Siglent SDS1x04X-E BodePlot II (SFRA) features and testings

[rf-loop]

If you comment in this thread, please only things about BodePlot II in SDS1x04X-E models.  There is lot of topics for other things with this scope! And if not, make new.



I will show here some of my tests and observations about Siglent new BodePlot version II for SDS1104X-E and SDS1204X-E.
(BodePlot is also known as FRA or SFRA, Sweep Frequency Response Analysis).

New version (II) is so advanced in this price class that it is best to introduce in separate  thread what handle only BodePlot version II and nothing else.


I will soon but slowly collect here some tests, details etc, depending how I have time and how do my outworn eyes condition allow.

-----------
Previously there have been BodePlot version I.
New Bode Plot II  is first time published with FW 6.1.33
Imho it is really big step to right direction.

There is some new functions and also many other improvements. It is more like totally new design, not like "facelift".

BodePlot II is now an amazingly good and versatile tool especially considering the Price Range.
I have also feel that they will still develop it better later.

Although it is astoundingly good it still needs some finishing and UI fine tuning to improve operating ergonomics. Yet, even if it were not developed any more, it is a good real tool.

It is not just an extra line to the sales brochure. It is a real tool with significant real value. But it need Siglent Signal generator.

BodePLot II have same basic limits as also first generation BodelPlot.

SFRA max frequency range is 10Hz - 120MHz (or less if used Siglent generator max is less).
With its frequency selective detector and automatic Channel Gain system it have well over 110dB max dynamic range.
 
1 channel for phase reference/DUT input level reference. (user can define what channel is for this)
3 channels for DUT outputs with Phase and Amplitude. (user can define channels)

Minimum span is 500Hz and maximum full available frequency span.
In Linear (step) sweep mode it can use 10 - 500 data points for span in use. (with 500Hz min span, 1Hz step)
In Linear step sweep mode user set center and span and steps in this span.

In Decade (log) (step) sweep mode it can set for 2  - up to 275204396* points/decade what can give up to ~500pts for used span. In this mode user set start and stop frequency and points/decade.
*with higher frequencies and narrow span this points/decade value is big. Just select "max" using virtual panel and you have around 500 points with this frequency span.


It can run in two modes.
Channel Gain:Auto mode ( more dynamic range specially over ~100kHz. Slower sweep than with Hold mode. It automatically continuously adjust scope input V/div   setting (using also fine adjust) during sweep, depending DUT output level, for optimal result)
 
Channel Gain:Hold mode (faster but max dynamic range is less, this mode keep channel V/div settings constant (remember first adjust best V/div so that signal do not clip in any point (for manual setup I recommend max 8div p-p for better level accuracy)

Sweep mode can be Simple (constant level) or Variable level. In variable level mode it use user defined table for sweep level profile. Profile can have max 10 data pairs, for freq and level. Between table points do linear interpolation for levels.

It need perhaps BodePlot II  User Reference Manual  (not these simple nonsense User Manuals). Except that after every FW it is obsolete..
There is so many things, what may confuses an inexperienced user, or if experience is from different device and imagines it to work like another. Of course it can be used with a try-oops-try method but many times it leads to frustration and then barking that it works wrong.
 
But, as told, there is also things what Siglent need develop better and also fix some bugs.

In all my tests and comments things are based to SDG1032X/62X using with SDS1x04X-E. (with this I can not do tests up to 120MHz)
SDG FW version 01.01.33R1
and oscilloscope  FW 6.1.33

About dynamic range. There is not clear simple answer. It depends..
If think DUT what do not have  signal amplification answer is bit more simple.

Example if use SDG1062X and whole bandwidth 10Hz to 60MHz.
BP II maximum setting for SDG level is +19.5dBm. (6Vp-p to 50ohm)
But SDG1000X  can not this over 10MHz. (Under 10MHz it can around +24dBm but BodePlot can not set this, max is 19.5dBm)
SDG 1000X maximum over 10MHz is +17.96dBm (5Vp-p to 50ohm)

Now, if with these facts and when DUT can only attenuate 0 or more we can use roughly up tp 110dB dynamic range when over 100kHz. Using 17.9dBm input it is better say that dynamic range is 100dB
(there is some frequency what may have noise level is down to -120dB (ref 17.9dBm as 0dB) but base noise level is not constant and in worst places it is near -100dB, in this test Ref go ti Ch1 and Ch2, 3 and 4 terminated.)

Note: Under 50kHz~100kHz frequencies dynamic range is less and bit weird, later about this.. Also displayed base noise level do not correlate this. These tests take time when use full resolution and I want keep the testing accuracy at a somehow good acceptable level. Also it is more complicated what I initially thought, so I need carefully think what and how I test. Also with full resolution and over many decades one sweep take time when run it in channels gain auto adjust mode.).

Here images
10Hz - 100kHz and 10kHz to 60MHz (ref 17.9dBm ref 0dB)
Signal (DUT in/reference Ch1  17.9dBm, Ch2, 3 and 4 terminated except positions where 0dB attenuated reference signal connected to  channel for check level)





How fast it do these. When all three DUT out channels are on and channel Gain Auto mode and Log sweep and max freq resolution. Both of these BodePlots takes over 20 minutes.

Of course now when channels can adjust gain automatically and if signal level is higher also dynamic range rise. But in turn if need use lower level reference/DUT input signal it also reduce available dynamic range, noise floor stay in its position.
And then if want use fixed channel gain (Channel Gain Hold)  of course then this kind of dynamic range is not available at all. (I will show some tests later but now need find some accessories after accidentally burned some terminators and step attenuator  )

Example if there is selected 10mV/div. Abs max signal level is ~100mVpp  and in practice better stay <80mVpp. Now attenuate it 40dB. What you can see. Just do not use fixed (channel gain Hold mode) if need maximal possible dynamic range. There is noise and there is 8bit ADC. (I do not know what method Siglent use for pick up signal out from noise but it looks like there is some method in use(?)  in BP II.




Here two BodePlot images -- not as I want do it (due to my burned step attenuator)  but also with these it can somehow see.


Channel Gain:Hold mode


In this image channel gain is in Hold mode. I have adjusted levels so100Hz give good visual level on screen and stay in most linear range.
It can see that (of course) around -40dB it goes to troubles. Signal is these so small. Still it can detect phase but with eyes if look normal YT display can not see "nothing" but reference signal and Ch4 around straight line and signal barely visible if know what to look. (If use example over 64 trace average there can see just around one bit (on screen 2 pixel) height rectangle duty depending about small posssible offset)

And then corresponding  YT images from 100Hz and 800kHz


Here is YT image with same setup when freq is start point 100Hz



This YT image is in same setup when freq is 800kHz. As can see - what you can see...
if voltage is divided by 100 it means 40dB drop. Look BodePlot, dynamic range is very limited.
We have 8 bit ADC. Can you use every single bit. No, you can not. This is real world, not from ideal world what do not exist.




Channel Gain:Auto mode


In this image channel gain is in Auto mode. When it sweep and signal drops it also rise sensitivity using fine steps. (as can see in image where it have reached 2.06mV/div when signal is attenuated ~55dB so it can see well.

And then corresponding  YT images from 100Hz and 800kHz


Here YT image how it is in 100Hz sweep start point when BP2 have adjusted levels automatically.



It looks like this when frequency is 800kHz and BP2 have automatically adjusted levels.
During sweep it continuously watch and adjust levels. Here can see level is roughly 4div height p-p and level related to reference can easy measure, visually roughly -55dBref as can see also in bodeplot (and of course phase)

With this method it can reach relatively high max dynamic range with basic 8bit imperfect ADC.

With higher frequencies, for more accurate phase.
You need compensate setup cable delays and channels skew.




note: some image names have error
BP2-ChannelGainAUTO-RC-LP-100k-1MHz
BP2-ChannelGainHOLD-RC-LP-100k-1MHz
100k is wrong and need be 100Hz

[rf-loop]

Here is some kind of partial "menu map" for BodePlot II main functions and settings.
Not including all details and settings.

Also one note for better use.

If you have Siglent signal generator what have two identical output channels.
(example SDG1000X, 2000X and so on)

For better performance with higher frequencies over audio range frequencies up to BodePlot II maximum.
Set generator for zero phase shift and zero delay and zero level shift channel tracking mode. So that Ch2 follow Ch1.

Take Generator Ch1 to BodePlot reference input (in scope menu reference input  is named as DUT input) default is scope Ch1 but you can also select other Ch.)

Take Generator Ch2 to DUT input and DUT output to oscilloscope DUT output channel (default is Ch2).
Also take care about signal pathways delays. They need be equal (cables travel time)  for better Phase accuracy.
Note: If need best phase accuracy then it is best to use same ADC channels for DUTout and DUTin/ref. Ch1/Ch2 pair or Ch3/Ch4 pair.

If you have one channel generator or if you just want use one  generator channel setup or 2ch generator what do not have enough accurate tracking outputs and/or if you work with low frequencies  you can split (even with simple T-split) this generator output channel to DUT input and BodePlot reference input channel ( in scope menu it is named DUT input)
In this case impedance matching is bit more difficult and so on but no problem with very low frequencies.
Naturally if you have real "calibration grade" splitter and set of cables, feed thru terminators etc it can do well but this kind of splitter have many dB power loss.
Of course in simple cases and with high impedance systems you can also use normal probes but as we know 1x probes freq response is what it is and so on and so on.
If BodePlot get garbage in its output is also garbage. So it is good practice first make test run also with null DUT just for your reference and for check cables/probes setup is ok. 

Here BP II simplified menu map.  (1500x2200 png)
Too big size for display in forum but you can download it for full res.




How to connect signal generator, DUT and oscilloscope (two connection principles (picture)).

If you do sensitive measurements (and specially with low frequencies). Take care oscilloscope self-calibration is fresh. Before BP it is best to keep oscilloscope running 30 minutes for thermal full equilibrium. Shut also Quick-Cal off (utility menu). If you use SAG1021 generator run also zero adjust procedure before use BPII.
Also if you do very sensitive tests with high accuracy needs, IF possible, select signal level so that during test it do not cross over voltage bands in your dynamic range needs is not high.
Voltage bands:

Code: [Select]

I     500uV/div - 118mV/div. 
II    120mV/div - 1.18V/div.
III   1.20V/div -   10V/div.
When during sweep  automatic variable gain cross over relay selected voltage band there may exist tiny step error/glitch in amplitude and/or phase.


There is two main methods. Using single channel generator for reference and DUT input or using two separate channel.  Before start BodePlot with two channel generator method user need set generator so that channel 2 is tracking channel 1 (same phase, same freq, same level). After this setting done in SDG then use BodePlot just normally and set "stimulus" level as need. In two channel method image default DUT is 50ohm I/O device.
Note that in BodePlot there is two methods fore measure dut level things. Vout/Vin and Vout. User need think which one is better to use. Two channel method system look Vin  what is now not same as DUT in specially if there is poor impedance matching and reactive DUT input. Also with one channel method user need take care about impedance matching and carefully think possible differences between scope DUTin/reference channel level and real DUTinput level. Low frequencies, audio etc, are not so critical.





As can see  in "Config" menu there is selections what I have marked with numbers 5 and 6.

Usually we use Sweep type "Simple" (5). It is also factory default.
This sweep type use constant user defined generator output level (defined in "Stimulus" menu).

Other Sweep type is named "Vari-level" (6)
User can define how generator output level change during sweep. User can make this frequency-level profile. There can set 4 separate profiles, A, B, C and D. Each profile have 2 - 10 nodes. Every node have frequency and level. First node is sweep start and last node is sweep end.
Between nodes system use lin/log interpolation for level. On the screen line is straight independent of if your selected horizontal scale is logarithmic or linear.

Before user can find all settings and how to make profile it may need some exercise.
Best practice with profile node table is imho. First think what is your sweep end frequency.
Open Edit and set it for 2 nodes. Edit this 2 for wanted end frequency. After then select how many nodes you want. When you turn selection from 2 to 10 it copy this 2 to all nodes from 2 to 10.
Why this. There is one principle. Lower node number can not have higher frequency than next node.
After then you can turn start freq (node 1) how you want and its generator output level level.
Then next node and next and so on. If you find you do not need all 10 nodes then just reduce these last nodes what do not need. Turn adjust knob carefully. If you remove too many then you need edit these again.
In Vari-level mode if you open "stimulus" menu there is not level selection (aka "Stimulus"). (your level selection is in this table)
But there you can select what units you use, Vpp. Vrms, dBm and so on, also system load impedance. If you have real load 50 ohm then use it. If your real load is example 600 ohm then select it. Only when set value  and real value match then measurements and scales can be true.

/example images/


I will update later some more details here.

[rf-loop]

Here  some kind of dynamic range test.
Why 455kHz.
Because this is one very common IF filters frequency range and with these filters need enough dynamic range because stop band is important, least for me - as old "radio man".

Of course also other frequencies are important but as can see in some previous image where is base noise over 10Hz to 60MHz there can not see extremely bad things.
I will make later more deep tests about dynamic range but due to lack of time and also due to fact that these tests takes long time when also want somehow trusted results - tests are now under work. Also it is somehow possible that later some FW change these things, specially what I have found under ~100kHz frequencies.
I think it is not really wise to talk about dynamic range if we do not define this level what is our reference level where from we count dB's. In some Keysight data sheet they use reference level 0dBm.  If use example Siglent BPII max set value  for generator it is 19.5dBm. From this level it have up to >110dB dynamic range. If there also can exist amplification and due to this DUT output is also over its input, then dynamic range can be even more. But then, whole range is not usable over whole 10Hz - 120MHz frequency range.   So it need carefully think how we talk about dynamic range.  Data sheets have very limited information about these things. Even Keysight higher end models data sheet, what I found, about BP are totally poor and leave many many open questions. So I do not know how to tell these so that things are comparable.

It need also note that channels cross talk is quite low (in this price range least good).
Also  mostly with this kind of work need just one channel. And if select reference(Dut in)  Ch1 and DUTout Ch4 then also reference cross talk is quite low to DUT out signal channel (distance to Ch4 os more than defaul DUT out Ch2)
Note: If need best phase accuracy then it is best to use same ADC channels for DUTout and DUTin/ref. Ch1/Ch2 pair or Ch3/Ch4 pair.




1. First test setup. Generator output level 18dBm from both channels. (Gen Ch2 is set for tracking Gen Ch1)




2.G Ch1 to scope Ch1 (50ohm term) as reference, Ch3 no signal, 50ohm termination (base noise) and G Ch2 to scope Ch4 via hp step attenuator set and 50ohm feed thru in scope input.




3. Same as 2. but attenuation changed between 0 to 120dB




4. This is how it looks in normal YT display.  Reference 18dBm, attenuated 110dB (Ch4) and empty terminated channel (Ch3)  (do not care this small ~<100uV internal dcoffset. It is fully nonsense. If internal offset DAC is 16 bit it means roughly 1-2 step)





5. One example about Data table. And with this table I can see all steps from 0dB - 80dB is very close (more like +/-0.1dB).  -90dB and -100dB still something like  +/- 0.3dB or better and then  -110 dB  is clearly less accurate (noise) +/- 1-2dB. (but if estimate average it is still perhaps inside +/- 0.5dB. Of course these are not absolute accurate.


Reference 18dB may drift bit, then step attenuators repeating accuracy, steps tolerances etc (exept that these are really good ones and I do not like use these for this kind of "try this and that-oops" things)

All cables (exept inside attenuator package)  RG223 (better shielding for keep external noise out)

As can see, this BodePlot II  is "bit better" than.....not just only checkbox feature in sales advertisement.



Now we can think some amazing numbers what you see in these pictures.
If you want simple calculator this can use:
https://www.analog.com/en/design-center/interactive-design-tools/dbconvert.html
Note if you are not familiar with these: Vpeak (Vp)  is not Vp-p (Vpp) 1Vp (Vpeak) is 2Vpp (V peak to peak).

Lets take first this 18dBm signal (18dB over 1mW, 0dBm is 1mW).
How much it is if think voltages (system is now 50ohm)

Signal what go to attenuator and is DUTin/ref is of course sinewave. 18dBm  is 2.51Vpeak, 5.02 Vpp
(This is why ChannelGain Auto adjust it around 1V/div with this level)

Now there was 110dB attenuator as DUT. After attenuator there is -92dBm (18 - 110)
How much -92dBm  is in the scope normal YT display if think peak to peak height.  It is  15.9uVpp

And as can see it can go even more low least in some frequency bands if you look this topic first message first two images (noise level) and then in this message (Reply #2)  images 2 and 3.

Siglent most sensitive setting is 500uV/div what is rare in oscilloscopes but now this BodePlot II takes all out from it and even more. If think 500uV/div and if ADC is ideal 8bit its range roughly 5000uV (5120)
it means ADC one step is 20uV. How it can measure 15.9uVpp signal and even less.. 






 
How Siglent did it. I really do not know (If I guess something, perhaps it explain why it is bit slow sweeping, specially when ChannelGain Auto mode is used).
I can only say - astounding.


But next two test images (there is more but I think this is enough for guessing and may give some tip...
(frequency selective receivers what listen only DUTinput/Ref frequency  in DUTout signal channels.)

For this test there is two totally independed and totally unsynched generators. One run normally in BodePlot control and one is separate generator what do not sweep at all.
There is one test with 120kHz carrier with AM modulation. Mod depth 20% and Mod req 10kHz signal and one with just 3MHz 10dBm carrier.

This test nearly proof that they use "freq selective" method what listen only this frequency what is DUT input/ref frequency from BodePlot controlled generator when it step sweeps (something like SA with tracking generator).  This also explain better (one side) about how it can measure so weak signal from DUT. Even when it is barely visible when input is 500uV/div and averaged 1000 times (image 4).
But this can not use for superheterodyne receiver testing/fun-playing between example antenna and IF freq. Because it do not listen other than sweep Generator freq... seriously.. no matter.
But, alone this, imho, can not explain how it goes with 8bit ADC to these small levels.
They must do something more for enhance vertical resolution.
When take all sensitivity and resolution out from BodePlot II  it is slow. Of course if not want all out from this it can set for more fast (set less resolution, Channel Gain Auto off (Hold mode) and so on.)



Special test setup for images 6 and 7. (this is how you DO NOT normally use BodePlot.)

Special test setup for images 6 and 7. (this is how you DO NOT normally use BodePlot.)
Numbers as in image 7 test. Image 6 test 3MHz carrier no modulation.




6.  DUT out signal (Ch4) is not at all from DUT input signal. It is 3MHz 10dBm carrier from separate generator



7.  DUT out signal (Ch4) is not at all from DUT input signal. It is 120kHz AM modulated signalfrom separate generator. Mod depth 20% and freq 10kHz.  There can see these frequency slices what it use.
Of course this is not SA. These images only give information about how this Siglent BodePlot II works. (Note: freq slices size also depend used freq and span and steps etc.)

I like it even if it's not perfect and even when it have some "bad manners". Now I'm not anymore sad after I long time ago sold out my old R&S lady SWOB. This is far better. As long as do not need more than this freq range.

[rf-loop]

More about dynamic range (my generator what I can use at this time for this have max freq 60MHz)


Picture 1
Test setup in next image (dual channel generator setup for 50 ohm system).

Made many tens of test sweeps using different attenuations etc for
produce final estimate about level/dynamic range. Also changed 0-11dB attenuator unit in atytenuator package after all previous tests (I found it was unreliable, some internal contacts not reliable)

It is difficult to talk BodePlotII dynamic range itself because it depends input/reference level and it adjust channel gain automatically. But then there comes also channels cross talk. When reference level high it can also see in channels what measure DUT output(s)

As we know oscilloscope have wide range V/div settings. From 10V/div down to 500uV/div. Think 10Vdiv. If take 63Vpp (6div p-p) signal in 50ohm system it is roughly 40dBm.
Then if go down to 500uV/div and again 6div p-p  it is 3mVpp. Level is now -46.5dBm
But it can easy go down to 63uVpp and now it is around -80dBm. It can go even more low. Freq area where is example typical receivers IF filters it can easy go to -90dBm and it is 20uVpp. It can go  even more down, to 10uV/div,  when ref is 0dBm but in this case there is lot of noise and there is not available average what can delete some amount of noise.


When think dynamic range it also good to remember that ADC is here 8bit.
If we have ideal theoretical 8bit ADC,  what do not exist at all in practice, whole range is 48dB
16bit ideal ADC can give 96dB range.

So, here is one of first preliminary result.


Picture 2
This result is based to set of tests using 0dBm reference signal/DUTin signal and DUT in tests can see in Pic 1.  Channel Gain: "Auto", Display Amplitude mode is "Vout". Sweep 100Hz - 60MHz. 43 points/decade (give 250 data points for whole sweep. (if use max then it give 500 points but it is slower and do not give any more info for this kind of tests.. Also I have all BP sweeps data.csv where from is easy to check measured levels in every point and also it can recall back to scope BPII and all is displayed as in original BP sweep. In stop mode BP can also pan and zoom vertically and horizontally if need.)
In image there is also marked SAG1021 and SDG1000X these max levels what can use under BodePlot II control.
Naturally if your DUT have example amplifier or what ever what rise level over these limits it is possible and BodePlot II Channel Gain Auto leveling take care about it. In this image 40dBm (50ohm system) is just for give some tip. 40dBm is of course 10W (bit over 6.3div p-p  in oscilloscope screen with 10V/div range (1x)). 40dBm to -90dBm (130dB) is quite a lot.


Next here is example what do not push it to dynamic range limits at all but still good example about performance. Many entry level price scopes with BodePlot can not do this, not even close.
But frequency resolution is pushed to its limits. ~1Hz step is bit too rough for this. This down peak is so steep and sharp. Even more if resonator loading is optimal - now it is not.


Picture 3
I hope Siglent improve this data table frequency column resolution, for example Hz.. (example there is well enough room for all digits down to Hz, even when frequency is maximal 120.000000 MHz and even if use dot and kHz or MHz.
It can do it because it can send this frequency command to generator with full resolution, even using under 1milliHz res)
As can see in data table.


Picture 3a
This data is always bacround behind image when you run BPII and look BPII traces, independent of if you open and look this "List" or not. Displayed "List" is reduced resolution from this full data in BPII acquisition memory. This full resolution is what you save when you save it to USB flash. This can recall also back to scope and scope show this old BodePlot just with full details and for further analyze with scope if need. So, if you do some serious work, saving these is even more important and useful than just screen capture.
(BTW, previously max was 501 data points, now there is 500. Previously one step was 1Hz when use 500Hz span. Now step is 500/499 Hz. Why they change this. If want 100 steps it need 101 points...Siglent!!!....






Also it need draw trace even when there is 1Hz wide peak and not reduce it depending its neighbor points or least user selectable smoothing on/off. (I know it do this "noise reduction". It can see if carefully watch example how it draw base noise trace when it is doing sweep with high freq resolution. In some (rare) cases it is nice if this "smooth" can turn off.



Picture 4

BodePlot is based to internal data table. They call it as "List".
If part of draw is out of screen you can shift vertically up and down or change vertical scale scale. All data is there, even if it is out of display. This can do in Display Amplitude menu in BP Run mode or BP stopped. (When it is stopped, also horizontal scale can adjust, even change display between lin and log.)


Attached also this 84kHz bode plot .CSV file. This forum do not accept .csv... shame. 
So if you look it as txt use example notepad++ or just download and remove .txt and open it with software what understand very common  .CSV (comma separated variables) format text  (example LibreOffice calc, OpenOffice calc, Exel or what ever)


Next 14 images (bit boring images. Bit more xplanation for this image set is coming as soon as I have again time
But main idea is (it was tiny part of my test series before I make this preliminary Picture 2.)
In these images A to P  generator output is constant +10dBm  and then there is attenuator between Gen out and scope channel 4 (FFT and BP) (50ohm system).
In BodePlots reference/DUTin (Ch3 selected for this) is constant 10dBm.) 
I forget change BP amplitude scale for dBm but in this case dB scale is same, just think there is also m)
When dB scale show 10dB in mean it measure 10dBm level. (I have set system for measure DUT out level ("Mode Vout") instead of DUTout/DUTin ("Mode Vout/Vin")
After BP with 110dB attenuation 10dB signal I shut BP of and look how FFT see this signal (BP was randomly stopped just with this frequency)  As can see it do not have any difficulties to show this -100dBm signal, with more YT trace average it can go more down of course. (FFT is also lot of improved in this FW 6.1.33 version)
Set begin with 0 attenuation. So there is 10dBm signal. List on and List off. Then every 10dB steps to 110dB attenuation.



Atten 0, level 10dBm, BP List on


Atten 0, level 10dBm


Atten 10dB, level 0dBm


Atten 20dB, level -10dBm


Atten 30dB, level -20dBm


Atten 40dB, level -30dBm


Atten 50dB, level -40dBm


Atten 60dB, level -50dBm


Atten 70dB, level -60dBm


Atten 80dB, level -70dBm


Atten 90dB, level -80dBm


Atten 100dB, level -90dBm


Atten 110dB, level -100dBm


FFT, Same situation as in previous last image. Atten 110dB, Ch4 level -100dBm

Last attachment is one data "list" CSV about 10dBm level BB sweep, (remove .txt)

[rf-loop]

Reserved for continue (because  images, attachments limit in one msg)

[rf-loop]

Reserved fo continue

[rf-loop]

Reserved for future

[rf-loop]

Some corrections and eta (Reply #2)



Updated:
(Reply #3)
preliminary level range ("dynamic range")

Added 84kHz XTAL resonator BP
Added full resolution saved data table partial picture (also downloadable full table as .CSV)

Added just one tiny sample set of images, just for example, about testing level range/dynamic.

I have now some lack of time make furher tests etc. (later more)

If you comment in this thread, please only things about BodePlot II.  There is lot of other topicks for other things with this scope! So we can avoid situation wheree no one can find anything if all is mixed mess. And if not, make new.

[rf-loop]

Here is "fun" comparison.

SSA image is from 3 years old example for show some SSA3000X features, not specially made for testing filter. (not perfect test setup)
SSA TG - Filter - SSA IN
Span 60kHz

Same with SDS1000X-E  BodePLot II
SDG1000X - Filter - SDS1x04X-E BP II
Same 60kHz span.  Roughly same kind of test setup (not perfect setup as also situation what was with SSA)

Note bit different vertical scale. SSA scale is 8dB/div and SDS scale is 10dB/dif also test signal level different.

Only big difference is sweep time. (note that also SSA was bit slow due to settings)
Of course BodePlot II is more slow due to very different process. Scope do freq steps and after every step it control channel gain (necessary due to 8bit ADC) and do frequency selective level measurement and here is maximal 500 measuring points in use.  SSA working principle is different.
BPII is over 20 times slower (with these settings) if set BPII bit differently it can be much faster and still image do not have big difference. In this image resolution is maximal and also channel gain in auto mode.

But as can see Siglent BodePlot II is more than just salesmans checkbox feature in shining advertisement brochure.. "Scrap a toys and buy real tool."

Also this stop band level is not at all limited by BodePlot, it is limited by DUT and setup (where external cross talk over filter (no shielding between in and out etc).


Siglent SSA3000X


SDS1x04X-E BodePLot II
Channel Gain Auto.






Next image is example what happen when Channel Gain Auto is OFF (Channel Gain Hold mode)
Now BP II sweeps much more fast. But as can see roughly around -48dB from ADC full range it fall in noise.
If you are working with more than bit over 45dB dynamic range you need use C G Auto mode.
Note that filter highest peak is adjusted using oscilloscope and SDG so that it is just bit under 8div peak to peak in oscilloscope YT display. (there is around 2 div more but normally we use scope YT based  to max display not max ADC.

If you are working with some simple filters or other DUT what do not need over 45-48dB measurement dynamic you can use also C G Hold mode if you carefully first set suitable DUTout channel V/div what do not lead to signal clip in highest level in BodePlot sweep range what is intention to do.



BodePlot II, Channel Gain Auto OFF (Channel Gain Hold mode selected)
Limited dynamic range.
For this I have first used normal oscilloscope mode with this test setup and using SDG to find filter highest and adjusted channel for 56mV/div. (important is that it is below ADC clip level) It give just 8div p-p sine in filter highest point. After then turned BodePlot II on and selected Channel Gain Hold mode so it do not touch this channel sensitivity during sweep.


So if you need more dynamic range keep Channel Gain Auto (it is also default)

If do not need higher dynamic and if need more fast sweep you can turn Channel Gain to Hold mode.
You need set suitable channel V/div your self in oscilloscope mode before open BodePlot II
Important: If you use Channel Gain Hold mode. You need take care that signals do not clip (ADC maximum is 10.2div p-p) but due to better linearity please keep all scope input signals maximum level 8div peak to peak during whole frequency range in use.

[Performa01]

Frequency range 1 – 120MHz with wideband amplifier.

This test should reveal the dynamic range over the full bandwidth of 120MHz that can be achieved when feeding the DUT with an amplified signal. The DUT is a combination of Inline and step attenuators in this example:


Setup

As can be seen from the block diagram, channel 2 of the DSO has been used. I don’t think there’s a significant difference with regard to crosstalk effects, but we get the advantage of near perfect phase accuracy when both reference and output channel belong to the same ADC.

EDIT: Forgot to metion that 50 ohm through terminators have been attached to Channels 1 and 2 of the SDS1104X-E. Channels 3 and 4 have been left unterminated.

The maximum output from the Marconi TF2175 is about +25dBm. The generator level has been set to -3dB, which results in an amplifier output of +24dBm. This is only 4.5dB more than what could be achieved with the SDG6052X alone (at the max. Bode Plot II setting of +19.5dBm). Consequently, this is not even enough to compensate for the signal loss caused by the splitter, hence we cannot expect a significant increase in dynamic range. Nevertheless, this test should be interesting …

Note that the Step attenuator has a constant insertion loss of 1dB.

First a reference measurement without amplifier and DUT, i.e. just the splitter alone, to make sure that there are no unexpected errors in the measurements:


SDS1104X-E_FRA_Ref_Ch2

We can see that the difference between channels 1 and 2 are negligible, so we get an excellent amplitude accuracy within +/-0.1dB for the entire frequency range from 10Hz up to 120MHz.

Now a reference measurement with the DUT and 0dB attenuation (+1dB insertion loss), which shows the frequency response of the TF2175 in the range from 1MHz to 120MHz:


SDS1104X-E_FRA_TF2175+5080.1_0dB_Ch2

Note that we now see the gain of the amplifier (27dB) minus the insertion loss of the step attenuator. The result should be +26dB, but a small deviation can be expected, because the amplifier gain is not 100% accurate of course.

The output level is almost spot-on, the TF2175 frequency response isn’t perfectly flat, yet easily within +/-0.5dB from 1.5MHz up to 120MHz.

Up to 40dB attenuation, there is no major change in the frequency response:


SDS1104X-E_FRA_TF2175+5080.1_40dB_Ch2

At 60dB attenuation, the level is a bit higher than expected, but considering the attenuator tolerances, which are no metrological devices after all, the result is still amazingly close to the expected level.


SDS1104X-E_FRA_TF2175+5080.1_30+30dB_Ch2

At 80dB, the bode plot isn’t very smooth anymore, but still pretty accurate:


SDS1104X-E_FRA_TF2175+5080.1_30+50dB_Ch2

At 90dB noise becomes an issue, level accuracy finally starts to suffer and the total error reaches +0.8dB, where an unknown part of that has to be attributed to the attenuators and the test setup:


SDS1104X-E_FRA_TF2175+5080.1_50+40dB_Ch2

At 100dB I’ve switched the vertical scale from 1dB/div to 5dB/div. We can see that accuracy is still not too bad, particularly in the range 2MHz to 60MHz. With these settings, the noise of the unterminated channels 3 and 4 is clearly showing as it’s rising with frequency:


SDS1104X-E_FRA_TF2175+5080.1_50+50dB_Ch2

Even at 110dB things don’t look too bad up to some 20MHz. If we can tolerate +/-5dB error, this could still be used up to 40MHz. Above that frequency, the signal starts touching the noise floor:


SDS1104X-E_FRA_TF2175+5080.1_50+60dB_Ch2

Finally we take a look at 120dB, where the result could still be of some use up to at least 2MHz:


SDS1104X-E_FRA_TF2175+5080.1_50+70dB_Ch2

As we can see, the usable dynamic range with +18dBm input level to the DUT can be as much as 120dB below a couple of MHz, but is limited to 100dB if we want to utilize the full frequency range up to 120MHz.

Level accuracy has been excellent down to -60dB and was still very acceptable down to -100dB over the full frequency range. We should keep in mind that we usually need best accuracy when characterizing the passband, where the level is high anyway.

Signal level measurements in the stopband usually need not be that accurate and a couple of dB plus or minus doesn’t really matter – all the more so as component tolerances and test setup can have a much higher influence on the results. At levels that low, double shielded cables and high quality connectors should be used throughout in order to get meaningful results.

120dB dynamic means quite a lot. It is the difference between 1V and 1µV, and also 1W and 1pW. Look at the specification of your cables and make sure that the screening efficiency is sufficiently high (I have used cables specified to >105dB up to 2GHz) and keep input and output cables well separated –particularly don’t run them in parallel at close distance.

[rf-loop]

Signal level measurements in the stopband usually need not be that accurate and a couple of dB plus or minus doesn’t really matter – all the more so as component tolerances and test setup can have a much higher influence on the results. At levels that low, double shielded cables and high quality connectors should be used throughout in order to get meaningful results.

Yes - but then also no. It is true that usually stop band level accuracy is not so important but stop band itself can be  very important or even more important - often overlooked or underestimated. Why we do band pass filters, example IF filters for receivers. Because we want block all but pass band signals and there this is simple - more is better. Filter stop band performance is very important when we talk example good analog HF radio. High end and state of art class analog radio can do only if IF filters stop band is excellent with good filter shape. In some most bottom level radios like example just home radios they can be nearly what ever. But in state of art receivers IF filters stop band performance is critical. How ever nice and excellent and perfect and expensive things there is but if IF filters stop bands are crap it is end of game.

Now we have under 1000 euro entry level oscilloscope SFRA system (BodePlot II) what can do lot of for measure also these things - but still not down to excellent IF filters stop band, not even close if we use "normal" test signal levels (without extra tools and even then - bit difficult). Not at all up to state of art class, not even high end class, not even real professional class (shift right price tag decimal point some steps) but still it leave competitors far behind.

Using more low test signal levels still we knock our nose to noise wall very fast.  We are still working with ~<7bit (oh yes digital data bus width is 8bit) ADC oscilloscope in 500 price  group.  And when thinking this, its  BodePlot II performance is amazing in this and also in bit upper price group.



Here is one image what give some tip how this frequency selective sweep works and how it adjust selectivity depending frequency. It also may give partial explanation about this noise floor levels (and dynamic range changes)

During this BP sweep there is fixed frequency generator what generate lot of harmonics (square wave) to DUTout channel (in this case I have made Ch1 as DUTout channel and Ch4 is ref/DUTin.
Most left narrow peak is 230kHz fundamental.

As can see BPII frequency sweeping selective level detector ("sweeping selective level meter") change its RBW filter width depending frequency when it sweep. (if want more deep inspection it need sweep using many separate frequency bands for get enough freq resolution)
As can see there with fundamental freq, ~230kHz BW is narrow (scale is linear), then it step bit more wide, again bit more, then more and finally after ~3.5MHz*) it have ~300kHz width. If go higher frequencies then width again change. (stepping filter width somehow proportionally with sweep freq).

*) Note reply #3 picture 2  change after 3MHz.

[Performa01]

Inter-Channel Crosstalk up to 120MHz.

I’ve stated before that I don’t think that crosstalk would be significantly different for the various channels. Now I’ve actually tested this and the screenshot below shows the crosstalk between channel 1 and all other channels:


SDS1104X-E_Bode_Crosstalk_+18dBm_Ch1

The reference signal is +18dBm, swept from 1kHz to 120MHz. Channel 1 was fed through a high quality double shielded low loss cable with >105dB screening efficiency up to 2GHz and a HP10100C through terminator (50 ohms). All other channels were terminated by the usual 50 ohms BNC end terminators.

As we can see, there is not much difference indeed, except for the frequency range 60kHz – 900kHz, where channels 3 and 4 are actually significantly better.

We can conclude that for passive structures, the dynamic range is limited by the crosstalk effects as listed below:

-   9.5MHz: 110dB
-   55MHz:  100dB
-   100MHz:  90dB
-   120MHz:  87dB

We can see that there is actually a significant benefit in using an external amplifier for boosting the signal into a passive DUT. In a previous posting I have demonstrated how the combination of +27dB amplifier and step attenuator could provide some 100dB usable dynamic range up to 120MHz. This hints on an overall improvement of at least 10dB.

EDIT: Corrected the reference level from +13dBm to +18dBm.

[rf-loop]

Ch-Ch cross talk depends also what channels are in use. 
Also base noise level depends (some amount) frequency due to selective level measurement bandwidth what is stepping more wide when frequency rise. But still effect is marginal. One BW change is roughly around 3.5MHz and this can see if carefully look. We can perhaps see many steps if there is trace average function but there is not.


Here is some individual examples pick-up from previously made quite big set of tests.

There is 4 equal test pairs  so that CH1 is DUTin and CH2, 3 and 4  DUT out.  There is no DUT. These channels are terminated and sensitivity 500uV/div.
Top image DUTin get quite constant level  from start to stop freq. Level is 17.9dBm
Bottom image all same but DUTin signal attenuated 20dB using external accurate HP step attenuator.
Test generator is SDG1062X.  All cables M17/84 with Suhner's of course. Signal from generator to DUTin channel terminated using Tektronix 011-0049-01 feed thru.

Next image CH2 is DUTin, then CH3 and last CH4 and other channels as DUT out.

I have used Amplitude measurement/display mode Vout (instead of default Vout/Vin) so display scale is also dBm  not dB vs DUTin.
Channel Gain is in Hold mode (instead of Auto)
(this setup is ok for this purpose)

In images can clearly see that different channels have different cross talk.
Around 0dBm DUTin cross talk start mostly fall in to noise.

As can see CH1 is quite good for DUTin channel and it is also BodePlot II default DUTin.
(note, channel cross talk may be different in individual scopes)

If think example typical IF filters around 100kHz or around 450kHz or 1.4MHz  it is good to take into account. Depending of course test setups. Typical filter input level is not so high so  there is not so much dynamic range available due to level of noise floor. But still example if max input level is 0dBm there is still over 80dB range with quite good level accuracy and over 90dB with reduced accuracy.  When DUTin signal level is 0dBm or less it do not rise other channels floor or effect is insignificant. If input level need be more low it is all away from dynamic range, noise floor do not change.

Not bad at all.


Note display scale. It is NOT dB vs DUTin signal. It is abs level in dBm!

[KungFuJosh]

I setup a bode plot using an SDG2042X and the SDS1104X-E. The DUT is a supercardioid microphone. I'm assuming I set this up wrong...just because. The test has been running for over 12 hours and is still going.

The settings are in the screenshots. Control is via network. I'm using a 50ohm terminated coax between the SDG and SDS on ch1. DUT is connected with a 10x Pico TA131 probe unterminated on ch3.

Side note: the frequency response for the mic is supposed to range between 40Hz to 16kHz.

[Rerouter]

Sound correct to me, It never stops, it just sweeps across the range endlessly until you stop it. that way you can use it to adjust circuit parameters and see the result

[OwO]

It should be possible to increase dynamic range by adding a variable gain amplifier before the inputs right? You could pre-characterize the VGA at all gain steps and all frequency points during zeroing, and then during measurement it could auto-range at each frequency point.

[Rerouter]

I was corrected earlier on another thread that it does uses the fine channel gain to extract the most vertical resolution it can from each frequency,

The full VGA gain and offset table is stored on the device, even the steps that do not end up being used, so they may be using them, but that would only net you a marginal improvement. I'm not yet clear on if ERES can be used in combination,

[KungFuJosh]

After screwing around more, I mean experimenting, I realized my amplitude was too high for the DUT (first screenshot). The rest of the screenshoots are me "experimenting" more. At least they look cool, right?

How can I reconcile the 50R terminated load on CH1 with a HiZ load on CH3?

For the bode plot to be accurate, do the waveforms need to match well? I tried offsetting the phase, which was about 120 degrees off, but that was never perfect, and the amplitude never matched. The closest was setting the amplitude to 90mVpp.

Thanks,
Josh

[Performa01]

It should be possible to increase dynamic range by adding a variable gain amplifier before the inputs right? You could pre-characterize the VGA at all gain steps and all frequency points during zeroing, and then during measurement it could auto-range at each frequency point.

As explained in the opening posting, the Siglent Bode Plot offers the AGC feature already. Do you really think that 100 to 120dB dynamic range is not enough?

I'm not yet clear on if ERES can be used in combination,

No. The Bode Plot forces the Acquisition mode to "Normal" - and thankfully so, because the DFT filtering in the Bode Plot detector is far superior to the Eres lowpass filter, which might limit the upper bandwidth in an undesirable way.

How can I reconcile the 50R terminated load on CH1 with a HiZ load on CH3?

This is an oversight in the Bode Plot module. The next version will allow to specify separate load impedances for the reference and output channels.

Always specify the output impedance. In your setup it would be high-Z (as it is represented by the x10 probe) and for frequencies that low you could even connect the reference channel without a 50 ohms inline termination. The result will be correct as long as the reference channel sees the same amplitude as the DUT input.

For the bode plot to be accurate, do the waveforms need to match well? I tried offsetting the phase, which was about 120 degrees off, but that was never perfect, and the amplitude never matched. The closest was setting the amplitude to 90mVpp.

It normally should be just a single signal, fed into both the reference channel of the scope and the DUT input. For frequencies above some 10MHz, proper 50 ohms termination and a resistive power splitter is highly recommended.

You can also use the dual channels of the signal generator instead, but then it has to operate in Channel Coupled mode for amplitude, frequency and phase.

For higher frequencies, matched cable lengths are important to minimize the phase error.

[KungFuJosh]

How can I reconcile the 50R terminated load on CH1 with a HiZ load on CH3?

This is an oversight in the Bode Plot module. The next version will allow to specify separate load impedances for the reference and output channels.

Always specify the output impedance. In your setup it would be high-Z (as it is represented by the x10 probe) and for frequencies that low you could even connect the reference channel without a 50 ohms inline termination. The result will be correct as long as the reference channel sees the same amplitude as the DUT input.

Considering this oversight, it doesn't seem functional at all in my current setup.

For the bode plot to be accurate, do the waveforms need to match well? I tried offsetting the phase, which was about 120 degrees off, but that was never perfect, and the amplitude never matched. The closest was setting the amplitude to 90mVpp.

It normally should be just a single signal, fed into both the reference channel of the scope and the DUT input. For frequencies above some 10MHz, proper 50 ohms termination and a resistive power splitter is highly recommended.

You can also use the dual channels of the signal generator instead, but then it has to operate in Channel Coupled mode for amplitude, frequency and phase.

For higher frequencies, matched cable lengths are important to minimize the phase error.

I'm reading that as "yes, they should match." I'm using both outputs from the AWG as described earlier in this thread. One is direct to the scope CH1 with a 50ohm adapter, and the other output goes to my DUT, which then goes through a 10x probe to CH3 on the scope.

