 points
    Hey Manos, might be best to start building TINA ckts to make sure we know what we are talking about - but, if I understand you points

    1. Yes, all the input Vcm capacitances will add in parallel
    2. If you want a output offset adjust, yes - tie all the Rg resistors to a single OPA858 op amp output to get an adjust to the final summed outputs.
  • 0 in reply to Michael Steffes
    Intellectual 850 points
    Hi Michael,

    I have already tried connecting OPA858S in parallel and simulation works pretty well.
    Input noise density (nV/rtHz) went down to 1.38nV from 2.45nV after connecting 4 in parallel.
    I wasn't sure whether the capacitances sum up when paralleling but i have just got my confirmation that they do.

    I placed a 1k resistor between the input source output and the +IN of the opamp and run AC analysis which gave me a BW of 277.8MHz (-3dB) with a gain of 7V/V which after calculating by hand was 1/sqrt(2pi * 1K * 0.6p) = 265.3MHz which is pretty close.

    After paralleling 4 opamps, i placed the same 1k resistor between the input source output and the +IN of the opamp and run AC analysis which gave me a BW of 65MHz (-3dB) with a gain of 7V/V which after calculating by hand was 1/sqrt(2pi * 1K * 2.4p) = 66.3MHz which is pretty close.

    Thank you very much for your time Michael.

    My best regards
    Manos
  • 0 in reply to Emmanouil Tsachalidis
    Guru 48395 points
    Great Manos, I also see now the OPA858 is a gain of 7V/V stable so you really can't use that to drive the combined Rg elements for offset control. For a very wideband unity gain output impedance, look at the BUF602.

    If you need better DC precision there, maybe the unity gain stable OPA820 or OPA656 op amps for that buffer stage.
  • 0 in reply to Michael Steffes
    Guru 48395 points
    Of course that active offset control is another path for input noise to the outputs. tough problem
  • 0 in reply to Michael Steffes
    Intellectual 850 points
    Hi Michael,

    Since the OPA858 is the JFET opamp with the lowest noise i will stay with that (i guess) because due to the very high BW i can have a relatively high gain without limiting BW below my requirements. Regarding the 7V/V it suits my needs as the attenuator (R1*C1==R2*C2) at 1x was configured to have an attenuation of 16.9dB. So using the OPA858 at the min gain of 16.9dB (7V/V) where it is stable i have a unity gain between the probe input and the OPA858 output. Now, the offset control will be driven (by an OPA858) through a really low resistor and then a relatively big capacitor so that the RC pole is at a frequency low enough to minimise noise performance degradation. I will run that scenario in TINA to see how it goes.

    Thanks again Michael
    Manos