Rigol HDO1000 and HDO4000 12bit oscilloscopes launched in China

[tv84]

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

I would bet it is mirrored! That should say something to those who know this stuff.

[Fungus]

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

I would bet it is mirrored! That should say something to those who know this stuff.

Yep. See video three posts up ^^^

[tv84]

Yep. See video three posts up ^^^

Sorry, I spoke too soon. Then I saw the rest of the posts.

Shouldn't the FFT stop at 100kHz if you have the 200kSa/s ? That would solve the picture at 200 kSa/s and save Rigol's honor. 

But then, why does the 400kSa/s show the same mirroring?   

These are the moments where you need an engineer next to the GUI programmers or in the betatesters group...

I wonder how such a bug passed the FFT QC tests?!? 

[rf-loop]

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

You can't deduce that from the data you've received so far.  I'm not surprised - for some reason.

[Fungus]

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

You can't deduce that from the data you've received so far.  I'm not surprised - for some reason.

I've been told I'm not allowed to deduce anything from forum posts or screen grabs.

I thought I'd ask a simple, direct question instead.

[mawyatt]

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

I would bet it is mirrored! That should say something to those who know this stuff.

Yes the display is mirrored around 100KHz, which implies a 200KSPS rate.

If one is viewing a center frequency of 100KHz, then the FFT should create a sample rate that supports a reasonable result around 100KHz, which 200KSPS is not. As mentioned this sample rate should not be allowed to be this low and the DSO should produce a much higher sample rate which evidently it does not, unless user intervention to "force" the lower sample rate and incur Nyquist sub-sampling artifacts, but hopefully knowing such artifacts will occur.

This is a serious FFT bug, that could easily fool a user into "thinking" these sub-sampled artifacts are real when in fact they are not.

scander36 has done a great service in showing these FFT results, which initially were to show the channel and ADC linearity by means of a Two Tone IMD signal and subsequent FFT, but this low sampling rate issue was uncovered.

Best,

[Fungus]

Shouldn't the FFT stop at 100kHz if you have the 200kSa/s ?

You mean simply display nothing on screen above that frequency?

That's a good question. Logic says "yes"... but it would look weird and I'm sure people would complain just as loudly. 

I don't know what other 'scopes do in this situation - I don't own them all. It seems suspicious that they claim an "FFT frequency" at double the sample rate so I'd guess "2x upsampling followed by a low pass filter" to try and minimize the effects of mirroring. If this is the case then there's a bug somewhere...

(yes, I know "2x upsampling" already involves a low pass filter so it can be a single step in practice...)

[mawyatt]

Shouldn't the FFT stop at 100kHz if you have the 200kSa/s ?

You mean simply display nothing on screen above that frequency?

That's a good question. Logic says "yes"... but it would look weird and I'm sure people would complain just as loudly. 

I don't know what other 'scopes do in this situation - I don't own them all. It seems suspicious that they claim an "FFT frequency" at double the sample rate so I'd guess "2x upsampling followed by a low pass filter" to try and minimize the effects of mirroring. If this is the case then there's a bug somewhere...

(yes, I know "2x upsampling" already involves a low pass filter so it can be a single step in practice...)

Here's as close as we can get to the Rigol HD at 100KHz with 200KSPS and 10K memory depth with the SDS2000X+ at 125KHz with 250KSPS and 16K memory depth.

Note, it does not display anything past 125KHz, immediately conveying to the user that this is "forbidden Nyquist frequency zone", and FFT display is not to be trusted!!

Edit: This makes one wonder if the Rigol HD uses the lower sampling rates to show a quicker user FFT response? Also, this shows that a significant amount of effort has been expended in the SDS2000X+, likely by some folks on here, to produce a quality FFT function & display that is user friendly and reliable.

After all, what good is a DSO function that can't be relied upon to produce reliable results?

Best,

[skander36]

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

Same thing.

It is more a memory lack problem.
This is happened only at 10kpts and lower. If you pay attention to the memory and you will intervene to alocate more memory (100kpts and over) this ghost will disapear.

[skander36]

Shouldn't the FFT stop at 100kHz if you have the 200kSa/s ?

You mean simply display nothing on screen above that frequency?

That's a good question. Logic says "yes"... but it would look weird and I'm sure people would complain just as loudly. 

I don't know what other 'scopes do in this situation - I don't own them all. It seems suspicious that they claim an "FFT frequency" at double the sample rate so I'd guess "2x upsampling followed by a low pass filter" to try and minimize the effects of mirroring. If this is the case then there's a bug somewhere...

(yes, I know "2x upsampling" already involves a low pass filter so it can be a single step in practice...)

Here's as close as we can get to the Rigol HD at 100KHz with 200KSPS and 10K memory depth with the SDS2000X+ at 125KHz with 250KSPS and 16K memory depth.

Note, it does not display anything past 125KHz, immediately conveying to the user that this is "forbidden Nyquist frequency zone", and FFT display is not to be trusted!!

Best,

Same on Rigol at even lower samplerate.

[2N3055]

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)

Why are you doing this? Now you want to derail topic about Rigol ...

Of course mirroring is a thing. You didn't say that. You said spectral leakage...Which is not same. That is why I said you should stay out of this discussion. You are just repeating some terms you saw on internet without understanding.
I don't need Teletubbies video for that, thank you. I know it from school...
OTOH, that video is actually not bad for someone new to topic...

Also it is obvious that user had great problems not being able to actual set FFT parameters effectively.
That is why there were dozens of posts...

[skander36]

Also it is obvious that user had great problems not being able to actual set FFT parameters effectively.
That is why there were dozens of posts...

If you are talking about me, I remember that I set the parameters as requested. This is why I request a specific setup.
If it was after me I can present a nice looking FFT and say : "Look how good it is!"

[2N3055]

Shouldn't the FFT stop at 100kHz if you have the 200kSa/s ?

You mean simply display nothing on screen above that frequency?

That's a good question. Logic says "yes"... but it would look weird and I'm sure people would complain just as loudly. 

I don't know what other 'scopes do in this situation - I don't own them all. It seems suspicious that they claim an "FFT frequency" at double the sample rate so I'd guess "2x upsampling followed by a low pass filter" to try and minimize the effects of mirroring. If this is the case then there's a bug somewhere...

(yes, I know "2x upsampling" already involves a low pass filter so it can be a single step in practice...)

FFT must not plot anything above Nyquist because there is nothing. That is mathematically correct.

Upsampling followed by lowpass filter is nonsense that doesn't achieve anything. Frequencies folding back are because of samples having specific sampled content at sample time... You cannot do anything afterwards. It's entropy, data is lost...

Why would 2X upsampling involve lowpass filter? Is there some new DSP opposite universe that I don't know about?

Could you please explain, because I'm genuinely confused about this?

[2N3055]

Also it is obvious that user had great problems not being able to actual set FFT parameters effectively.
That is why there were dozens of posts...

If you are talking about me, I remember that I set the parameters as requested. This is why I request a specific setup.
If it was after me I can present a nice looking FFT and say : "Look how good it is!"

Problem is that scope does not let you set these parameters the way FFT should be set.
Also you say problem is only with 10kpts or less..
But that is a big problem, because number of points shouldn't matter. FFT is routinely done on 1024 or 512 points for that matter..
Scope at one point shows both 200 ksps/s and 400ksps for FFT and shows wrong data. No setting should produce that. Where do you set FFT points ?? Is there a setting for that?

[2N3055]

Yep. See video three posts up ^^^

Sorry, I spoke too soon. Then I saw the rest of the posts.

Shouldn't the FFT stop at 100kHz if you have the 200kSa/s ? That would solve the picture at 200 kSa/s and save Rigol's honor. 

But then, why does the 400kSa/s show the same mirroring?   

These are the moments where you need an engineer next to the GUI programmers or in the betatesters group...

I wonder how such a bug passed the FFT QC tests?!? 

It should stop at 100kHz. And it is not happening all the time.. If you put scope in Auto memory mode, it will have high enough sample rate and you won't see this..  It wasn't tested properly, it was rushed to market... I refuse to believe Rigol engineers don't know this. They were not given enough time.

[skander36]

Also it is obvious that user had great problems not being able to actual set FFT parameters effectively.
That is why there were dozens of posts...

If you are talking about me, I remember that I set the parameters as requested. This is why I request a specific setup.
If it was after me I can present a nice looking FFT and say : "Look how good it is!"

Problem is that scope does not let you set these parameters the way FFT should be set.
Also you say problem is only with 10kpts or less..
But that is a big problem, because number of points shouldn't matter. FFT is routinely done on 1024 or 512 points for that matter..
Scope at one point shows both 200 ksps/s and 400ksps for FFT and shows wrong data. No setting should produce that. Where do you set FFT points ?? Is there a setting for that?

