Rigol HDO1000 and HDO4000 12bit oscilloscopes launched in China

[rf-loop]

Like this?

No.

Is it better to first read and after then do.

[skander36]

Then I need more clear info.
You said "and just input single sinewave, say example 100.5kHz.  "
You want to be at 400 kSa/s?
Edit- or with cursors?

[rf-loop]

Then I need more clear info.
You said "and just input single sinewave, say example 100.5kHz.  "
You want to be at 400 kSa/s?

No.. I said:
"But, FFT window info tell 400kSa/s. If this is real truth then it must not fold back (alias). Is it (400kSa/s) real truth or is it just Rigolism.
If they have different sampling for FFT it is then nice to see what is display if normal oscilloscope and freq.axis displays are both displayed same time and just input single sinewave, say example 100.5kHz."

Now please:
Set oscilloscope Horizontal 2.5ms/div, length 10ksample, and you can see it tell sampling speed is 200kSa/s
So oscilloscope top display info tell (H) 2.5ms/div (A) 200kSa/s 10kpts (D) 0.0s  also set trigger near zero and mode Edge (as it was before)

Now keep this oscilloscope displayed.

And open FFT.  (both visible, time base display (aka oscilloscope) and freq. base display (aka FFT in this case)
FFT info as it was earlier, 400kSa/s

FFT Center 100kHz.

Input sinewave 100.5kHz (or 101kHz)

Now in your last FFT window there was also 50kHz span.

You can set bit more narrow span for better visibility for this case now. Say example 10kHz span.

And, if it is now possible with Rigol display both together. Oscilloscope view and FFT view on same screen. (overlaid or in separate windows depending what is possible in Rigol)


ETA: oh ok you did it allready. Try also open oscilloscope screen together with FFT.
And narrow FFT span to 10 or 5kHz.

Is it fun what you see in FFT window   

I think Rigol need urgently do some rework.

 

[skander36]

Then I need more clear info.
You said "and just input single sinewave, say example 100.5kHz.  "
You want to be at 400 kSa/s?

No.. I said:
"But, FFT window info tell 400kSa/s. If this is real truth then it must not fold back (alias). Is it (400kSa/s) real truth or is it just Rigolism.
If they have different sampling for FFT it is then nice to see what is display if normal oscilloscope and freq.axis displays are both displayed same time and just input single sinewave, say example 100.5kHz."

Now please:
Set oscilloscope Horizontal 2.5ms/div, length 10ksample, and you can see it tell sampling speed is 200kSa/s
So oscilloscope top display info tell (H) 2.5ms/div (A) 200kSa/s 10kpts (D) 0.0s  also set trigger near zero and mode Edge (as it was before)

Now keep this oscilloscope displayed.

And open FFT.  (both visible, time base display (aka oscilloscope) and freq. base display (aka FFT in this case)
FFT info as it was earlier, 400kSa/s

FFT Center 100kHz.

Input sinewave 100.5kHz (or 101kHz)

Now in your last FFT window there was also 50kHz span.

You can set bit more narrow span for better visibility for this case now. Say example 10kHz span.

And, if it is now possible with Rigol display both together. Oscilloscope view and FFT view on same screen. (overlaid or in separate windows depending what is possible in Rigol)


ETA: oh ok you did it allready. Try also open oscilloscope screen together with FFT.
And narrow FFT span to 10 or 5kHz.

Is it fun what you see in FFT window   

I think Rigol need urgently do some rework.

10 kHz span - 2,5ms/div and 2 ms/div

[2N3055]

Then I need more clear info.
You said "and just input single sinewave, say example 100.5kHz.  "
You want to be at 400 kSa/s?

No.. I said:
"But, FFT window info tell 400kSa/s. If this is real truth then it must not fold back (alias). Is it (400kSa/s) real truth or is it just Rigolism.
If they have different sampling for FFT it is then nice to see what is display if normal oscilloscope and freq.axis displays are both displayed same time and just input single sinewave, say example 100.5kHz."

Now please:
Set oscilloscope Horizontal 2.5ms/div, length 10ksample, and you can see it tell sampling speed is 200kSa/s
So oscilloscope top display info tell (H) 2.5ms/div (A) 200kSa/s 10kpts (D) 0.0s  also set trigger near zero and mode Edge (as it was before)

Now keep this oscilloscope displayed.

And open FFT.  (both visible, time base display (aka oscilloscope) and freq. base display (aka FFT in this case)
FFT info as it was earlier, 400kSa/s

FFT Center 100kHz.

Input sinewave 100.5kHz (or 101kHz)

Now in your last FFT window there was also 50kHz span.

You can set bit more narrow span for better visibility for this case now. Say example 10kHz span.

And, if it is now possible with Rigol display both together. Oscilloscope view and FFT view on same screen. (overlaid or in separate windows depending what is possible in Rigol)


ETA: oh ok you did it allready. Try also open oscilloscope screen together with FFT.
And narrow FFT span to 10 or 5kHz.

Is it fun what you see in FFT window   

I think Rigol need urgently do some rework.

10 kHz span - 2,5ms/div and 2 ms/div

Something is wrong here. FFT shows RBW of 39,99 mHz (0.03999 Hz), but FFT plot for the peaks is 500 Hz wide... At that magnification it should be just a single vertical line..

Also scope samples at 200kS/s but FFT states 400 kS/s. 
Why do you keep scope at stupidly small 10 kpts?
Set the scope memory to AUTO, and then look at FFT.

If ADC is sampling at 200kS/s 101 kHz is violating Nyquist..
And FFT is upsampling aliased data...

On my scope I couldn't even make it do something so wrong until I made custom math transformation.
I made it sample at fixed 250 kS/s sample rate, fed it 125 kHz and 126 kHz signal.
But since FFT (correctly) refuses to show anything above Nyquist, I had to manually create math upsample from 250 kS/s to 500 kS/s, and then did FFT of that. And lo and behold, here is wrong plot in all it's glory...

Then I went and set sampling and memory to auto, and gave it a bit more data and got different picture...

All of this drills down to few points being trumpeted here (but hey it's all a propaganda, you know!!).
1. Manual settings of a scope is as dangerous as it is versatile. Auto is your friend as far as memory management is concerned. Unless you need to set it specifically for some reason, but make sure you understand repercussions.
2. If you want to set something manually, with FFT you want to set manual sample rate, not memory size.. That way you control aliasing, and it doesn't change all the time with timebase. Otherwise keep it AUTO and let FFT downsample if it needs. That way you loose RBW (unless in very long timebases) but no aliasing. After you see propper picture of signal, then you can tweak sample rate/memory size or whatnot to try to optimize what you see.
3. This scope is upsampling in FFT without warning or any explanation that what it's doing might be wrong. Bad implementation of FFT. They meant to make it more like SA, but ignored math that goes with it. So it is neither good FFT or good SA. If you want to understand what I'm saying, just look for all the write ups on FFT implementation on R&S RTB2000 and all associated pro et contra...

[skander36]

Then I need more clear info.
You said "and just input single sinewave, say example 100.5kHz.  "
You want to be at 400 kSa/s?

No.. I said:
"But, FFT window info tell 400kSa/s. If this is real truth then it must not fold back (alias). Is it (400kSa/s) real truth or is it just Rigolism.
If they have different sampling for FFT it is then nice to see what is display if normal oscilloscope and freq.axis displays are both displayed same time and just input single sinewave, say example 100.5kHz."

Now please:
Set oscilloscope Horizontal 2.5ms/div, length 10ksample, and you can see it tell sampling speed is 200kSa/s
So oscilloscope top display info tell (H) 2.5ms/div (A) 200kSa/s 10kpts (D) 0.0s  also set trigger near zero and mode Edge (as it was before)

Now keep this oscilloscope displayed.

And open FFT.  (both visible, time base display (aka oscilloscope) and freq. base display (aka FFT in this case)
FFT info as it was earlier, 400kSa/s

FFT Center 100kHz.

Input sinewave 100.5kHz (or 101kHz)

Now in your last FFT window there was also 50kHz span.

You can set bit more narrow span for better visibility for this case now. Say example 10kHz span.

And, if it is now possible with Rigol display both together. Oscilloscope view and FFT view on same screen. (overlaid or in separate windows depending what is possible in Rigol)


ETA: oh ok you did it allready. Try also open oscilloscope screen together with FFT.
And narrow FFT span to 10 or 5kHz.

Is it fun what you see in FFT window   

I think Rigol need urgently do some rework.

10 kHz span - 2,5ms/div and 2 ms/div

Something is wrong here. FFT shows RBW of 39,99 mHz (0.03999 Hz), but FFT plot for the peaks is 500 Hz wide... At that magnification it should be just a single vertical line..

Also scope samples at 200kS/s but FFT states 400 kS/s. 
Why do you keep scope at stupidly small 10 kpts?
Set the scope memory to AUTO, and then look at FFT.

If ADC is sampling at 200kS/s 101 kHz is violating Nyquist..
And FFT is upsampling aliased data...

On my scope I couldn't even make it do something so wrong until I made custom math transformation.
I made it sample at fixed 250 kS/s sample rate, fed it 125 kHz and 126 kHz signal.
But since FFT (correctly) refuses to show anything above Nyquist, I had to manually create math upsample from 250 kS/s to 500 kS/s, and then did FFT of that. And lo and behold, here is wrong plot in all it's glory...

Then I went and set sampling and memory to auto, and gave it a bit more data and got different picture...

All of this drills down to few points being trumpeted here (but hey it's all a propaganda, you know!!).
1. Manual settings of a scope is as dangerous as it is versatile. Auto is your friend as far as memory management is concerned. Unless you need to set it specifically for some reason, but make sure you understand repercussions.
2. If you want to set something manually, with FFT you want to set manual sample rate, not memory size.. That way you control aliasing, and it doesn't change all the time with timebase. Otherwise keep it AUTO and let FFT downsample if it needs. That way you loose RBW (unless in very long timebases) but no aliasing. After you see propper picture of signal, then you can tweak sample rate/memory size or whatnot to try to optimize what you see.
3. This scope is upsampling in FFT without warning or any explanation that what it's doing might be wrong. Bad implementation of FFT. They meant to make it more like SA, but ignored math that goes with it. So it is neither good FFT or good SA. If you want to understand what I'm saying, just look for all the write ups on FFT implementation on R&S RTB2000 and all associated pro et contra...

I ALWAYS USE AUTO SETING FOR MEMORY.
In some specific cases  I use Maximum available memory on that scope.
Auto means that when I use some specific setting as was requested this value is autmatically adjusted. So I'm not keeping myself. I was adjusting only timebase, as sample rate is automatically adjusted with memory.
It was always in auto position.
I just want to help mawyatt to see how is looking a two tone FFT on this scope following his setup. I don't want to show something from my side.

[rf-loop]

I want still stay in this particular case because (even if user setup is not at all best possible for this and not--- also if we look previous two tone mess)

All is in this image.

@skander36 previous image

There is now ONE signal, 100.5kHz to input, is it as I requested.  (naturally with its own some harmonics etc but these do not matter here now at all)
FFT display two signals!!!!!! 100.5kHz as is ok and if FFT real samplerate is 400kSa/s ( f_Nyquist 200kHz)
But there is also 99.5kHz signal with quite same level as real input signal 100,5kHz peak.


