Products > Test Equipment

Cheap bode plot device...

<< < (5/6) > >>

tautech:

--- Quote from: jasonRF on August 24, 2022, 02:00:12 pm ---Those Siglent results look very nice!   Channel separation on that scope is excellent.  Perhaps unless someone writes their own code to control instruments, I suspect that is the most cost-effective way to get automated Bode plots up to those high frequencies. 

--- End quote ---
mawyatt's Bode plot examples are done with the SDS2000X Plus whose capability follow on from when Siglent introduced Bode plot in the release SDS1004X-E DSO's that unlike SDS2000X Plus models which have an internal AWG the 1000X-E models can pair with any Siglent AWG via LAN or USB or control Siglent's SAG1021I USB powered AWG for the Bode plot stimulus.
https://siglentna.com/product/external-arbitrary-waveform-generator-2/

The 1000X-E bode plot feature is examined in depth here:
https://www.eevblog.com/forum/testgear/siglent-sds1x04x-e-bodeplot-ii-(sfra)-features-and-testing-(coming)/

Warpspeed:
Some excellent suggestions there Jason.

My problem is bulk noise swamping the low amplitude signal, mainly high amplitude switching spikes producing broad band hash.
I will continue on with my tracking filter as its already about half built.

The idea is that I have an  LC oscillator tunable between 100Khz and 140Khz. There is also a 100Khz crystal oscillator using a high quality reference crystal. These two sinusoidal signals go into an analog multiplier which produces a nice clean 0 to 40Khz output after some low pass filtering.

This 0 to 40Khz drives an injection transformer and becomes the testing frequency.

The two noisy recovered signals go into two identical channels. Each consists of another analog multiplier fed with the recovered signal (0 to 40Khz) and the output of the LC oscillator (100Khz to 140 Khz).  That then passes through a crystal bandpass filter at 100Khz which has a -6db bandwidth of about 3Hz and pretty steep attenuation either side.

The filtered 100Khz signal then go into a third analog multiplier again mixed with the LC oscillator frequency to get back down to baseband  0 to40Khz.
Its all fairly simple.  So far I have both the 0 to 40Khz test signal and the 100 to 140Khz carrier frequency generated.
I have breadboarded the filters and up/down frequency conversion, but need to get a proper board made before I can do any serious test and measurement  and get some performance figures.

The bandwidth is so narrow, its not really possible to sweep the frequency, it would take forever to cover any reasonable bandwidth.  But I can turn the knob on my LC oscillator to any frequency of interest and the whole thing settles down in a couple of seconds and I should be able to make pretty noise free measurements.

I may need some switched gain/attenuation ahead of the first analog multiplier, but it may also be workable without.  Its been a bit of work, but not an expensive project. Results so far have been very encouraging.

Someone:

--- Quote from: Warpspeed on August 24, 2022, 08:59:30 am ---Trying to resolve millivolt signals below volts of switching spikes and a very noisy ground is just not practical with either an oscilloscope or something like the HP3575A.
--- End quote ---

--- Quote from: Warpspeed on August 24, 2022, 10:54:09 pm ---My problem is bulk noise swamping the low amplitude signal, mainly high amplitude switching spikes producing broad band hash.
--- End quote ---
It is common/routine to measure sub milivolt phasor signals in switchmode power supplies, to characterise the control loop(s).

Here is a good reference going into some of the dynamic range limitations:
https://fscdn.rohm.com/en/products/databook/applinote/ic/power/switching_regulator/fra_phase_margin_appli-e.pdf
small signals which scopes are able to measure, as jasonRF mentions above you need more data to improve the filtering of the noise. Just as a narrow bandpass filter has a long settling time, the fundamentals apply in analog or digital. Now that cheap scopes have 10M or 100M of sample memory its often easier to use (offline) digital processing rather than analog hardware. There can be multiple frequency selective steps, modern scopes have high-resolution, averaging, and sometimes other filtering available each with their own multiplicative effects through the processing chain.

Warpspeed:
No doubt its possible, but the wide available bandwidth of an oscilloscope is pretty much self defeating when dealing with very broad spectrum high amplitude noise.
There is also the problem of judging exact phase and amplitude difference between two fuzzy waveforms even after a huge amount of sample averaging.

I will continue on with my project just out of sheer curiosity.  It may be the old fashioned way of doing things, but it should still work for me.
Radio receiver technology does pretty much the exact same thing, pick out one very weak signal out of a whole spectrum of chaos.

mawyatt:
@Warpspeed,

Understand the "curiosity" aspect, see note below :)

Since you are using a tracking filter to select the signal of interest, have you considered "Synchronous Sampling" or a "Commutating Filter" techniques. These are easily implemented with standard CMOS and tune with the clock.

Best,

Navigation

[0] Message Index

[#] Next page

[*] Previous page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod