Rigol DHO804 Test and Compare Thread

[iMo]

.. i heard DHO800 FFT not good? i provided facts not fart talk, tell me what SA price has -120 to -140dB dynamic range? this is possible because Rigol proved themselve by providing API documentation and working PC SW example since DS1000E, and it got better and better. i havent seen one single sig-fanboy proved otherwise (siglent has better FFT? duh!) thats why i stick with Rigol brand... my Anritsu MS8609A cant reach -100dB let alone -140dB..

How they do that -120 to -140dB dynamic range with a 12bit ADC?
PS: provided you mean the "noise floor" - how they could achieve such a low n.floor?

[Fungus]

Resting stats is something that you do all the time. It went over your head that to get good measurements, you need to measure several time to assure all is right. You do that by keeping scope running, and resetting stats several times.

Please provide a concrete example of when manual resetting is better than simply seeing (eg.) the RMS of the last 100 acquisitions on a 'scope that's constantly running.

I can easily think of when sliding window is much more useful, eg. when you're fiddling with a trim pot and don't want to keep reaching over to the 'scope to reset the stats every time you turn it to see what happens.

I'm struggling to think of an example where manual resetting is better.

[Fungus]

How they do that -120 to -140dB dynamic range with a 12bit ADC?
PS: provided you mean the "noise floor" - how they could achieve such a low n.floor?

That's just the way FFT math works.

It's like averaging of datapoints to get extra bits from the ADC but in the frequency domain.

[2N3055]

Resting stats is something that you do all the time. It went over your head that to get good measurements, you need to measure several time to assure all is right. You do that by keeping scope running, and resetting stats several times.

Please provide a concrete example of when manual resetting is better than simply seeing (eg.) the RMS of the last 100 acquisitions on a 'scope that's constantly running.

I can easily think of when sliding window is much more useful, eg. when you're fiddling with a trim pot and don't want to keep reaching over to the 'scope to reset the stats every time you adjust it.

I'm struggling to think of an example where manual resetting is better.

Don't be mad if I don't, I explained several times.
Just think about it more, you will understand.
Every scope I have, have reset button directly inside stat window.
LeCroy and Siglent also have Clear sweeps that flushes full acquisition memory for same reason..

[2N3055]

.. i heard DHO800 FFT not good? i provided facts not fart talk, tell me what SA price has -120 to -140dB dynamic range? this is possible because Rigol proved themselve by providing API documentation and working PC SW example since DS1000E, and it got better and better. i havent seen one single sig-fanboy proved otherwise (siglent has better FFT? duh!) thats why i stick with Rigol brand... my Anritsu MS8609A cant reach -100dB let alone -140dB..

How they do that -120 to -140dB dynamic range with a 12bit ADC?
PS: provided you mean the "noise floor" - how they could achieve such a low n.floor?

Noise floor of scope is defined by ADC inherent noise (analog noise in input driving circuits that drive actual conversion circuitry) and quantization noise that is defined by resolution and sampling imperfections.
But on a scope you also have a front end before ADC, and that has a preamp that amplifies input signal for ADC.
Scope noise floor will be combination of those two.
So if you have a good, low noise preamp, your bottom can start from -140dBm but because dynamic range is 70db, top of scale will be -70dBm. Of course that would be ideal, tha cannot ever happen in prractice.

You can say that by using preamps/attenuators you have a 70dB window you can slide up and down: from -100 dBm to -30dBm. Or from -40dBm to +30dBm. But never from -100 dBm to +30dBm, cause that would mean 130dBm dynamic range...
That would mean 21,5 bit converter running at GHz speeds...
If you know how to make one, there is a Nobel prize waiting...

[ebastler]

Resting stats is something that you do all the time. [...]

Please provide a concrete example of when manual resetting is better [...]

Gentlemen, I am confused now. I thought you had found that manually resetting the statistics is possible with the DHO800? The user manual seems to say that the operation can even be assigned to the Quick button on the front panel if you need to use it often.

If I got this right and the scope does offer the choice of manually resetting stats -- what is the debate about?

[Fungus]

Don't be mad if I don't, I explained several times.

All you explained is that you like to press reset, not that it's necessary.

Just think about it more, you will understand.
Every scope I have, have reset button directly inside stat window.

Argument from authority. Got it.

Anyway: You're avoiding the question. Can the Siglent do all this segmention/stats stuff?

[2N3055]

Resting stats is something that you do all the time. [...]

Please provide a concrete example of when manual resetting is better [...]

Gentlemen, I am confused now. I thought you had found that manually resetting the statistics is possible with the DHO800? The user manual seems to say that the operation can even be assigned to the Quick button on the front panel if you need to use it often.

If I got this right and the scope does offer the choice of manually resetting stats -- what is the debate about?

He said it has it and that is good enough for me. Fact that it can be added to quick button is great. Thank you for that data.
Argument, I don't know, Fungus seems to insisting resetting stats is not necessary because Rigol is doing sliding buffer stats.
I don't understand  how that is connected.

All in all Stats work, you can reset them, they work on segmented buffers...
 

[2N3055]

Don't be mad if I don't, I explained several times.

All you explained is that you like to press reset, not that it's necessary.

Just think about it more, you will understand.
Every scope I have, have reset button directly inside stat window.

Argument from authority. Got it.

Anyway: You're avoiding the question. Can the Siglent do all this segmention/stats stuff?

Are you deliberately provoking me today because I try to be polite even to you?
Argument from authority? You bet your ass it is. Are you saying you are smarter and know better than Kesight, R&S, LeCroy, Picoscope, Siglent and many more.... I know I wouldn't dare to say something like that for myself

If you don't understand why you would need to reset stats all the time while doing measurements, than I cannot explain you. It should be point of pure logic not a data point to learn by heart.
As I said, you will remember my words, once you actually learn how to use statistics to your benefit.

And I'm not avoiding the question I already said it before: it can.
That is why I'm asking, for COMPARISON...
If Rigol didn't do it it would be a minus for it.
Since it can that is good for Rigol.
Get it?

[2N3055]

.. i heard DHO800 FFT not good? i provided facts not fart talk, tell me what SA price has -120 to -140dB dynamic range? this is possible because Rigol proved themselve by providing API documentation and working PC SW example since DS1000E, and it got better and better. i havent seen one single sig-fanboy proved otherwise (siglent has better FFT? duh!) thats why i stick with Rigol brand... my Anritsu MS8609A cant reach -100dB let alone -140dB..

How they do that -120 to -140dB dynamic range with a 12bit ADC?
PS: provided you mean the "noise floor" - how they could achieve such a low n.floor?

Some literature to scratch the surface...

https://www.edn.com/dsos-and-noise/

[Fungus]

If you don't understand why you would need to reset stats all the time while doing measurements, than I cannot explain

I don't see any difference between resetting and simply waiting 2 seconds for the window to slide along.

(apart from the physical movement of your arm)

[2N3055]

If you don't understand why you would need to reset stats all the time while doing measurements, than I cannot explain

I don't see any difference between resetting and simply waiting 2 seconds for the window to slide along.

(apart from the physical movement of your arm)

Two hints.
How deep is sliding window ?
What timebase?

[Mechatrommer]

If those are made on your PC from data pulled from scope, that is just funny.
We are arguing FFT implementation on a scope is half cooked, and to disprove me you post images of FFT that you made on PC because FFT on scope is crap...
What a great irony is that?

so you didnt see the picture of built-in FFT i posted?



312K FFT points (625Ksamples) i'm providing both i'm not sure what siglent marketing hype refering to "points"? do you have any better siglent screenshot FFT? you asked people to do you favor by providing screenshots, but you didnt provide anything yourself. people are having hard time understanding what you want. in any case, i was showing that FFT of DHO800 is much much better than previous DS1000Z. i initially thought this thread is comparing new rigol vs older rigol, but it turn out differently, re-reading OP, maybe he want to compare between rigol and siglent, but his rigol hasnt arrived yet? so lets wait, i hope its going to be a thorough one including pulling data to PC, i havent seen that be done with siglent scope and its code example.

and to the rest replying i dont have time quoting each one... my point was, raw 16bit data can be pulled to PC from Rigol DHO800 scope, can this be done with Siglent? if not, dont bother claiming it has better built-in FFT, because it crap anyway compared to proper SA. i also pointed out, with higher data point count, cleaner FFT can be achieved, i dont know your theoritical formula for X bit scope's noise floor, but FFT i've shown has cleaner floor at -120dB although there are some spurious below -90dB, we should be able to see low level signal existence in -60 to -80dB range, since spurious are below that, and where there are no spurious in freq domain, we can see signal of much lower level, too bad my rfgen can only go -40dBm output and not the high end pure sine one. we can safely say imho anything above -90dB are real signal to be analyzed. even built-in FFT above also showing -120dB floor, does FFT algorithm lies?. and its lower RBW too with more datapoints only can be achieved by pulling data out of Rigol to PC, i believe even the holy grail siglent built in FFT is "blobby" (leaky) due to large RBW. and through VISA driver API, i can automate it into program instead of copying it into CSV in pendrive and then play math in Excel or Matlab, that is tedious and slow work. from some stand, i can say any entry level dso's built in FFT are useless anyway, so there is not much point highlighting it. thats why i developed my own SW, even 1Mpts siglent is not enough for me. i was looking its programming guide i believe i have it stored somewhere, but i havent seen a working SW/code example. ymmv.

.. i heard DHO800 FFT not good? i provided facts not fart talk, tell me what SA price has -120 to -140dB dynamic range? this is possible because Rigol proved themselve by providing API documentation and working PC SW example since DS1000E, and it got better and better. i havent seen one single sig-fanboy proved otherwise (siglent has better FFT? duh!) thats why i stick with Rigol brand... my Anritsu MS8609A cant reach -100dB let alone -140dB..

How they do that -120 to -140dB dynamic range with a 12bit ADC?
PS: provided you mean the "noise floor" - how they could achieve such a low n.floor?

Some literature to scratch the surface...
https://www.edn.com/dsos-and-noise/

thats not literature... this is literature... https://dsp.stackexchange.com/questions/86124/units-of-6-02n-1-76-as-an-fft-noise-floor and this https://www.eetimes.com/fft-plots-provide-insight-to-a-d-performance/ but i wont bother digging deep. i already have 10Mpts FFT, with markers system and zoom out zoom in feature. so thats what i will use.. what you need to prove to me now is why FFT algorithm lies?

[iMo]

The FFT picture above tells me the real spectral lines of your signal (11 lines) are positioned in -30dB to -90dB band and they come from the FFT. All other spectral components you may see there is a math junk..

[2N3055]

If those are made on your PC from data pulled from scope, that is just funny.
We are arguing FFT implementation on a scope is half cooked, and to disprove me you post images of FFT that you made on PC because FFT on scope is crap...
What a great irony is that?

so you didnt see the picture of built-in FFT i posted?

312K FFT points (625Ksamples) i'm providing both i'm not sure what siglent marketing hype refering to "points"?

I didn't, sorry.  But you don't understand what I'm asking you just seem to want to insult me because I dare to question your precious..

You say 312K FFT points (625Ksamples). OK. Where do those numbers come from? Where does it say on screen? It should otherwise those are just some squiggles..

And as for other response I provided a link to introductory text on topic of FFT on scope and implementations..
You provided links to what can be calculated with FFT if you dig deeper..

My link is relevant to his question, your link is relevant to what you are doing on your PC.. Good read and good articles but not on topic per se.. They are a bit too far down the rabbit hole...
I already told you, I admire what you can do with data on PC. Open separate topic and discuss it there in depth. But when FFT on scope as implemented is discussed, your offline PC analysis is off topic and only serves to confuse people.
I did mention several time that users (and that is you) reported that data transfer to PC i quite decent and that already some users are doing offline analysis on PC..

[Fungus]

How deep is sliding window ?

Whatever you want. (eg. In my segmentation test I set it to '4')

It remembers the setting, even across power cycles.

What timebase?

Doesn't matter. You can easily look at the number and see if it's still settling or not.

If it's taking its sweet time about settling down then feel free to manually reset it (or, more sensibly, adjust the window size so you don't need to reset it on the next change to your circuit).

[2N3055]

The FFT picture above tells me the real spectral lines of your signal (11 lines) are positioned in -30dB to -90dB band and they come from the FFT. All other spectral components you may see there is a math junk..

Some of that grass will be spurious tones from ADC nonlinearities, and some are folded back signal that was aliased.

[Fungus]

If it's taking its sweet time about settling down then feel free to manually reset it (or, more sensibly, adjust the window size so you don't need to reset it on the next change to your circuit).

Also: If it never settles down then you know the system isn't stable.

Your manually-reset stats might hide that valuable information from you.

[Mechatrommer]

The FFT picture above tells me the real spectral lines of your signal (11 lines) are positioned in -30dB to -90dB band and they come from the FFT. All other spectral components you may see there is a math junk..

very good! now at least some people looked closer! its possible i think to subtract dso's noise esp the larger one exist in three places there using some power combination formula, iirc its some form of  "pythagorean" or squared root rms formula, but i need to read paper and formulation again first to back this up and apply it in real application, but thats for later. with built-in scope's FFT, we can only "eyeball".

Some of that grass will be spurious tones from ADC nonlinearities, and some are folded back signal that was aliased.

same rule applied to any FFT calculating machine, siglent, rigol, tek, hp R&S whatever. is at the mercy of sampling HW system. to get rid aliasing entirely, you need brickwall filter, and that is a theoritical filter one cant achieve in real world. unless you are far crippling the sampling mechanism with much much lower BW filter. 8th order approximation also not that flat, how good front end filter for both rigol and siglent?

[2N3055]

How deep is sliding window ?

Whatever you want. (eg. In my segmentation test I set it to '4')

It remembers the setting, even across power cycles.

What timebase?

Doesn't matter. You can easily look at the number and see if it's still settling or not.

If it's taking its sweet time about settling down then feel free to manually reset it (or, more sensibly, adjust the window size so you don't need to reset it on the next change to your circuit).

Hehehe, grasshopper, set stat length of 1000 and then you'll get it..
4 averages are not stat, really... 

[2N3055]

If it's taking its sweet time about settling down then feel free to manually reset it (or, more sensibly, adjust the window size so you don't need to reset it on the next change to your circuit).

Also: If it never settles down then you know the system isn't stable.

Your manually-reset stats might hide that valuable information from you.

You don't understand much do you?

[2N3055]

Some of that grass will be spurious tones from ADC nonlinearities, and some are folded back signal that was aliased.

same rule applied to any FFT calculating machine, siglent, rigol, tek, hp R&S whatever. is at the mercy of sampling HW system. to get rid aliasing entirely, you need brickwall filter, and that is a theoritical filter one cant achieve in real world. unless you are far crippling the sampling mechanism with uch uch lower BW filter.

Stop being rude and aggressive. I explained something, replying to his "math junk" comment to which I disagree. I didn't say it was "Rigol problem".
Of course every system has same. There are no ideal systems. We discussed this many times.
At his point I don't have any opinion on how well it performs. Thanks to your data I will take a look and then will comment.

[iMo]

OK, "math junk" - I meant "none existent in the source signal, created artificially by any internal mechanism"..

[TimFox]

Otherwise known as "artifacts".
While studying physics at the University of Chicago, I had friends in the archaeology department who thought artifacts were a good thing (as opposed to rocks carved by natural processes).

[2N3055]

Otherwise known as "artifacts".
While studying physics at the University of Chicago, I had friends in the archaeology department who thought artifacts were a good thing (as opposed to rocks carved by natural processes).

Usually called spurious tones in SA world... When in spectrum mode I call them so.. Seems appropriate..