Best 350MHz scope in a hackable world (Siglent SDS2104X Plus or Rigol MSO5072)

[Martin72]

Don´t forget to check it out against the siglent with 10bit  and 10 plus 3 bit Eres..

[nctnico]

Watching the video and looking at some tests I've done myself it seems that what the HiRes mode does is increase the dynamic range. It seems that I can get much finer detail in a 5V peak-to-peak signal in the Rigol than in the Siglent. I need to do more tests, as that somehow contradicts what Martin said about measuring power supplies ripple, where the Siglent performed much better.

More noise does help with oversampling to get more bits from an ADC so it is plausible that the Rigol actually performs better where it comes to Hi-res mode. It is one of the cases where turning the bandwidth limit OFF helps to see more tiny details in a signal.

[Fungus]

	Vpp	StdDev
MSO5074 HiRes	1.14mV	184uV

How many samples were being averaged there? If the 20Mhz limiter is on then you can easily go to 32x (or even 64x).

Sorry, I don't understand the question.

What that video calls "decimation interval" at 0:43.



Does the 'scope let you choose the number of samples in that interval? If so, how many was it using?

[Fungus]

Watching the video and looking at some tests I've done myself it seems that what the HiRes mode does is increase the dynamic range.

Dynamic range is mathematically the same as "less noise". It's what I've been trying to get at for many posts now.

The Rigol has sample rate to burn. Surely the HiRes mode can be used to reduce noise and still keep the 350MHz bandwidth.

[Fungus]

In terms of noise there is no discussion, but IMHO, noise is not everything, not even the more important thing depending on what you are working with.

Noise is important but not as much as the Siglent owners would have the Rigol owners believe (IMHO). It's like a computer benchmark or a car's MPG rating - an easy number for reviewers to look at and publish but ultimately useless in the real world.

If you're genuinely interested in noise then you need an external amplifier to boost the signal to a level where both Rigols and Siglents can see it easily, eg.  If I were a power supply designer then I'd have a signal amplifier permanently connected to channel 4 of my 'scope. They cost an order of magnitude less than the difference in price between a Rigol and a Siglent.

Plus: I'm still not convinced there isn't a setting anywhere on the Rigol that will reduce the noise for people who don't measure it very often, eg. HiRes mode or a combination of math functions.

To the others: Why is this even on topic and not just me being a dumbass? Because most of the Rigol vs. Siglent discussion has centered around noise. Sorry for not taking it as gospel that low noise is sooooo desirable.

[kahuna0k]

I've found no way to set the decimation interval or any other option related to HiRes mode, I don't think it is exposed in the interface, probably something hard coded, hopefully dependent of the bandwidth. I know that higher dynamic range should be come with a lower noise, but I think that, depending on the source of the noise, we could have some amount of noise at very low voltages, but high dynamic range. For example it could resolve tens of milivolts on a 10V peak to peak signal (around 10 bits of dynamic range) but it could still have 1mV of noise below that.

[nctnico]

Plus: I'm still not convinced there isn't a setting anywhere on the Rigol that will reduce the noise for people who don't measure it very often, eg. HiRes mode or a combination of math functions.

To the others: Why is this even on topic and not just me being a dumbass? Because most of the Rigol vs. Siglent discussion has centered around noise. Sorry for not taking it as gospel that low noise is sooooo desirable.

This takes me back to the time where I owned an Agilent DSO7104A. This scope is very noisy (even with the bandwidth limit on). Most of the time I had hi-res on because the noise would obscure details of the signal. However the hi-res mode isn't compatible with all acquisition modes and when you compress a signal you get all kinds of aliasing artefacts. All in all having to use hi-res all the time is a nuisance.

[2N3055]

We came at the point where random words are being said and no correlation to any real data and concepts is being done.

Lets make it clear:

1. More noise is more noise. Hires uses LOW amount of noise, just enough to make bit flicker up and down at the edge of the A/D threshold, and then takes sequence of 4,8,16 32 consecutive  samples and averages it. Superimposed noise will have effect of creating residual voltage (noise should average to 0 but it won't because sub bit voltage threshold will skew distribution up or down, proportional to voltage) that will yield higher resolution than A/D used.  If you drown signal in 2 bits worth of sampling noise, all you will see will be noise.  For Hires to work A/D has to be monotonic and to have very good DNL. Otherwise you just get noise... Hires works well if front end is noisy and A/D is perfect. If noise is artefact from A/D converter, Hires doesn't work that well.  That is why I suspect that it might be that  in new Rigols, A/D is the one being the source of noise, and that would explain poor Hires implementation.

2. Rigol doesn't have "sample rate to burn". Hires works such that you need to filter down to effective sample rate that is 20x less to gain 2 bits (roughly). That means you get effective rate of 400MS/s to gain 4x less noise (if it works well) and you get 200MHz bandwith.. 4X reduction will still be a bit more noise than SDS2000x+ running at full speed and at higher sample rate at that time. When same tricks are available to both scopes, one with better starting position will always win... GIGO....

3. The video that was referenced (MSO5000 vs R&S2000 hires something) is very confused and misguided video showing something else than what is discussed here. They are not comparing noise, and they are not comparing dynamic range. Rigol's new integrated front end chips seem to have one good characteristics: they seem to have faster overdrive recovery than front end in RTB2000. So when you push the signal off the screen, amplifier doesn't distort so much as the one in RTB2000. Also in some ranges, scopes have different offset capabilities. Rigol simply found a voltage range where they could work at higher amplification than RTB2000 so signal was reaching A/D in more favorable way in Rigol. And left side of signal on Rigol is still distorted, it just recovers real fast (kudos for that, real good job Rigol) so most of the signal after first two edges looks nice. But it is still distorted, just much less. And you can see at all times how much more noise there is on Rigol compared to R&S, despite Hires being on. They also did the trick to show it in colour grading mode, so big, thick yellow line is not so obvious.

New Rigols have more noise. Period. No amount of bullshit will change that.

What is important to end user is fact: does it matter for your end use?
If you're not looking at signals at milivolt levels at 100 MHz+ bandwith probably not. MSO5000 has some other nice features, it's retrigger rate is faster so it will look more analog like on the screen etc etc.
It is perfectly good scope for general purpose use. It has 4 decodes, 4 math channels, etc etc. So there are things where it has strengths..

You just have to chose wisely what is important to you.
Analog performance, LeCroy type of concept? Go with SDS2000X+.
Digital work, decoding, interactive work? MSO5000 will do just fine.

Not really sure?  Both will do...

[Martin72]

New Rigols have more noise. Period. No amount of bullshit will change that.

Word.

You just have to chose wisely what is important to you.

Word.

Analog performance, LeCroy type of concept? Go with SDS2000X+.

That´s why I´ve changed to it, others may not need it and would be happy with the 5000, which costs less - win-win for them.

The video that was referenced(...)

Is from rigol to show rigols performance against another brand.....hmmmmm
The opposite were the same hmmm, better a neutral person who got access to both of them should compare something between.

[nctnico]

3. The video that was referenced (MSO5000 vs R&S2000 hires something) is very confused and misguided video showing something else than what is discussed here. They are not comparing noise, and they are not comparing dynamic range. Rigol's new integrated front end chips seem to have one good characteristics: they seem to have faster overdrive recovery than front end in RTB2000. So when you push the signal off the screen, amplifier doesn't distort so much as the one in RTB2000.

No. The problem in the video is that they choose the memory length in the RTB2000 too short so the sampling rate is too low. You are looking at a filtered (hi-res also filters), undersampled signal (look at the samplerate of the RTB2000). Remember that what is on screen in the video is a stopped acquisition which is then enlarged. Overdrive recovery is not an issue in this case.

[2N3055]

3. The video that was referenced (MSO5000 vs R&S2000 hires something) is very confused and misguided video showing something else than what is discussed here. They are not comparing noise, and they are not comparing dynamic range. Rigol's new integrated front end chips seem to have one good characteristics: they seem to have faster overdrive recovery than front end in RTB2000. So when you push the signal off the screen, amplifier doesn't distort so much as the one in RTB2000.

No. The problem in the video is that they choose the memory length in the RTB2000 too short so the sampling rate is too low. You are looking at a filtered (hi-res also filters), undersampled signal (look at the samplerate of the RTB2000). Remember that what is on screen in the video is a stopped acquisition which is then enlarged. Overdrive recovery is not an issue in this case.

No look again.

Scope is in RUN mode, and distorted one also have offset applied.  Check screenshots attached from 1:40 on...
You can clearly see how scope reacquires new sample after vertical sensitivity change..

Sample rate is low, but 305 kS/s is more than enough to capture 25000 points for the duration of screen, so plenty of oversampling for that screen..
No, that is simply overdrive recovery, where Rigol decided to skew test to particular point where they chose something that was favorable for Rigol.

Problem is that they didn't use  RTB2000 screen zoom function, because with it's lover noise, higher real A/D resolution and HIRES, it could have easy shown that BETTER than Rigol because you could have avoided distortion completely.  Because Rigol also distorted those first two edges on the left.  It looks nicer but it still wrong...

So that is shady marketing... If I where them I would pull that video. It makes them look sleazy.
If they wanted to show some advantage, I'm sure they could find some real benefits....

[TurboTom]

I can't help but wonder about the noise figures of both mentioned scopes. I guess we've got to approach the noise figures differently to make them more comparable and to better understand where they originate from. @ncnico made the reasonable suggestion to test noise with the input BNCs capped off (to eliminate interference) but not terminated or shorted.

I assume that none of the mentioned scopes have a "proper" 50 ohms input (MSO5k obviously not) with a properly designed 50 ohms divider signal path. In this class of scopes, there's usually the "generic" 1Mohm divider and a 50R terminating resistor that can be connected in parallel with the BNC via a relay if selected. So, if the input is left unterminated, it should "see" in the highest sensitivity ranges an impedance of 1Mohm with a few picofarads in parallel. The thermal noise of 1Mohm at 20°C and an observed bandwidth of 20MHz is approx. 570µV RMS. If any of the the scopes reports less noise under these conditions, it's got to be cheating.

Depending on the coupling of the input, the impedance "seen" by the first amplifier stage may differ a lot with the input terminated. It may actually range from 50R (direct coupling) to probably several tens or maybe even hundreds of kiloohms.

Siglent's noise represents an input impedance with a terminated input of round about 3kohms while Rigol's is more close to 100kOhm.

So, if I may ask someone who's got access to both these scopes if he may try another configuration for a noise measurement:
TB 1µs, sensitivity set to the last "genuine" range, i.e. without digital magnification (I guess that's 2mV/div on the MSO), input impedance 1Mohm, 20MHz bandwidth limit enabled, BNC if possible capped off (even some aluminium foil wrapped aroud may do).
I'ld like to compare the following measurements: Sampling mode (1) normal and (2) peak (to understand if there's some additional "smoothing" going on in normal sampling mode). Measure functions: Vpp and STD (AC RMS) averaged. Maybe a few sensitivity comparisons would also be interesting, i.e. how the noise measurement changes when the input range is changed, maybe between 5mV/div and the lowest range available.

I guess there's much more "signal shaping" going on in these scopes than we can imagine...

[Martin72]

In this class of scopes, there's usually the "generic" 1Mohm divider and a 50R terminating resistor that can be connected in parallel with the BNC via a relay if selected.

Have a look:

https://www.flickr.com/photos/eevblog/49964995676/in/album-72157714555911588/

So, if I may ask someone who's got access to both these scopes if he may try another configuration for a noise measurement:

Can do this on my siglent only, but not today for several reasons...

[nctnico]

3. The video that was referenced (MSO5000 vs R&S2000 hires something) is very confused and misguided video showing something else than what is discussed here. They are not comparing noise, and they are not comparing dynamic range. Rigol's new integrated front end chips seem to have one good characteristics: they seem to have faster overdrive recovery than front end in RTB2000. So when you push the signal off the screen, amplifier doesn't distort so much as the one in RTB2000.

No. The problem in the video is that they choose the memory length in the RTB2000 too short so the sampling rate is too low. You are looking at a filtered (hi-res also filters), undersampled signal (look at the samplerate of the RTB2000). Remember that what is on screen in the video is a stopped acquisition which is then enlarged. Overdrive recovery is not an issue in this case.

No look again.

Scope is in RUN mode, and distorted one also have offset applied.  Check screenshots attached from 1:40 on...
You can clearly see how scope reacquires new sample after vertical sensitivity change..

If that is the case I'll take your word for it. I just skipped through the video to get to the conclusion. To me it doesn't make any sense to do such a comparison with the scopes in run mode so I assumed the maker of the video would have done the same. The whole point of hi-res is to be able to zoom in on tiny details of a signal in stop mode once you run out of vertical offset range.

[Fungus]

All in all having to use hi-res all the time is a nuisance.

Sure, I don't imagine you'd need it for looking at 5V signals.

But if you only occasionally look at mV signals and it works for noise reduction then for occasional use, I don't see $500 worth of problem. Especially not when you can buy an amplifier for $50 and get better results than the Siglent.

[Fungus]

We came at the point where random words are being said and no correlation to any real data and concepts is being done.

Lets make it clear:

1. More noise is more noise.

I don't think anybody's arguing other wise.

What I'm arguing is that the difference between Siglent noise and Rigol noise isn't as big a deal as people are making out. If mV measurements are really your passion then you'll own a suitable signal amplifier to boost the signal.