With AWG CH2 locked to CH1, it doesn't appear that there's a way for me to have the waveforms match phase and amplitude on the scope.

Am I wasting my time with this? Is there any real benefit to the Bode Plot for audio equipment (specifically microphones and tube amplifiers)?

Thanks,
Josh

[Electro Fan]

Hi rf-loop,

This is kind of off topic, but first thanks for all of your many contributions to the forum.  Your explanations are very helpful and your documentation is very impressive.  Any chance you can share what editing program you use to annotate screen shots? 

Thanks!

[Performa01]

How can I reconcile the 50R terminated load on CH1 with a HiZ load on CH3?

This is an oversight in the Bode Plot module. The next version will allow to specify separate load impedances for the reference and output channels.

Always specify the output impedance. In your setup it would be high-Z (as it is represented by the x10 probe) and for frequencies that low you could even connect the reference channel without a 50 ohms inline termination. The result will be correct as long as the reference channel sees the same amplitude as the DUT input.

Considering this oversight, it doesn't seem functional at all in my current setup.

Quite obviously, the load impedance is irrelevant for voltage/current measurements, so this is only applicable when dBm units for power measurement (mainly in RF applications) are used.
For professional audio, dBu is standard, but Vrms, dBV or even arbitrary dB might be just as useful, depending on the actual task.

This means, you will of course not see any difference as long as you don’t select dBm units for the measurement. It is also not a problem if you just look at the DUT output (Pout) – you just need to specify the load impedance to get a correct power measurement.

To cut a long story short, there is only a problem when measuring the power gain, i.e. Pout/Pin and the load impedance of input and output are not the same. This could yield a maximum error of 3dB. As stated before, this is a known issue and will be fixed at some point.

For the bode plot to be accurate, do the waveforms need to match well? I tried offsetting the phase, which was about 120 degrees off, but that was never perfect, and the amplitude never matched. The closest was setting the amplitude to 90mVpp.

It normally should be just a single signal, fed into both the reference channel of the scope and the DUT input. For frequencies above some 10MHz, proper 50 ohms termination and a resistive power splitter is highly recommended.

You can also use the dual channels of the signal generator instead, but then it has to operate in Channel Coupled mode for amplitude, frequency and phase.

For higher frequencies, matched cable lengths are important to minimize the phase error.

I'm reading that as "yes, they should match." I'm using both outputs from the AWG as described earlier in this thread. One is direct to the scope CH1 with a 50ohm adapter, and the other output goes to my DUT, which then goes through a 10x probe to CH3 on the scope.

With AWG CH2 locked to CH1, it doesn't appear that there's a way for me to have the waveforms match phase and amplitude on the scope.

Sorry, but I don’t quite understand your problem? Are you saying you can’t get a perfect channel tracking for amplitude, frequency and phase on your AWG?

Am I wasting my time with this? Is there any real benefit to the Bode Plot for audio equipment (specifically microphones and tube amplifiers)?
 

Bode Plot is perfectly suitable for audio applications – it is just a matter of the frequency range.

Here is an example for a simple ~20kHz RC lowpass filter:


SDS1104X-E_FRA_CTC_100nF_Mylar

Here’s another example; this is a discrete transistor amplifier with 60dB gain (measurements shows less because for this test there is also a ~14dB attenuator at the amplifier input). This is with the SDS5000X, but the Bode Plot application has the same core as on the SDS1004X-E:


SDS5104X_FRA_R41_-A14+G60dB_Ch2

So this is perfectly suitable to analyze and document the frequency/phase response of a tube amplifier.

Your question about the microphone is a different story. Measuring the frequency response of a microphone is about the most demanding task of all and at least as much of a challenge as antenna measurements in telecommunications.

The Bode Plot of the SDS1004X-E is not ideal for such tasks, because it cannot apply a correction curve, hence cannot be calibrated. The requirements for the test setup are a true challenge and very expensive on top of that.

The DUT would then usually look like this:


Mic_FRA

Since a microphone is a transducer that has no electrical input, we need another transducer to convert the electrical signal to sound pressure (some sort of reference loudspeaker). The latter will certainly not have a flat response, so you need to know its detailed frequency/phase response in order to apply the required correction to the measurement results. Such a calibrated sound pressure source together with the anechoic chamber would be prohibitively expensive for a hobbyist I think.

[rf-loop]

In some special cases perhaps also this kind of principle is possible?  (as my nickname tell I'm more RF.... )
(of course it need well known ref Mic etc...  and very carefully istalled so it do not mess Microphone under test and it get as perfectly same as Mic under test get.)

[Performa01]

In some special cases perhaps also this kind of principle is possible?  (as my nickname tell I'm more RF.... )
(of course it need well known ref Mic etc...  and very carefully istalled so it do not mess Microphone under test and it get as perfectly same as Mic under test get.)

Thanks for pointing this out. It is of course viable, but now requires an calibrated measuring microphone, which makes the setup even more expensive. Just like reference sound pressure sources, measuring microphones do not have a perfectly flat response, but are calibrated, i.e. their response is precisely known and the corresponding correction can be applied to the measurement results.

EDIT: I think your point is that the sound pressure source needs not be calibrated, but it still needs to be high quality to come close to a point shaped source without partial oscillation of the membrane to give a uniform sound field.

[KungFuJosh]

In some special cases perhaps also this kind of principle is possible?  (as my nickname tell I'm more RF.... )
(of course it need well known ref Mic etc...  and very carefully istalled so it do not mess Microphone under test and it get as perfectly same as Mic under test get.)

Thanks for pointing this out. It is of course viable, but now requires an calibrated measuring microphone, which makes the setup even more expensive. Just like reference sound pressure sources, measuring microphones do not have a perfectly flat response, but are calibrated, i.e. their response is precisely known and the corresponding correction can be applied to the measurement results.

EDIT: I think your point is that the sound pressure source needs not be calibrated, but it still needs to be high quality to come close to a point shaped source without partial oscillation of the membrane to give a uniform sound field.

Very interesting. It appears that you're suggesting that the test should use a live capsule on the microphone. I often test the mic prior to installing a capsule. So the mic would have a ~72pF dummy load instead of the capsule, and then I insert a signal directly to the mic circuit.

Shouldn't this be significantly simpler to test the circuit that way?

Thanks,
Josh