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: High Speed Substructing Op-Amp followed with comparator to 400mV

Part Number: OPA858
Other Parts Discussed in Thread: LMV861, , THS3217, THS3215, OPA356, LMV7219

i'm trying to find a suitable substracting OpAmp that will allow me to substract two DC signals that represent the output power of a PA, these signals are Vdet &Vref (coming from PA diodes)
i want to substruct between them and to compare via comparator to the value 400mV
the thing is that the PA works in pulses, thus i want that the substructing time will be for e.g. 20nSec while the pulse is about less than 100nSec
i though that i need to look for high slew rate and little setting time but things didn't work for me. i add here few simulation pics

i simulated with LMV861 the result is ok but its taking too much time.

input is 0.8V as Vdet to V+ 0.4V to V- as Vref

below is 

below is OPA858, fast enough but substracting result not really equels to Vdet-Vref , and why i need this 50 ohm to stabalization

input is 600mV to V+ to V- its 400mV

OPA858 not works well

i would be glad if you could recommand me to chose the right parts for this "simple" issue, we will going to make 1000 per year.

  • Morning Arye, 

    Several issues here -

    1. The OPA858 is not unity gain stable so your 100ohm across input takes care of that

    2. The absmax total supply is 5.5V so your +/-3.3V is excessive

    3. The outputs are not RR so you need some headroom to the supplies for correct operation, 

    Here is a start on this, just the inverting side for now where those caps are another way to shape noise gain to get stable operation - this looks good, but the non-inverting path might not look as good, I scaled the Rf up to 200ohm as the 100ohm you had is a heavy load to the output. The non-inverting side gets the noise gain peaking those inverting side caps provide for stability. The total R into the non-inverting side is 200ohms again where that 15pF cap equalizes the peaking on that side, 

    In any case the non-inverting side did have the noise gain peaking the caps cause - Using 2 100ohm on that side I then added a 15pF to ground to equalize that peaking, worked very well it seems - here is a step on the non-inverting side, the rolled off signal at the V+ pin is sharpened back up by the compensation. 

    The other option might be to just use the D2S part of the THS3215 or THS3217 disabling the output stage power amp. That's what they do with a high input impedance - is to subtract two signals from each other. Intended for DAC output interface to get signal ended, might be useful for what you are doing - more power needed and split supplies of course. This is a V9 file, hopefully you can open it of the non-inverting setup above, 

    OPA858 gain of 1 subtractor non-inverting side.TSC

  • thank you very much for your inputs, i have found yestarday some other OpAmp that might be suitable enough, what is your opinoin on my simulation?5224.Pwr_det_n_Comparator_Ok_model_OPA355_V20.TSC

  • Hi Arye,

    Michael has provided some really valuable feedback here and some pretty elegant possible solutions. It seems like we are jumping around a bit in bandwidth and maybe have some confusion about the eventual load. I also see you are looking at part with shutdown/enable. Do you need that functionality as well? Is the comparator a must-have for some reason? (already available unused in dual package etc.) There are lots of options here so if you have any more specifics about the application please let us know.



  • Hi Sam,

    i have also simulated OPA356, i don't need the enable.

    maybe i didn't understand Michael's solution?

    i'm looking for the most tiny (XY) solution that could provide me indication that my PA transmitts high power during pulse.

    for that i though to substract two DC voltages: Vdetector - Vreference (Voltages coming from PA's internal diodes, pulse is very fast)

    When Pulse OFF:  (No Transmition Mode) the DC voltages are: vdetector = 400mV, Vreference = 400mV thus

    Vdetector - Vreference = 0V  < 400mV thus the comparator provides '0' output 

    When Pulse ON: (Transmition Mode) the voltages are: Vdetector = 900mV, Vreference = 400mV thus

    Vdetector - Vreference = 500mV > 400mV thus the comparator provides '1' output



  • Hi,

    I apologize for misunderstanding your question.

    Is the detector reference voltage fixed? 

    If so, it seems like what you are describing is possible with just the comparator. You can account for the detector reference voltage when you set the comparator reference threshold, skipping a step. 

    Here I have used a ramp input to demonstrate the threshold triggering at 400mV and the recovery lag.

    Choosing the right comparator will come down to the duty cycle (how quick the next pulse arrives) to allow for the device to recover. High speed comparators are not my area of expertise but I can forward this question if you would like.

    Am I better understanding your application? Was this helpful?



  • Hi Sam,

    Your solution seems to be a good direction, it might somehow to simplify to only comparator, but the thing is that the substraction between 

    Vdetector-Vreference provides a solution for temperature changes\offsets, using only comparator will require from us to define a constant Vref according to some worst case scenario, itts good direction, i need to check this out. in such case i need to find the tiny possible comparator that can follow input signal of < 50 nSec duration. 

    would be glad to find something smaller than LMV7219

    thank you, sorry for providing not clear enough question (also for Michael)  

  • Hi Ayre,

    what you say is confusing me:

    for that i though to substract two DC voltages: Vdetector - Vreference (Voltages coming from PA's internal diodes, pulse is very fast)

    You aren't having DC voltages but fast pulses, right?


  • The PA Output power is pulsed and thus the Vdet and Vref 2 PA DC out voltages changing due to PA output power pulses 

  • So arye, I had kind of hoped you would put your stimulus sources in front the circuit I attached. Could you post a TINA file for that first circuit you showed so I could pull those sources over to what I had done. 

  • i think i already attached the first simulation file, havn't i?

  • I see pictures but no TINA file, use Insert file

  • I found it, putting those in front of where I had stopped shows this. Yes you have some pretty fast edges here - not sure you need something as fast as the OPA858 but you do need some speed I suspect. The two sides of this circuit looked pretty flat seperately, but that big overshoot is likely coming from phase mismatch - the easiest way to get rid of that is add a postfilter

    Yes, here is a post 80MHz RLC Bessel filter - looks pretty good I think - little insertion loss here but could be recovered in the differencing stage. 

    And this last file, 

    OPA858 gain of 1 subtractor with postfilter.TSC