If you are suggesting that there are too few setting available, I agree with you, but this setup case was a particular one, with two tone combined, that revealed some limits on this aspect.
Rigol could make more settings available to the user.

[Fungus]

You mean simply display nothing on screen above that frequency?

That's a good question. Logic says "yes"... but it would look weird

Note, it does not display anything past 125KHz, immediately conveying to the user that this is "forbidden Nyquist frequency zone", and FFT display is not to be trusted!!

Cool!

(So Rigol's "fix" is easy...  )

[Fungus]

Of course mirroring is a thing. You didn't say that. You said spectral leakage...Which is not same.

It's been a while since I studied this. I got my terms mixed up. Sue me.

That is why I said you should stay out of this discussion. You are just repeating some terms you saw on internet without understanding.

Weird that none of the "experts" mentioned mirroring before I did.

Just sayin'...

[skander36]

... If you put scope in Auto memory mode, it will have high enough sample rate and you won't see this..

This is one of the problems, my scope was in AUTO memory mode, as i said I use most of the time auto setting, but maybe this is not a good habit. You should pay attention all the time at this parameter as it can change over a lower limit. The automatic memory management is not so reliable and in some cases, like this, it will not allocate the correct memory back to the necessary number.

[2N3055]

Of course mirroring is a thing. You didn't say that. You said spectral leakage...Which is not same.

It's been a while since I studied this. I got my terms mixed up. Sue me.

That is why I said you should stay out of this discussion. You are just repeating some terms you saw on internet without understanding.

Weird that none of the "experts" mentioned mirroring before I did.

Just sayin'...

No, don't be sorry. It is OK. Honest mistake is honest mistake.. it is grand of you to admit. Kudos!

Actually aliasing was mentioned several times. It includes "mirroring" or "folding" or downsampling... But I agree, it has been long and protracted over few pages.. again, easy not to connect...

[mawyatt]

Weird that none of the "experts" mentioned mirroring before I did.

Just sayin'...

Think GF mentioned "mirroring" in #687, and we mentioned aliasing earlier.

Best,

[dreamcat4]

what i am finding the most amusing about all this is we were simply just trying understand adc linearity / imd. and then trying to use this specific scope feature measurement around fft we hit up these bug(s).

so then you inevitably end up wondering if we had picked any other tests on other scope features at random we wouldn't have seen the same levels of bugginess / things yet to be fixed.

therefore am struggling to feel sufficient levels of trust to spend those 1k usd (for me a lot, for others not). not until a much later timeframe on this model. maybe to come back and see after +6 months, then another +12 months from now. right? so end of 2023.

in meantimes i keep a lookout for any possible rtb2004 deals. which of course are few and far between or simply dont exist. nearest i got to see was a full locked no options for 1300. but didn't seem worth it compared to other pricing, or to the have to pay for the missing options on top.

but my other hope is that siglent might decide to replace some of their current lineup with newer model. in response to these new rigol. whichever - am not too deep into choosing and 1 camp to ignore the others.

but what i liked was rudi's deep dive video series for comparing the r&s against the siglent. at least that gives me a frame of reference. he put a lot of time and effort into those video series + document.

just am not seeing the point to bash these rigol hdo series over it. not until their team have fully a generous extra time to fix and improve on what is by all accounts a yet to be finished product. but if they want my money... they're gonna have to work for a bit longer at it. at least they seem on the right track. it just depends how well they respond and will be addressing these issues.

[tv84]

It wasn't tested properly, it was rushed to market... I refuse to believe Rigol engineers don't know this. They were not given enough time.

I expect a rushed DSO to have GUI problems. I have difficulty accepting that an erroneous FFT implementation,  that is there since day 1, is due to the fact that it was rushed to market. Proper FFT limits should be in place since the beginning.

These guys were programming a DSO, not a drawing tablet. If the physics of signal processing isn't correct why bother with the menus layout? Or they started with the menus because the libraries existed already?

[Fungus]

Anyway ... I think it's clear what the problem is.

Also the workarounds: Always use a high sample rate, don't break Nyquist!

And what Rigol can do about it: Either fix their upsampling or, even better/simpler, truncate the FFT display.


Maybe @Dave could do a video calling them out. 

[tautech]

Maybe @Dave could do a video calling them out. 

Please no.
The ex Rigol seller found this.