Rigol's New DHO800 Oscilloscope unbox & teardown

[2N3055]

The offset alone is the smaller problem - if at all one would care about the offset drift.

For the accuracy it is hard to compare just the specs. This are not hard numbers that one can measure on a few units, especially not when they come out new. The numbers are more like rough estimates with possibly company and engenier related bias. For the scale factor one anyway often has 1:10 probe and the probe alone can have more uncertainty.

What may matter and could be measured would be the linearity: with a good sine wave look at the harmincs in the FFT. Even this may need a few instruments for averaging individual variation.
Good linearity may be relevant for using the 12 bit scope as a low cost spectrum analyser substitute. Here the extra bits at the ADC really help.

You are more than correct: linearity (and related distortions) are most important.
Also you are more than correct it is complicated.

That being said, scopes have no quick ZERO button to instantly remove 0 errors for measurement. So inaccuracy is a problem if you want to use it to do DC measurements. For long term drift (as equipment ages) you have self cal.

As for probes, there comes in fractional probe factor. Siglent supports it on touchscreen scopes, others too. So you can set your probe to be 10.03X if you calibrate it to be so and sneak up to very high gain accuracy.

Unlike meters, scopes have more complicated front end layouts with VGAs that makes accurate gain harder. More moving parts, so to speak and more permutations...

Logic is not exactly the same as for a meter.

[ebastler]

But in addition Dave also demo'd pretty well a couple of bugs or not working, [...]

However as for the unfixed bugs I do agree completely and that does indeed still remain the biggest concern for me overall. Especially when looking at how long the other DHO1000/DHO4000 series have already been out for.

Which bugs in the DHO800/1000/4000 are you mostly concerned about? I might have missed something since I fast-forwarded through a few sections of the review video. The issues I saw were more along the lines of "wouldn't it be nice...", e.g. no possibility to use more of the vertical screen real estate for trace display. Plus the crash of the UI, of course, although at least with graceful recovery.

[Kleinstein]

What we should see in practice then?
For example with by 50ohm shorted BNC channel input at 1mV range (at a specific sample rate and BW).
What AC.RMS (or standard deviation or peak-peak) is to expect with 8bit and 12bit scopes?
Or what 12bit/8bit ratio of the noise should we see?
So far - based on the videos - I do not see the ratio to be really 12bit/8bit..

PS: for example my borrowed 20y old DS1062CA shows 440uVpp noise terminated with 50ohm (1mV/div range, BWlimit ON, 100ms/div timebase)..

The noise test with maxium gain is mainly looking at the amplifier noise. Chances are the ADC noise is relatively unimportant in this case.
To look at the ADC part and compare different ADC one should look at a range like 20 mV/div or maybe 100 mV/div, just before the relay kicks in the extra input divider.  I would not expect the full extra 4 bits or a ratio of 16 - that may happen to some of the early more noisy DSOs. I would definitely hope for better than 8 bit ENOB. There is still the question with the sampling rate and it may depend.

[2N3055]

What we should see in practice then?
For example with by 50ohm shorted BNC channel input at 1mV range (at a specific sample rate and BW).
What AC.RMS (or standard deviation or peak-peak) is to expect with 8bit and 12bit scopes?
Or what 12bit/8bit ratio of the noise should we see?
So far - based on the videos - I do not see the ratio to be really 12bit/8bit..

PS: for example my borrowed 20y old DS1062CA shows 440uVpp noise terminated with 50ohm (1mV range, BWlimit ON, 100ms/div)..

This is something that was explained in 12 bit Siglent postings but somehow gets forgotten when discussed here..

Firstly, there is input referred noise of input amplifiers (analog front end). Analog noise is connected to BW. More BW more noise.

Then there is ADC noise that is also partly analog noise (input buffers) and quantization noise (noise introduced while digitizing). Quantization noise will be related to ENOB (effective number of bits).

You can have 8 Bit scope with lover noise than 12 bit scope, at low ranges, for instance if you pair very low noise front to very high  ENOB (for a 8 bit) ADC compared to high noise front end connected to 12 bit ADC with not so stellar ENOB.

ENOB is always less than architectural resolution. Here it is that good 8 bitters will have 7.5 Bit ENOB, but good 12 bitters will have 10-10.5 bits.

And here comes the interesting part.  Noise is about lover part of signal. Hence term of noise floor. Basically, noise floor is a measure what smallest signal we can measure reliably at current RANGE of vertical sensitivity.

At large V/div, where analog front end contributes little noise, noise will be dominated by quantization noise of ADC and basically equal to it's ENOB. As you go more sensitive, front end will contribute more to noise, to the point of it being dominant source of noise...

At 1mV/div (or less) there won't be 4X noise floor difference between 8 and 12 Bit that ADC ENOB difference would suggest.

So for measuring very low signals, preamp quality comes first.

What 12 bit gives you is not ultimate low noise at low end (that is function of preamp) but dynamic range (that is how ADC are calling it anyways). If for instance , in FFT, 8 bit scope can measure 1mv signal clearly at 100mV/div, a 12 bit scope can do the same at say 400mV/div.

That means that at the same time, on same screen 8 bit scope can show both 1 mV and 390mV peak, but a 12 bit one will be able to show 1mV and 1,58V signal at the same time. But that is FFT, frequency domain look.

In time domain, waveform display standard scope mode with 12 bitter (compared to good low noise 8 bitter) you will see same trace at high V/div,  a nicer thinner trace in some intermediate ranges and increasingly same thickness of traces as you go to sensitive ranges (where preamps noise dominates), provided similar preamps.

[iMo]

So, in order to characterize the 8bit vs. 12bit we have to have some testing procedure, taking into the account the above. Hopefully the designers tried to lower the noise of the AFE when applying their new 12bit ADCs.

We need perhaps to do FFT (also mind the various FFT params, like number of points, windowing function, sample rate, etc) at the minimum voltage range (like the 1mV/div) with say 100uVpp signal (like sine 100uVpp with a frequency = samplerate/8), and then at higher range, like 100mV/div (or higher) again with 100uVpp signal, all done at 50ohm. And use the ratio of the noise floors and the ratio of signals amplitudes for scope A vs. scope B as a "measure". Or something of that kind..

[dreamcat4]

well yes, this is what i was alluding to in my prior (but too-brief) comment. that dave's demonstration of fft mode shows how it generally performs in that mode compared to the other scopes on his table.

and that fft (for dsos) is a bit of a shortcut to stress test the general or total / combined noise performance of a scope. without examining too deeply, that it is easy domain to throw a variety of signals, both high and low ranges. and see how well it performs.

i don't presume to think it's going to replace a more thorough and comprehensive testing. but it does give a valuable insight into the combined performance of multiple stages put together. and help to point to potential signs of weakness (to then mull over / think about, and consider reasons for why).

again it's not meant to be a replacement for a comprehensive testing approach, of each individual pieces and functions.

[nctnico]

Did nobody else watch Dave's video? No comments on the pulse response vs. the Siglent? How about the FFT speed?
These things are supposed to be the Siglent's differentiators vs. the "toy" oscilloscopes that Siglent owners love to turn up their noses at but this new Rigol is much better.
PS: Looks like Siglent also gets the "fat traces" award now.

Same conditions too, the Leo Bodnar pulse gen connected direct to the scope 1M input.
And yes, the Rigol traces are thinner for the same bandwidth and sample rate. I'd love to have a full screen waveform option to make use of the resolution available. And if they could map it to the full HD HDMI output that would be mind blowing.

Trace thickness differences can be due to different display resolution and rendering. Some DSOs fatten the traces up deliberately so they are visible on the display.

[nctnico]

Personally I'm missing accuracy tests. Not all 12-bit scopes are the same and accuracy in measurements (and offset) is important. This is usually what makes this huge price differences in ADCs!

The DHO800 has 1% accuracy, where the SDS2000 HD has 0.5% for example. Offset error is also about double.

The SDS2000 HD is also ~5x the price 

Yes, and as I mentioned this is part of the price differences in ADCs, and a price I'm willing to pay because I'm interested in overall accuracy..

Just don't get into the false idea that all 12 bit scopes are the same and Rigol has somehow beat everyone here.

Agreed. What would be interesting, is a comparison between ENOB versus frequency for various 12bit scopes. For 12 bit resolution at high samplerates you need an extremely low jitter clock. I strongly doubt any of the 12 bit scopes reaches 12 bit ENOB at sample rates over a couple of hundred MHz.

Another factor that comes into play is the trigger jitter. If you want to look at small signals and clean them up a little bit, then averaging will depend heavily on having a low trigger jitter.

[nctnico]

That means that at the same time, on same screen 8 bit scope can show both 1 mV and 390mV peak, but a 12 bit one will be able to show 1mV and 1,58V signal at the same time. But that is FFT, frequency domain look.

Not necessarily FFT only. That would defeat the purpose of having 12 bits in the first place. IMHO 12 bit should be usefull to get more detail from a signal. Recently I did some measurements using a 10m Ohm current shunt using my 12 bit Yokogawa scope.



Vertical zoom is 100x and probe attenuation factor is set to 100. So 1mV = 1mA. The effective sensitivity is 50uV/div. So I'm measuring an 8mA pp signal using a 10milli-Ohm shunt here.

[2N3055]

That means that at the same time, on same screen 8 bit scope can show both 1 mV and 390mV peak, but a 12 bit one will be able to show 1mV and 1,58V signal at the same time. But that is FFT, frequency domain look.

Not necessarily FFT only. That would defeat the purpose of having 12 bits in the first place. IMHO 12 bit should be usefull to get more detail from a signal. Recently I did some measurements using a 10m Ohm current shunt using my 12 bit Yokogawa scope.



Vertical zoom is 100x and probe attenuation factor is set to 100 so the effective sensitivity is 50uV/div.

I was simplifying for like to like comparison.

How much can you expand signal captured and stopped will be, like I said, connected to what is ENOB at that particular range.
Most gain, as also I pointed out numerous times before, will be in situations where we are looking at signal that had large dynamic range (large difference between highest peaks and smallest details we would like to see) and at ranges where ENOB of the scope is maximized.

Situations where we want to look at small signal riding on top of a DC or slower moving other signal, for instance.

I would again remind that I disagree with usage of word "zoom" or "zoom in" or "zoom out" for a practice where you change vertical sensitivity or time base while looking at signal in STOP mode.

It confuses the heck out of people because they confuse it with two other uses of word zoom:
- dedicated Zoom view mode (most scopes have that in different implementations).
- display mode where in main display window you set it to show only part of vertical range that is actually captured. Some R&S scopes support that.

I would use phrase "magnify" for that practice, or something in that venue..  I'm not insisting on it but just pointing out history of confusion with this practice..

[nctnico]

I think you'll have a hard time to find consensus amongst the oscilloscope manufacturers. For example:  the image I posted is a zoom window while Yokogawa is using both 'zoom' and 'magnification' for looking at a signal in more detail. And there are more terms which are not consistent between various oscilloscope manufacturers so in the end you'll need to refer to the manual.

[Njk]

BTW, concerning the topic, only the ADC (RT1642IQ) is registered with the China national IP administration. The two other chips of the "chipset" are not registered (or registered not on Rigol's account).
https://www.cnipa.gov.cn/art/2023/7/14/art_164_186285.html?xxgkhide=1

[Martin72]

Agreed. What would be interesting, is a comparison between ENOB versus frequency for various 12bit scopes. For 12 bit resolution at high samplerates you need an extremely low jitter clock. I strongly doubt any of the 12 bit scopes reaches 12 bit ENOB at sample rates over a couple of hundred MHz.

Siglent and lecroy give values between 8.4 and 8.7, depending on the bandwidth.
Rigol simply writes "greater than eight"...
While user "Wolfgang" had determined the ENOBs for the Rigol DHO4000 and R&S MX04 on his website, at 10Mhz:

https://electronicprojectsforfun.wordpress.com/to-enob-or-not-to-enob/

[mawyatt]

In DHO series only ASIC is analog front end ASIC. Their ADC is not ASIC as much as local ADC source. That ADC can be replaced with any other ADC, and if you get good one at a good price, you can match them.

Actually believe the Rigol 12 bit ADC is more of an Application Specific ADC than not. Reasoning is that Analog Devices and TI ADC design targets a much broader range of applications than just the DSO, whereas Rigol may have reduced the initial design requirements, efforts, and cost with just the DSO product target. This allows relaxed specifications for the ADC in many areas and thus reduces the initial design task and costs. Of course we don't know for sure, and don't generally like to speculate, but do have some background here and some "hints". Would be very interesting to find the actual "wafer yield", but that's usually a highly regarded parameter and not going to be disclosed by Rigol or the IC fab.

Rigol still needs to utilize a boatload of these ADCs to recover the NRE, and likely the DSOs under discussion are the recipients of the lower grade ADCs from the chip testing fallout, where the higher grade ADC chips go the higher class DSOs.

Sure one could replace the Rigol ADC with a TI or AD 12 bit ADC and be done, but this would have a higher Recurring Cost and not achieve this attractive price point IMO.

ENOB, Linearity and so on will revel just how good these ADCs are. However, without a better FFT as R&S and Siglent have implemented, some of this might be difficult to properly access.

All this might not matter to the audience these lower cost DSOs are targeting and just having a 12 bit ADC with low noise front end, good screen resolution, and HDMI is enough. Being able to "see" finer waveform details and a nice Hi-Rez display may be all that matters to many whether they are "real" or accurate isn't important!!

A good set of tests would be in order. The classic 2-Tone IMD and another interesting test would be to utilize the math functions to evaluate precision waveforms and view the numerical results. Things like waveform average, RMS(SD), DC Levels and so on might hint at just how good the input channels behave over the ranges. We did such when we got our 1st SDS2000X+, and were pleasantly surprised how well it performed.

