Rigol's New DHO800 Oscilloscope unbox & teardown

[Martin72]

One less parameter to juggle. 

You can´t set the size "manually", like on other scopes.

[nctnico]

One less parameter to juggle. 

You can´t set the size "manually", like on other scopes.

You can, but you drive the number of FFT points through the RBW setting on the Rigol (this is not uncommon, FFT on an R&S scope works in a similar manner). Setting the RBW at least gives you some information about what to expect from the FFT in terms of frequency resolution.

[Fungus]

If you don't understand why it is necessary to have control over number of bins, that is proof you need to learn a bit more before you are qualified to discuss quality of FFT implementation.

To make it clear, there is nothing wrong if scope has auto set of number of bins. It can really be nice for some quick look-see. But it has to have manual control too. If you are doing anything that you have to compare two results you need to be able to make parameters same between two measurements otherwise you cannot compare them 1 to 1.

Number of bins usually has to be "as many as possible".

You can easily compare two measurements by making sure the same "resolution" is shown at the top of the screen. That's based on the sample rate so it can be controlled just as manually as the number of bins can. Maybe it's not the way you're used to working, but it can be done. To me it seems less of a problem than Siglent's "zoom" philosophy but Siglent owners will defend that one to death.

(b): The proof of the pudding is in the eating! The Rigol seems to be comparing favorably with a monster R&S in that little shootout.

(c) Pending Dave's confirmation, there definitely seems to be something wrong with the Siglent. I'm very sure he'll have triple-checked the FFT settings before posting that video. Place your bets.

[Fungus]

You can, but you drive the number of FFT points through the RBW setting on the Rigol (this is not uncommon, FFT on an R&S scope works in a similar manner). Setting the RBW at least gives you some information about what to expect from the FFT in terms of frequency resolution.

Yes, it's much easier to just adjust the horizontal timebase and see the resolution visually than to set a number of bins in a menu somewhere then try and find a timebase that works well with that number of bins.

Like I said: One less parameter to juggle (and possibly get wrong).

This seems like a sensible decision on an entry level 'scope. That together with the pinch-zoom and touch screen movement probably means you hardly ever have to use the setup dialog.

[2N3055]

author=Fungus link=topic=385309.msg5077336#msg5077336
Number of bins usually has to be "as many as possible".

No,no and no. You use as much as you need.
For instance, you might have results from a simulation on a PC. And you set scope to as close settings as you can to compare....
If you are trying to plot BW plot you might want wider bins to get a nice looking plot instead of "comb" to make measurements with cursors easier..
And many other reasons..

Look it is similar to multimeter autoranging. Autoranging IS useful for general probing. But there is no meter that does not allow you to lock manual range. Because when you need it you just do..

[Kleinstein]

Where is the size setting ?

One less parameter to juggle. 

What really counts is the resolution.

Every time I start to feel bad that I said something bad to you, you keep doubling down on wrong stuff..

If you don't understand why it is necessary to have control over number of bins, that is proof you need to learn a bit more before you are qualified to discuss quality of FFT implementation.

To make it clear, there is nothing wrong if scope has auto set of number of bins. It can really be nice for some quick look-see. But it has to have manual control too. If you are doing anything that you have to compare two results you need to be able to make parameters same between two measurements otherwise you cannot compare them 1 to 1.

With the FFT resolution they have it the wrong way around: the FFT resolution is set by 1 over the recording time. So with a fixed number of points (e.g. 1 M maximum) the higher the sampling rate the shorter the interval and the lower the resolution. It can thus make sense to intentionally reduce the sampling rate to get even better resolution for the FFT. With 1 M points one usually has quite good resolution, but in extreme cases one may still want more. If not directly as a parameter one could use the horizontal scale to adjust the sampling rate - though this may interfere with a parallel display in the time domain.

[2N3055]

Where is the size setting ?

One less parameter to juggle. 

What really counts is the resolution.

Every time I start to feel bad that I said something bad to you, you keep doubling down on wrong stuff..

If you don't understand why it is necessary to have control over number of bins, that is proof you need to learn a bit more before you are qualified to discuss quality of FFT implementation.

To make it clear, there is nothing wrong if scope has auto set of number of bins. It can really be nice for some quick look-see. But it has to have manual control too. If you are doing anything that you have to compare two results you need to be able to make parameters same between two measurements otherwise you cannot compare them 1 to 1.

With the FFT resolution they have it the wrong way around: the FFT resolution is set by 1 over the recording time. So with a fixed number of points (e.g. 1 M maximum) the higher the sampling rate the shorter the interval and the lower the resolution. It can thus make sense to intentionally reduce the sampling rate to get even better resolution for the FFT. With 1 M points one usually has quite good resolution, but in extreme cases one may still want more. If not directly as a parameter one could use the horizontal scale to adjust the sampling rate - though this may interfere with a parallel display in the time domain.

Thank you!
Yes, you need separate control of timebase (sampling speed) and number of bins..

[nctnico]

Where is the size setting ?

One less parameter to juggle. 

What really counts is the resolution.

Every time I start to feel bad that I said something bad to you, you keep doubling down on wrong stuff..

If you don't understand why it is necessary to have control over number of bins, that is proof you need to learn a bit more before you are qualified to discuss quality of FFT implementation.

To make it clear, there is nothing wrong if scope has auto set of number of bins. It can really be nice for some quick look-see. But it has to have manual control too. If you are doing anything that you have to compare two results you need to be able to make parameters same between two measurements otherwise you cannot compare them 1 to 1.

With the FFT resolution they have it the wrong way around: the FFT resolution is set by 1 over the recording time. So with a fixed number of points (e.g. 1 M maximum) the higher the sampling rate the shorter the interval and the lower the resolution. It can thus make sense to intentionally reduce the sampling rate to get even better resolution for the FFT. With 1 M points one usually has quite good resolution, but in extreme cases one may still want more. If not directly as a parameter one could use the horizontal scale to adjust the sampling rate - though

Actually, one should use the horizontal scale (timebase) to set the base FFT frequency span by forcing a certain samplerate. No matter how the FFT is controlled through the GUI, that is the most crucial setting (next to forcing the memory length if there isn't a seperate setting for the number of FFT points).

[Njk]

I assumed you knew what you were talking about but I just double-checked the manual and all the parameters seem to be there. Which ones are missing?

Basically, the FFT settings are the same as in DS1054Z, nothing is new except more potent hardware. Where are a features like Spectrum View (https://www.tek.com/en/blog/new-approach-frequency-analysis-oscilloscopes-part-1)?

[mojorizing]

The UltraAcquire mode looks like a poor-man's seqmented memory mode. It's mentioned in Dave's Full Review around the 40 minute mark.  This from the DHO4000 manual, page 60:

The UltraAcquire mode is not available when any of the following functions is enabled:
cursors, decoding, Search, Zoom, Pass/Fail test, waveform recording, power analysis,
reference waveform, roll mode, slow sweep mode, and XY mode. To switch to the
UltraAcquire mode, please ensure that all of those functions are disabled.
• In UltraAcquire mode, the functions mentioned above (except for roll mode and slow
sweep mode) are disabled. When roll mode or slow sweep mode is enabled, the
acquisition mode is automatically switched to "Normal".

[Martin72]

It´s a marketing gag with nice looking graphics.

[mawyatt]

Thank you for summarizing it much better than what I ever could.

As for my "not really ASIC" comment for Rigol ADC I thank Asmi and you for education. I'm not in ASIC business so my use of terminology will not be correct. What I wanted to say is that while Rigol's ADC obviously is ASIC as a type of product because it is custom made IC that was made by company for their specific use, ASICs are usually made for a very specific function: a switching matrix for a high speed interconnect, a complete integrated sound engine for a synthesizer, or something like that.
In this case, Rigol made "just another ADC", a chip for which there is alternative. Of course that assessment is based on my limited knowledge from what Rigol has published. Maybe they did combine ADC and some sort of DSP to linearize response or something like that.. In which case it would be really single use application specific chip for scopes that could not be replaced easily. And would allow them to have simpler BOM and maybe make significant saving in FPGA resources and such...

But from what I see they still have same architecture as before with ADC-FPGA-App processor. So it seems any saving would be just in price of chip. I personally think it is more about control of sourcing the chip than good price (which is just a bonus).

You are welcome!!

The ADC, or FPGA, is likely the single most expensive IC component in the DSO, and at this price point every $ in BOM matters. Without a custom in-house ADC, one is restricted to available chips from AD, TI and maybe some others, and don't think that anyone can get these ADCs at a price that would fit this price point.

We must remember that TI, AD and others have developed and produced high end performing ADC chips that satisfy many market segments, these incurred much higher development costs, and likely occupy more precious silicon area, involve extensive testing and so on. Whereas Rigol was able to isolate the important parameters that are more meaningful to the DSO market, trim the architecture to reduce the up front development costs, reduce the Si area, and so on.

Would wager that if we got ahold of one of these ADCs where we could do detailed isolated standalone chip testing, the overall performance would be mediocre in general compared to TI and AD. Things like no missing codes, monotonicity, differential & integral linearity, ENOB, aging, drift, temp, PSRR, and so on would not compare well with the TI and AD counterparts. Many of these parameters aren't that important in a DSO application, I mean who really cares if the ADC has a stuck lower bit, or non-monotonic around the LSBs, and so on. Now consider using this ADC as feedback in a complex high speed control system, stuck bits and non-monotonic behavior would be a disaster indeed!!

I strongly doubt that. For use in a DSO you need a very good ADC as well. Maybe even more so compared to a control system. Otherwise you'll get all kinds of weird distortions that will show up in a signal. In a control system you can work around it through signal processing (calibration) but when visualising a signal (either time or frequency domain), it is likely that faults do show up and there is very little you can do to cover things up.

Obviously you have no clue about feedback systems and employing ADCs as the feedback mechanize mechanism. On high gain high precision feedback systems almost the entire system precision performance is dictated by the feedback, not the forward path, this is like typical Op-Amp territory. A little control theory will show this is true and if one involves a high speed ADC as the "feedback" mechanism, then almost the entire system performance is dictated by that very ADC in the feedback, far more demanding than the open-loop use of a DSO.

Now consider the SDR, this really leverages the ADC, and most of the radios performance is dictated by the core ADC. Also the SA use, the linearity is paramount and one must remember not all SA utilize the simple FFTs as these DSOs employ, many utilize various DSP modes including the Chirp Z Transform, blends of hardware and so on.

So if one thinks that these DSOs are a demanding application for an ADC, they really need to go back and review where ADCs are utilized in todays world.

Best,

[mawyatt]

I don't know what Rigol has done to get their 'own' ADCs. It could even be that they are buying dies from TI or AD and put these into their own packages with an easier to route pin order. Or maybe they wanted some pre-processing or an interface that allows for a lower BOM cost at the FPGA side. If you look at the ADCs from TI, you'll see that these typically have a JESD204B interface. This uses several high speed, differential pairs to transport the data. At the FPGA side you'll need expensive transceivers. Looking at Dave's teardown, it seems that the ADC Rigol uses has a parallel style interface. Likely a DDR style interface using the standard SERDES inside the FPGA I/O blocks which is a much lower cost solution IF you can place the FPGA close to the ADC.

Rigol wouldn't be advertising their custom 12 bit ADCs if they were simply repackaging a lower grade ADC from AD or TI, this would lead to total embarrassment in the marketplace when discovered!! They may have employed TI, or AD to consult in the ADC design, but this is unlikely as TI and AD usually don't get involved in this for their key IP products. We're reasonably confident that Rigol did develop their own ADC, maybe with some "outside" help, or "commissioned" said development, but will leave it at that!!

Anyway, Rigol has introduced a low cost DSO employing some custom ICs that has created lots of interest and discussion, so we are all in for a treat as this emerges!!

Best,

[Fungus]

We're reasonably confident that Rigol did develop their own ADC, maybe with some "outside" help, or "commissioned" said development, but will leave it at that!!

The fact that it took them a few iterations to get it working at this level also supports this, ie. that they didn't simply buy an ADC from somebody..

[Fungus]

The UltraAcquire mode looks like a poor-man's seqmented memory mode. It's mentioned in Dave's Full Review around the 40 minute mark.

UltraAquire has a few interesting tricks.

This video compares both modes (normal segmented and UltraAcquire):

[mawyatt]

We're reasonably confident that Rigol did develop their own ADC, maybe with some "outside" help, or "commissioned" said development, but will leave it at that!!

The fact that it took them a few iterations to get it working at this level also supports this, that they didn't simply buy an ADC from somebody..

Yep, pretty obvious to those "skilled in the art" that this is true!!

Best,

[Fungus]

Looks like there's a firmware update this week:

https://www.eevblog.com/forum/testgear/new-rigol-hdo1000-12-bit-dso-bugs/msg5077489/#msg5077489

(that's for the 1000-series but it will probably appear for the 800/900 too)

[2N3055]

The UltraAcquire mode looks like a poor-man's seqmented memory mode. It's mentioned in Dave's Full Review around the 40 minute mark.

UltraAquire has a few interesting tricks.

This video compares both modes (normal segmented and UltraAcquire):

It is just a visual mode. None of measurements decoding or anything really of what analysis it does have does not work in either of the modes...

[PELL]

Ummm, I don't know what you guys are arguing about, but I just got my shiny new DHO804 and have some fun with it. 

Got this baby at 2299 RMB (320 US dollars) 

[MathWizard]

I watched Dave's review, and if I didn't already have some scopes, I'd probably get 1. The DHO4000 is out of my price range, so is the SDS2000HD. I'll get some 12-bit scope down the road, so probably the DHO1000 or similar Siglent 1000HD

[iMo]

..I also wonder why some of you are so obsessed with FFT. What is it good for (in this category)? Many old scopes have it too. Practically useless. At least to what I've seen. You need it often? Get a spectrum analyzer! You need it once a year? Dump the data and use Octave..

The properly working FFT is the cherry on the top.. It actually depicts the real parameters of the scope as the resulting spectra depend on the scope parameters which EEs may consider important. And yes, in past the produced spectra were of little use, but with a higher resolution and a better FFT implementations (what does it mean better here?) the FFT could become rather useful tool (ie the spectrum analyzer for lower frequency areas). And hopefully we will get the Octave installed in that scope soon

[Kleinstein]

People who buy a budget scope usually don't have a spectrum analyzer. So it is real plus if the scope can also offer at least some SA functions. From what is shown in Daves video the FFT implementation is quite good: quite fast and with a large number of points the dynamic range is also not that bad. Here the 12 Bit ADC may really help for a little more dynamic range - it can depend on the gain setting though as it is not only the ADC to cause noise.

It is not just for low frequencies - it allready covers the usualy radio IF frequencies (e.g. 10.7 MHz , 455 kHz and even some 45 MHz used in old days analog TV).

There is another plus (especially for a beginner) of the scope over a real SA: the inputs are usually way more robust.

[tautech]

...........
- they are bringing to market 12 bit scopes with good analog performance to low cost market.

Are you sure ?
Go back a few pages and look at the disturbing display from a Bodnar pulser where fast edge artifacts seem to be hidden.

That's assuming that what the Siglent scope shows is the ground truth. In Dave's quick test and comparison, the R&S RTB2004 showed the same pulse shape as the two Rigols -- https://www.youtube.com/watch?v=S8jrpCoZyx8&t=1937s. Do you have more data to confirm that the Siglent gets it right?

Dunno about you but I trust Dot mode and when I see a waveform from a Bodnar pulser without over/undershoots I am very suspicious of inaccurate waveform reconstruction.

Even the bottom of the heap 50 MHz SDS1052DL+ and right to current flagship model SDS6204A, both in Dot mode show us what should not be hidden as they also do in Vector mode.

[iMo]

..
It is not just for low frequencies - it allready covers the usualy radio IF frequencies (e.g. 10.7 MHz , 455 kHz and even some 45 MHz used in old days analog TV)..

And the talented EEs here may build a DIY down converter for the higher frequencies as well (like 45-XXXXMHz)..

PS: ie. the usual SDR DVBT USB dongles have got ~3MHz baseband BW and an 8bit ADC.. 

[ebastler]

Dunno about you but I trust Dot mode and when I see a waveform from a Bodnar pulser without over/undershoots I am very suspicious of inaccurate waveform reconstruction.

Even the bottom of the heap 50 MHz SDS1052DL+ and right to current flagship model SDS6204A, both in Dot mode show us what should not be hidden as they also do in Vector mode.

Well, ringing on that scale is not present in the actual output of the pulse generator. (As shown in the scope traces taken at 20 ps/div, https://www.leobodnar.com/shop/index.php?main_page=product_info&cPath=124&products_id=375).

Is it just a "philosophical" matter how scope designers lay out their front end low-pass filters then? The Siglents seem to use a very sharp cutoff, which results in ringing being displayed on the scope which is not actually present in the input signal, while Rigol and R&S seem to apply a smoother cutoff. Which avoids adding "artificial ringing", but may also mean that real high-frequency components near the nominal cutoff are suppressed somewhat more.