0
\$\begingroup\$

I thought I got familiar with noise calculations and RMS nv/rthz conversions, hence I went forward to test my understanding. I am comparing some opamps based on their noise levels. So I wanted peak to peak noise, for an opamp ADA4528 at 0.1 to 10 Hz band, and they quoted the same in datasheet as 99 nVpp. so I wanted to verify this and I went on to their noise density graph at page 18. which shows noise density to be 6nV/rtHz. at 0.1 to 10 Hz and its nice rectangle so I went on and multiplied 6 nV/rtHz with sqrt(9.9 Hz) to get 18nVrms. so to get output noise I multiplied it with gain (100) and everything shattered. :P Where am I going wrong. in datasheet it says 99nvpp at gain of 100 and I get nowhere near it by calculation. Please help.

\$\endgroup\$
1
\$\begingroup\$

I think, what you have overlooked is, that when you basicly multiply the 10Hz bandwidth with the 6nV, you get an rms voltage.

According to Paul Horowitz, experiments have shown, that the peak-peak-voltage is about 6 times the rms-voltage. So when you mulitply the 18nVrms by 6, you get about 99nVpp.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ The x6 does not come from experimentation. It comes from noise having a Gaussian distribution. The Vrms is a single standard deviation. To cover 99.7% of the possible values 6 standard deviations is a good estimate - but occasionally values will fall out of this range as well. \$\endgroup\$ – EasyOhm Apr 26 at 9:18
1
\$\begingroup\$

Here's a link to the ADA4528.

There a nasty noise spike at ~200kHz.. make sure you stay away from that.

So first off you (almost) always see noise refereed back to the input, regardless of the gain. The gain of 100 just tells you the conditions under which the noise was measured. (So if you wanted to confirm this you would measure the noise density at a gain of 100 and then divide the output noise density by 100 to get the noise at the input.)

A typical opamp will have both a flat noise spectrum out above some low corenr frequency and 1/f noise below that corner. You can't quantify 1/f noise with a single number, so the spec sheet usually quotes a total p-p noise in some bandwidth. (Go to page 19 of the spec sheet to see the noise in the 0.1-10 Hz bandwidth. It's about 80 nVp-p)

The ADA4528 does not show any 1/f noise that I can see. Here's an opamp with 1/f noise. See graph on page 4, bottom left hand corner. The 1/f noise corner for this opamp is between 100 and 1kHz.

I hope that helps.

Edit (an addition): From my somewhat limited experience the 1/f noise corner for opamps is not well controlled and there can be order of magnitude variations between devices. The number quoted in the spec sheets is usually a worst case number. The high frequency white noise density, on the other-hand, is usually very consistent part to part. (~10% variation maybe.. this from only a limited selection of opmaps.)

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ "There a nasty noise spike at ~200kHz"-inherent disadvantage of auto zero amps "The ADA4528 does not show any 1/f"-one greatest advantage of auto zero amps. since I am interested in only 15Hz bandwidth, so i only care about 0.1Hz to 10Hz noise. In many data sheets they just give noise density at 0.1 - 10Hz, so that's input noise, and to get o/p noise I need to multiply it with gain.RIGHT? \$\endgroup\$ – Sajid Oct 1 '14 at 17:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.