New 2ch pocket DSO+SG - Zeeweii DSO2512G

[jebem]

By the way, did someone figured out at what exact vertical sensitivity setting will the scope start to exhibit the amplitude/bandwidth relationship issue?
Or the issue happens based only on the signal amplitude independently of the scope vertical sensitivity setting?

It's simple: the relay clicks when you switch from 200mV to 500mV vertical sensitivity in 1x mode or 2V to 5V in 10x mode and vice versa.
That makes the difference here afaics.
Actually I think the problem is not so dramatic. Up to ~50MHz it works anyway.
If you want something seriously better, you have to pay $100 more at least.
But it sure is good to know the limitations. It should be documented like the FY6900-100 does for the output voltage >20MHz.

Thank you, Aldo!

That makes more sense now.

The symptom is not so much a "amplitude/bandwidth" issue, but more of a "vertical attenuation/bandwidth" issue.

And this behavior  is common to find on several other low cost scopes, as they do not really use a proper active PGA circuit.
They choose to use simple ladder attenuator with resistors and some capacitors for frequency compensation, followed y a simple, although fast, operational amplifier IC, that is required to buffer the input and drive the ADC IC

For my needs on analog receivers, this scope is good enough. It will be able to read small RF signals up to 100 - 120 MHz with enough amplitude precision.

[gitm]

Aldo, I had a better, simpler idea in showing you how the slew rate affects the waveform and can limit bandwidth.  I fed the 20MHz square wave with 3 different amplitudes of 1Vpp, 2Vpp, and 3Vpp (nothing else changed) to both the 2512G and the MSO5204.  Pictures are attached, file names indicate what their contents are.

For the 2512G you'll note how at 1Vpp it is recognizable as a 20MHz square wave.  At 2Vpp is starts rounding off into a near perfect, but distorted sine wave. At 3Vpp I really doubt you could tell it was supposed to be a square wave if you didn't know it was supposed to be a square wave.

For the MSO5204, all three amplitudes are clearly recognizable as square waves.

[tatel]

So gitm put his finger on the wound and it turned out that they had painted it with a lipstick?   

Just guessing... that 36 MHz Fc is reached with just one channel active?

Well, 18 MHz with both channels active would be enough for me, even if it's in estu ta larri mode. More so if that Fc is greater with small signals. Good to know. But will that Fc stand up to even larger signals, say, 12 Vpp?

[Aldo22]

Hi gitm!

For the 2512G you'll note how at 1Vpp it is recognizable as a 20MHz square wave.  At 2Vpp is starts rounding off into a near perfect, but distorted sine wave. At 3Vpp I really doubt you could tell it was supposed to be a square wave if you didn't know it was supposed to be a square wave.

Thanks for testing!
I don't have an amplitude-adjustable SG for this voltage range.

But I have a different conclusion!

It is as I said before (also to jebem above).
The difference is with the "relay click", i.e. with the vertical sensitivity changing from 200 mV to 500 mV (in 1x mode).
Between 2 and 3 volts there is no difference in details in your pictures, the signal is just stretched because of higher amplitude (See attachment).
And the remedy for me is also clear.
Just take 10x probes and you're at 200/300mV at the inputs.
That's probably how it's supposed to be used.
It's a toy after all.

EDIT: Attached the resampled 3 V image H: 320px -> 240px. It's exactly the same signal shape as in the 2 V picture.

[Aldo22]

Btw. I'm experimenting with attenuators for direct connection without probes.
si5351 module and -20dB attenuator in the attached picture.
Probably not a flawless solution (Maybe you will explain why), but the square wave is clearly recognizable with relatively steep edges.

[tatel]

But I have a different conclusion!

@aldo22: That probably means it's bursting the seams on two sides. Anyway, I don't care. If this can get near 20 MHz bandwidth under real usage conditions, I wouldn't call it a toy. There was a time when people had to get by with much less than this, or even without any oscilloscope at all. We are lucky, for less than 100 bucks we can get our hands on something that was once a wet dream.

@gitm: should you have some spare time, would you mind provide us with screenshots illustrating the results of any more tests you could do? I for one would really appreciate it.

[Aldo22]

@aldo22: That probably means it's bursting the seams on two sides.

I do not understand what you mean.
Could you please use a little less "poetic" and more concrete language? 
I'm not a native English speaker and enough gets lost in translation anyway.

[tatel]

Of course. I mean it could have a double failure a) slew rate; b) attenuation.

I'm not sure about attenuation distorting the waveform, tough.

[Aldo22]

Of course. I mean it could have a double failure a) slew rate; b) attenuation.

I'm not sure about attenuation distorting the waveform, tough.

I don't see any connection with what I said.

I meant: The problem is not primarily the input voltage but the "vertical sensitivity" setting which at a certain point switches a relay.
At this point, it probably switches to a different way of processing the input signal (attenuation, amplification, filtering etc. I don't know, jebem seems to know how it works).
If the input voltage is in a range where both settings (200mV and 500mV at 1x) can fully display the signal, you will see a degradation in signal quality just by switching the vertical sensitivity from 200mV to 500mV without the voltage at the input changing.

Accordingly, the signal quality does not change when going from 2V to 3V at the input if the vertical sensitivity setting remains the same.

So if you make sure that the input voltage stays so low that the signal "fits" into the 200mV sensitivity screen, you have the full bandwidth.

I hope I made that clearer now.

[gitm]

Hi gitm!

For the 2512G you'll note how at 1Vpp it is recognizable as a 20MHz square wave.  At 2Vpp is starts rounding off into a near perfect, but distorted sine wave. At 3Vpp I really doubt you could tell it was supposed to be a square wave if you didn't know it was supposed to be a square wave.

Thanks for testing!
I don't have an amplitude-adjustable SG for this voltage range.

But I have a different conclusion!

It is as I said before (also to jebem above).
The difference is with the "relay click", i.e. with the vertical sensitivity changing from 200 mV to 500 mV (in 1x mode).
Between 2 and 3 volts there is no difference in details in your pictures, the signal is just stretched because of higher amplitude (See attachment).
And the remedy for me is also clear.
Just take 10x probes and you're at 200/300mV at the inputs.
That's probably how it's supposed to be used.
It's a toy after all.

EDIT: Attached the resampled 3 V image H: 320px -> 240px. It's exactly the same signal shape as in the 2 V picture.

Glad you're enjoying the journey.  Exploring is fun.  Remember when you asked if I could articulate in a comprehensible way what I meant by the BW definition including "waveform distortion"?  You tell me, did I?  BTW, you never said it had to be pain free. 

A side note: some op amp datasheets recognize distortion as being the limiting factor for their gain bandwidth product.  Food for thought.  Many op amp datasheets will list harmonic distortion figures and graphs.  And if you want to chew on that thought, check out this 10 minute video from Texas Instruments.... https://www.ti.com/video/4078676441001

Re: your setup with 20dB attenuator: signal at the scope input is Small Signal less than 1Vpp.  You'll find this on op amp datasheets as the "Small Signal Response" usually shown with a hand selected best figure (not only do scopes lie but some datasheets lie/mislead too) and/or small_signal_response/frequency graph.  But it can and often does differ from "Large Signal Response".

But, looking at your screenshot, I think I see reflections somewhere on the setup.  Just a guess, but they might be between the signal source and the input to the attenuator where there is a discontinuity in impedance.  The run between attenuator to scope input looks short enough that reflections might smear out becoming less noticeable.  Again, nothing more than a guess. 

If you really want to know, you could try decreasing sq. wave frequency (keeping the edge rise time and maxing amplitude) to make the signal stretch out and measure the ringing frequency with cursors.  I've done this with the 2512G using the internal signal generator at 1kHz and successfully measured ringing up to 90MHz.  You don't need vertical accuracy to measure that ringing frequency and all signals will devolve to sine waves anyway.  Once you've got a decent ringing frequency estimate, you can calculate it's wavelength (remember that the cable/connectors have a velocity factor -- on part datasheets for cable or if you don't have it assume 0.67 to get you in the ball park).  The wavelength will be in the ball park for the length of run that ringing originates, assuming only one thing is ringing.

ETA: I forgot to mention that a "length of run" has two ends and reflection can occur on one or both ends depending on impedance mismatches.  Frequency can be "interesting" as a result.  W2AEW (I think is a member of this board?) has the best video explanation I've seen on it https://youtu.be/g_jxh0Qe_FY

Peace out.



 

[Aldo22]

Hello gitm

Thank you for your reply and the video links.
I will have a closer look later.

Just a quick note regarding the attenuator experiment.
The si5351 is a pretty nasty fellow.
The signal is not as soft as that from your DG1062Z @20MHz.
And then we are certainly no longer in the range where the DSO2512G can show a "decent" square wave. I would say, that is about up to 10MHz.
With a si5351 @20MHz it's already quite challenged and still shows clearly no sine wave.
Check out how this looks on a "real" scope at Adafruit @13.5 MHz.
It looks very stable, but pretty rippled as well. I think it does not look that much different on the toyscope @20MHz.
https://learn.adafruit.com/assets/18759

[jebem]

Of course. I mean it could have a double failure a) slew rate; b) attenuation.

I'm not sure about attenuation distorting the waveform, tough.

I don't see any connection with what I said.

I meant: The problem is not primarily the input voltage but the "vertical sensitivity" setting which at a certain point switches a relay.
At this point, it probably switches to a different way of processing the input signal (attenuation, amplification, filtering etc. I don't know, jebem seems to know how it works).
If the input voltage is in a range where both settings (200mV and 500mV at 1x) can fully display the signal, you will see a degradation in signal quality just by switching the vertical sensitivity from 200mV to 500mV without the voltage at the input changing.

Accordingly, the signal quality does not change when going from 2V to 3V at the input if the vertical sensitivity setting remains the same.

So if you make sure that the input voltage stays so low that the signal "fits" into the 200mV sensitivity screen, you have the full bandwidth.

Based on the published test conditions from several members, I am inclined to conjecture that the bandwidth limitation issue would be related primarily to a less than adequate attenuator circuit, that is inserted at specific input sensitivity positions to reduce the signal to acceptable levels to feed the OpAmp that in turn drives the ADC.
Other factors will contribute to the final behaviour, of course, like poor (cloned) analog IC quality and/or implementation.

Again, I am not expecting low end scopes to feature active PGA or VGA active circuits.
What I have seen from other similar and even higher priced scopes is exactly the classic resistive with capacitor compensation non-linear networks used as attenuators.
That said, I am not dismissing the RC topologies, but they must be properly designed and tuned to ensure an acceptable flat response. Doing this for a 20MHz scope is one thing. Trying to do it for a 100MHz scope looks like a big challenge.

See here some discussion on analog front-end for scopes:
https://www.eevblog.com/forum/projects/diy-oscilloscope-(yet-again)/


 

[gitm]

Hello gitm

Thank you for your reply and the video links.
I will have a closer look later.

Just a quick note regarding the attenuator experiment.
The si5351 is a pretty nasty fellow.
The signal is not as soft as that from your DG1062Z @20MHz.
And then we are certainly no longer in the range where the DSO2512G can show a "decent" square wave. I would say, that is about up to 10MHz.
With a si5351 @20MHz it's already quite challenged and still shows clearly no sine wave.
Check out how this looks on a "real" scope at Adafruit @13.5 MHz.
It looks very stable, but pretty rippled as well. I think it does not look that much different on the toyscope @20MHz.
https://learn.adafruit.com/assets/18759

No, that does not look stable to me; it looks like classic ringing.  [See attachment] However, it appears to be CMOS logic levels and as long as the logic low/high signal doesn't bounce outside the CMOS levels, which does vary with logic family, it is irrelevant can be ignored.  Can't determine the rise time from the graph, but ring frequency is about 33MHz and it is damped -- classic ringing.   Overshoot is 740mV or about 22.3% which is substantial.

Si5351 datasheet shows typical 1ns rise time with 8mA drive (which is pretty fast) and the recommended characteristic impedance is 85 ohms.  So yeah, there's an impedance mismatch + fast rise time and.... ringing.  You could look at adding an impedance matching network to its output before attachment to the cable to help with that.  Lots of online impedance matching calculators with easy diagrams.

The DG1062Z is not as bad as the impression I've apparently given you.  I set up the DG1062Z with sq. wave @15.533MHz 3.2Vpp (accidentally, instead of the 3.32Vpp, took captures, realized but didn't want to go back and do it over again.  Too much time into this already.)    I have two captures, one with 50 ohm termination and the other with 1meg ohm termination. [See attachments]. With both, the rise time is about 10ns.  For 50 ohm, the avg. overshoot is 3.9% (std_dev=0.72%)  and for 1meg ohm the avg overshoot is 4.2% (std_dev=0.78%).  I would say that is "better" than the si5351 for some applications and "worse" for others.

[wasedadoc]

But, looking at your screenshot, I think I see reflections somewhere on the setup.  Just a guess, but they might be between the signal source and the input to the attenuator where there is a discontinuity in impedance.  The run between attenuator to scope input looks short enough that reflections might smear out becoming less noticeable. 

Aldo22's photo si_atten.jpg above shows his 5351 connected directly to the attenuator which is directly connected to the scope.  A few centimetres of adaptors but no cables.  Distance much too short to have impedance mismatch reflections impacting the frequencies of interest.

[gitm]

I agree.  I didn't take a close look at the picture and just assumed the si5351 was connected out of the picture through the wires shown.  Looking at the associated screenshot and trying to guess the period, it looks like ~10ns by eyeball mark-1.  Which would put the length at about 15 inches (38cm). 

However, the AdaFruit capture really, really looks like ringing.  If not trace length then some resonance/oscillator somewhere.

@Aldo, can you decrease the frequency to stretch out the signal but still keeping the rise time?  It would be interesting to see if damps down and dies away or if continues to oscillate.

[Aldo22]

Hi gitm!

@Aldo, can you decrease the frequency to stretch out the signal but still keeping the rise time?  It would be interesting to see if damps down and dies away or if continues to oscillate.

Do you mean like this? (5MHz, -20dB).
It keeps oscillating, but otherwise does not look too bad.
With the -10dB attenuator, it looks a bit smoother afair.

[wasedadoc]

^^ That is NOT ringing.  Ringing only occurs after a transition.  The waveform shown above contains pre-shoot.

Please lower the frequency even more. 100kHz. But keep the timebase at 20ns/div.

[Aldo22]

Please lower the frequency even more. 100kHz. But keep the timebase at 20ns/div.

This is the lowest frequency I can set with this setup. 370kHz

And now I have to go to bed. It's 11 PM here. 

[gitm]

^^ That is NOT ringing.  Ringing only occurs after a transition.  The waveform shown above contains pre-shoot.

Please lower the frequency even more. 100kHz. But keep the timebase at 20ns/div.

I agree that there might be pre-shoot in that capture, but the si5351 datasheet makes no reference to having pre-shoot, pre-emphasis, de-emphasis, or TxEQ.  I also haven't found anything online to indicate it does. That with Aldo's second capture (370 kHz) doesn't seem to show it which leaves me wondering.  And the adafruit capture doesn't seem to have pre-shoot as far I can tell which, of course, is not dispositive.

Perhaps Aldo, when he comes back to the forum, could capture on the falling edge just to see if it appears there immediately preceding the falling edge.   Maybe another capture setting the timebase to capture two cycles.

[Aldo22]

Thank you.
Whatever I try with the si5351, there is some ripple.
I don't really have any other option anymore than to order new accessories from Ali.
I tried with dummy load, -10/-20dB attenuators, direct, with coax, 250MHz probes with hook and with ground spring in 1x or 10x mode and many combinations.

But I found another cheap signal generator in the closet that I didn't even know could do more than a few kHz. A PWM generator.
You see in the attachment, that it shows less ripple.
The disturbance at the beginning is probably what you call "ringing".

And that of course brings me to the question of how to tell if a signal is representing "reality" (whatever that may be) or not.
Especially with cheap equipment, where you don't know if you can trust the signal source or the measuring device.

I mean, maybe the nice looking polished wave is not really there at all?

Maybe the wild looking wave is more like what the signal source is producing?

Is there any way to tell without expensive equipment?

EDIT: See also: https://www.eevblog.com/forum/testgear/new-2ch-pocket-dsosg-sigpeak-dso2512g/msg4863599/#msg4863599

[Aldo22]

Well, actually I still have a TinySA and (toy-) FFT in the DSO2512G to get some more information.
Maybe I can see the reason for the difference?
So I looked at the si5351 and the PWM generator at 5MHz with -20dB Attenuator (Necessary for TinySA).

There is a difference, but does it mean anything?

[gitm]

...

And that of course brings me to the question of how to tell if a signal is representing "reality" (whatever that may be) or not.
Especially with cheap equipment, where you don't know if you can trust the signal source or the measuring device.

I mean, maybe the nice looking polished wave is not really there at all?

Maybe the wild looking wave is more like what the signal source is producing?

Is there any way to tell without expensive equipment?

Ah, the 64 dollar question.  I think a lot of people starting on this journey wrestle with this question.  Sometimes clever testing techniques can give surprisingly good results (think 4-wire resistance measurements).  Sometimes custom testing fixtures, checking results against alternate methods, specially designed test circuits.  All a lot of effort.
 
A signal generator is generally useful and might be worth investing in.  If nothing else you could use it to a do Bode plot of the DUT, again comparing it to a reference of the same signal generator/scope but with conventional signal capture methods.

You could look other cheap gadgets, like at a poor man's pulse generator or build one, but either way you would need to characterize that device against the DG2512G using conventional signal capture methods (coax, probing, etc.). You would then have a "standard" to compare future tests against.  That si5351 was supposed to be your standard, but there's some ugliness going on with it.  You may have to try several alternates to get one that is acceptable. 

Re: the FFT in DG2512G, are you able to read any of the frequencies out of the display?  I can't.

[Aldo22]

Thank you!

Re: the FFT in DG2512G, are you able to read any of the frequencies out of the display?  I can't.

No, I don't know how to interpret that either. It is a gadget and not documented further.
But since the one with the PWM generator is recorded with the same time base, I think you can see a bit more even harmonics with si5351 in the right part?

[gitm]

Yeah, thanks for confirming.  What happens when an unexpected blip shows up in that FFT?  Or it's not a square wave signal with nice predictable 3rd, 5th, ... harmonics?

And the scale is another problem with it.  What are the units?  Log(v) or dBV.  It doesn't say.  So yeah, not useful.

[Aldo22]

Yeah, thanks for confirming.  What happens when an unexpected blip shows up in that FFT?  Or it's not a square wave signal with nice predictable 3rd, 5th, ... harmonics?

And the scale is another problem with it.  What are the units?  Log(v) or dBV.  It doesn't say.  So yeah, not useful.

Yes, for me is like a analog voltmeter without a scale. Maybe it is a bit useful if you know from experience where 5 or 10 volts are approximately.

Maybe someone who has more experience with FFT can interpret it.