Anyway, fun discussions going on.

Best,

[iMo]

..Vertical zoom is 100x and probe attenuation factor is set to 100. So 1mV = 1mA. The effective sensitivity is 50uV/div. So I'm measuring an 8mA pp signal using a 10milli-Ohm shunt here.

While looking at your picture - the p-p noise on top of your signal is something like 5uVpp. That is ~1uVrms..  With a typical BW say 20MHz it is something hard to believe..

[iMo]

..ENOB, Linearity and so on will revel just how good these ADCs are. However, without a better FFT as R&S and Siglent have implemented, some of this might be difficult to properly access..

You may export the data in a csv afaik, and do the FFT in the PC..

[nctnico]

..Vertical zoom is 100x and probe attenuation factor is set to 100. So 1mV = 1mA. The effective sensitivity is 50uV/div. So I'm measuring an 8mA pp signal using a 10milli-Ohm shunt here.

While looking at your picture - the p-p noise on top of your signal is something like 5uVpp. That is ~1uVrms..  With a typical BW say 20MHz it is something hard to believe..

For that particular picture I had averaging enabled but the bandwidth is also limited to several tens of kHz. The Yokogawa oscilloscope I used for this measurement is specifically targetted at relatively low frequency & high resolution measurements. I just wanted to highlight that having '12 bit' on the badge is only part of the story.

[ptluis]

I think it's a waste of time and mental resources to debate which brand is best, because the truth is that there is no perfect equipment.  One may do better than the other at a certain task, or have more functionality.  I'm not a fan of any, I want equipment that is as functional as possible when I need to use it in a certain area.  I own several brands and in fact I never used just one, because as I said before, if the X oscilloscope performs the current task better, that's the one I choose.  Brands are complements to each other.

[Fungus]

Some DSOs fatten the traces up deliberately so they are visible on the display.

Which ones?

[Fungus]

Knowing Rigol's history, there will be tons of bugs that takes years to fix. People get all exited about new equipment and then are super happy with every firmware release that fixes problems/bugs that should not have been there in the first place. ... I'm sceptical.

When do you estimate Siglent will get around to fixing their FFT? Dave saw some serious problems with it in his video.

See video at 32:17: https://youtu.be/S8jrpCoZyx8?t=1937

(and look at how slow the Siglent FFT is by comparison...    )

[mawyatt]

..ENOB, Linearity and so on will revel just how good these ADCs are. However, without a better FFT as R&S and Siglent have implemented, some of this might be difficult to properly access..

You may export the data in a csv afaik, and do the FFT in the PC..

Yeah, but nicer to just have the results displayed directly on the DSO. We tend to prefer instruments that don't require a PC for use, and have a nice 16 bit Pico 4262 that rarely gets used because of such.

Best,

[nctnico]

Knowing Rigol's history, there will be tons of bugs that takes years to fix. People get all exited about new equipment and then are super happy with every firmware release that fixes problems/bugs that should not have been there in the first place. ... I'm sceptical.

When do you estimate Siglent will get around to fixing their FFT?

To be honest: I don't care. I don't see myself buying anything from Siglent again.

[Njk]

Yeah, but nicer to just have the results displayed directly on the DSO. We tend to prefer instruments that don't require a PC for use, and have a nice 16 bit Pico 4262 that rarely gets used because of such.

That's the famous Rigol FFT. But the video shows that the noise grass is somewhere between -100 and -110 dB, which is not bad, I think. Better than with a 8-bit scope. BTW if one is going to come up with the numbers that quantitively characterizes the analog part of the scope, it'll make sense to upload the points to PC for more close inspection. Otherwise it's difficult to objectively compare it with that of the competitors.

[2N3055]

Knowing Rigol's history, there will be tons of bugs that takes years to fix. People get all exited about new equipment and then are super happy with every firmware release that fixes problems/bugs that should not have been there in the first place. ... I'm sceptical.

When do you estimate Siglent will get around to fixing their FFT? Dave saw some serious problems with it in his video.

See video at 32:17: https://youtu.be/S8jrpCoZyx8?t=1937

(and look at how slow the Siglent FFT is by comparison...    )

Are you really that thick or you are paid to spew bullshit?
What is wrong with that FFT? It is slower because device has slower CPU. That is not a bug.
A 8 bit converter dynamic range is bug??

Muppet.

[Fungus]

When do you estimate Siglent will get around to fixing their FFT? Dave saw some serious problems with it in his video.

See video at 32:17: https://youtu.be/S8jrpCoZyx8?t=1937

Are you really that thick or you are paid to spew bullshit?
What is wrong with that FFT?
Muppet.

One of these FFTs isn't like the others...can you see which one?




Screenshots from Dave's video at the time I linked to above.