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.

OPA552: OPA552 0.1Hz - 10Hz noise density

Part Number: OPA552
Other Parts Discussed in Thread: OPA827, OPT101

Hi,

In our datasheet, we do not give out noise density between 0.1Hz to 10kHz. Do you have the related level? Thansk!

Jimmy

  • Well Jimmy, 

    If you have the flatband value and the 1/f corner frequency (where the flatband has increased 1.41X) you can use this equation to get an equivalent flat noise that will integrate to the same power as the actual 1/f shape, 

    The OPA552 shows this plot where we can get 14nV flat and 120Hz 1/f about. 

    Then putting those numbers in with F1 = 0.1Hz and F2 = 10Hz will give an equivalent flat spot noise of 106nV (that will integrate to the same power as the 1/f shape)

    Then taking that times the square root of (10-0.1)Hz will give an integrated noise of 332nVrms. Normally we use a 6X crest factor so this will be a 0.1Hz to 10Hz input noise Vpp =2uVpp. Take that times the noise gain for an output number. 

    I derived that equation long ago and put it into page 7 of this comprehensive app note on noise

    http://www.ti.com/lit/an/sboa066a/sboa066a.pdf

  • hey Jimmy, i went on to test this against a few data sheets that have 10sec noise plots with the 0.1Hz to 10Hz filtering - in any 10sec span, it is not likely the Vpp will exactly hit +/-3sigma (which is what that 6X is doing) but the results were close, but the flow I showed you is of course approximate - dealing with noise afterall

  • Morning Jimmy, 

    So this topic is current for me as it is the topic of my next AudioXpress article. Just to test a little more to make sure I don't have some old error in here, I stepped through the attached using the OPA827 which has some good info on this topic. Looks like the flow works very well matches sim data pretty much exactly. The spec line for 0.1Hz to 10Hz Vpp noise is showing 250nVpp as a typical. I quess if you use that 6X multiplier on the RMS noise as what you should expect in any 10sec test span (with that 0.1Hz to 10Hz high gain active filter bandlimiting the noise) that kind of makes sense. 

    OPA827 low frequency noise analysis testing.docx

  • Happy Easter, Michael!

    I usually take a 6.6x multiplying factor according to a crest factor of 3.3:

    Kai

  • Very nice noise calculations can also be seen in the old Burr-Brown datasheet of OPA101:

    Kai

  • Well Kai, 

    Any number you use is approximate - The OPA827 typ spec looks to be 6X. 

    I ran through a number of examples (for myself) where data sheets showed a specific 10sec (0.1Hz to 10Hz bandlimited) time waveform. Interesting stuff - any particularly sweep is going to have a random maximum Vpp where they rarely exceed a 6X Vpp number - one of the more interesting ones was the ADA4622 example in this app note - the time waveform shown is actually quite lot lower Vpp that calculations would suggest. The other problem I was having was estimating the 1/f corner frequency - a lot of stuff shows asymptotic curves intersecting - I usually just try to find the 1.4X flatband freq. Having measured countless parts in the lab, that 1/f model is only approximate and many parts transition oddly onto that 1/f curve. 

    I think one of the real dangers here is thinking that Vpp in this limited bandwidth is solid number - very very very approximate given the wide variation in part to part 1/f profiles I have seen. 

    https://www.analog.com/media/en/analog-dialogue/volume-51/number-2/articles/understanding-and-eliminating-1-f-noise.pdf

    The OPT101 is mainly a photodiode noise discussion, I did not see this 1/f to Vpp stuff. 

  • Correct, no 1/f to Vpp calculations but integrations over frequence bands which isn't simple either.

    Kai

  • Thanks for your detailed reply!