Rigol DHO804 Test and Compare Thread

[ebastler]

SDS2000X Plus while a 500 MHz design uses two 2GSa/s ADCs however when 3 channels are activated automatic BW limiters are engaged to ensure Nyquist remains met with sufficient sampling rates.

Ah, thanks for pointing that out. That is a clean solution; I wish Rigol did the same in the DHO900. Maybe they will add it in a future firmware update; the required lowpass filter settings are probably all on board.

Hey, maybe that's the real reason why they have a 70 MHz entry model -- that happens to be the Nyqist bandwidth in the worst case of using all DHO900 channels. 

[tautech]

SDS2000X Plus while a 500 MHz design uses two 2GSa/s ADCs however when 3 channels are activated automatic BW limiters are engaged to ensure Nyquist remains met with sufficient sampling rates.

Ah, thanks for pointing that out. That is a clean solution;

No it's the only proper solution when insufficient sampling is available for 4 channel operation.
These auto BW limiters get clearly displayed in each channel tab when applied just like the manually engaged 20 MHz BW filter.
SDS1204X-E & SDS5104X don't use this solution as they provide sufficient sampling to meet Nyquist when all channels are active.
OTOH SDS2504X Plus and SDS2504X HD do use this solution for 350 MHz max when 4 channels are active.

SDS1204X HD will use a similar solution to maintain sufficient sampling rates whereas SDS1104X HD provides sufficient sampling rates to not need BW limiters.

[Fungus]

Yes, but it DOESN'T sample at that rate, not when you use it sensibly.

Apparently that sensible use can only be expected from the more advanced users of more expensive oscilloscopes. Tautech explained this with reference to the SDS2000X plus series once: 

Rubbish. Most people simply won't be connecting that much stuff to it. Not connecting stuff doesn't take knowledge.

Besides: It will only be a problem if the incoming signals are over Nyquist, which they won't be if they're coming from Arduinos and/or breadboards with dupont cables.

To me the biggest limitation will be the small screen for showing all that stuff. Anybody looking to connect that many wires should probably go for an MSO5000 which costs about the same.

I can tell you're not going to let this go though so let's not argue that.

I also predict that an awful lot of DHO900 sales will be for a couple of channels with 250Mhz analog bandwidth and/or the built-in AWG, NOT for the LA.

(especially given the price of the LA probes)

[ebastler]

Apparently that sensible use can only be expected from the more advanced users of more expensive oscilloscopes. Tautech explained this with reference to the SDS2000X plus series once: 

Rubbish.

I realize that this topic is close to your heart, but please calm down a bit. The yellow thing in my comment is a smiley, ya know?

Besides: It will only be a problem if the incoming signals are over Nyquist, which they won't be if they're coming from Arduinos and/or breadboards with dupont cables.

Sampling of the digital channels at 625 MSa/s is perfectly fine. To look at analog signals in parallel -- whether analog I/O of the system under test or digital signals that need close scrutiny -- I would use proper analog probes, and then the scope's violation of the Nyquist limit may become a problem.

Especially for the use case of scrutinizing a digital signal to look for overshoots, runt pulses etc., I do see the low sampling rate on the analog inputs as a real limitation. What good is this if you have a much lower sampling rate on the analog channels than on the digital inputs? You could only do this on a single analog channel -- which is better than nothing though.

Still, the sampling rate limitations would not be a total show stopper for me. As mentioned earlier, a DHO914 including Rigol's rather nice logic probe is still significantly less expensive than e.g. a Saleae Logic Pro 16, and has better specs in the digital and analog domain.

I can tell you're not going to let this go though so let's not argue that.

As mentioned a couple of times: The thing I really resent Rigol for is that they don't disclose the sampling rate impact on the analog channels when using the LA. Let's assume it is just an oversight and they will come out with a corrected data sheet quickly. Umm, right... 

[UK]

Haven't you found that it's a bit weird that most of Rigol's threads here look more like advertising of Siglent than a strict topic discussion?
Why in Siglent threads not always talking about Rigol?!

[TurboTom]

...

Thus it seems the max r/w bandwidth into those external drams is those ~625Mwords/sec (or the twice when interleaved r/w). That is somehow in sync with the ~667MHz max clock of the Zynq FPGA there (assuming the lowest grade FPGA).

The other DSOs I saw have got usually 4x drams where you can share the r/w bandwidth even better.

That's actually not quite correct: The speed is valid for the two ARM Cortex A9 cores which don't shovel the raw ADC data to the sample memory. The FPGA logic (Artix7-equivalent with 74k logic cells) with its parallel architecture can perform much faster than that. The RAM that's used is specified at 2.133 GT/s, which at the utilized 16 bit wide interface will easily match the 1.25GSa/s data rate of the ADC plus some overhead. All of this, combined with the 3.3MB block RAM contained inside the FPGA makes me wonder if the two additional RAM foorprints are actually populated in the DHO900 series, considering the observation of the sample rate loss when enabling the digital inputs...

This wouldn't be the first time that the BOM of production versions of test gear has been significantly "economized" compared to pre-production units which afterwards limits performance and expandability considerably (SDG6000X...). And Rigol was "Clever" enough not to publish any MSO-related performance data

[ebastler]

[...] makes me wonder if the two additional RAM foorprints are actually populated in the DHO900 series

They are, at least in @hubertyoung's DHO924 unit: https://www.eevblog.com/forum/testgear/rigols-new-dho800-oscilloscope-unbox-teardown/msg4976278/#msg4976278

That is an early unit from July though. It is also the donor of the vendor.bin which has been used by various '804 owners to upgrade their scopes, by the way.

[2N3055]

Haven't you found that it's a bit weird that most of Rigol's threads here look more like advertising of Siglent than a strict topic discussion?
Why in Siglent threads not always talking about Rigol?!

You are in the wrong thread.

This is a comparison thread.

But apparently there should be no comparison or even discussing features and capabilities of different equipment ..
 

[TurboTom]

[...] makes me wonder if the two additional RAM foorprints are actually populated in the DHO900 series

They are, at least in @hubertyoung's DHO924 unit: https://www.eevblog.com/forum/testgear/rigols-new-dho800-oscilloscope-unbox-teardown/msg4976278/#msg4976278

That is an early unit from July though. It is also the donor of the vendor.bin which has been used by various '804 owners to upgrade their scopes, by the way.

Yes I'm aware of that. And the relatively early time that Hubert got his specimen makes it very likely that his was a pre-production unit. Just like Shariar's (Signalpath) SDG6000X that he reviewed and that contained the four dedicated sample memory chips while mine doesn't: https://www.eevblog.com/forum/testgear/siglent-sdg6000-series-awg_s/msg3545119/#msg3545119

[Fungus]

Haven't you found that it's a bit weird that most of Rigol's threads here look more like advertising of Siglent than a strict topic discussion?
Why in Siglent threads not always talking about Rigol?!

You are in the wrong thread.

This is a comparison thread.

But apparently there should be no comparison or even discussing features and capabilities of different equipment ..
 

He's talking about Rigol threads in general.

I don't remember a single Rigol thread ever which was any different.

[2N3055]

Haven't you found that it's a bit weird that most of Rigol's threads here look more like advertising of Siglent than a strict topic discussion?
Why in Siglent threads not always talking about Rigol?!

You are in the wrong thread.

This is a comparison thread.

But apparently there should be no comparison or even discussing features and capabilities of different equipment ..
 

He's talking about Rigol threads in general.

I don't remember a single Rigol thread ever which was any different.

And Siglent related threads are no different. Instead of comparing notes to be better educated and demand more from all manufacturers for benefit of all..

And even if in that case this should still be the wrong thread. For facts about thread post in that thread... Or open a complaint thread...

But I simply refuse those accusations, at least on my part.
How many posts I have in DHO hacking thread ?
How many posts in any (non comparison) DHo based thread where I mention Siglent (if it was not to answer the question)...
How many times in this very thread I said something about DHO800 was good and right once I saw how it was done?
Have I ever said that DHO800 was crap because it did not have as many function as SDS2000X HD for instance?
Or did I repeatedly said that being a simple scope and in that price range is quite OK?

Really, go look... I might be abrasive sometimes, but facts are facts.

It is not even about brands anymore, it's simply lynch mob mentality... No facts needed....

[ebastler]

[...] makes me wonder if the two additional RAM foorprints are actually populated in the DHO900 series

They are, at least in @hubertyoung's DHO924 unit

Yes I'm aware of that. And the relatively early time that Hubert got his specimen makes it very likely that his was a pre-production unit.

Would that imply that @hubertyoung's DHO924 can possibly sustain its analog sampling rate even when the LA is activated? (Provided that its current firmware and FPGA configuration can still detect and use the extra RAMs, which may be a long shot.)

I'll send him a PM and ask, in case he does not monitor all those DHO threads...

[Mechatrommer]

Why in Siglent threads not always talking about Rigol?!

if every Rigol owners are like me... i have to admit Siglent spec is somewhat better than Rigol, except some minor aspects (such as me insist on downloadable data through usb cable), so we cant argue in their thread otherwise we just shooting our own foot...but since Siglent are more expensive, because of the price alone, we tend stick to Rigol, or there is no real need to switch (upgrade) to Siglent because we can manage with our existing Rigol dso. to get a picture of what we (i) feel when i see new siglent dso model appearing is like... oh this scope is sooooo good, but when i look at the price... oooh ok, no wonder why that scope is soo good. when we look at new Rigol, ok this rigol is much better than previous one, and it cheapest bang per buck! lets go get it! and hack it to maximum spec it can be hacked to. i bought 804 is for a reason, 814 is just few dollar more expensive, i can afford if i want to but.... we know rigol's "tradition" since dave made this forum, and it seems it hasnt changed.

edit: fair enough siglent is designed properly. not to mention its 2-10X the price.. but it will never see oscillating hi bw opamp with excessive feedback capacitance. If you havent seen it in its aliased form, you wont love rigol

[TurboTom]

...

Would that imply that @hubertyoung's DHO924 can possibly sustain its analog sampling rate even when the LA is activated? (Provided that its current firmware and FPGA configuration can still detect and use the extra RAMs, which may be a long shot.)

I'll send him a PM and ask, in case he does not monitor all those DHO threads...

Well to be honest, I don't know  . All I can tell that when applying logic and common sense, tripling the memory and data rate available to the Zynq as seen on the pre-production DHO900 vs. the DHO800 and still having to cut sampling rate by half when enabling the digital channels doesn't make much sense. My take was that the two additional ram chips identical to the main sample memory chip were there to provide storage for sampling the 16 digital channels (would be a good fit since the GDP2BFLM-CA DDR3 RAM is organized 16 bits wide) and for the AWG to store arbitrary waveforms (14 bit DAC, hence also a good fit...).

But when looking at the cuts Rigol made regarding analog sample rate when using the digital channels makes me believe that all the sampled data (analog/digital) has to be transfered via the same memory bus. The meagre arbitrary waveform size of 16kSa that the AWG can actually process can easily be placed in the Zynq's internal block ram and so won't affect sampling performance at all. A single GDP2BFLM-CA provides storage for 256M x 16 bits so it would be easily possible to put there 2x50MSa of data plus double buffering / buffer switching. So my educated guess is that Rigol could actually have arranged the DHO800/900 with just a single DHO800 hardware configuration and only add the AWG piggyback board for the (S) versions. Certainly, the engineers at Rigol had a different concept in mind, but very likely, the bean counters got in their way...  I may be wrong tough, in which case the engineers didn't do their work properly. 

[Fungus]

My take was that the two additional ram chips identical to the main sample memory chip were there to provide storage for sampling the 16 digital channels

That's still unknown.

Are those two chips installed in production DHO900s?

A Google image search for "Rigol DHO900 teardown" comes up blank.

[asmi]

The RAM that's used is specified at 2.133 GT/s

RAM can work that fast, but memory controller in lower-end Zynqs can not. DDR3 can only go up to 533 MHz on a PS side (hard IP), and 400 or 533 MHz on PL (fabric) side (depending on the speed grade and if it's ran at 1.5 or 1.35 V).

[iMo]

The RAM that's used is specified at 2.133 GT/s

RAM can work that fast, but memory controller in lower-end Zynqs can not. DDR3 can only go up to 533 MHz on a PS side (hard IP), and 400 or 533 MHz on PL (fabric) side (depending on the speed grade and if it's ran at 1.5 or 1.35 V).

There is a single dram populated (see the teardown thread) at the Zynq.
2 dram pads are empty.
How they push 1.25Gsa/s into it then?

[asmi]

There is a single dram populated (see the teardown thread) at the Zynq.
2 dram pads are empty.
How they push 1.25Gsa/s into it then?

You direct the question to the wrong person. How would I know? I'm not the one who designed it. 16 bit DDR interface at 533 MHz can theorerically push 533 * 2 * 16 = 17056 Mbit/s, while 1.25 GS/s @12 bit requires 15000 Mbit/s, so it looks to have enough margin for overhead. I wonder what kind of connection does it have to application processor, Zynq-015 has 4 multi-gigabit transceivers which allow implementing up to PCIE 2.0 x4 link, which should provide for 20 Gbps of link bandwidth.

[ballsystemlord]

If @Martin72 is still accepting requests, do you have the capability to test how close the Oscilloscope is to an NIST calibrated DMM on a DC signal and AC sine wave?

Thanks!

EDIT: To be clear I'm asking for 2 separate tests. *Not* an AC sine wave imposed on a DC signal.

[Martin72]

I am just warming up the siglent sdm3065X, the rigol scope and the dmm ref.plus source...

[Fungus]

Place your bets:

[Fungus]

Can you also measure difference between min/max voltage to eliminate offset errors?

(It should show min/max/delta in stats)

[Martin72]

Too late...
No, can repeat it, but first the measures I´ve already made.
The results are really good for a scope...
And it makes a difference when measuring dcv what scale you have set. At 200mV/div. it´s more precise than when setting 1V/div.
I´ve added DVM and counter, the counter was set to maximum digits (6).
The sdm3065x was simultaneously connected to the reference, here are the results(left sdm3065x/right rigol (no DVM):

DCV: 4.99937V/4.9980V(-0.03%)
ACV 100Hz: 4.99351V/4.9939V(+0.008%)
ACV 10khz: 4.99277V/5.0104V(+0.35%)

[Martin72]

Here is another picture with the fancy table.
First I was surprised that the counts stop at 1000.
But a look in the manual has then given the all-clear, you can set the counter from 2 to 100000.

[Fungus]

The results are really good for a scope...