the eternal question about rigol vs siglent

[Fungus]

Note that the probe was deliberately set to 1x mode to get these Rigol-hatin' results. This makes the noise 10x worse than you'll see in real life (in reality, 1x mode is something you might never use in an entire lifetime of 'scoping). Divide those Vpp and RMS numbers by 10 for realioty.



(Edit: Why all the 'history' and 'playback' modes in the Siglent? Does that improve the signal? Why not use the same modes on the Rigol?  Suspicious... )


nb. Dave did a whole video on the subject of 1x mode and the pitfalls/traps is has:

[tautech]

Note that the probe was set to 1x mode to get these Rigol-hatin' results. This makes noise 10x worse than what you'll see in real life (1x mode is something you might never use in a lifetime of 'scoping).

Oh come on, pull the other one. 

Have you just turned up here and never checked for mV ripple on a PSU ?

[Fungus]

Note that the probe was set to 1x mode to get these Rigol-hatin' results. This makes noise 10x worse than what you'll see in real life (1x mode is something you might never use in a lifetime of 'scoping).

Oh come on, pull the other one. 

Have you just turned up here and never checked for mV ripple on a PSU ?

The sort of person who needs to accurately measure mV ripple on a PSU shouldn't be arguing over the differences between a $350 and a $500 oscilloscope, they should be looking at a professional tool.

[Rerouter]

Bit confused there fungus, measuring power supply ripple to multiple mV is well within the capacity of just about every scope I have used, You may need to cross check your measurements with an external wavegen if you need fractions of a single mV,

Both the scopes showcased by the OP have at least 1 full division of resolution at 1mV, I cannot speak for the rigols noise. but the siglents is well low enough to measure better than 300uV if you spend a few minutes setting up the measurement.

[Fungus]

Both the scopes showcased by the OP have at least 1 full division of resolution at 1mV

Yep, and probably more with the right combination of display settings.

You normally want the BW limiter turned on in 1x mode, too, but the person who made that image didn't bother with realistic settings.

I cannot speak for the rigols noise. but the siglents is well low enough to measure better than 300uV if you spend a few minutes setting up the measurement.

So the question becomes: Is the purchaser of the 'scope likely to do that?

If so, spend more money and get the Siglent.

If not, get a Rigol and a decent multimeter (or whatever) instead. You'll be better off.

This "eternal" argument only makes sense if they both cost the same amount of money.

[Rerouter]

there both entry model scopes, its not a question of whats unlikely, more, can it do what you want to do until you feel you have earned back its value through using it.

was it likely that I would use the trigger out of my scope to fill the role of a thermostat until the part arrived? No, Am I happy I could do it, Yes

If you want GHz scopes with crazy options for not much coin, you go dig yourself an ebay hole looking for them second or third hand, Heck for a high grade multi meter I got a tektronics 3 way mainframe and 3 DM5XX modules working for under $80, again is searching for these things worth it to you, and can the tool do what you want to accomplish without too much fiddling.

In my book, if you need to document things, the siglent may be more convenient, you can just spam the save as PNG and save .bin file buttons from the web interface. Vs the intermediary step of using a USB drive. the lack of an RTC is a pain, but there are commands to set the time and date for the session you have it powered up so things get time stamped correctly.

Again this is without me having used the rigol, I almost certainly have some level of bias,

[HalFET]

I've not tried the Bode plot on a Siglent yet.  The Instek version is a joke.  When I attempted a sweep from 13.3 to 13.6 MHz  of a 40 MHz 3rd overtone crystal, I got 2 data points.  One can do better with the MDO SA function, but it takes the square root of the power specrtrum, so the series resonance is a peak.

However, a person with the requisite mathematical skills and the ability to transfer data to a PC can match the performance of *any* instrument  up to the BW of the DSO and signal generator.  My "To Do" list has writing the software for that close to the top.  Fixing the Tek 11801 sampling scope I just bought and testing my SD-22 sampling heads are at the top with  setting up my 8753B and 85046A close behind.

Surprisingly, the cleanest and quickest solution is often still just writing down the points and sketching it yourself. And especially for beginners this is important, it's good to know what's actually going on. That way you can point out when something goes haywire.

A hacked DS1054Z will give you 100 MHz.  A stock Siglent 1202X-E will give you 200 MHz as will a hacked SDS1104X-E.

A hacked DS1054Z is much closer to 200Mhz than 100Mhz. The Siglent is also well over 200Mhz.

eg. 1.7ns rise time on a DS1054Z (as measured by TurboTom with his rubidium source):


As mentioned earlier, both those numbers are well into the area where you need to really know what you're doing in terms of probing and connecting the cables. You can't simply poke at a circuit with the supplied probes and get those results.

Bandwidth shouldn't really be a factor when deciding between these two 'scopes, both have enough for everyday use, neither has enough for "serious" work.

Why do you need a rubidium clock for a fast rise time pulse? But anyway at 100 MHz you don't have to be that careful with probing,  the thing is that both of these scopes their response curves won't be flat once you go above 75% of the "design bandwidth". Grab a decent RF generator and sweep at a constant power, you'll see interesting things happen. Do make sure to hook it up properly though! In practice both these scopes are limited to practical signals of around 70 MHz anyway, anything above that and you lose too many harmonics to still see what's truly going on. That third harmonic is really nice to have if you want to actually see something.

Welcome sir rf-loop, I was expecting your appearance in this false marketing show! Let me start by saying that Chinese companies still haven't learned that engineers are capable of reading through bullshit datasheets anyway, so mentioning higher numbers doesn't actually convince them. People buy HPAK, Keysight, R&S and others because their equipment actually meets the specifications they list on their datasheet. In fact, every single HPAK scope I've bothered to gauge the performance of significantly outperforms the datasheet. Meanwhile, the Rigol and Siglent units on the other hand barely make ends meet. Furthermore, the major manufacturers seem to be capable of running spellcheck on datasheets, that way you don't end up with "channe:" and other fun things like that. In short, both Siglent and Rigol have craptacular datasheets.

Rigol DS1000Z data sheet:
Vertical Scale
(Probe ratio is 1X) 1 mV/div to 10 V/div.

And same data sheet claim 12 bit resolution (High res) what is bullshit. After High res, data resolution is still 8bit and nothing else and "8 - 12bit" is only for reduce noise on the screen.
If look example some Tektronix they also use double bytes for keep high resolution averaging result data.

They do not tell 1 and 2mV/div is derived from 5mV/div. (I have owned DS1000Z and also used it and 4 years ago proofed these 1 and 2mV are not full vertical resolution what was one show stopper together with enormous signal noise and total lack of measurements horizontal resolution "highly decimated data like just from screen")

What does the probe ratio have to do with this? Do you actually understand what you're reading on datasheets?

And their 12 bit can actually be 12 bit, it's called oversampling and you can implement it in an oscilloscope. You can directly group samplings of the ADC and average them out (or better yet, interpolate them properly), this will decrease the uncertainty of the measured voltage significantly and grant you extra resolution. A better tactic which is a bit more difficult to analyse is using multiple waveform captures in a row and averaging those out. In that case the actual resolution also depends on the repeatability of the trigger and timebase stability. But I'll agree it's bad practice to just slap a number on it. However, Siglent does pretty much the same thing for other values on their datasheet (i.e. delta error between channels or when making phase measurements) so you don't get to use this as an argument against Rigol.  You could have easily pointed out that their DC gain and offset definition is less precise, but they don't try to cut it up into as many chunks as Siglent to make the unit look more impressive than it is.

Siglent SDS1000X-E full resolution range starts from 500uV/div.
How useful it is, it is other question but it is not digital magnification from less sensitive range.







Images are from:
https://www.eevblog.com/forum/testgear/siglent-sds1204x-e-released-for-domestic-markets-in-china/msg2102305/#msg2102305

And these images prove exactly nothing, the settings were cherry picked to get that result. And testing the noise on scope frontends with no short probe or BNC short attached to them is a bit of a joke given their input impedance.    All you show is that they have a different way of displaying noisy signals. Furthermore, the Keysight unit shows the quantisation quite clearly, as it should. That's the scope holding up a sign with the following message: "Hey, we're running into the limit here, I can still show you some data but you'll have to figure out what it actually means." Neither the Swiglent or Riggedol pass the test by this metric.

SDS1104X-E and generator SDG1032X
With Siglent you car run this using up to 500 steps (501 points)

Minimum BW is 500Hz with minimum step 1Hz

exactly same XTAL using 5kHz BW

500 steps, 10Hz steps.

Individual points in list.

And of course Siglent FRA ("bode plot") have 3 simultaneous channels (+ref)
Here this test setup was really poor due to device under test etc. Only what is tell is that there is 3 channel.

Again, what is the error on this measurement? You keep showing fancy plots but can never quite explain the numbers behind them. I wouldn't be amazed if you have 10%, or more, error on that measurement you're showing.

Have you just turned up here and never checked for mV ripple on a PSU ?

A scope is the wrong tool for that job... Try again please.

Bit confused there fungus, measuring power supply ripple to multiple mV is well within the capacity of just about every scope I have used, You may need to cross check your measurements with an external wavegen if you need fractions of a single mV,

Both the scopes showcased by the OP have at least 1 full division of resolution at 1mV, I cannot speak for the rigols noise. but the siglents is well low enough to measure better than 300uV if you spend a few minutes setting up the measurement.

Your measurement is meaningless though. It shows something on the screen, sure. What you're actually seeing is a good question, my bet is that it might have more in common with the cosmic microwave background than with the actual signal.

It'd be fun if someone rewrote the firmware of these scopes to indicate the uncertainty bounds on the measured signals, I think people would be unpleasantly surprised.

[gf]

You normally want the BW limiter turned on in 1x mode, too, but the person who made that image didn't bother with realistic settings.

The question is, whether the BW limit really makes a difference for the Rigol's noise floor, does it? On my Hantek 6074BD the BW limit reduces only noise present in the captured signal, but not the noise floor of the scope.

The most obvious reason is that the noise floor is dominated by ADC noise, and not by noise from the frontend (at least the dominant component of the noise floor is obviously introduced after the BW-limit filter). Hereby it is noteworthy that this scope makes full use of the HMCAD1511's "digital gain" feature. 2mV/div is not achieved by higher analog gain, but 2mV/div is in fact the same as 100mV/div in the frontend, with 50x digital gain in the ADC.

[ Fully utilizing the HMCAD1511's features certainly enables the simple design of this frontend: Two relay-switched 10:1 attenuators at the input (-> 100:1/10:1/1:1) => FET buffer => ADC driver (fixed gain of only ~2) => HMCAD1511. No need for variable gain amp or analog multiplexers. ]

As far as i know, both, the Rigol and the Siglent use the HMCAD1511 as well, but I don't know whether either of them also makes (full) use of the ADC's "digital gain" feature as well. A BW-limited analog pre-amp stage can IMO well result in a lower noise floor than using digital gain of HMCAD1511 (the latter is likely only as usefull a claimed by the datasheet, if full BW is required, so that a BW-limited pre-amp is not an option).

[Rerouter]

Ok, I'll give you at least what I can, Turning off the nearby electronics doesn't change it that much, So I'll call it my baseline.

Turning the thing on, in my 24C room, a few weeks since I last ran self-cal, image 80 is what I get, all inputs are open, and set to DC 1M with a 20Mhz limit, you can see the offset does drift a little with time and temp.

Leaving it for 30 minutes to "warm up" does not change it by any significant amounts. running self cal brings the dc offsets back in line. it can be trimmed better based on the DAC in the front ends, and I have made Siglent aware. Image 81,

In this configuration, (100ms Tdiv, 500uV Vdiv, 7M samples)
all channels have 500-560uV of peak to peak noise, standard deviation between 50-54uV,

Adding 50 ohm terminators to the front BNC jacks reduces that to about 460uV peak to peak, and about 42uV standard deviation.
These measurements are made with the channel VGAC at a gain of 25db I cannot say if that is anything you can use in your math, and I can ramp that up to about 35db gain if it helps your comparison.

[Rerouter]

Scaling up the VGAC gain still gives practically the same values, So i would also pin it on ADC being the noise source,

at VGAC 35db gain, i measure 560uV of peak to peak noise, and a standard deviation of 54uV

Putting on the 50 ohm terminator brings it down to 450uV of peak to peak noise, and 43uV standard deviation.

I'm also happy that previously I misread the specs of the VGAC, so max VGAC gain is actually more like a 200uV range.

[Fungus]

What do you get without the 20Mhz limiter?

[gf]

What do you get without the 20Mhz limiter?

Exactly, it would be interesting whether the noise floor of this scope does profit from the BW limter, or not (as said, my scope does not).

[Fungus]

FWIW my Rigol gives this with nothing connected to the inputs:



If I turn off the BW limiter the VPP goes up by about 100uV (20%). Offsets are there because the 'scope has only been on for a minute or so.

I have no idea what the Siglent boys were doing to get this:

[Rerouter]

without 20Mhz limiter, and unlocked to 200MHz, All channels 880-920uV peak to peak, 83-84 stdev,
Fitting a 50 ohm terminator is 840-860uV pk-pk, and 81 for standard deviation.

Bumping the gain up to 35db,
1M inputs,
890uV pk-pk, 90 stdev
50 ohm terminator,
860uv pk-pk, 87 stdev

[HalFET]

Scaling up the VGAC gain still gives practically the same values, So i would also pin it on ADC being the noise source,

at VGAC 35db gain, i measure 560uV of peak to peak noise, and a standard deviation of 54uV

Putting on the 50 ohm terminator brings it down to 450uV of peak to peak noise, and 43uV standard deviation.

I'm also happy that previously I misread the specs of the VGAC, so max VGAC gain is actually more like a 200uV range.

This is bad measurement to gauge a scope's performance. You can quite easily doctor these things to the point that it's entirely meaningless, especially once you start playing with the acquisition settings. Since you folks keep insisting, let me demonstrate with a DSOX2012A. Please note that my point here is to actually show what I'm doing, you can perfectly hide what you did if you really insist on it. This is what you get if you put the scope in its 10 mV/div range with a 100 ms/div time base immediately after turning it on:


Lets now turn on high-resolution mode:


But lets say we want to push it even further, lets turn on averaging (8 waveforms):


But since we're doing useless pointless measurements, why not go for the entirely ridiculous settings:


Now I could have hid what I did there and have claimed that Keysight makes scopes with a 26 µV noise floor. Could have probably gotten it closer to zero if I had bothered to find a short BNC terminator. Also note how the ADC on an Agilent scope doesn't need to run nearly as fast, which might give you a bit of a hint about what's actually going on here. Which is to say: pointless measurement.

[Fungus]

I have no idea what the Siglent boys were doing to get this:

Wait, yes I do! I fiddled around and found I can get their display if I zoom out and don't adjust the persistence. Here's two views of the exact same thing with different horizontal timebase:





They look completely different! In reality it's just the way the data is presented on screen. The first picture is the more correct display and also the most natural one in use.

(zooming in on signals is the most natural thing to do when you're using a 'scope).

nb. It's well known that the DS1054Z calculates RMS using the on-screen data so that also explains their "suspicious" numbers on screen. They either don't know how to use a Rigol or they're willfully ignoring that detail.

[gf]

If I turn off the BW limiter the VPP goes up by about 100uV (20%).

But when I see 1.36mV pp vs. ~450uV pp in the two screenshots, then it is rather a factor 3 (and not just 20% more)?
So obviously the BW limiter does help here.

[Rerouter]

I gave both as reference, the VGAC comes before the ADC, i have the datasheet for the chip, and A way of reading and writing to its gain register,

https://www.analog.com/media/en/technical-documentation/data-sheets/ad8370.pdf

After self cal, 500uV range is a High gain code of 45, I was raising this to a high gain code of 127,

Acquisition has no averaging or eres turned on, no persistence, just dots mode, with things left to sample about 10 times before I collected my ranges over lets say 5 sampling windows.

My limited understanding has it that if increasing the gain of your preamp by 2.5 times and there is no change the noise measured, then it must be after that point that the noise is occurring, The input terminator was to try and isolate the contribution of noise from the 1M input impedance.

please if you feel there is a better way to determine the noise source in the front end chain, I would be interested,

Edit: Attached is an image of the noise after I manually set the VGAC, all values measured are off by 8db, or a scaling factor of 2.51, as the scope UI doesnt know that I rescaled it. This was how I was measuring those values,

[Fungus]

But when I see 1.36mV pp vs. ~450uV pp in the two screenshots, then it is rather a factor 3 (and not just 20% more)?

The 1.36mV isn't my number - see my later post on display zoom for explanation of where that number likely comes from.

The 20% comes from enabling/disabling the BW limiter without changing horizontal timebase.

[Fungus]

If I was really interested in looking at PSU ripple (ie. not background/ADC noise) I'd be using either averaging or hires modes. That's what they're for.

[gf]

Averaging makes only sense, IMO, if the signal is periodic and if you have a stable trigger. Then it decreases the random noise and retains the non-random components of the signal.

[ Well, it may be non-peridoc, if the waveform in the time window around the trigger point is the same at each acquisition. ]

Hi-res, on the other hand, is basically oversampling + decimation with a box filter. So it does neither require a periodic signal, nore a stable trigger.

[gf]

I gave both as reference, the VGAC comes before the ADC...

Just a brief calculation:

In order to raise 500uV/div (which is about 5mV pp full-scale) to the ADC's 2V pp full-scale input voltage, a gain of 400V/V is required. If the VGA is limited to ~50V/V, then either an additional amp stage were required, or the ADC still needs to be operated with 8x digital gain. But only 8x - at this "low" gain level the HMCAD1511 noise is supposed to be still pretty good (see https://pdfs.semanticscholar.org/6063/513875078a4f4f6759895c1982d31fe5f53a.pdf).

[Fungus]

Averaging makes only sense, IMO, if the signal is periodic and if you have a stable trigger. Then it decreases the random noise and retains the non-random components of the signal.

That's why I said either/or.

Ripple is quite often periodic though. 

[rf-loop]

Rigol DS1000Z data sheet:
Vertical Scale
(Probe ratio is 1X) 1 mV/div to 10 V/div.

What does the probe ratio have to do with this? Do you actually understand what you're reading on datasheets?

If you did not understand what means sentence "Rigol DS1000Z data sheet:"  then I'm just so sorry.  Perhaps it need bend using iron wire. If I write what reads in Rigol data sheet and then you wonder like teenager in school why it include what is printed in data sheet... It is there because I quoted 1:1 Rigol data sheet. Ask Riglol why they write it.  (But I understand also why they write it there, sorry if you not.)

Is it now better:

Vertical Scale  (Probe ratio is 1X)  1 mV/div to 10 V/div.

As can see in attached image.

If I quote data sheet it is not my habit to selectively remove or edit something what is there.
But yes, I have not fully understood every single databook and datasheet what I have read during over 5 decades hobby+work with electronics but even after this, many things and machines still works in industry even in mil sector even today what I have done. Perhaps only good luck...or perhaps also because I read data sheets beforehand and not only after somethings goes wrong. "first brain, then muscle"
But it feels very nice that today I'm retired because If I need read seriously for work today's many data sheets etc...responsibility has been passed on to the reader just as the listener has full responsibility when the politician speaks. 

[HalFET]

Averaging makes only sense, IMO, if the signal is periodic and if you have a stable trigger. Then it decreases the random noise and retains the non-random components of the signal.

[ Well, it may be non-peridoc, if the waveform in the time window around the trigger point is the same at each acquisition. ]

Hi-res, on the other hand, is basically oversampling + decimation with a box filter. So it does neither require a periodic signal, nore a stable trigger.

Of course averaging doesn't make sense in most cases, but then again, neither does this measurement. The premise of using a plain scope without an amplifier for a low noise measurement is bonkers to begin with.

Rigol DS1000Z data sheet:
Vertical Scale
(Probe ratio is 1X) 1 mV/div to 10 V/div.

What does the probe ratio have to do with this? Do you actually understand what you're reading on datasheets?

If you did not understand what means sentence "Rigol DS1000Z data sheet:"  then I'm just so sorry.  Perhaps it need bend using iron wire. If I write what reads in Rigol data sheet and then you wonder like teenager in school why it include what is printed in data sheet... It is there because I quoted 1:1 Rigol data sheet. Ask Riglol why they write it.  (But I understand also why they write it there, sorry if you not.)

Is it now better:

Vertical Scale  (Probe ratio is 1X)  1 mV/div to 10 V/div.

As can see in attached image.

Then quote a relevant part, like the measurement uncertainty. All you did was say that it has a 1 mV/div to 10V/div range, which is entirely meaningless for this discussion.

If I quote data sheet it is not my habit to selectively remove or edit something what is there.
But yes, I have not fully understood every single databook and datasheet what I have read during over 5 decades hobby+work with electronics but even after this, many things and machines still works in industry even in mil sector even today what I have done. Perhaps only good luck...or perhaps also because I read data sheets beforehand and not only after somethings goes wrong. "first brain, then muscle"
But it feels very nice that today I'm retired because If I need read seriously for work today's many data sheets etc...responsibility has been passed on to the reader just as the listener has full responsibility when the politician speaks. 

Are you using Google translate? :|