This display is IMPOSSIBLE if somethings is not really messed up in Rigol system. If input is as requested.
Also look this noise "signature"... it reveals more than Rigol hopes..  (think about 100kHz center)  

IF
FFT samplerate is 200kSa/s as also main timebase is. Then this display is still messed up BUT in this case it need display 99.5kHz peak alone and signal trace need end to 100kHz points so 100.5kHz is out of FFT. (But 100.5kHz is folded back from f_Nyquist 100kHz  so result is 99.5kHz)
But because it also display 100,5kHz it looks like FFT samplerate is not same as main timebase 200kSa/s. If FFT samplerate is 400kSa/s as displayed info is. How it  generate this 99.5kHz. Even if 100,5kHz input signal have extremely high harmonixs this display is still not possible.



Now I want ask from @skander36

Are you really sure there is only one 100.5kHz sinewave signal going to Rigol input as I requested earlier? (not dual tone)

If yes, then Rigol have severe problem.  (also without this answer I know... )

[gf]

The FFT trace looks horizontally mirrored at 100kHz, even the noise seems mirrored. Strange.

[skander36]

Yes, is a single wave on the signal.
Indeed, the wave at slower timebase look as an modulated wave.
But it is the wave below.

 

[2N3055]

I ALWAYS USE AUTO SETING FOR MEMORY.
In some specific cases  I use Maximum available memory on that scope.
Auto means that when I use some specific setting as was requested this value is autmatically adjusted. So I'm not keeping myself. I was adjusting only timebase, as sample rate is automatically adjusted with memory.
It was always in auto position.
I just want to help mawyatt to see how is looking a two tone FFT on this scope following his setup. I don't want to show something from my side.

No need to be defensive. I didn't say you have no clue or something offensive.
But look at that screen. Scope says it is sampling with 200kS/s and 10kpoints.  Why?.
I was actually hoping that is because it was set to manual setting... Because if it isn't, then something is indeed wrong..

Scope will limit number of points at 5 or 10 ns/div. Because of maximum sample rate.
When you are at ms/div, scope (at AUTO) should lower sample rate only when it runs out of memory. But it doesn't say 100 Mpts... it says 10kpoints..

Why is the scope using only this much? It has 100MPoints per datasheet..
And in time domain it shows 200 kPts/s and 400 in FFT..

As I have shown, it is obvious that DHO1000 is actually sampling at 200kHz, therefore aliasing 101kHz to 99kHz and all kinds of wrong things that stem from this.

Why is it doing this?
What kind of weird timebase is 2.5ms/div? (what happened to 1-2-5-10 sequence?)
Is FFT forcing sample rate (but does it wrongly)?
Can you disable FFT, set it to 2.5ms  (or 2ms) /div and see what sample rate and sample memory uses then. Then enable FFT and look if it changes. Did you try to change RBW for FFT? Can you even do that? I cannot see in manual that RBW is among setable parameters.. Maybe scope autosets sample rate and RBW was too wide for timebase and it started changing things.

I know you are trying to show something that Mike asked you for. But I assure it is in your interest to research this weird behaviour into detail. What you see is not right (either a bug or not being used correctly, maybe for the lack of good documentation..) and you should be aware that I would not trust FFT on that scope a single bit, until we get it to show correct graph.

[rf-loop]

Yes, exactly. It was what I start wonder because some peaks looks like f_Nyg fold backs in original previous dual tone test. So we do not need anymore wonder this previous dual tone test mess.

Attached @skander36 picture (with my mod)

Left side bottom trace is mirror from right side. 

So we can now sure say Rigol FFT is broken. Whole mess.

[2N3055]

Yes, is a single wave on the signal.
Indeed, the wave at slower timebase look as an modulated wave.
But it is the wave below.

And your scope still shows 10kpoints.. On every single image your scope shows fixed 10kpoints.
 It is set for manual 10kpoints obviously... Fix that and try again..

[rf-loop]

Yes, is a single wave on the signal.
Indeed, the wave at slower timebase look as an modulated wave.
But it is the wave below.

And your scope still shows 10kpoints.. On every single image your scope shows fixed 10kpoints.
 It is set for manual 10kpoints obviously... Fix that and try again..

How ever but Rigol FFT is broken. Period.

[skander36]

I ALWAYS USE AUTO SETING FOR MEMORY.
In some specific cases  I use Maximum available memory on that scope.
Auto means that when I use some specific setting as was requested this value is autmatically adjusted. So I'm not keeping myself. I was adjusting only timebase, as sample rate is automatically adjusted with memory.
It was always in auto position.
I just want to help mawyatt to see how is looking a two tone FFT on this scope following his setup. I don't want to show something from my side.

No need to be defensive. I didn't say you have no clue or something offensive.
But look at that screen. Scope says it is sampling with 200kS/s and 10kpoints.  Why?.
I was actually hoping that is because it was set to manual setting... Because if it isn't, then something is indeed wrong..

Scope will limit number of points at 5 or 10 ns/div. Because of maximum sample rate.
When you are at ms/div, scope (at AUTO) should lower sample rate only when it runs out of memory. But it doesn't say 100 Mpts... it says 10kpoints..

Why is the scope using only this much? It has 100MPoints per datasheet..
And in time domain it shows 200 kPts/s and 400 in FFT..

As I have shown, it is obvious that DHO1000 is actually sampling at 200kHz, therefore aliasing 101kHz to 99kHz and all kinds of wrong things that stem from this.

Why is it doing this?
What kind of weird timebase is 2.5ms/div? (what happened to 1-2-5-10 sequence?)
Is FFT forcing sample rate (but does it wrongly)?
Can you disable FFT, set it to 2.5ms  (or 2ms) /div and see what sample rate and sample memory uses then. Then enable FFT and look if it changes. Did you try to change RBW for FFT? Can you even do that? I cannot see in manual that RBW is among setable parameters.. Maybe scope autosets sample rate and RBW was too wide for timebase and it started changing things.

I know you are trying to show something that Mike asked you for. But I assure it is in your interest to research this weird behaviour into detail. What you see is not right (either a bug or not being used correctly, maybe for the lack of good documentation..) and you should be aware that I would not trust FFT on that scope a single bit, until we get it to show correct graph.

For showing a slightly better graph I need to manually set the memory, not auto.
I have show also a timebase of 2 ms that show better.
I was keeping 2,5 because it has attract the attention to RF-loop and indeed none of my scope show this at 2,5ms.

[2N3055]

Yes, is a single wave on the signal.
Indeed, the wave at slower timebase look as an modulated wave.
But it is the wave below.

And your scope still shows 10kpoints.. On every single image your scope shows fixed 10kpoints.
 It is set for manual 10kpoints obviously... Fix that and try again..

How ever but Rigol FFT is broken. Period.

Well I agree that it is not right. But I'm trying to help a person figure out what so they can maybe report it or at least be aware they have a problem and where it is...

On my scope, if I force main acquistion to 200Msps/s, FFT won't show anything more than 100kHz in FFT..
That way it shows you BW of the process.. If you have aliasing, you have it, but if initial sample was 200kSps/s, nothing over 100kHz makes sense..

[2N3055]

I ALWAYS USE AUTO SETING FOR MEMORY.
In some specific cases  I use Maximum available memory on that scope.
Auto means that when I use some specific setting as was requested this value is autmatically adjusted. So I'm not keeping myself. I was adjusting only timebase, as sample rate is automatically adjusted with memory.
It was always in auto position.
I just want to help mawyatt to see how is looking a two tone FFT on this scope following his setup. I don't want to show something from my side.

No need to be defensive. I didn't say you have no clue or something offensive.
But look at that screen. Scope says it is sampling with 200kS/s and 10kpoints.  Why?.
I was actually hoping that is because it was set to manual setting... Because if it isn't, then something is indeed wrong..

Scope will limit number of points at 5 or 10 ns/div. Because of maximum sample rate.
When you are at ms/div, scope (at AUTO) should lower sample rate only when it runs out of memory. But it doesn't say 100 Mpts... it says 10kpoints..

Why is the scope using only this much? It has 100MPoints per datasheet..
And in time domain it shows 200 kPts/s and 400 in FFT..

As I have shown, it is obvious that DHO1000 is actually sampling at 200kHz, therefore aliasing 101kHz to 99kHz and all kinds of wrong things that stem from this.

Why is it doing this?
What kind of weird timebase is 2.5ms/div? (what happened to 1-2-5-10 sequence?)
Is FFT forcing sample rate (but does it wrongly)?
Can you disable FFT, set it to 2.5ms  (or 2ms) /div and see what sample rate and sample memory uses then. Then enable FFT and look if it changes. Did you try to change RBW for FFT? Can you even do that? I cannot see in manual that RBW is among setable parameters.. Maybe scope autosets sample rate and RBW was too wide for timebase and it started changing things.

I know you are trying to show something that Mike asked you for. But I assure it is in your interest to research this weird behaviour into detail. What you see is not right (either a bug or not being used correctly, maybe for the lack of good documentation..) and you should be aware that I would not trust FFT on that scope a single bit, until we get it to show correct graph.

For showing a slightly better graph I need to manually set the memory, not auto.
I have show also a timebase of 2 ms that show better.
I was keeping 2,5 because it has attract the attention to RF-loop and indeed none of my scope show this at 2,5ms.

Thank you for explaining.. Yeah that is not good FFT implementation... Please don't mind my persistence, I'm just trying to figure this out.  Thanks for patience...
Best,
Sinisa

[mawyatt]

I just want to help mawyatt to see how is looking a two tone FFT on this scope following his setup. I don't want to show something from my side.

Sorry been away awhile helping friends daughter with Calculus 3 upcoming exam.

Thanks for the effort and such, we are all just trying to learn something here about this new HD DSO. In my case this was to investigate the input channel and 12 bit ADC linearity by means of the Two Tone IMD method.

Think what's been uncovered here is the FFT may have some shortcomings and the HD is not behaving as expected wrt the FFT.

Your efforts are appreciated and have enabled this learning experience, otherwise we would all still be in the "fog" as to how this new HD performs, especially the FFT!!

Best,

[rf-loop]

Yes, is a single wave on the signal.
Indeed, the wave at slower timebase look as an modulated wave.
But it is the wave below.

And your scope still shows 10kpoints.. On every single image your scope shows fixed 10kpoints.
 It is set for manual 10kpoints obviously... Fix that and try again..

How ever but Rigol FFT is broken. Period.

Well I agree that it is not right. But I'm trying to help a person figure out what so they can maybe report it or at least be aware they have a problem and where it is...

On my scope, if I force main acquistion to 200Msps/s, FFT won't show anything more than 100kHz in FFT..
That way it shows you BW of the process.. If you have aliasing, you have it, but if initial sample was 200kSps/s, nothing over 100kHz makes sense..

Yes I know you kindly try advise user for better adjustments and this is of course more than good and kind.

But now we  can also see that Rigol FFT have some very severe class bug-bugs.  So using it may lead to severe misleading about signals under test, specially to user who perhaps do not have enough knowledge and experience to analyze its errrors and then avoid its effect.

If real samplerate is example 200kSa/s FFT trace need end to 100kHz.
But now it looks like with this setting it mirrors, as can see in my modified image (right side horizonta mirror under left side. Mirror point 100kHz.

If samplerate is 200kHs it must not display at all this 100.5kHz.  It must only display 99.5kHz peak. But it display both. This problem first looks like that there is same real samplerate for FFT as for oscilloscope but...  this do not explain this at all, because it still display 100.5kHz also. Just wondering what is this mess what it is doing. If I have this machine on table I know how to test further for analyze how it behave.

I can only say. Rigol need repair it. 
User need be careful and do not believe what is display until he know more things about how it behave exactly in different situations. As can see this previous dual tone test, it was one fun example how it can fool user.

[Fungus]

If I have this machine on table I know how to test further for analyze how it behave.

Police, help!

Somebody is having an opinion about an oscilloscope without owning it!

[rf-loop]

If I have this machine on table I know how to test further for analyze how it behave.

Police, help!

Somebody is having an opinion about an oscilloscope without owning it!

It is not my problem if you do not understand anything about these things.

[Fungus]

It is not my problem if you do not understand anything about these things.

The first thought that came into my head when I saw those screenshots wasn't that the sky was falling.

My first thought was "spectral leakage", followed by "I wonder what window function they're using?"

If those questions didn't enter your head before you made a million anti-Rigol posts then... 

I'll leave it to the people who own them to figure out what's going on.

Me? I suspect the FFT math is correct (it's not exactly rocket science) but the sampling rate and buffer size is completely unreasonable, ie. It's an edge case.

Is it a problem? That depends on how easy it is to get into that edge-situation with this 'scope.

[2N3055]

It is not my problem if you do not understand anything about these things.

The first thought that came into my head wasn't that the sky was falling.

The first thought was "spectral leakage", followed by "I wonder what window function they're using?"

If those questions didn't enter your head before you made a million anti-Rigol posts then... 

Seriously, if you don't understand, don't enter the discussion...
Spectral leakage has nothing to do with the fact that sampling rates and sampled data and analysed graph are not correlating...

It's all right to not know something..
But you're not helping HDO1000 users by pretending there is no problem.

And even if we presume basic math behind this FFT is correct, how do you use this thing? Several basic parameters are not available and you have no control over process.

So FFT at this stage needs more work. Which is a shocker... Whole scope is basically in alpha stage. Documenting all the problems and reporting them to Rigol is only way for users of HDO ever getting a scope that actually works well...

So if you want to help, go buy HDO1000, and we'll work together to debug it, time permits...
Putting head in a sand is no help to ACTUAL HDO users. Fixing bugs is...

And if we were Siglent shils (people involved into this FFT troubleshooting), we would sit back, laugh and enjoy this clusterfuck and NOT help Rigol fix stuff they don't know by themselves...

Robert is a nice guy, and I would like to help. You should find somebody to help too. It feels nice, it's a worthy cause..

[dreamcat4]

ok so in simple layman's terms, for these observed results:

which of the following *could be true i.e. we cannot be sure yet, maybe we can at least rule out some of these possible reasons?

* there is an issue with the analog linearity, i.e. adc or in hardware. that is not fixable

* there is a firmware bug, or other issue with the firmware that is either (as mentioned) possible windowing FFT, or misleading, not documented, not clear in the UI

* there was a problem taking measurements or in testing methodology, the assumptions or interpretation of what they mean given the setup + settings input into the machine

so for example you might answer 1+2, but not 3
or 1,2, or 3

Sorry this probably is not a very helpful quesitron. A better line of enquiry would be to consider what next different tests could be done, and/or with other sorts of test equipment.

Just as a novice I am asking to try better to understand the implications here. For example if it could be a firmware bug, then I guess it would make sense that Rigol should need to be asked / should answer if they will address in firmware. Things like this. If that isn't an overstep to be suggesting like that...

[Fungus]

It's all right to not know something..

FFT mirroring is a thing. Here's a Teletubbies video:

(skip to 5:33...)

But you're not helping HDO1000 users by pretending there is no problem.

I'm not pretending there's no problem, just that we need to know every last setting ... and the same settings tried on other devices.

And even if we presume basic math behind this FFT is correct, how do you use this thing?

You make sure that Nyquist is high enough for the signal you're looking at.

(Which you're supposed to do anyway)

[Fungus]

Question: What happens if the input signal is 99.5kHz?