Products > Test Equipment
Siglent SDG1032X sine distortion at 1 kHz ?
<< < (4/11) > >>
rf-loop:

--- Quote from: Martin72 on December 25, 2022, 11:09:54 pm ---
--- Quote from: rf-loop on December 24, 2022, 02:01:01 pm ---I'm lazy and do not want calculate total from harmonics. ;)

--- End quote ---

Hopefully the siglent will do it for us someday. ;)

What your measure concerns:
Using flat top makes sense to me, as it got the best amplitude accuracy.
Also using averaging.
I see the values displayed in the screen upper right..
To be honest, in all the years I´ve used fft as it comes, adjusted as long as the output seems plausible to me, without thinking about it.
Something I want to change, so I´ve read the application notes from lecroy about setting up the fft.
Span = half of samplerate, OK got it.
"Delta"-f = 1/time per division *10 divisions, OK got it...
But there must be more when I look at your screenshot.
Your timebase is set to 5ms, so delta-f should be 1/50ms = 20Hz, but it isn´t, it is 23.84Hz....Why ?
Is it because of the amount of fft points, we can choose up to 2M ?
And what about delta-f in general, the lower the better ?

--- End quote ---

"Span = half of samplerate, OK got it."
FFT full span (0Hz to FFT max frequency) = sampling frequency / 2
As you can see in my image displayed span is not at all this FFT full span.  There is displayed 20kHz span just because I have adjusted it for this span. This particular FFT full span is in this case 25MHz   
(but if user know what he is doing he can also look higher frequencies...  If I now feed her 50.001MHz signal I can also see 1kHz but lets leave this story to other time and place...)


"Your timebase is set to 5ms, so delta-f should be 1/50ms = 20Hz.... "
Please try to unlearn this wrong habit...  ;)

Lets look it right way.
"Your timebase is set to 5ms, so delta-f should be 1/50ms = 20Hz, but it isn´t, it is 23.84Hz....Why ?"

Δf = FFT sampling frequency / FFT length
In this case (my picture)  sampling frequency is 50MSa/s (look FFT sampling not time domain sampling)
FFT length is 2097152 pts.

50000000Hz / 2097152 = 23,84185791 Hz.   So we can display informantion Δf = 23.84Hz

"And what about delta-f in general, the lower the better ?"

There is no single simple answer at all. Because it depends.... better for what?
Reducing FFT points give less resolution (higher Δf) and more speed again depending our othewr settings... t/div, memory length, time domain sampling speed (in some cases FFT use same sampling speed and on other cases it use samplerate what is decimated from time domain samplerate what also can be decimated from real ADC samplerate.
So there is many things what affect each others. What is optimal setup for different needs - simply answer is: Lowest Δf is not always best but some times best is as as low Δf as possible. 


But naturally resolution depends what is Δf (aka Bin width) and what is used FFT window function.
If we think normal spectrum analyzer we talk resolution using RBW (And if we dive to bit more deep we need also know filter shape factors)
In spectrum analyzer RBW is defined (mostly) so that it mean filter -3dBc width. So when you want more or less resolution you adjust RBW
In this kind of FFT here, you adjust resolution by changing Δf and also FFT window function and here need understand that this is bit more complex depending what user need in different cases. Here I have used window "FlatTop" because it give best level accuracy because its top is wide and there is minimal if even any level drop between these single FFT points. (starting from 0Hz there is FFT points just using Δf  interval up to max freq what is sampling freq/2)

Roughly rounded it can say that Flat Top window "RBW" width is 3.7*Δf  (and shape factor is roughly 2.5) (-60 dBc width roughly bit over 9*Δf)
Flattop window scallop loss is roughly 0 so it is level accurate also with signals between FFT points (what exist  Δf intervals starting from 0f).


Other windows have different "RBW" width and different shape factor and scalloping loss 

What are optimal FFT settings depends what we are doing. In some case we need speed but not high resolution... in some case we want look some carrier with weak modulation so we need perhaps maximal resolution and perhaps still good level accuracy for measure modulation levels.

Example if look 7.05MHz 0dBm carrier with 0.2% AM modulation there need "all resolution what can get" (SDS2kX HD) for look and try measure this for get somehow acceptable level accuracy for this modulation level. (-60dBc peaks 120Hz distace from carrier).

So there need select 20MHz sampling freq for FFT (full span is then 10MHz) and 2Mpts FFT length. So we get Δf  9.54Hz. Now we can roughly think Flat Top window RBW is <40Hz  (perhaps near 35Hz (-3dBc))Then set center to carrier and some example 1 or 2k span. With FFT FlatTop window it give roughly -60dBc level for modulation peaks. Of course if modulation depth is more, example 10 or 50% it goes very easy. (as we now know Flat Top window "width" at -60dBc is around 85Hz we have enough resolution for measure these 120Hz modulation peaks.)


ETA: oh yes Performa01 answer allready when I have been slowly writing answer multitasking with many other things with family etc. ;)
Martin72:
Thank you both !
gf:

--- Quote from: Martin72 on December 24, 2022, 02:09:43 pm ---Ah, dBm... |O

--- End quote ---

No need to bang the head. The unit (dBV, dBm,...) does not really matter. What matters is the relative amplitude of the harmonics w.r.t. the fundamental, which is usually denoted dBc.

BTW, another thing which does influence the SRDF and harmonic distortion is the utilization of the DAC's and ADC's full scale range. SFDR of a DAC or ADC is usually specified in dBFS units, i.e. relative to a full-scale sine wave. I.e. the specified numbers apply to full-scale sine waves. For lower amplitudes the numbers are correspondingly worse.
2N3055:

--- Quote from: gf on December 26, 2022, 11:30:59 am ---
--- Quote from: Martin72 on December 24, 2022, 02:09:43 pm ---Ah, dBm... |O

--- End quote ---

No need to bang the head. The unit (dBV, dBm,...) does not really matter. What matters is the relative amplitude of the harmonics w.r.t. the fundamental, which is usually denoted dBc.

BTW, another thing which does influence the SRDF and harmonic distortion is the utilization of the DAC's and ADC's full scale range. SFDR of a DAC or ADC is usually specified in dBFS units, i.e. relative to a full-scale sine wave. I.e. the specified numbers apply to full-scale sine waves. For lower amplitudes the numbers are correspondingly worse.

--- End quote ---
Respectfully, it matters because distortion won't be same at different amplitudes. Relative dB will be the same of course...
markone:

--- Quote from: rf-loop on December 24, 2022, 02:01:01 pm ---SDG1000X series sinewave charasteristrics (THD, harmonics, non harmonic spurs) are specified for 0dBm signal level.

Here 1kHz sine out from SDG1000X  and level 0dBm (load 50ohm).
Test using SDS2504X HD
FFT window of course "flat top"
For better visibility time domain average 512 and freq domain average 8. (for reduce random noise level)

I'm lazy and do not want calculate total from harmonics. ;)
Also only 8 markers (latest public FW).




--- End quote ---

Just out of curiosity, what are (and where are specified) DANL and Phase Noise numbers for SDS2504X HD in FFT mode ?
Navigation
Message Index
Next page
Previous page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod