Siglent SDS2000X Plus

[Bad_Driver]

This feature is very handy.

You can build your own probes (e.g. for digital check points with a 1 kohms resistor in the probe and the internal 50 ohms resistor as divider and set the attenuation value to 21:1)
and get the right readings.

Or you have a LNA with an amplification of about 100 and you calibrate it with a known input signal with the help of the channel attenuation factor. 
You can e.g. also include the correction for Gaussian noise distribution because we can not do any further math operations with the RMS readings from the measurements.
(or have I overseen something??)

This feature makes the readings of Vpp and RMS (and other vertical values) much easier and should work.

Has anybody else checked his device and can repeat this error?

[Bad_Driver]

Siglent services has agreed, that I found an unknown bug. They could reproduce this behavior.
They will take care of it.

[Martin72]

Test it with my HD, which got two user defined profiles per channel, no problems.

[tautech]

Test it with my HD, which got two user defined profiles per channel, no problems.

Yes as does the beta firmware for X Plus, 2 custom probe profiles, probe check and a few other fruits and fixes.

[Martin72]

Interesting to see that more and more benefits of the HD will be also avaible on the plus...

[tautech]

Interesting to see that more and more benefits of the HD will be also avaible on the plus...

They were already in 6kA. 

[Martin72]

And on the 5000 I guess.
Strong Scope lines from siglent...by the way:
Rigol ? Are you still here ? 

[tautech]

And on the 5000 I guess.

Yes IIRC, just sold my 5054X…..delivered yesterday.

[Bad_Driver]

One question that came into my mind this night:

How calculates the scope RMS?

[2N3055]

One question that came into my mind this night:

How calculates the scope RMS?

What do you mean?
Root mean square.. That is how? What exactly would you like to know?

[Bad_Driver]

Compared to a true RMS voltmeter as the HP3400 that uses thermo-coupling - how good is  a digital scope with non sinoid signals (e.g. pink noise)??

[thinkfat]

Limited, as always, by the sample rate and dynamic range. And accuracy, of course. Whether the waveform is sinusoidal or not doesn't concern an oscilloscope. It's just a DSP.

[2N3055]

Limited, as always, by the sample rate and dynamic range. And accuracy, of course. Whether the waveform is sinusoidal or not doesn't concern an oscilloscope. It's just a DSP.

Sample rate doesn't matter for RMS. Folded alias spectrum will simply be calculated properly in a wrong place... it's just energy integrated..
On 2000X HD, with 0,5% accuraccy, you will get pretty good results, as long as scope's own noise floor does not get in a way, like you said dynamic range...
So it will actually be quite wideband and decent accuracy, compared to old true RMS meters....

edited typo

[mawyatt]

Limited, as always, by the sample rate and dynamic range. And accuracy, of course. Whether the waveform is sinusoidal or not doesn't concern an oscilloscope. It's just a DSP.

Sample rate doesn't matter for RMS. Folded alias spectrum will simply be calculated properly in a wrong place... it's just energy integrated..
On 2000X HD, with 0,5% accuraccy, you will get pretty good results, as long as scope's own noise floor gets in a way, like you said dynamic range...
So it will actually be quite wideband and decent accuracy, compared to old true RMS meters....

Makes one wonder how good the new HD will compare to the KS34465A or DMM6500 which use computational techniques for RMS evaluation. Recall we did some tests awhile back that showed the X+ version did quite well within frequency ranges where the meters were most accurate.

Best,

[bdunham7]

ICompared to a true RMS voltmeter as the HP3400 that uses thermo-coupling - how good is  a digital scope with non sinoid signals (e.g. pink noise)??

Even the basic SDS1104X-E with a 10X probe and the 20MHz BW limiter is at least the equivalent of if not better than the HP3400A everywhere except at the lower limit of the HP3400A sensitivity where the noise floor of the scope takes over.  There are some slight differences in how you have to use them (the HP3400A can take peak voltages up to 10X FS and still be fairly accurate) but the DSO will properly calculate both TRMS_AC+DC (RMS) and TRMS_AC (StDev) simultaneously.  Or, if you like, you can AC couple it.

[Performa01]

In theory, an RMS voltmeter using thermal heating could have a high bandwidth and good accuracy. Yet the venerable HP 3400 is only specified up to 10 MHz and has its best accuracy of 1% at 1 MHz. At 10 MHz it can already be 5% off. The same applies to low frequencies below 50 Hz.

By contrast, a modern DSO can measure much higher frequencies; the SDS2354X Plus with 500 MHz option should be able to measure up to at least 200 MHz with no more than 5% additional error.

The table below shows my measurements with various waveforms at 1 MHz. The bandwidth of the white noise has been set to 10 MHz. The accuracy of the signal source is better than 1%. While the SDS2000X Plus cannot quite compete with the sweet spot of the HP 3400 at 1 MHz, its accuracy will remain fairly constant over a wide frequency range. As a side note, the 12-bit SDS2000X HD is noticeable more accurate.

The time base has been set to 10ms/div in order to get a lower bandwidth limit of 10 Hz for the measurements, so there’s a bit of 1/f noise included. The sample rate was 200 MSa/s and the 20 MHz bandwidth limiter has been used. No 10-bit acquisition mode, because this cannot improve the accuracy.

Waveform	f [Hz]	Ampl. [V]	Vpp [V]	AC-RMS [V]	RMS [V]		Crest [-]	Delta [%]
Base Noise	20,000E+6	33,330E-3	33,330E-3	4,896E-3	5,190E-3		3,211E+0	-99,481%
DC	0,000E+0	29,370E-3	29,370E-3	7,173E-3	1,012E+0		14,504E-3	1,246%
White Noise	10,000E+6	2,633E+0	12,684E+0	1,151E+0	1,151E+0		5,510E+0	15,106%
Pulse 10%	1,000E+6	2,098E+0	2,133E+0	598,539E-3	1,029E+0		1,036E+0	2,929%
Sine	1,000E+6	2,929E+0	2,929E+0	1,028E+0	1,028E+0		1,425E+0	2,780%
Square	1,000E+6	2,087E+0	2,124E+0	1,032E+0	1,032E+0		1,029E+0	3,220%
Triangle	1,000E+6	3,513E+0	3,513E+0	1,029E+0	1,029E+0		1,707E+0	2,906%

Please note that I’ve used the RMS measurement throughout, since the DC offset is not an issue at low sensitivities (vertical gain settings >5 mV/div). At very low levels and waveforms that don’t contain a DC component, we should use the Stdev measurement (= AC-RMS) in order to get rid of any unwanted DC offset in the DSO-frontend.

Not all measurements make sense for all waveforms. Amplitude measurement is only valid for square and pulse, AC-RMS (Stdev) differs from RMS as soon as a waveform has a DC component, which applies to the base noise, DC and pulse waveform.

Delta denotes the difference to the nominal value, and this is of course meaningless for the base noise. As expected, we get the best accuracy for DC, followed by a sine wave and it is worst for white noise, where the 8-bit acquisition system appears to approach its limits.

Attached is an example for the pulse measurement.

SDS2354X Plus_RMS_Pulse10%_1MHz_1Vrms

Compare this to the SDS2000X HD:

https://www.eevblog.com/forum/testgear/siglent-sds2000x-hd-12bit-(published-for-chinese-domestic-market-only)/msg4309894/#msg4309894

[Wintel]

And on the 5000 I guess.
Strong Scope lines from siglent...by the way:
Rigol ? Are you still here ? 

Rigol just released the HDO1000 and HDO4000 12-bit series yesterday.

HDO4000:
- 12-bit 4GSa/s ADC
- 4 analog channels, up to 800 MHz bandwidth
- 18 μVrms Front ends noise floor
- 100 μV/div ~ 10 V/div Vertical scale
- Up to 500 Mpts record length

https://rigol.com/Images/HDO4000_DataSheet_PDF_zh%20_tcm4-4912.pdf

[Martin72]

Where did you find it ?
I think it would be better to start a new thread about it
When you do this, I got a comment for it concerning the samplerate that drops down to 1GSa/s when using all channels.

[Bad_Driver]

In theory, an RMS voltmeter using thermal heating could have a high bandwidth and good accuracy. Yet the venerable HP 3400 is only specified up to 10 MHz and has its best accuracy of 1% at 1 MHz. At 10 MHz it can already be 5% off. The same applies to low frequencies below 50 Hz.

By contrast, a modern DSO can measure much higher frequencies; the SDS2354X Plus with 500 MHz option should be able to measure up to at least 200 MHz with no more than 5% additional error.

The table below shows my measurements with various waveforms at 1 MHz. The bandwidth of the white noise has been set to 10 MHz. The accuracy of the signal source is better than 1%. While the SDS2000X Plus cannot quite compete with the sweet spot of the HP 3400 at 1 MHz, its accuracy will remain fairly constant over a wide frequency range. As a side note, the 12-bit SDS2000X HD is noticeable more accurate.

Thanks, this was very helpful for me!
Makes a "low f" FFT (10 Hz....100 kHz) with the SDS2000X+ sense??
After buying a SSA I'm out of the "SDS-FFT business" for a while but the SSA starts at 100 kHz and is not useable for low f spectra's.
See my attached try. I'm missing a logarithmic scale for "f" 
 

[Performa01]

Makes a "low f" FFT (10 Hz....100 kHz) with the SDS2000X+ sense??

I’ve said it before and I’ll repeat it again that a general-purpose oscilloscope is neither a dynamic signal analyzer nor an audio analyzer, even when it can provide more than 8 bits of resolution. The guaranteed dynamic range of an 8-bit acquisition system is limited to about 49 dB and below 1 MHz the strong 1/f noise might cause further restrictions. Most importantly, the linearity of a 570 MHz wideband system cannot be the same as an audio-analyzer whose bandwidth only needs to be far less than one megahertz.

Nevertheless, the SDS2000X Plus and its FFT can be quite useful even at very low frequencies. Look at the first two screenshots, where I’ve measured a 10 Hz sine to 1.07% accuracy and a 100 kHz sine to 0.72% with the very same settings.

SDS2354X Plus_Meas_Sine_10Hz
SDS2354X Plus_Meas_Sine_100kHz

The accuracy of level measurements in this frequency range is generally quite good. The following screenshot demonstrates the FFT measurement of a 200 mVrms (= -14 dBV) sine signal in the range of 0-100 kHz. As can be seen, the measurement is spot-on around 1 kHz to 20 kHz and the error is barely exceeding 1% otherwise.

SDS2354X Plus_FFT_FR_100kHz

The next three screenshots show some distortion measurements for 33 Hz, 440 Hz and 20 kHz. The source is a low distortion (-96 dBc) sine wave, and we get about -60 dBc for the strongest harmonic at these frequencies. That’s certainly better than what we could normally expect from an 8-bit system, yet not nearly enough to characterize a high-quality audio system.

SDS2354X Plus_FFT_THD_33Hz_VBW1kHz
SDS2354X Plus_FFT_THD_440Hz_VBW20kHz
SDS2354X Plus_FFT_THD_20kHz_VBW100kHz

After buying a SSA I'm out of the "SDS-FFT business" for a while but the SSA starts at 100 kHz and is not useable for low f spectra's.
See my attached try. I'm missing a logarithmic scale for "f"   

Yes, we don’t have a logarithmic X-axis. It certainly could be implemented, yet the bin-width in any FFT is constant over the entire frequency span, hence a linear X-axis is the natural fit. Consider a range from 1 Hz to 100 kHz. For the lowest decade of 1-10 Hz we might want to have a bin-width of 0.1 Hz, whereas for the highest decade of 10-100 kHz a bin-width of 1 kHz should be plenty. The FFT has to be setup for the narrowest bin-width, so we had to use at least 200 kSa/s / 0.1 Hz = 2 Mpts FFT-length, which happens to be the maximum the SDS2000X Plus can deliver. So yes, it’s possible, but not very common.

Sorry, I’m not quite sure what you’re trying to demonstrate with your screenshot.

Is this the noise floor and some spurious signals of your instrument? Markers without corresponding table don’t make much sense…

A noise floor around -140 dBV looks rather impressive – that’s about 100 nVrms. What are you intending to do? Connecting a moving coil pickup system to the scope directly? Even then you should get a couple of hundred microvolts, hence see signals around -80 dBV.

I have checked the same frequency span up to 2 kHz on my SDS2354X Plus with slightly different settings and got a similar spectrum with the strongest spur of about -107 dBV at the mains frequency of 50 Hz.

SDS2354X Plus_FFT_NF_VBW2kHz

EDIT: description of the noise spectrum corrected.

[Performa01]

The FFT was (and still is) a hot topic in that other thread:

https://www.eevblog.com/forum/testgear/functional-comparison-of-rs-rtb2000-siglent-sds2000x-and-keysight-dsox1000/msg4304743/#msg4304743

There I’ve pointed out that the SDS2000X Plus can show two FFT plots simultaneously – and thanks to the formula editor the source data can be math results.

LISN: The SDS has a formula editor, which, among other things, allows FFT on the sum/difference of two input channels with only one math channel. That’s probably the trick that Keysight does implicitly to get that functionality out of a single math channel.

Since FFT is a math channel, the SDS can also show two FFTs at the same time. As a consequence, SDS could show one FFT on the common mode signal and another one on the differential mode signal simultaneously.

I’ve tried this and fed the very same pulse with 3 ns rise time and 10 ns fall time and a repetition rate of 160 kHz into channels 2 and 4 of the SDS2000X Plus simultaneously. The upper window shows an overview of the entire record, whereas the bigger window below shows the zoom view of the time domain capture as well as the two FFT traces. The red trace is for the differential signal and the light blue trace for the sum signal. So, in my simulation we have a lot of common mode noise and the differential signal is surprisingly weak, i.e. poor man’s differential probing provides a quite good CMRR of ~46 dB which is ultimately limited by the native resolution of the 8-bit system.

There’s a couple of screenshots showing the spectrum up to 1 MHz, 10 MHz, 100 MHz and 500 MHz – sorry, I forgot to disable the unused markers…

SDS2354X Plus_Dual_BW1GHz_Z1MHz
SDS2354X Plus_Dual_BW1GHz_Z10MHz
SDS2354X Plus_Dual_BW1GHz_Z100MHz
SDS2354X Plus_Dual_BW1GHz_Z500MHz

Yet another screenshot of the 500 MHz spectrum, this time with 10 dB/div:

SDS2354X Plus_Dual_BW500MHz_Z500MHz_Zoom

[Bad_Driver]

Makes a "low f" FFT (10 Hz....100 kHz) with the SDS2000X+ sense??

I’ve said it before and I’ll repeat it again that a general-purpose oscilloscope is neither a dynamic signal analyzer nor an audio analyzer, even when it can provide more than 8 bits of resolution. The guaranteed dynamic range of an 8-bit acquisition system is limited to about 49 dB and below 1 MHz the strong 1/f noise might cause further restrictions.

My interest is not any kind of audio stuff. For audio I have a 192kHz/24 Bit Terratec USB audio interface with analyzing software.

I'm interested in noise of power supplies and low noise amplifiers. That why I asked for RMS measurement with the SDS.
I try to compare the FFT results of the audio interface with the SDS FFT (which takes very, very long @ low frequencies).

But I appreciate your help very much, I learned so many useful things in this forum during the last years!

[Performa01]

I'm interested in noise of power supplies and low noise amplifiers. That why I asked for RMS measurement with the SDS.

I think the results are not too bad. The average DMM usually provides barely any better accuracy for AC and has a very limited bandwidth on top of that.

Just make sure to use the "Cycle" measurements of the DSO like I've demonstrated in my last posting, so you get valid results even when the period of the signal does not happen to be an integer fraction of the screen width. Of course this will only work for cyclic waveforms and doesn't apply to any form of noise.
 

I try to compare the FFT results of the audio interface with the SDS FFT (which takes very, very long @ low frequencies).

Well, it takes a long time to record very low frequencies. In your screenshot you've used a timebase of 5 s/div, so the record length is 50 seconds and consequently the lower bandwidth limit is 20 mHz. What's the lower bandwidth limit of your audio interface?

This way, your acquisition takes already 50 seconds for a single record and the additional ~2 sec. processing time for the 2 Mpts FFT can be almost ignored. The PC with the audio interface will compute the FFT at least ten times faster, but if a FFT down to 20 mHz is required, it has to acquire 50 seconds worth of data all the same.

Ideally, the parameters should match. There's little use in having a lower bandwidth limit of 20 mHz when at the same time the FFT-length and sample rate are configured in a way that the bin-width is 50 mHz. In general it is wise to stick with a maximum of 20 Mpts record length, configured in the Acquire menu, and use 200 Mpts only when you really need records that long. Of course, this also has to do with the required upper bandwidth limit, but it's hard to believe that an application requiring a lower bandwidth limit in the millihertz range requires a high upper bandwidth limit at the same time. All this is ultimately dictated by the maximum FFT-length, and 2 Mpts is pretty long anyway.

[points2]

Hi,
Did anyone open his SDS2000x+ to enhance its perf by tweaking this or that ?

I know it's a pretty weird question...
but I just ask because I'm used to tweak some PCBs on my hifi gear (full digital, no analogue parts) and in that kind of stuff (digital is real-time flow of data, where PDN of the PCBs plays a key role, as well as precise & stable clocking), when you lower the noise of a SMPS, and change an quartz or a XO oscillator by a good TCXO or OXCO (& PS accordingly) => music is nicer.

Based on that "experiences" (don't laugh please  ), I "think" that :
- the main PS into the SDS is a switching PS, thus ripple & noise and so on related to the SMPS
- on the PCB, that I didn't open yet, we can expect regulators (switchers) to get voltages x/y/z to this & that parts
- clock : 1 or several clocks ? I guess it's not a top-notch OCXO. Thus, we have a basic XO or a good TXCO (what kind of output ? no clue...) ?
well... a few basic tweaks that can enhance the hardware, and why not enhance the results.

Does anyone agree with that and has already done some experiments (successfull & non-lethal is preferred  ?
Rgds

[tautech]

Hi,
Did anyone open his SDS2000x+ to enhance its perf by tweaking this or that ?

I know it's a pretty weird question...
but I just ask because I'm used to tweak some PCBs on my hifi gear (full digital, no analogue parts) and in that kind of stuff (digital is real-time flow of data, where PDN of the PCBs plays a key role, as well as precise & stable clocking), when you lower the noise of a SMPS, and change an quartz or a XO oscillator by a good TCXO or OXCO (& PS accordingly) => music is nicer.

Based on that "experiences" (don't laugh please  ), I "think" that :
- the main PS into the SDS is a switching PS, thus ripple & noise and so on related to the SMPS
- on the PCB, that I didn't open yet, we can expect regulators (switchers) to get voltages x/y/z to this & that parts
- clock : 1 or several clocks ? I guess it's not a top-notch OCXO. Thus, we have a basic XO or a good TXCO (what kind of output ? no clue...) ?
well... a few basic tweaks that can enhance the hardware, and why not enhance the results.

Does anyone agree with that and has already done some experiments (successfull & non-lethal is preferred  ?
Rgds

So you think you're working at a level beyond the designers, good luck with that.
These instruments are already some of the lowest noise in their class and not some Yaigol Project knock together.

But if you must, then you must..............