Siglent SDS1104X-E and SDS1204X-E Mixed Signal Oscilloscopes

[radiolistener]

BW is dictated by the -3dB point, nothing else !
Significant signal attenuation is evident once you go past the rated frequency

Actually, the bandwidth is just the reciprocal value of the rise time and it doesn't depends on 3 dB.
The bandwidth and the rise time is just the same value, but expressed in different measurement units 

Max 3 dB attenuation is a requirement which must comply within the bandwidth.
Actually, frequency response may be even flat within the bandwidth of oscilloscope.
But it should not exceed 3 dB level, because this is the maximum limit for oscilloscopes.

The issue with the screenshot above is sample rate... it's too small for so high frequency.
There are just 28 points on entire screen and it leads to high measurement error, because there is no way to restore the waveform with sufficient accuracy. In order to get better accuracy, there is need more high sample rate. At least 2 GS/s or even more...

As you can see, there is 1.02 ns measured by oscilloscope. But this is impossible for 100 MHz input circuit. This is definitely measurement error, because input signal is out of range for oscilloscope specification...

[toli]

Its actually not completely accurate. This equation you are using of 0.35/rise time is only true if you assume a linear 1st order LPF response. It can be easily extracted from the capacitor charging equation by finding the time difference between 10% to 90% for the exponential function.
While this was true in the past, and might be true for these low cost scopes (although I doubt it is, have to verify their frequency response to do this), this isn't true for modern scopes of higher quality. They will normally have different filters/frequency response to get maximum flatness within the pass band, and then it'll drop much faster. There, this equation isn't true as it no longer behaves like a simple single pole LPF.

[radiolistener]

this isn't true for modern scopes of higher quality

I know about these modern high-quality oscilloscopes, that use 0.4 instead of 0.35 to determine the bandwidth.
I read their application notes about it.
But I believe this is just a marketing in order to write a little better specs, than it really has. 

There is no term "quality" in mathematics, so you cannot just replace one constant with another...

The rise time doesn't depends on frequency response flatness at all. It depends on sine function.
For example, let's calculate the rise time for sine wave with some frequency.
The rise time should be measured for specific threshold, usually it's 90%.

We need to measure the period of sine wave for value range [-0.9; +0.9]. Where sine wave has value range [-1; +1].
Let's split it on two equal parts [-0.9; 0] and [0; +0.9]. In that form, we can write it as equation:

sin( ω * t ) = 0.9

where t is is a half rise time period and 0.9 is a threshold.
From which we can find that:

t = arcsin( 0.9 ) / ω

So, now we can find rise time:

Rise time = 2 * arcsin( 0.9 ) / ω,

ω = 2 * pi * frequency, so:

Rise time = 2 * arcsin( 0.9 ) / (2 * pi * frequency) = arcsin( 0.9 ) / (pi * frequency) = period * arcsin( 0.9 ) / pi = period * 0.3564337

As you can see there is 0.35 multiplier. But it depends on threshold level.
And you can play with threshold level in order to get "better" specs for marketing purposes and to hide your bad real bandwidth 

You can find some pictures here: http://www.hit.bme.hu/~papay/edu/Lab/RiseTime.pdf

[toli]

Rise time is a measure of response to a step input. For a sin wave input there's no point of talking about rise time, its a single tone and the -3dB BW is all you need there. It is an inverse of the exponential function that describes the capacitor charging for a first order LPF that accounts for the 0.35 value in the equation.
No need to write it again, it can be seen in multiple places online such as this tech note:
https://www.k-state.edu/edl/docs/pubs/technical-resources/Technote2.pdf
just have a look at (15)-(25).
This is why modern scopes have values of 0.4-0.45, the roll-off of the filters causes it to behave differently than a single pole system. No need to try and put it down to marketing, scopes that use these filters will follow these number quite closely.

[radiolistener]

No need to write it again, it can be seen in multiple places online such as this tech note:
https://www.k-state.edu/edl/docs/pubs/technical-resources/Technote2.pdf
just have a look at (15)-(25).

This is exactly the same as I wrote above. But this is more complicated calculations with ln instead of sin.
But results is exactly the same, for threshold 90% it's also 0.35.

Rise time is just a time period for sine wave from low threshold 10% to high threshold 90%.
Where sine wave has a frequency equals to the bandwidth.

This is why modern scopes have values of 0.4-0.45.

They have the same 0.35 values, because sine function doesn't depends on oscilloscope quality and modern level   
But they use a little different threshold level for measurements.
In order to get specification which looks a little better than specs from other manufacturer

[toli]

I don't want to steer this topic too much off course, but this is simply not correct.
When you have multiple poles and zeros in the transfer function, the time domain response in no longer a single simple exponential term, this can be found from the inverse Laplace transform of the frequency domain transfer function. In that case the equations that you try to solve for 10% and 90% will give a different number than in the example I've linked above.
Again, no need to claim its sales issue or isn't, just go try it for yourself. You will see that the rise time for a step input doesn't match the 3dB point for sin input with this 0.35 ratio on these maximum flatness scopes.

[radiolistener]

Again, no need to claim its sales issue or isn't, just go try it for yourself. You will see that the rise time for a step input doesn't match the 3dB point for sin input with this 0.35 ratio on these maximum flatness scopes.

You can found that these modern oscilloscopes actually has a little smaller real bandwidth than specified in their specification.
If you ask manufacturer why there is a difference. You will get an answer about coefficient 0.4 instead of 0.35, because of modern technology, super-flatness and multiple poles transfer functions...   

I should say that we're talking about other oscilloscope manufacturer.
At least my Siglent SDS1102X has it's honest 100 MHz bandwidth. 

[Performa01]

Bandwidth is universally defined by the -3dB amplitude drop, unless otherwise specified, e.g. “0.5dB bandwidth”. This also applies for oscilloscopes. The bandwidths of the Siglent SDS1000X-E series have been accurately measured, so no need for speculation.

The 100MHz frontend in the SDS1100X-E has a fairly constant bandwidth of
  74MHz @ -1dB
110MHz @ -3dB
141MHz @ -6dB
for all vertical gain settings from 500uV/div up to 1V/div (not measured above).

The measurements are shown in the “SDS1104X-E Review 50-70” document that can be found here:

https://www.eevblog.com/forum/testgear/siglent-sds1104x-e-in-depth-review/msg1371775/#msg1371775

Attached to the same posting, there is also the “SDS1000X-E Bandwidth” document, which also covers the topic of sample rate.

The result in short: 2.5 times the sample rate is enough for an accurate reproduction of the bandwidth limited input signal, or put differently, 500MSa/s is enough for a 200MHz input signal. Yet there are no explicit anti-alias measures and even if so, it would be impossible to implement an ideal brick-wall filter. So aliasing can be an issue when more than two channels are in use and the sample rate drops to 500MSa/s. Yet it requires pulses with fairly fast edges to produce strong enough high frequency content above Nyquist to cause real problems.

As member toli has pointed out, the filter transfer function determines the relationship between rise time and bandwidth. Apart from that, a sine wave that doesn’t reach the full amplitude anymore (which in itself already proves that the bandwidth limit has been exceeded) can of course show rise times faster than specified and what would be related to the -3dB bandwidth by whatever halfway realistic factor.

Look what the SDS2304X can do. Does that mean it has a bandwidth of nearly 600MHz?


SDS2304X_Fake_RT

[radiolistener]

Look what the SDS2304X can do. Does that mean it has a bandwidth of nearly 600MHz?

I think if it has so low rise time and flat response (< 3 dB deviation for entire 600 MHz range), then it really can satisfy with 600 MHz bandwidth requirements. Just rise time is not enough, it should has at least some amplitude accuracy. 3 dB is about 30-40% error, it's too much for measurement device...

[toli]

I have received my SDS1204X-E (+MSO option) this week, and now over the weekend finally had a chance to try and play with it.
I have installed latest fw on both scope and MSO, as well as latest operating system. However, even following a very short experience with it (~1 hour), I have run into some bugs. At first, turning on the serial decode option made the scope freeze. After a minute of no response I've restarted it and enabled the decoder again. Now the decoder works but looks like there are a few problems with it:
1 - it will occasionally miss the entire sequence. So I can see the waveform and its just fine, but the decoder fails to recognize it. Running the same sequence over the SPI line again fixes it.
2 - The scroll option doesn't work at all for the list of decoded messages of the SPI decoder (didn't try I2C). I can turn the knob forever, and even punch in a number, does nothing for the list.
3 - The MISO/MOSI lines at the bottom of the screen don't scale in the horizontal scale with the waveforms when I change the time base after data has been captured. It just stays fixed at what it was when the waveforms were captured.

Anyone else gave it a try and seen similar issues? Can you scroll through the list of decoded messages?

Thanks.

EDIT:
To try and understand the cause of the problem I've modified the decoder to use the analog channels instead of the digital channels, and now I don't observe the same issues.
So there is definitely something wrong with the software/integration of the digital channels. Performa01/tautech/anyone else who has direct contact with Siglent, can you please forward this to the Siglent team and see what input they can give you? This is with latest fw on both units.
I can try and forward this to Siglent via the local distributor like I did with the fw bug I had in the SSA3021X, but it'll take much longer than the solution I've seen provided to problems reported from within this forum.

EDIT2:
A small video demonstrating the problems (2+3, I didn't capture problem 1 in this video):
https://drive.google.com/file/d/1Ar6GOoOX9GTlZQr1wLTwaW_rOyJ16NaU/view?usp=sharing
I have it in higher res in case anyone wants to, but its ~70MB uncompressed. Decode 1 is digital channels, Decode 2 is analog channels. I only use CH1 for clock, so MISO/MOSI in the analog channels case are constantly 0. CS is used as clock timeout for both decoders for an even comparison, but I did try with CS connected and observed similar results.

[ian.ameline]

So it is possible to hack an SDS1104X-E to be a SDS1204X-E.

You need to patch the OS update so that the root password is known -- another member of eevblog has done this (check over in the ADS file format thread - https://www.eevblog.com/forum/testgear/siglent-ads-firmware-file-format/125/ )

telnet to the scope, and log in as root. And execute the following commands and then power cycle the scope;

mount -o remount,rw ubi2_0 /usr/bin/siglent/firmdata0
cd /usr/bin/siglent/firmdata0
mv bandwidth.txt bandwidth.bak

[Mr. Scram]

So it is possible to hack an SDS1104X-E to be a SDS1204X-E.

You need to patch the OS update so that the root password is known -- another member of eevblog has done this (check over in the ADS file format thread - https://www.eevblog.com/forum/testgear/siglent-ads-firmware-file-format/125/ )

telnet to the scope, and log in as root. And execute the following commands and then power cycle the scope;

mount -o remount,rw ubi2_0 /usr/bin/siglent/firmdata0
cd /usr/bin/siglent/firmdata0
mv bandwidth.txt bandwidth.bak

Interesting. It'd make the Siglent a lot more competitive compared to the DS1054Z.

[BillB]

So it is possible to hack an SDS1104X-E to be a SDS1204X-E.

@janekivi @ian.ameline
If you have a beer fund to which I may contribute, let me know! 

[iMo]

You have to mess with capacitors in the front-end as well, it seems (only 3.54Vpp/100MHz with 4Vpp input).

[rf-loop]

Signal source Hewlett-Packard 8642B (calibration outdated)
0.8m Suhner RG223, Tektronix 011-0049-01 50ohm feedthru (Note 1),
Acg mode Normal, 1GSa/s, Channel 1, 50mV/div, Sin(x)/x, Coupling DC, Trigger rising edhe level 0mV
Level measurement: Measure Amplitude, statistics on, read average after >1000 count.

0dB Ref: 1MHz Sine, 600mVp-p (212mVrms) Ref Signal height p-p 6 division as is normal old thumb rule recommendation for oscilloscope freq response measurements (ref Tektronix)
Result:
-0.5dB @  70MHz
-1.0dB @ 114MHz
-1.5dB @ 178MHz
-2.0dB @ 197MHz
-2.5dB @ 216MHz
-3.0dB @ 231MHz

-3.5dB @ 244MHz
-4.0dB @ 269MHz
-4.5dB @ 280MHz
-5.0dB @ 289MHz
-5.5dB @ 301MHz
-6.0dB @ 316MHz
-10 dB @ 384MHz

Note 1:
50 ohm Feed Thru connected to oscilloscope Hi-Z input is never true 50ohm impedance termination over whole frequency range. This is of course because it is connected to oscilloscope input what have inductive and capacitive reactances. Most importtant mismatch in this case with these frequencies is of course scope input capacitive reactance. I have not available suitable normalizer.
But measurement is only for show that it is in that ball park what was expected.

[ian.ameline]

Mine has pretty flat response from 1 Mhz to 120Mhz after the hack... (Using coax, 50ohm terminated directly from a signal generator) - much flatter than before.

I wonder of someone who has a real 1204 can plot frequency response using 50 ohm terminated co-ax.

[rf-loop]

Mine has pretty flat response from 1 Mhz to 120Mhz after the hack... (Using coax, 50ohm terminated directly from a signal generator) - much flatter than before.

Yes, freq response shape is some amount different and flatness is better. It can also see in Performa01 review where is some compare with SDS1104X-E and SDS1202X-E.



Here  risetime using Tektronix 284 (around 70ps risetime)
Signal pathway: TEK284 out, Gen-Rad GR874-N adapter, Suhner N-BNC, bit under 30cm RG223, Tek 50 ohm FeedThru, Scope input BNC.
Scope: display dots, persistence 30s, acg normal, Sin(x)/x on,

[tmbinc]

The only thing the "bandwidth.txt" changes are the FIR filter constants. The 20MHz limit is switched in the frontend, but the frontend is not switched for 100 MHz vs. 200 MHz. (Now of course it could be that the SDS1204X-E uses different components in the frontend but I doubt that somehow).

So from what it looks like, they "just" apply some gain so move the -3dB points at the cost of more noise...

Here are the FIR filter banks. They are selected based on bandwidth (100MHz vs. 200MHz) and sample rate (1G, 500M, 250M i think) but I haven't figured out all details.

Code: [Select]

FIR[0][0] =  [ 0.00079346  0.00073242 -0.00085449 -0.00634766 -0.01455688 -0.01586914  0.00582886  0.06085205  0.13955688  0.21099854  0.23999023]
FIR[0][1] =  [ 0.00079346  0.00073242 -0.00085449 -0.00634766 -0.01455688 -0.01586914  0.00582886  0.06085205  0.13955688  0.21099854  0.23999023]
FIR[0][2] =  [ 0.          0.00125122  0.0032959   0.00231934 -0.00872803 -0.02700806 -0.02737427  0.02441406  0.13296509  0.24938965  0.29998779]
FIR[1][0] =  [ 0.00161743 -0.00170898 -0.02911377  0.01168823  0.27908325  0.48001099   0.27908325  0.01168823 -0.02911377 -0.00170898  0.00161743]
FIR[1][1] =  [ 0.00161743 -0.00170898 -0.02911377  0.01168823  0.27908325  0.48001099  0.27908325  0.01168823 -0.02911377 -0.00170898  0.00161743]
FIR[1][2] =  [ 0.          0.0065918  -0.01742554 -0.05477905  0.26596069  0.6000061  0.26596069 -0.05477905 -0.01742554  0.0065918   0.        ]
FIR[2][0] =  [ 0.          0.00576782 -0.03933716  0.1055603   0.85598755  0.1055603 -0.03933716  0.00576782  0.          0.          0.        ]
FIR[2][1] =  [ 0.          0.00576782 -0.03933716  0.1055603   0.85598755  0.1055603 -0.03933716  0.00576782  0.          0.          0.        ]
FIR[2][2] =  [ 0.         -0.00161743  0.00845337 -0.02197266  0.03717041  0.95599365  0.03717041 -0.02197266  0.00845337 -0.00161743  0.        ]
FIR[3][0] =  [-0.00039673 -0.00271606 -0.00787354 -0.01071167  0.00460815  0.05377197  0.13238525  0.20831299  0.23999023  0.          0.        ]
FIR[3][1] =  [-0.00039673 -0.00271606 -0.00787354 -0.01071167  0.00460815  0.05377197  0.13238525  0.20831299  0.23999023  0.          0.        ]
FIR[3][2] =  [ 0.0015564   0.00097656 -0.00469971 -0.01818848 -0.02166748  0.02157593  0.12612915  0.2461853   0.29998779  0.          0.        ]
FIR[4][0] =  [-0.00082397 -0.01571655  0.00924683  0.26473999  0.48001099  0.26473999  0.00924683 -0.01571655 -0.00082397  0.          0.        ]
FIR[4][1] =  [-0.00082397 -0.01571655  0.00924683  0.26473999  0.48001099  0.26473999  0.00924683 -0.01571655 -0.00082397  0.          0.        ]
FIR[4][2] =  [ 0.00314331 -0.00939941 -0.04336548  0.25228882  0.6000061   0.25228882 -0.04336548 -0.00939941  0.00314331  0.          0.        ]


[rf-loop]

So from what it looks like, they "just" apply some gain so move the -3dB points at the cost of more noise...

Of course noise level rise, this is normal price of higher BW, you meet this fact everywhere... in theory and in practice.
Basic fundamentals.
Power spectral density of "white" noise is constant over frequency but this is of course true only for white noise (what I have never seen in my life) but no need think it,
This mean that noise power is proportional to bandwidth. If  measurement bandwidth is double, detected noise power will double (an increase of 3 dB) --- in theory and in practice! (if noise is enough "white" in context)
If do not go more deep to details but it is extremely natural that noise level rise whjen BW rise.
Totally other case is if we rise BW and build circuit what have less noise using differrent components and circuit what have less noise. But again, if we then rise this "low noise" circuit bandwidth, same happen, noise rise.

Also it can see quite well here. Around 3dB rise.

Measured noise (1us/div, normal mode)  inputs 1M, 500uV/div, protected from external EMI using BNC cap.



Old test (in this thread previously)

This new test have same settings so they are quite highly comparable and also same individual scope.


If we look how much it rise and take randomly example CH4. Level (Vrms ¹)  rise 3.14dB
And if we look how our BW shape and BW width change,  it can say that it is just as expected.
1. V Stdev = V rms - possible DC offset

[tv84]

The "official" BW configurations of the SDS1004X-E model are 50, 70, 100 and 200Mhz.

As said previously, the fw has references to 250 and 300 MHz inside. But, among many other references, that has nothing to do with this scope.

Tests were made with 250M and 300M and proved that the scope didn't recognize these BWs and reverted to the basic 50MHz config.

[tautech]

The "official" BW configurations of the SDS1004X-E model are 50, 70, 100 and 200Mhz.

Maybe but no 50 MHz version exists for market.
Only the Chinese market get to see 70 MHz versions in the X-C range.

[tv84]

Maybe but no 50 MHz version exists for market.
Only the Chinese market get to see 70 MHz versions in the X-C range.

OK, let's correct that:  The only BW configurations of the SDS1004X-E model are 50, 70, 100 and 200Mhz.

These are guaranteed by the current fw and have been tested.

[insine]

I'm wondering if these coefficients are the same for every scope or maybe they also include factory calibration individual for every unit.

[rf-loop]

The 20MHz limit is switched in the frontend, but the frontend is not switched for 100 MHz vs. 200 MHz.

Can you prove that this is a fact. Where are evidences?
Have you analyzed 4 channel models analog front end signal pathway from AD8370 to ADA4932?

[BillB]

Looking at Performa01's review of the probes that come with the 1104 vs the 1204, the charts and graphs provided seem to show that they have similar 3dB points and rise times, but the PP215s have a much wider 1dB BW.   

Does it make sense to spend the hundred bucks and bump up to the PP215's from the PP510's?