Low Cost PCB's Low Cost Components

Author Topic: Rigol DM3058 DC voltage filter and noise  (Read 465 times)

0 Members and 2 Guests are viewing this topic.

Offline jp430bb

  • Contributor
  • Posts: 20
  • Country: us
Rigol DM3058 DC voltage filter and noise
« on: July 28, 2017, 10:23:26 AM »
The Rigol DM3058 has an option called "Filter" for DC voltage measurement. It can be set to on or off.  I'm measuring a DC signal of order +/-30mV with >1MOhm source impedance and significant AC noise superimposed.  The filter seems to help the meter get reasonable DC readings; without it, the meter is very inaccurate.  The downside is that enabling the filter results in a very slow response time to steps in the signal, like >5s.  I plan to try some external filter caps to cut down on the AC noise, so that I can get by without enabling the DM3058 filter. 

Does anyone have any information on what the DM3058 filter does and how it works?  The manual mentions something about improving the DC voltage accuracy when AC components are present, but it doesn't go into any detail. 

Also, there was a thread on this forum in 2015 discussing RFI being emitted from a DM3058 inputs.  Has anyone else experienced that? 
 

Offline alm

  • Super Contributor
  • ***
  • Posts: 1085
  • Country: 00
Re: Rigol DM3058 DC voltage filter and noise
« Reply #1 on: July 30, 2017, 06:42:22 PM »
From the manual, the filter appears to be (surprisingly) an analog low-pass filter, as was common until the late eighties. Basically the same as the capacitor you are proposing. Both will have a slow rise time with a large source impedance (RC time constant goes up). Lowering the source impedance (e.g. buffering the signal, or measuring at a lower impedance node) would help.

Increasing the integration time (slower measurement speed) may increase normal mode rejection, depending on the frequency. If the problem is that the DC reading is noisy, a moving average function may reduce the noise yet give a better step response that is independent from source impedance. This is how most modern DMMs implement filtering. To further optimize step response, you could implement a +/- X% window around the average: if the new reading is outside this window, reset the average and start from this point. Since I do not believe the Rigol DMM supports this, you would have to do it in post processing (if you record the data on a computer).
 

Offline jp430bb

  • Contributor
  • Posts: 20
  • Country: us
Re: Rigol DM3058 DC voltage filter and noise
« Reply #2 on: August 02, 2017, 11:32:42 AM »
I'm experimenting with buffering the signal with a JFET-input op amp, and that approach looks promising.   :-+
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 3311
  • Country: de
Re: Rigol DM3058 DC voltage filter and noise
« Reply #3 on: August 03, 2017, 02:52:26 AM »
Analog filtering at the input can be a good path, but it makes step response slow. Digital filtering like sliding average (many Keithley meters offer this) is in some respect kind of making you feel the noise is lower than it actually is. So it can be kind of misleading. The two types of filter are also different in what they are good for: the analog filter is effective to suppress higher frequency background, like > 1 kHz, so to improve the suppression of higher line harmonics and similar background. The sliding average is effective against some of the low frequency noise, like < 1...10 Hz. As a third option some meters offer to adjust the integration time - this is more in between. Effective at the low frequency and to a limited degree also for the higher frequency part.

Ideally an integrating ADC offers a reasonable good suppression on high frequency noise. However some real world ADCs can get upset with too much amplitude at higher frequencies. This is why some DMMs offer optional analog filtering before the ADC. It is usually optional as settling can be slow.

With low voltage signals a purpose made amplifier can be sometimes better. However it depends on the source impedance. A JFET OP might be a good option, though offset drift can be an issue here.
 
The following users thanked this post: alm


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf