| Electronics > Metrology |
| DC-accurate Low-pass-filter |
| << < (11/13) > >> |
| Kleinstein:
Today one would ideally implement the 0.1-10 Hz filter to get the standard low frequency noise digitally. So the analog part with amplifier and ADC would have a somewhat larger BW (e.g. 0.01 to 1000 Hz) and the actual limits are set in the digital domain. This could also to an FFT or similar analysis. Even with a full analog filter it would help to have the accurate filter later, after some gain. The analog filters usually have some extra noise in the transition region that can be reduced this way. Normally 0.1 to 10 Hz is in a range where one would not use inductors, but active RC based filters. It is only for ultimate DC precision where an inductor may get interesting again, as such a filter could handle the capacitor leakage better, possibly work with electrolytic capacitors. This would not be for noise measurement, but more ref. filtering (known to be difficult) or a PWM DAC. For the scope preamplifier it depends one the scope that is used. The usual configuartion is to have the higher ranges with a divider in front and one should aim for the highest range that does no use the divider - which range this is depends on the model. The lowest range that uses the divider may not be useful and have relatively high noise again. |
| EC8010:
--- Quote from: Kleinstein on December 30, 2024, 12:39:58 pm ---Even with a full analog filter it would help to have the accurate filter later, after some gain. --- End quote --- That's what I initially thought, but it turned out that if you want to measure the noise of a bench power supply, there are narrow (but high amplitude) spikes of digital clatter at high frequencies that can overload your amplifier, so for practical measurements it's worth having the filter before the amplifier. In the end, it's all about dynamic range and trying to reduce the real world's dynamic range to fit within the very limited dynamic range of the oscilloscope's ADC (usually about 50dB, no matter what claims are made for 16 bit sampling). So although doing the filtering in digits is attractive, it leaves the ADC exposed to a very large bandwidth of noise and therefore possible overload. Of course, if you had a dedicated low frequency digitiser, you might well have much more dynamic range than 50dB and could afford to do the filtering in digits. I've only measured using a Tek MSO54, but I don't imagine there will be a huge difference between oscilloscopes at that level. I measured the 'scope's self-noise, then calculated what external gain was needed to make its noise negligible compared to that of the amplified DUT. For a lot of work, x100 is adequate, but for really quiet noise sources, x1000 is needed. As you point out, you really want to operate the oscilloscope at 50mV/div rather than 1mV/div. The "standard" 0.1Hz - 10Hz filter was done because it was practical and FFT measurements at the time were impractical. But that's no longer the case and we don't need to be constrained by such a crude method. A noise spectrum is far more informative than a single digit measurement via 0.1Hz -10Hz filter. By the way, all of my measurements started as a means of measuring and taming voltage reference noise. |
| coppercone2:
Right I forgot about my little dynamic signal analyzer with 50KHz bandwidth, I was thinking about how much it could effect RMS (I also have that thermal RMS meter). I think that thing could be set to like 2KHz, and it went down to some 1Hz ? That DSA one DOES have some nice filters. You can turn off its AA filter but that makes it do some crazy aliasing on the display . I need a clone I might throw in some other ceramic res into a digikey cart, I only bought 1 per decade for looking at it with the impedance analyzer last year because there was no good information but alot of information on carbon composition.E24 ceramic resistor might be worthy to buy I was visually looking at the impulse response visually and forgot about Q, if you do Q it does seem to dial in on 0.7 at ~417 ohms. But it does still have a "bump" on impulse when the Q is 0.7. I guess there is no right answer. At 510 ohms, there is no bump/ringing, but the Q is 0.57 For your values I got 0.62 Q factor BTW. But I am not taking into consideration the parasitic of the components I do have the jim williams 100KHz active (analog) filter circuit that i built too, but I think that one has a HP of 5Hz built into it with the sallen key filters. HOWEVER I may have added a design change that I can just use the LT1028 gain stage without the filter on a auxillary output BNC, its possible that I might be able to use that box for amplifying this signal. It is in a steel enclosure, but it is hard wired with a 330uF polymer capacitor input. I suppose I should just build another amplifier, its fun. About where to put inductor, EC8010 I got something to stop malfunctioning for a customer before by doing exactly that, kick the inductor out side the circuit. Its like a dog. It gets rid of stuff the actual electronics don't like. Industrial system stopped misbehaving. It was in a larger chemical facility that was having a issue with some data card. I thought its like putting a guard dog outside in the yard instead of having it in the living room. It would probobly be better if they let me have two inductors, aka leave the old one alone and add a new one, but it was a real cheap place, so thankfully I could just move it and it stopped doing whatever bad thing it was doing. I never heard about any measurements, but its for sure a valid strategy that the inductor might serve a better purpose on the input before the amplifier, because the complaint went away, and those people loved to complain. Very fortunate that someone put a inductor in the circuit, because if I could not just move it, then I can't have it on the circuit, because its a new expensive part! I don't know what kind of insanity that could have turned into if I was not fortunate to have that inductor there on the circuit board |O |
| EC8010:
--- Quote from: coppercone2 on December 30, 2024, 05:25:25 pm ---But it does still have a "bump" on impulse when the Q is 0.7. I guess there is no right answer. At 510 ohms, there is no bump/ringing, but the Q is 0.57 --- End quote --- Yes, Q = 0.57 = Bessel response, optimised for impulse. But the LC filter before pre-amplifier/oscilloscope is just there to protect the two from overload. Its impulse response doesn't really matter. What matters (slightly) is the in-band response before the roll-off. But it's easy to measure that and correct for it in the spreadsheet. |
| coppercone2:
Ok I windowed it nice and I guess I can see how from a FFT prospective something around 400 ohm seems the best for flatness but also not cutting off early but also not having a big bump. At the bessel point it looks alot different. |
| Navigation |
| Message Index |
| Next page |
| Previous page |