New 2ch pocket DSO+SG - Zeeweii DSO2512G

[4thDoctorWhoFan]

To be safe, I would use some attenuation externally.

Thanks.
Unfortunately I don't have any SMA-SMA attenuators and I can imagine that some tinkering interferes with the signal.

dBm means power (like watts) right?
If I have no idea what the signal generator is doing, how do I measure it?
Do I measure the voltage and then calculate something against 50 ohms impedance?
Sorry for noob questions.

Certainly, you are able to adjust the output of a signal generator.  The O-scope is what you can use to measure the output of the signal generator.
Perhaps the attached chart will help you.
For instance 0dbm = 1mW (milliwatt) = .224Vrms = .316Vp (Peak) = .632Vp-p (Peak to Peak).  The scope we have been talking about can measure the frequency along with the Volts rms, peak or peak to peak.

[Aldo22]

Certainly, you are able to adjust the output of a signal generator.  The O-scope is what you can use to measure the output of the signal generator.
Perhaps the attached chart will help you.
For instance 0dbm = 1mW (milliwatt) = .224Vrms = .316Vp (Peak) = .632Vp-p (Peak to Peak).  The scope we have been talking about can measure the frequency along with the Volts rms, peak or peak to peak.

Ah, thanks!
I wasn't so wrong that these values are calculated with reference to 50 ohms.
Then for a limit of +10dBm 707mV RMS would be the maximum, right?

Thank you!

[4thDoctorWhoFan]

It's hard to read the settings of your TinySA in the picture.
What is the level output set to on the TinySA?  Did you adjust the level output as you increased the frequency or did it remain the same?

Everything default: -7.0dBm
An no, I did not increase the level output of course.
I didn't change anything except the frequency.

Thank you!

Using -7dbm input, I was able to get a 97Mhz bandwidth in normal mode.  The waveform becomes a little shaky around 70Mhz but still useable.
Using the same -7dbm input in overclocking mode, I was able to achieve 120Mhz.  Pretty impressive actually.

Remember, this is at -7dbm (100mVrms).  With my unit, anything over 300mVrms, the bandwidth drops to about 35Mhz.

[4thDoctorWhoFan]

Certainly, you are able to adjust the output of a signal generator.  The O-scope is what you can use to measure the output of the signal generator.
Perhaps the attached chart will help you.
For instance 0dbm = 1mW (milliwatt) = .224Vrms = .316Vp (Peak) = .632Vp-p (Peak to Peak).  The scope we have been talking about can measure the frequency along with the Volts rms, peak or peak to peak.

Ah, thanks!
I wasn't so wrong that these values are calculated with reference to 50 ohms.
Then for a limit of +10dBm 707mV RMS would be the maximum, right?

Thank you!

Correct.
Although, I would NOT want to push the upper power limits on any device which is why you should use an external attenuator or make sure you are below the maximum input level.

[JeremyC]

Okay, I figured out the problem.
The scope cannot handle amplitudes higher than about 300mVrms and maintain bandwidth specs.

I retested the scope using your -7dbm (100mvRMS) settings and I got the same results you did.  The bandwidth specs was much better.  However, once I got above 300mVrms input level, the bandwidth dropped like a rock like in my initial test with 1Vrms.

Not sure why this is the case.  Perhaps someone here with more knowledge can shed some light.

Did you use sine wave in the tests?

[Aldo22]

Using -7dbm input, I was able to get a 97Mhz bandwidth in normal mode.  The waveform becomes a little shaky around 70Mhz but still useable.
Using the same -7dbm input in overclocking mode, I was able to achieve 120Mhz.  Pretty impressive actually.

Yes!
And now there is also an explanation why there are such different opinions about the bandwidth: the amplitude.
Maybe with 10x or 100x probes less a problem?

[4thDoctorWhoFan]

Okay, I figured out the problem.
The scope cannot handle amplitudes higher than about 300mVrms and maintain bandwidth specs.

I retested the scope using your -7dbm (100mvRMS) settings and I got the same results you did.  The bandwidth specs was much better.  However, once I got above 300mVrms input level, the bandwidth dropped like a rock like in my initial test with 1Vrms.

Not sure why this is the case.  Perhaps someone here with more knowledge can shed some light.

Did you use sine wave in the tests?

Yes, sine wave.
My source is a HP 8648C.

[tatel]

And now there is also an explanation why there are such different opinions about the bandwidth: the amplitude.

Not really an expert but this first time I heard about an oscilloscope whose bandwidth depends on signal amplitude. 35 Mhz would be good enough, though.

[tatel]

A couple videos about DSO154 behavior.

https://www.eevblog.com/forum/testgear/zeeweii-toyscopes-how-do-they-do-it/msg4828307/#msg4828307

Do you would say it works good up to the advertised frequency (18 MHz IIRC)?

So wasedadoc perfectly described how aliasing works, one can see how measured frequency is decreased, not increased, in 1 MHz steps on the sine wave video above 20 MHz

However, we were wrong, it doesn't seem there is a frequency counter, it looks as if it's counting over the sampled (aliased) signal?

Aliased frequencies can be differentiated from the real ones from the fact that amplitude is lower?

DSO154 is said to do 40 MSa/s, very interesting to see what it does at 20 MHz frequency

Square wave video is even funnier

Does DSO2512 behave the same way?

[Aldo22]

Do you would say it works good up to the advertised frequency (18 MHz IIRC)?

You know that I am not an expert, but I would say it works as I would expect it from my experience with the DSO2512.

It can keep up counting the frequency within the advertised bandwidth and has ~-3dB at 18MHz.
Maybe the amplitude drop comes bit earlier than his big brother's.
I've shown in the screenshots above that there is very little amplitude drop even at 90MHz in normal mode.
Beyond the advertised bandwidth, the frequency count is incorrect. That is the same with the DSO2512.

from my little experience, a square wave looks more or less decent up to 1/10 of the bandwidth.
It can be distinguished from a sine wave up to about 1/3 of the bandwidth.
That would be ~1.8MHz and ~6MHz for the DSO154 and that's what I'm seeing in the second video.
It is ~9.6MHz and ~40MHz for the DSO2512.

Thanks!

[tatel]

I guessed this is the first time I'm getting a motive to do a FFT and so see how these harmonics that make the square wave to dance as if it were high on amphetamines. Probably it's academic but here they are.

[tatel]

Do you would say it works good up to the advertised frequency (18 MHz IIRC)?

You know that I am not an expert, but I would say it works as I would expect it from my experience with the DSO2512.

It can keep up counting the frequency within the advertised bandwidth and has ~-3dB at 18MHz.
Maybe the amplitude drop comes bit earlier than his big brother's.
I've shown in the screenshots above that there is very little amplitude drop even at 90MHz in normal mode.
Beyond the advertised bandwidth, the frequency count is incorrect. That is the same with the DSO2512.

from my little experience, a square wave looks more or less decent up to 1/10 of the bandwidth.
It can be distinguished from a sine wave up to about 1/3 of the bandwidth.
That would be ~1.8MHz and ~6MHz for the DSO154 and that's what I'm seeing in the second video.
It is ~9.6MHz and ~40MHz for the DSO2512.

Thanks!

I agree. However I would like to know what our more experienced fellow forum members would have to say.

If, as said, DSO2512 does sample at 200/100 MSa/s and does sin(x)/x interpolation, that would add up with the results? 100/2.5 = 40 MHz.

However I doubt it's being tested with both channels activated?

Is it sampling at 5x?

Do you think it works good enough up to 20 MHz with both channels activated?

Would be DSO2512 good enough to do some low-end but serious work or is it just a toy?

Anyone?

[Aldo22]

I agree. However I would like to know what our more experienced fellow forum members would have to say.

If, as said, DSO2512 does sample at 200/100 MSa/s and does sin(x)/x interpolation, that would add up with the results? 100/2.5 = 40 MHz.

However I doubt it's being tested with both channels activated?

I don't quite understand how you come up with these numbers.
I have already shown the different "bandwidths" of this device here before.
By this I mean the limit up to which it can still measure the frequency, not the absence of "artifacts".

2 channel/standard mode: ~48MHz
2 channel/OC mode: ~60MHz
1 channel/standard mode: ~96MHz
1 channel/OC mode: ~120MHz

The 40 MHz you mentioned refer to the square wave. It is the state you see on the DSO154 at 6MHz.
The DSO2512 is specified with 500Ms/s. I guess in reality it is probably like this (Or they lie!  ):

2 channel/standard mode: 200Ms/s
2 channel/OC mode:250Ms/s
1 channel/standard mode: 400Ms/s
1 channel/OC mode: 500Ms/s

Would be DSO2512 good enough to do some low-end but serious work or is it just a toy?

Opinions differ on this question. 

It's enough for me for the time being.
It's the first and only oscilloscope for me, so I have no comparison
I'd say it's a lot better than no oscilloscope and a little more usable than the DSO154.
I don't use it professionally, just as a hobby. (Arduino etc.)

I would say if you want to buy a new handheld 2-channel toy-or-not oscilloscope, then there is nothing that costs less than the DSO2512 (I paid 77.50 CHF including delivery)
I don't have a dedicated work desk for these devices, so I need to be able to easily stow them in a drawer. Therefore handheld.

[tatel]

The 40 MHz you mentioned refer to the square wave. It is the state you see on the DSO154 at 6MHz.
The DSO2512 is specified with 500Ms/s. I guess in reality it is probably like this:

2 channel/standard mode: 200Ms/s
2 channel/OC mode:250Ms/s
1 channel/standard mode: 400Ms/s
1 channel/OC mode: 500Ms/s

OMG. I wouldn't say that square wave on DSO154 is useless at 6 MHz, but to me it's more disgusting than a bunch of phlegm. What a toothache. Please see FFT screenshots. At 6 MHz, only first higher odd harmonic fits into the bandwidth. So I think it's aliasing for sure.

About sampling rate, I think you are about just right. Still I would love to know what people would have to say about this. radiolistener? Your opinion would be appreciated. Also anyone's, of course

EDIT: My numbers come from what people I respect said. If DSO2112 does sample at 100 MSa/s with both channels activated, that would be 100/2.5=40 if using sin(x)/x

[Aldo22]

OMG. I wouldn't say that square wave on DSO154 is useless at 6 MHz, but to me it's more disgusting than a bunch of phlegm. What a toothache. Please see FFT screenshots. At 6 MHz, only first higher odd harmonic fits into the bandwidth. So I think it's aliasing for sure.

This is a square wave at 40MHz on the DSO2512.
I think it is clearly distinguishable from a sine.
This is the limit I'm talking about.
Above that, the square wave gets really ugly and at some point looks like a sine.

[tatel]

OMG. I wouldn't say that square wave on DSO154 is useless at 6 MHz, but to me it's more disgusting than a bunch of phlegm. What a toothache. Please see FFT screenshots. At 6 MHz, only first higher odd harmonic fits into the bandwidth. So I think it's aliasing for sure.

This is a square wave at 40MHz on the DSO2512.
I think it is clearly distinguishable from a sine.
This is the limit I'm talking about.
Above that, the square wave gets really ugly and at some point looks like a sine.

Yeah, the bunch of phlegm is also distiguishable from a sine wave. That's why I said isn't useless.

Matter is: do you see your waveform as stable as that while watching "live" your scope?

My screenshots at 6 MHz look way better than what I see "live", that's why I decided to post a video instead.

To me, DSO154 is good to do square waves up to 1 MHz. 2 MHz at most. Above that I consider it just "useable". Keeping my eyesight as good as age allows, and keeping any aches out of my head, matters.

BTW, do you have AC coupling on? What for? No need, DC coupling will work perfectly in mos cases. To me, AC coupling main benefit is to delete DC component while looking at, say, PSU ripple. That way you can set your Volts/div at a low setting and see that ripple much better. Otherwise, if PSU output is, say, 12 V, you have to set your  Volts/div at say, 2 Volts/div. That make ripple not so easy to see.

EDIT: Can DSO2512 do math? Even if it's just ch1-ch2?

[Aldo22]

Just briefly, I have to go out afterwards.

To me, DSO154 is good to do square waves up to 1 MHz. 2 MHz at most. Above that I consider it just "useable". Keeping my eyesight as good as age allows, and keeping any aches out of my head, matters.

That's what I said: 18MHz/10 = 1.8MHz.

EDIT: Can DSO2512 do math? Even if it's just ch1-ch2?

Not sure what you mean exactly. It has some toy-FFT and an XY mode.
There is a totally silly video about the XY mode. 



Yes, it's a toy! 

See you later!

[radiolistener]

Aldo22

did you found the reason why it shows incorrect measurements? si5351 module has amplitude 1.5 Vpp, but all your screenshots never shows correct value. This is very strange, did you tested it at low frequency (1-2 MHz and below)?

[Aldo22]

Matter is: do you see your waveform as stable as that while watching "live" your scope?

No, of course it dances nervously in this range.
But it shows - as smart people would say - the 3rd harmonic, if I'm not totally wrong.
And that is - if I'm not totally wrong - an indication of a certain bandwidth which has to be 3x higher than the 1st harmonic.
So in case of the DSO2512 that would be 40Mhz->120MHz.
If I have misunderstood, please correct me.

[Aldo22]

did you found the reason why it shows incorrect measurements? si5351 module has amplitude 1.5 Vpp, but all your screenshots never shows correct value. This is very strange, did you tested it at low frequency (1-2 MHz and below)?

Why do you think it has to be 1.5Vpp?
I can only measure that with my (toy-) oscilloscope and if it shows 3.x Vpp, then I have to believe it.
If you look at the "Adafruit Si5351 Clock Generator Breakout" page: https://learn.adafruit.com/adafruit-si5351-clock-generator-breakout/wiring-and-test
When scrolling down you see three oscilloscope screenshots. They show similar Pk-Pk values, like 3.64 Volt @10.7KHz.

Thank you.

[tatel]

Matter is: do you see your waveform as stable as that while watching "live" your scope?

No, of course it dances nervously in this range.
But it shows - as smart people would say - the 3rd harmonic, if I'm not totally wrong.
And that is - if I'm not totally wrong - an indication of a certain bandwidth which has to bee 3x higher than the 1st harmonic.
So in case of the DSO2512 that would be 40Mhz->120MHz.
If I have misunderstood, please correct me.

Well, maybe. Not really sure about that right now. But I for one would rather prefer a stable waveform. I think that unstability is some form of aliasing produced by the harmonics that doesn't fit the bandwidth, as wasedadoc said on the Zeeweii toyscopes" thread. I guess radiolistener would say that's because there's no antialiasing filter, and rightly so.

After some reading, it turns out that to increase the bandwidth can be cheaper than putting a good antialiasing filter; so no wonder some people says that even good tektronix oscilloscopes have aliasing. No worries about your "toyscope" then. If it does 20 MHz with both channels activated, it's good. If it does 40 MHz, then it's excellent at that price range. Didn't I said that it could turn out it's not that bad after all?

BTW a couple things have been learned while testing it.

Not sure what you mean exactly. It has some toy-FFT and an XY mode.

FFT is a math function, and a much more complicated one than just to substract channel 2 from channel 1, I think. That substraction would be quite useful, since it allows to take "poor man's differential measurements". Much more useful than that flapdoodle apple XY video.

IIRC, it has been said in this thread that DSO2512's FFT is not that far from Rigol's DS1054Z (which isn't particularly good, I'm told). But, hey, it's there as an added bonus. If there's FFT, then there could be some more simple math functions like that useful substraction.

Ciao!

[Aldo22]

Didn't I said that it could turn out it's not that bad after all?
BTW a couple things have been learned while testing it.

Yes, learning and trying to classify something is always interesting.

But fundamentally, I find the discussion somewhat questionable.
It's like measuring a Citroën Méhari against the qualities of a Porsche.
You can certainly find 500 good technical arguments why the Porsche is a better and safer car, but do I care?

In the end, I might decide in favor of the Méhari. 

FFT is a math function, and a much more complicated one than just to substract channel 2 from channel 1, I think. That substraction would be quite useful.

No, it does not have that as far as I know.
You can only set markers, as shown here, but that's something different.

[radiolistener]

Why do you think it has to be 1.5Vpp?
I can only measure that with my (toy-) oscilloscope and if it shows 3.x Vpp, then I have to believe it.
If you look at the "Adafruit Si5351 Clock Generator Breakout" page: https://learn.adafruit.com/adafruit-si5351-clock-generator-breakout/wiring-and-test
When scrolling down you see three oscilloscope screenshots. They show similar Pk-Pk values, like 3.64 Volt @10.7KHz.

Just because I have both - original blue Adafruit and Chinese magenta clone si5351 modules, both have the same 1.5 Vpp output. Chinese one has worse oscillator, which has significant frequency drift.

Regarding to the waveform from adafruit site, it seems that it shows waveform measured without dummy load. Probably they just put a probe into output pin. You can't measure amplitude of frequency above 1-3 MHz properly in that way. And even for lower frequency, you're needs to divide result by 2, because you're measuring open output, which amplitude is twice higher than loaded with matching load.

Since si5351 has square wave output with 1 ns rise time edges, it has harmonics up to 1 GHz and even higher (in practice si5351 harmonics are usable up to 1.2-1.3 GHz), so measure it without proper matching connection with termination dummy load can lead to a high amplitude error which will depends on a frequency and cable length.

For proper measurement you're needs to use direct coax cable connection with pass-through dummy load on oscilloscope side or use dummy load on adafruit module and use oscilloscope probe to measure amplitude. Direct coax cable with pass-through dummy load is preferred way, because oscilloscope probes response is not flat.

[Aldo22]

@radiolistener: Yes, with dummy load and coax I get amplitudes in the range you mentioned, but since the signal is so ugly (or the toy oscilloscope can't handle it), there are higher Pk-Pk values.

Thank you!

[radiolistener]

Yes, with dummy load and coax I get amplitudes in the range you mentioned, but since the signal is so ugly, there are higher Pk-Pk values.

I think the reason why it shows unstable waveform is because it has no antialias filter. Since si5351 module has pretty high harmonics up to 1.3 GHz, these harmonics leads to some instability. When you connect it without dummy load or with dummy load + probe it lieads to a high impedance mismatch and as result the cable/probe works as low pass filter and cut off high frequency components, so its going more stable.

Can you try to measure some known source for a sine wave with dummy load?

You can use some LPF filter with cut-off at about 20-30 MHz on the input to fix that issue. Do you have LPF filter for testing?

In order to use LPF filter, put it between pass-through dummy load and coax cable. Pass-through dummy load should be connected directly on device input connector.

There is needs to experiment with LPF filters with different cut-off frequency to check which cut-off frequency allows it to work stable with si5351 square wave output. You can try test LPF with the following cut-off: 10 MHz, 30 MHz, 50 MHz and 100 MHz.