Products > Test Equipment
REVIEW - Rigol DS2072 - First Impressions of the DS2000 series from Rigol
<< < (249/566) > >>
marmad:

--- Quote from: Galaxyrise on June 26, 2013, 05:53:39 am ---It's almost certainly working as intended.  I wouldn't call it a bug, but it is deceptive.

--- End quote ---

Just to elaborate: the way that the DSO SHOULD be implementing anti-aliasing is with oversampling which is then randomly decimated before display to avoid the appearance of a false low frequency component - or if the sample length is already deep, just random decimation when displayed.
Galaxyrise:

--- Quote from: marmad on June 26, 2013, 10:45:35 am ---Just to elaborate: the way that the DSO SHOULD be implementing anti-aliasing is with oversampling which is then randomly decimated before display to avoid the appearance of a false low frequency component - or if the sample length is already deep, just random decimation when displayed.

--- End quote ---
Right, we're agreeing with each other.  I'm saying the DS2000 seems to do the second part; it does sample->display randomization or something similar.  But it doesn't do the much more important signal->sample randomization.
Galaxyrise:

--- Quote from: rf-loop on June 26, 2013, 09:56:47 am ---Tell to Tektronix they have done it wrong. It is more like they have defined world of oscilloscpes (in history) and others have then followed....

--- End quote ---
All I've found from Tek so far is "In hi-res mode, the data is significantly oversampled, and then a boxcar average is performed in acquisition hardware to real-time average".  This implies to me that the averaging happens before data is written to sample memory, but it's still a little vague.

From reading this Agilent app note,  they state the ADC in their example is reading 20GSa/s but the waveform is 2.5GSa/s.  That sounds like what I was expecting: averaging in acquisition. 

This is not what the Rigol does. I don't think the DS2000 can oversample; that if the Rigol ADC is making 2G readings/s, then it's writing 2Gsa/s to sample memory.  Rigol's High Resolution averaging happens when displaying the sample memory. This is how Rigol achieves 2GSa/s, by moving everything to sample memory post-processing.

Rigol is approximating features of the big boy scopes, so they named them the same and put them in the same place... but they're not the same! And their approximation of Anti-Alias is nearly useless since their "vectors" algorithm already does a decent job preventing sample->display waveform aliasing. Calling them the same thing is definitely deceptive.
Wim13:

--- Quote from: Teneyes on June 26, 2013, 01:36:18 am ---I know the meaning of AC and DC coupling for the Signal input
But is the function of AC and DC coupling on the Trigger settings independent of the Channel signal coupling on the DS2000?

What happens with the Signal DC coupled, and Trigger AC coupled?

What happens with the Signal AC coupled, and Trigger DC coupled?

With the Trigger set to AC coupling;
the Trigger level (edge) varies with the trigger level knob, (orange voltage value)
but there is No orange Line --------------------.
Is that because the triggering subsystem cannot know the absolute DC level to put the line on the Display?  :-//

Am I explaining that OK.

--- End quote ---

Trigger on AC removes the DC component, example:

DC +10 Volt + sinus 1 Vtt,  then trigger level is about +10 volt at DC
lot of turning the trigger knob to 10 volts

if you turn trigger to AC, the trigger will trigger at 0 Volt, removing the DC component of 10 Volts.
you can turn trigger on and push the trigger button for 0 Volt.

So in AC mode the trigger signal is not the same level anymore as the displayed signal.
The DSO of course does not know the difference of the AC + DC componont of the signal, so
the trigger in AC mode does not know where the desired trigger level is on the screen. And the
orange line is meaningless.

marmad:

--- Quote from: Galaxyrise on June 26, 2013, 03:24:55 pm ---Right, we're agreeing with each other.  I'm saying the DS2000 seems to do the second part; it does sample->display randomization or something similar.  But it doesn't do the much more important signal->sample randomization.

--- End quote ---

No, we are not agreeing with each other. As far as my tests show, the Rigol does NOT do anything similar to what I mentioned at all.


--- Quote from: Galaxyrise on June 26, 2013, 03:56:02 pm ---This is not what the Rigol does. I don't think the DS2000 can oversample; that if the Rigol ADC is making 2G readings/s, then it's writing 2Gsa/s to sample memory.  Rigol's High Resolution averaging happens when displaying the sample memory. This is how Rigol achieves 2GSa/s, by moving everything to sample memory post-processing.

Rigol is approximating features of the big boy scopes, so they named them the same and put them in the same place... but they're not the same! And their approximation of Anti-Alias is nearly useless since their "vectors" algorithm already does a decent job preventing sample->display waveform aliasing. Calling them the same thing is definitely deceptive.

--- End quote ---

Sorry, man, but it seems as if you don't completely understand high-res or anti-aliasing.

Vectors on the Rigol don't do a damn thing towards preventing aliasing - as evidenced by many images already posted here. If you think they do, post an image which demonstrates this. Their Anti-Aliasing does NOT work - and is either a bug, unimplemented feature, or mistake.

But their High-Res works fine:

1) It's not approximating anything - it's doing High-Res PERFECTLY correctly. It's just math - whether you do it while sampling or after-the-fact - it doesn't make a bit of difference. No low cost DSO samples faster than their max. sample rate - they reduce the effective sample rate (and bandwidth) to deliver the simulated bits of resolution (see attached chart from Agilent 2000X manual). BTW, the Agilent is ALWAYS sampling at 2GSa/s - it just throws out samples instead of actually reducing the rate. And when you ask it to do High-Res, it just averages the samples it would have thrown away otherwise.

2) It's preferable to have post-processing done on data being moved to display memory - because then the sample memory contains the ORIGINAL samples! Which is want I want - I don't want already processed samples in memory! Why on Earth would you think that's better? I can always average the original samples again any time.

Navigation
Message Index
Next page
Previous page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod