OPA1656: Little more noise model reconciliation

Michael Steffes

Part Number: OPA1656
Other Parts Discussed in Thread: OPA810

With the TINA model corrected, can also go on to test this 20Hz to 20kHz integrated input noise Vrms in this spec line, this also gives very close agreement to the this number indicating the model probably needs to be updated to 2.9nV flatband and 36nV at 100Hz (or 1.51kHz 1/f corner)

Just putting a sense voltage meter on the input node, gives these two "total noise" curves from 20Hz to 20kHz which is the integrated noise through 20kHz. The corrected model shows 0.504uVrms with the original model having too high 1/f corner (but too low of flatband value) shows about 0.6uVrms

The corrected model yields a 1.51kHz 1/f corner, putting that into this equation from my old noise app note gives an equivalent flat spot noise for integration purposes of 2.9nV*1.23 = 3.57nV, then taking that times root(20kHz-20Hz) gives an integrated Vrms = 0.506uVrms, closely matching the TINA sim number.

This is on page 7 of this app note,

over 3 years ago

0 Michael Steffes over 3 years ago

Guru 47545 points

and to go a bit further, how about the 0.1Hz to 10Hz Vpp number of 1.9uVpp.

That can also be solved using the equivalent flat noise equation which actually does not end up matching up very well - but, this curve sure looks measured and gives that 1.9uVpp exactly? working backwards to a corrected 1/f corner to predict the measured 1.9uVpp, gives a corner at 2.58kHz - then using that for a 10Hz spot noise number gives 46nV, not the 36nV on the front page?? So now I am wondering about the spot noise plot - I think the 2.9nV is right, but those lower F numbers in the data sheet do not match up to the 1.9uVpp 0.1Hz to 10Hz plot? There is some part to part variation on 1/f region and that might explain the mismatch from the spot noise plot to the Vpp plot below - headline specs and plots, maybe should be a little more accurate?

Here I modified the TINA model to give 45nV at 10Hz, that integrated 20Hz to 20kHz result is now 0.55uVrms - actually a little closer to the reported 0.53uVrms in the spec line. I think I will stick with this for now -

flat voltage noise 2.9nV

10Hz noise at 45nV, or 1/f corner at 2.58kHz.

I kind of got going on this working on using the figure 51 for an imbedded buffer design - the audio guys I was talking to (while we all think the OPA1656 is a really good part) did note the low frequency noise was not that hot, really good for a JFET, but not that "ultra -low" kind of thing. Now, it turns out the model is in error and maybe the spot noise plot is a even optimistically low at lower F. Not much, but it seems to be there. I really like the phase lead in the Aol response - that actually is giving me 65deg phase margin in the composite of Figure 51!!!

And one more test against the OPA810 model, another JFET input a little faster, yes the OPA1656 is physically much lower noise over frequency it seems

0 Michael Steffes over 3 years ago in reply to Michael Steffes

Guru 47545 points

And of course, the OPA810 TINA model does not match its datasheet numbers either???? I guess no one checks these things it seems, I wonder what else is not accurate?

Sim data,