Siglent SDS2000X Plus

[inxkwit]

Hey there! I've been doing some level of analysis with voltage dividers and my logic 8 but I feel like it's time to upgrade. I'm heavily considering the SDS2000x plus line (I see the info about unlocking capabillities) but notice the lowest end model has been discontinued. Is this something siglent does normally, or could this indicate a newer line coming (I didn't find results in my searches)? In looking at their release schedule it typically seems pretty agressive, I think I'm just looking for some reassurance before I pull the trigger on the SDS2104X Plus.

Struggling to choose probes to pair, definitely going to grab the LA, but probes capable of sniffing around high frequency ic/ic communications would be nice. Seems the active probes for siglent aren't compatible with the sds2000x.

[Martin72]

I'm heavily considering the SDS2000x line (I see the info about unlocking capabillities) but notice the lowest end model has been discontinued.

From the 2000XPlus Line
I think you mean the X without plus...

[inxkwit]

Sorry I did mean the Plus. Edited for clarity.

[tautech]

Hey there! I've been doing some level of analysis with voltage dividers and my logic 8 but I feel like it's time to upgrade. I'm heavily considering the SDS2000x line (I see the info about unlocking capabillities) but notice the lowest end model has been discontinued. Is this something siglent does normally, or could this indicate a newer line coming (I didn't find results in my searches)? In looking at their release schedule it typically seems pretty agressive, I think I'm just looking for some reassurance before I pull the trigger on the SDS2104X Plus.

Struggling to choose probes to pair, definitely going to grab the LA, but probes capable of sniffing around high frequency ic/ic communications would be nice. Seems the active probes for siglent aren't compatible with the sds2000x.

Welcome to the forum.

2000X Plus range replaces the SDS2000X range which was developed from the SDS2000 range, released in 2013.

For active probes SDS5000X models and higher have Siglents S-Bus that supports their own probes and some Tek and LeCroy probes using Siglents active probe adapters.
Currently SDS2104X Plus is the best priced offering however doesn't support active probes unless they have their own power source and only a BNC connector.

Have a crash around in the EU/US website where once you click on Products, then on the far right of the dropdown is the Discontinued model list:
https://www.siglenteu.com/#
https://siglentna.com/

[inxkwit]

Thanks so much! Happy to be here. Thanks for the info! From what I've been reading I guess I want to minimize capacitance of my probes if I intend to work with high speed signals. Seems like the SP3050A don't present the greatest value?

Good idea, thanks! Did I miss it on the NA site or do they hide it for some reason?

[tautech]

Thanks so much! Happy to be here. Thanks for the info! From what I've been reading I guess I want to minimize capacitance of my probes if I intend to work with high speed signals. Seems like the SP3050A don't present the greatest value?

There are other ways of skinning the low capacitance cat !
Most of Siglents better scopes have high sensitivity inputs so for signals in your normal working range you can reduce tip capacitance by using 100:1 probes.

For shits and giggles some years back I did a short vid showing using 1:1, 10:1, 100:1 and 1000:1 probes on the 3V probe compensation output while using the little X-E's webserver but you get the idea......
https://youtu.be/W30AqcDQL18

[Martin72]

Hi,

After again several tryings and adjustments, I giving it up to reproduce the results 2N3055 was getting.
Maybe there is still a bug in the math functions or the sds2k+ are simply not capable to do this - Or it´s me who´s too dumb for it.
With the last I could live..
EDIT:
Am I the only one who owns a bodnar AND a sds2k+....Can´t believe it, so guys feel free to replicate the measures...

[BmaxTom]

Is it possible that the dx/dt (F1) function is not working correctly or is not calculating with the data displayed by C4? I think the maximum slope of C4 is not at -0.5ns but at 0.0ns.

Function d(Avg(C4))/dt (F2) fits better.

What I noticed while trying it out is that the formula editor always displays the last formula entered and not the current formula of function 1 or 2.

[gf]

Is it possible that the dx/dt (F1) function is not working correctly or is not calculating with the data displayed by C4? I think the maximum slope of C4 is not at -0.5ns but at 0.0ns.
Function d(Avg(C4))/dt (F2) fits better.
What I noticed while trying it out is that the formula editor always displays the last formula entered and not the current formula of function 1 or 2.

I checked the manual to see what d/dt is really supposed to do (see screenshot).
Do your F1 and F2 possibly use different dx for the finite difference approximation?

EDIT:

I find it actually a bit strange that delta t is specified in div units, where 1 div = 50 pixels.
(I'd rather expected to specify either number of samples, or seconds.)
And I also wonder whether 1 div = 50 pixels can be correct, if the screen has 10 div and screen width is 1024 pixels?

Sorry, was obviously an old version of the manual - the new one specifies dx in points.
Replaced screenshot. Still it sais 1 point = 0.01 div. So dx is still specified in 1/100 div units (and neither samples nor seconds)?

[gf]

I checked the manual to see what d/dt is really supposed to do (see screenshot).

If dx is (say) 4 samples, then the fnite difference approximation is equivalent to a convolution with the FIR kernel [ -0.25 0 0 0 0.25 ].
The corresponding transfer function in the frequency domain is attached.
For a true differentiation, the magnitude of the corresponding transfer function in the frequency domain were proportional to frequency.
As we can see, for the finite difference approximation this is only granted for frequencies up to about fs/20.
At higher frequencies, the applied transfer function has nothing to do with differentiation any more. The same also happended here (obviously dx happened to be 4 samples here).

So it is only suitable for low frequency signals (with highest frequency content < fs/20 fs/(5*dx), or for data that have been interpolated/oversampled by a factor of >= 20 5*dx.
EDIT: Generalized 20 to 5*dx, where dx is in samples.

[Martin72]

OK.....

And what´s the reason why the same formula "works" on 2N3055´s scope ? More samplerate?
I have the suspicion rf-loop was right

[inxkwit]

If my understanding is correct, for a passive probe to accept such high bandwidth it needs to be fairly high quality as it will be dealing with pretty significant power? I'm struggling to find much in the 500mhz 100/1000x range.

The P5100a seem like they tick a lot of the marks, but will I really benefit from such expensive probes? It does seem they can do around 10V at 500mhz which should be plenty for most signals I'm trying to identify and interact with, but am I missing something or is this the going rate for a high BW, low impedance, 100x probe?

If I'm going to be spending another 2.4k+ on a set of 4 probes, could I get a better scope that has more capable probes included?

[bdunham7]

but am I missing something or is this the going rate for a high BW, low impedance, 100x probe?

If I'm going to be spending another 2.4k+ on a set of 4 probes, could I get a better scope that has more capable probes included?

I think what you are missing is that 500MHz with a high-impedance probe is not something anyone normally tries to do.  (I'm assuming you didn't mean 'low impedance')  If  you relax your requirements to a more reasonable 200MHz/6pF, then the generic eBay P4250 or P2301C models will probably suffice.  If you must have 500MHz with a passive probe, you might look for a used P6156 which is a low impedance (500R, 1000R, 5K), low capacitance (<1pF), high BW (3.5GHz) model that works with your 50R scope input.  You have to deal with the fairly low input resistance, but at 500MHz the capacitance dominates anyway.

[inxkwit]

Whoops, totally was thinking capacitance, my bad. I'll look around at some of the other models you've mentioned, but figure it might be worth buying one pair that will cover max capabilities of the device and never have to worry about it.

[bdunham7]

but figure it might be worth buying one pair that will cover max capabilities of the device and never have to worry about it.

IMO definitely not worth it and you'll always have to worry about it no matter what you do! 

For one thing, even if you get those big$ probes, unless you have a special, exact way of grounding it at your source, your signal will be pretty distorted by ~200MHz anyway.

Your time and money would be better spent scrounging up some type of active probes. 

[inxkwit]

Awesome. I really appreciate all of the input. Will grab the siglent and take it for a whirl in the next week or two hopefully and go from there.

[gf]

OK.....

And what´s the reason why the same formula "works" on 2N3055´s scope ? More samplerate?
I have the suspicion rf-loop was right

At <= 20ns/div your interpolated sample rate was reported >= 50GSa/s, which is >= 25x oversampled.
As I said, with this amount of interpolation/oversampling, the d/dt approximation is OK (unless you would increase the step size significantly) -- and your d(C1)/dt trace looks fine at 10ns/div.
Next thing to investigate would be the next intermediate sub-expression Average(d(C1)/dt)), or d(Average(C1))/dt. I'd display traces for both ones, to see if it makes a difference.
Both should look like d(C1)/dt. Ideally all three ones should differ only be the amount of noise.

I'm still irritated that the number of FFT points cannot be set manually to the desired value. More points were actually essential. Is there possibly a limitation if the source of the FFT is a math sub-expression?

What also comes into my mind, btw: Which window function did you select for the FFT? Since you have only 512 points (and ~100MHz frequency spacing), don't use a window function which "smears" adjacent bins, but rather stay with rectangular window.

[Martin72]

Hi,

Tried all the FFT windows with more or less effects.
What the points concerns, there is, for me, a strange behaviour to see.
The shorter the timebase gets, the less memorypoints you got, the less FFT points you got too, all make sense.
But...
At 50ns you have 512 points, at 20ns 1024, at 10ns back to 512 ?
I didn´t change anything inbetween, just playing with the timebase.

I'm still irritated that the number of FFT points cannot be set manually to the desired value. More points were actually essential. Is there possibly a limitation if the source of the FFT is a math sub-expression?

You can set it up to 2 million points but it´s the same like setting the memory, with decreasing the timebase from ms to ns, less memory are avaible and the manual setting will be useless.

[gf]

You can set it up to 2 million points but it´s the same like setting the memory, with decreasing the timebase from ms to ns, less memory are avaible and the manual setting will be useless.

The question is how to get both, interpolation AND a larger number of FFT points?
You say that you need a slower timebase to get more FFT points, but then it does not interpolate any more
(And without interpolation, at 2GSa/s, the d/dt approximation is only good enough up to about 100 MHz)

[Martin72]

Thankyou, it make some things clearer for me.
Here the setting 2N3055 got on his scope at 10ns.
More samplerate(2.5x), more memory points(2.5X), more FFT points(4x).
It seems it´s enough to get the expected answer.

[gf]

According to the handbook, there also seems to exist an "Interpolate" math function.
Can it be used to interpolate explicitly, even when a slower timebase is used?
If yes, how long are the records generated by this function? (with 20x interpolation, they would need to be 20x longer than the input record -- if they are not truncated)

[2N3055]

Sorry guys,  I was busy..

I unfortunately don't have SDS2000X+, only that SDS6104H12 to try.
And also can try on one scope similar to 2000X+ but higher res ....  .
I'm not sure if any of those two behave exactly as SDS2000X+.

If you guys enable Averaging as an Acquisition mode, then scope will interpolate more points...
Try that. On both ones I have here it increased apparent sample rate for FFT..

Also gf is right.
There is "interpolate" math function, that it would be simpler if it was called "upsample".
It resamples current samples to higher sample rate by interpolating from real data. I guess name comes from that.
On that "scope similar to sds2000X+" I can upsample by factor of 20X.

So if you use Averaging acquisition mode, and math formula FFT(d(Interpolate(C1))/dt) with interpolate 20X I get this:



SDS6000H12 with it's 5GS/s seems to have sample rate high enough by default... But even with it this method helps..

[Martin72]

Hi 2N3055,

Quickshot....

Edit:

So if you use Averaging acquisition mode

On SDS2K+,  averaging is only as a math function avaible..

[2N3055]

Hi 2N3055,

Quickshot....

Edit:

So if you use Averaging acquisition mode

On SDS2K+,  averaging is only as a math function avaible..

Oh it shows that I don't have it... I forget they are not exactly the same...

I would say something is not right with that FFT...
It shows effective sample rate as 4 TS/s (4000 giga).
Could you try adjust it to be somewhere at 200-400GS/s so bin width is not half scope's bandwidth.
Play with upsample rate until you get bin width of 20 ish MHz..

Someone else with fast pulse gen and SDS2000X+ should try..
Maybe there is a bug lurking somewhere in there..

[Martin72]

Will do....
....a couple of hours later..

I would say something is not right with that FFT...

Yeah, this is what I got in mind since I´ve tried it and recognized the thing with the changing points (50ns 512pts, 20ns 1024pts, 10ns 512 pts).
Also we got a bug solved with the last (?) firmware, concerning FFT and memorysize.
Maybe there are something more to correct.
But I will check your suggestions in appx 12 hours..