[Fungus]

3. The video that was referenced (MSO5000 vs R&S2000 hires something) is very confused and misguided video showing something else than what is discussed here.

The single frame I extracted from it was useful (I hope).

(The person who called overlaying of multiple waveforms "Averaging mode" needs shooting).

[mawyatt]

For someone trying to decide on a DSO and having distilled the selections down to a couple candidates (difficult with all the stuff flying around on DSOs), it seems getting both candidates and doing an evaluation for what is important to the user is in the works

With a quality supplier available and shipping charges reasonable....and the supplier allows returns for no reason...., then doing your own side by side testing in your own environment is the best possible decision maker!! No vendor staged videos, specmanship data manipulation, fanboy or girl representations, biased views, or whatever, just the end user evaluation to their needs. The only additional cost would be the return shipping.

I'm not advocating this for all equipment purchases, but complex higher-end and cost devices like these DSOs seems a reasonable approach. Returned items take away from resell value, which reduce margins for distributors and OEMs, so very selective use advised. I haven't done this yet, mainly because I try and do my homework before the purchase, and when working (retired now) we never had to worry about new equipment as the factory reps were all too willing to bring over and leave the latest and greatest instruments for our evaluation. We always had plenty of time from rep to evaluate new equipment in our environment, which often led to a purchase later, so the "incentive" was in place. Most may not have this luxury, now I don't, so other means are required in the equipment decision process, and the self test comparison seems reasonable today with the on-line purchasing, and low cost shipping. 

Best,

[nctnico]

All in all having to use hi-res all the time is a nuisance.

Sure, I don't imagine you'd need it for looking at 5V signals.

On the Agilent DSO7104A I had to use it for all signals in order to be able to make cursor measurements.

[Martin72]

In this class of scopes, there's usually the "generic" 1Mohm divider and a 50R terminating resistor that can be connected in parallel with the BNC via a relay if selected.

In the picture of the teardown (the link in may last post) you can see a 24R9 Resistor soldered to the bnc input, is that part of a:

I assume that none of the mentioned scopes have a "proper" 50 ohms input (MSO5k obviously not) with a properly designed 50 ohms divider signal path.

?

Actually it´s not clear for me, why it is a 24R9 resistor there as I never make up my mind about proper 50Ohm Signal termination.

[nctnico]

3. The video that was referenced (MSO5000 vs R&S2000 hires something) is very confused and misguided video showing something else than what is discussed here.

The single frame I extracted from it was useful (I hope).

(The person who called overlaying of multiple waveforms "Averaging mode" needs shooting).

No because that is exactly what averaging does. It overlays multiple acquisitions to produce an average. The confusing part is that Rigol seems to be using the term averaging also for high-res. In a sense that is correct because oversampling in order to achieve more bits averages a number of subsequent samples from the same acquisition.

[TurboTom]

In this class of scopes, there's usually the "generic" 1Mohm divider and a 50R terminating resistor that can be connected in parallel with the BNC via a relay if selected.

In the picture of the teardown (the link in may last post) you can see a 24R9 Resistor soldered to the bnc input, is that part of a:

I assume that none of the mentioned scopes have a "proper" 50 ohms input (MSO5k obviously not) with a properly designed 50 ohms divider signal path.

?

Actually it´s not clear for me, why it is a 24R9 resistor there as I never make up my mind about proper 50Ohm Signal termination.

I guess there's another 24R9 on the opposite side of the PCB  -- Merry Christmas!

[Martin72]

Merry Christmas!

Same!

[TurboTom]

I couldn't stop to reason about the input noise and finally, I simply had to add to my previous statements about this issue my latest thoughts:

If the (standard high-impedance) input of a scope is left open, it's resemling a resistance of 1Mohm in parallel with a capacitance of a few picofarads, say 15pF. This RC-combination is basically a low pass filter of first order with a 3db frequency of 67kHz (in this case), and not the 20MHz bandwidth that I used to estimate the thermal noise. At a frequency band of 67kHz, the thermal noise would be round about 33µV rms (surprise, surprise...), so if it's arranged that way, Siglent's noise figure may be quite accurate, provided they don't change that much with an open input.

I guess, in the next days someone will help out with a few tests - thanks so much for this  . I haven't got any of these instruments and as yet, don't see the need to change this (though TEA tries to convince me...).

[Martin72]

and as yet, don't see the need to change this (though TEA tries to convince me...).

"Das Vergleichen ist das Ende des Glücks und der Anfang der Unzufriedenheit"

"Comparing is the end of happiness and the beginning of dissatisfaction"

(Kierkegaard)

Stay tough !