2

I have an example of a noise source, but the specific device is unimportant to the fundamental aspect of this question. I'm trying to measure it's level with an industry standardised method so that I can share and compare it in a meaningful way.

I can't imagine how to arrive at a singular figure such as Vrms. I used a digital 1 GSa/s oscilloscope with the 20 MHz bandwidth limiter set on. I got:-

   mVrms   Sa/S    Depth
   -----   ----    -----
    155     25 M      6 k
    236    250 k      6 k
    382     25 M    600 k 
    457    250 k    600 k

By virtue of the sampling theorem, I sampled the noise at 12.5 MHz and 125 kHz (trying to stay well away from the 20 MHz limiter). I also varied the memory depth between 6,000 and 600,000 samples. Vrms seems very much related to both settings. I measured a 'range' of ~3x for Vrms (457 mV - 155 mV).

It seemed easier with the old (green screen) analogue scopes. The had infinite horizontal resolution. I seem to be able to 'select' the Vrms I want when using a digital scope. (Is this progress?) A simple RMS value seems to exclude all concepts of crucial sample depth/sample rate. And different scopes have different sample depths/rates.

Q. How do I measure an 'industry standardised' noise Vrms that I can share with other people, given that Vrms ~ digital scope settings? What scope settings?

Supplemental:

It seems that you don't. Ti's note How to measure LDO noise suggests that spectrum analysers are used instead. I don't have a spectrum analyser.

I've seen Measuring noise with an oscilloscope, How to measure broadband noise (unanswered) and How to calculate RMS of total noise from digital Oscilloscope (FFT)? , but they don't seem to address the point of noise readings being related to scope settings.

Hope this isn't too much stuff, but these are the relevant screen shots for exactly the same noise source:-

7 8 9 10

Paul Uszak
  • 7,327
  • 5
  • 37
  • 69
  • I'm not understanding your question (Q). Noise is a random thing and if you wait long enough the maximum p-p value will be a volt or 10 volts etc.. The RMS may still only be 1 mV. – Andy aka Oct 12 '20 at 15:40
  • @Andyaka No, you got it spot on. The quoted noise is in Vpp units, as it is on [other](https://www.rigol.eu/Public/Uploads/uploadfile/files/ftp/DP/%E6%89%8B%E5%86%8C/DP800/EN/DP800_DataSheet_EN.pdf) power supply specs. So what 'standard' configuration is used to measure it, given that it changes depending on scope settings (as you say)? – Paul Uszak Oct 12 '20 at 15:56
  • It probably doesn't change that much with scope settings because even if the scope is undersampling (worst case scenario), it can still measure peak values if the "analogue" front end bandwidth is sufficient. Most p-p noise specs give p-p from RMS by assuming certain values of the standard deviation of the noise: https://www.allaboutcircuits.com/uploads/articles/what-an-electronics-engineer-needs-to-know-about-noise-part-ii_Mahdi_AAC_image3.jpg - this picture shows +/- 3 \$\sigma\$. – Andy aka Oct 12 '20 at 16:02
  • 1
    @Andyaka it seems to change a great deal, unless I've made a mistake (v.poss.) – Paul Uszak Oct 14 '20 at 12:05
  • The RMS in those four pictures is lowest 250 mV and highest 300 mV and the highest looks like it is aliasing so might be unreliable leaving 250 mV to 275 mV as the options. Is that not good enough? – Andy aka Oct 14 '20 at 12:33
  • 1
    @Andyaka Sorry Andy, you've misread them. All four are the same source on different scope settings, and they range 155 - 457 mVrms. They're summarised in the text table. Ah! 300 and 250 are the vertical scales in mV/div. – Paul Uszak Oct 15 '20 at 10:40

0 Answers0