Products > Test Equipment
Equipment to measure RC filter attenuation
<< < (5/5)
David Hess:
Take advantage of your DSO's external trigger to synchronously demodulate the signal of interest.  Attach the external trigger or another channel to the signal generator's output and use that as the trigger source.  Now when averaging is used on the DSO, everything except the signal from the function generator, including noise, is removed.
rf-loop:

--- Quote from: Someone on September 15, 2022, 10:43:21 pm ---
--- Quote from: rf-loop on September 15, 2022, 08:25:37 am ---But there is no single simple truth - as nowhere. Now of course it needs to define what we mean by dynamic range. In bottom level we need to define what noise level we accept - what level measurement uncertainty and so on, so on.

One limit is also maximum level what  depends about generator and other things..

If we think in some freq sweep DUT top level in some frequency point is  0dBm. Distance to FRA bottom noise level is over 100dB.
If DUT max out level is 10dBm then distance to FRA noise level is over 110dB... And so on.
If use oscilloscope internal 50ohm it set limit to maximum level. It isa then around 22dBm (because 1V/div max)
If use external 50ohm load and inputs 1M. Max is now 42dBm. Now FRA noise level is over 140dB down from this level.
--- End quote ---

It is very complex with no single truth as you say. Were your example plots using a 20 or 40dBm injection signal? the channels didn't seem to be high enough scale for that. Other effects come to play with increasing the stimulus signal, crosstalk of the scope channels, and leakage/crosstalk in the DUT. So it is not always adding more stimulus amplitude = more dynamic range.

It would be good to show an actual configuration achieving more than 120dB dynamic range, rather than just moving the reference value and promising it might could achieve it.

--- End quote ---


--- Quote ---Were your example plots using a 20 or 40dBm injection signal? the channels didn't seem to be high enough scale for that.
--- End quote ---
In these examples naturally not because my test signal to DUT input was 0dBm for look things down from 0dBm (previously told why).


--- Quote ---Other effects come to play with increasing the stimulus signal, crosstalk of the scope channels, and leakage/crosstalk in the DUT.
--- End quote ---
Now we need separate things what we are talking. I am talking more about FFT performance.  DUT leakage/crosstalk is what it is, I do not handle here DUT characteristics. Let's keep these separate.

Crosstalk inside FRA apparatus aka example inside Siglent some oscilloscope is of course part of FRA performance - characteristics and need take care for avoid traps. Every engineer can think how to arrange test setup and how to reduce these kinds of things and/or think how much these affect result. For high dynamic range needs, one possibility is reduce reference signal level so that it is not as high as DUT in level, just for reduce crosstalk effect.
In Siglent FRA all 4 channels (Reference channel aka "DUT in" all 3 DUT out channels) and have independent and continuous automatic gain control during sweep (every freq step check level) and this AGC range is 500uV/Div to 10V/div (1x level). (except when internal 50ohm then it is reduced to max 1V/div (1x levels))
Naturally example 40dBm level need external 50ohm load (if we are talking about 50 ohm systems)
In many cases FRA reference channel input level can attenuate for avoid cross talk affect inside oscilloscope between channels.
But in multichannel FRA there can be crosstalk between channels inside oscilloscope what may be more challenging.

One tiny example about crosstalk.

This is SDS2kXHD. (naturally whole crosstalk is much more things than just this.)
Channels 2 and 3 do not have signal and inputs also have metal cap (not shorting model) for isolate possible external noises.
These channels display just FRA noise floor in this setup.
Channel 4 is DUT out.  (Reference channel (C1) level is lot less than DUT out max peak)

Now if look CH2 and CH3 there can see crosstalk from Dut out Ch4 when DUT out peaks. If look Ch3 and Ch2 peaks distance to CH4 peak (~84090Hz) it is well over 125dB. Naturally cross talk is highly dependent about frequency. So, user need (know his equiment and) think this specially when do multi channel FRA or if need very high dynamic range so that reference channel cross talk to DUT out channel does not spoil the result.

In this image sweep step is 1Hz and span 500Hz. (For narrow things 1Hz is like 12" wrench on watchmaker table.)
Let's hope Siglent give some day more steps/more resolution. For example, 10kHz span resolution is only 20Hz.

User need have enough knowledge and experience for use these things and very importantly know his instruments performance and features.



--- Quote ---So it is not always adding more stimulus amplitude = more dynamic range.
--- End quote ---
Yes. Indeed, of course.

I think this place/thread is not the best possible place to go deeper and talk about only the performance and features of Siglent FRA, and of course that includes grains and chaff. Good and bad sides - as in everything that man does. It also depends a lot on the individual needs of the user - and the pocketbook - which can affect what inconveniences are tolerated.


Someone:

--- Quote from: rf-loop on September 16, 2022, 08:22:26 am ---One tiny example about crosstalk.
...
This is SDS2kXHD. (naturally whole crosstalk is much more things than just this.)
Channels 2 and 3 do not have signal and inputs also have metal cap (not shorting model) for isolate possible external noises.
--- End quote ---
Thank you for additional examples, I think crosstalk and feedthrough is more realistically measured with the probe/cable connected with the DUT but to/through a short. Higher end equipment is usually using differential measurement to get such large dynamic ranges. I do not think there is real demonstration of 120dB with such easy setup/equipment.
mawyatt:
Out of curiosity we did some Bode Plots of RC and an Active (2N3904) LPF using the Bode feature within the SDS2000X+ and SDG2042X AWG.

1) Screen capture #77 is with 100 ohm series R with 10uF Mylar Shunt C.

2)  #78 is with 10uF electrolytic and using Voltage Offset Feature to keep cap from "seeing" a negative voltage a low frequencies.

3)  #79 is with 10uF Polypropylene (Two 5uF in parallel).

4)  #80 is with 1uF Mylar.

5)  #81 is with 1uF Polypropylene.

6)  #82 is an "active" type LPF with 2N3904 acting to lower output impedance as "seen" by load, shunt C from Base is 0.47uF Mylar, R from Collector to Base is 1K and load is 1K on Emitter, Input is to Collector.

7)  #83 is same as 6 but with an additional 0.1uF Mylar shunt C across load.

Note, used the Bode Function Voltage Offset feature to Bias the Active Filter and supply load current (~5ma), so no need for additional equipment. Very handy feature indeed!!

Best,
nctnico:

--- Quote from: slugrustle on September 15, 2022, 12:46:25 am ---What do you mean by HiZ?  I only see 1MΩ and 50Ω input impedance options on the scope.  One of the filters I was testing had 110kΩ impedance at DC, hence preferring the 10MΩ load of a 10x probe vs. the 1MΩ scope input impedance.

--- End quote ---
A 10M Ohm probe isn't going to do much good either. These have enough capacitive load to seriously affect your readings by adding an extra pole to your filter; you need a different probing solution to accurately measure filters like that at frequencies over 100kHz. Think about incorporating an amplifier in your circuit that buffers the signal into a 50 Ohm output and feed that into an oscilloscope or network analyser.

BTW low frequency network analysers do exist. Like the Anritsu MS4630. Such equipment may also be called audio or dynamic signal analysers (but check the max. frequency; some only go to 100kHz).
Navigation
Message Index
Previous page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod