Magnova oscilloscope

[S13]

Hi,
First, your trigger is out of range. Then, when you apply a 50 MHz signal, what you “see” in the lower us/ms range could be anything, but it’s no longer the correct waveform.

I disagree. Even when using auto triggering in auto trigger mode (so not relying on a rising edge trigger for example), your waveform should still show a correct waveform. Like other people mentioned before me, in this particular case the observed behavior i is just a result from having a sample rate exactly equal to the wave frequency, so you get a perfect DC. 

Sure. I am talking about a different aspect. If/when a slow timebase is selected (as the result - low samplerate, etc., in addition/moreover - when the trigger is not set/active/activated), there is a lot of ''spare'' time for creation of ''signal preview''. If/when I am inspecting an unknown signal/shape, in such a case (for example), I have no clue that there is some high-frequency signal, too. Signals that cannot be displayed ''correctly'' (due to their too high frequency) should more or less appear as a "filled'' screen (from min to max amplitude). 

So you know perfectly why it happens, i think it is just a matter of accepting that this is the standard behavior of a digital sampling scope in normal sampling mode.
Alternatively, i dont know if the Magnova's has a Peak Detect sampling mode that would allow you to see that waveform regardless of sampling rate?

[rf-messkopf]

Sure. I am talking about a different aspect. If/when a slow timebase is selected (as the result - low samplerate, etc., in addition/moreover - when the trigger is not set/active/activated), there is a lot of ''spare'' time for creation of ''signal preview''. If/when I am inspecting an unknown signal/shape, in such a case (for example), I have no clue that there is some high-frequency signal, too. Signals that cannot be displayed ''correctly'' (due to their too high frequency) should more or less appear as a "filled'' screen (from min to max amplitude).

Well, at some point, the scope has to reduce the sampling rate, depending on the selected record length, an there will inevitably be aliasing, unless you filter in the analog domain. This would come at a cost, of course. E.g., the rise time gets worse.

What you can try to get your filled screen is switch to Peak Detect. Then you should see a filled band like on an analog scope. But even this fails when the sampling clock and the generator are perfectly snychronized. You can also select HiRes mode. Then the scope does not do simple decimation, but averages over the decimated samples. Your 50 MHz signal at 50 MSa/s would then simply disappear.

In theory a scope could do tricks like dithering of the sampling clock. Then aliasing components would be smeared out and look like noise. But also "legitimate" signals in the lower Nyquist band would be affected as well. In effect, you would make the scope deliberately worse, therefore this is generally not done.

[2N3055]

Hi,
First, your trigger is out of range. Then, when you apply a 50 MHz signal, what you “see” in the lower us/ms range could be anything, but it’s no longer the correct waveform.

Yup, some don't understand the trigger is all powerful !

Guys, I understand why it's happening. At the same time, I marked this as a bug because (again, my own subjective point of view) for such a segment (I am talking about price range) scope, there should be more advanced algorithms than ''let's reduce samplerate if/when timebase is increased''.

There are no "more advanced algorithms". It is magical thinking.
It is elementary school mathematics.
If you take 1e6 samples in 1s, that is one sample every 1µs, ie. 1e6 Points/s.
1 MS/s.

With that sample rate highest frequency you sam sample directly is 500 kHz. Nyquist.

If you feed it larger frequency, you will start down-converting input frequencies. You will see "mirrored" frequencies into 500 kHz BW.

And if signal source and scope are synchronous, what you see on the screen might be even more confusing.

What you expect is how analog scope behaves. Digital scopes are sampling systems.

What could help you get something similar on digital scope to what you expect is to put scope from Normal sampling mode to Peak Detect mode. You can use that when you don't know what you are looking at. Peak detect mode has it's different problems, so it is not what is usually used.
But you can also go through the timebase and inspect signal into detail in normal sampling mode. That is what people usually do.

[S13]

What you can try to get your filled screen is switch to Peak Detect. Then you should see a filled band like on an analog scope. But even this fails when the sampling clock and the generator are perfectly snychronized.

Shouldnt Peak Detect always run at maximum sampling rate of the scope, and then store min/max values for each sampling interval (in this case 1/50MHz)?
So Peak Detect should be a good solution as long as you are not inputting a wave frequency equal to the maximum sampling rate of the scope.

But I dont generally use Peak Detect that much, and i cant verify it on any of my scopes at this moment (dont have one with internal gen).

[kaspich]

Hi,
First, your trigger is out of range. Then, when you apply a 50 MHz signal, what you “see” in the lower us/ms range could be anything, but it’s no longer the correct waveform.

Yup, some don't understand the trigger is all powerful !

Guys, I understand why it's happening. At the same time, I marked this as a bug because (again, my own subjective point of view) for such a segment (I am talking about price range) scope, there should be more advanced algorithms than ''let's reduce samplerate if/when timebase is increased''.

There are no "more advanced algorithms". It is magical thinking.
It is elementary school mathematics.
If you take 1e6 samples in 1s, that is one sample every 1µs, ie. 1e6 Points/s.
1 MS/s.

With that sample rate highest frequency you sam sample directly is 500 kHz. Nyquist.

If you feed it larger frequency, you will start down-converting input frequencies. You will see "mirrored" frequencies into 500 kHz BW.

And if signal source and scope are synchronous, what you see on the screen might be even more confusing.

What you expect is how analog scope behaves. Digital scopes are sampling systems.

What could help you get something similar on digital scope to what you expect is to put scope from Normal sampling mode to Peak Detect mode. You can use that when you don't know what you are looking at. Peak detect mode has it's different problems, so it is not what is usually used.
But you can also go through the timebase and inspect signal into detail in normal sampling mode. That is what people usually do.

There could be ''timing noise'' added; also - some high/samplerate ''burst'' periods added, etc. There are a lot of solutions.
For sure, none of them are implemented in cheap China-made scopes. But, if Magnova acts 1:1 as Ri^ol, Ow%n, and shit like this - why pay 10 times more?
As you/someone mentioned before -  signal shape should remain correct!

[kaspich]

Hi,
First, your trigger is out of range. Then, when you apply a 50 MHz signal, what you “see” in the lower us/ms range could be anything, but it’s no longer the correct waveform.

Yup, some don't understand the trigger is all powerful !

Guys, I understand why it's happening. At the same time, I marked this as a bug because (again, my own subjective point of view) for such a segment (I am talking about price range) scope, there should be more advanced algorithms than ''let's reduce samplerate if/when timebase is increased''.

There are no "more advanced algorithms". It is magical thinking.

Not true.
Please check how very-high-frequency signals are captured/sampled.
Also, as I mentioned before - there could be peak/peak sampling frames added. Also, high-sample-rate frames could be added.
It's a question of ''quality'' of the software.

[kaspich]

I disagree. Even when using auto triggering in auto trigger mode (so not relying on a rising edge trigger for example), your waveform should still show a correct waveform.

Exactly! Without any exceptions!!!

[2N3055]

Hi,
First, your trigger is out of range. Then, when you apply a 50 MHz signal, what you “see” in the lower us/ms range could be anything, but it’s no longer the correct waveform.

Yup, some don't understand the trigger is all powerful !

Guys, I understand why it's happening. At the same time, I marked this as a bug because (again, my own subjective point of view) for such a segment (I am talking about price range) scope, there should be more advanced algorithms than ''let's reduce samplerate if/when timebase is increased''.

There are no "more advanced algorithms". It is magical thinking.

Not true.
Please check how very-high-frequency signals are captured/sampled.
Also, as I mentioned before - there could be peak/peak sampling frames added. Also, high-sample-rate frames could be added.
It's a question of ''quality'' of the software.

Oh well.. Could be, should be...

Please point to the oscilloscope that does exactly what you say. I am always curious to learn, and am glad to be show wrong if I learn something from that...

There are some "anti aliasing" methods, that for instance my Keysight has, but they work only on marginal cases. If I undersample 50 MHz signal at 3 Mpts/s it simply downconverts like any other scope.

[KungFuJosh]

I disagree. Even when using auto triggering in auto trigger mode (so not relying on a rising edge trigger for example), your waveform should still show a correct waveform.

Exactly! Without any exceptions!!!

This might be true if your trigger were set correctly.

Making this argument with the trigger set too high is the same as thinking there's a bug in the scope when the waveform looks weird using AM modulation with a trigger anywhere off the vertical center/50% point of the waveform. The attached images are all the same waveform.

[kaspich]

ok, guys.
Reputation test accepted.
Below you can see three pictures.
Magnova
Siglent SDS2000 series
20 years old Agilent 6000 series

As you see, the most incorrect representation: Magnova. 100% modulation. Frequency meter failed (F=2Mhz).
Siglent shows around 10% modulation; interference is clearly visible.
Agilent - interference = barely visible, peak detection works perfectly.

[2N3055]

And one of my Siglent scopes. 50MHz at 200ms/div.
Like I said, Peak Detect works well...

You can use equipment how it is designed to be used, or you can be stubborn and expect the equipment to adapt to you.
I haven't time for lamenting what would be if it would be. If it can be done, but in different way that what I am accustomed to, I change my habits.
World changes, and I adapt.
And my SDS6000A (the one from snapshot) is so superior in every aspect to old Agilent that I would not take the old one ever again.

I checked, and MSOX3000T does what you talk about. But that is because it was very necessary. It has so little memory, it start aliasing immediately as soon you go longer than 100µs/div.  I basically don't use MSOX3104T at all anymore, and am thinking of selling it..

[kaspich]

In my actual projects, I use my stuff to evaluate analog signals. This is one of the reasons why I purchased Magnova: 12-bit (and hiRes mode) low-noise frontend.
For digital signal observation (my own point of view), a usual 8-bit ADC is good enough. On the other hand, if Magnova has such a frontend (and a pretty low samplerate, compared to actual competitors, like Rigol, for example), I assume this scope is intended for analog signals, too.
Analog signals often mean - frequency, amplitude is unpredictable (and even non-repeating). So, all possible scenarious possible:
a. high frequency signals sampled at low samplerate (Fin>>Fsamplerate);
b. input frequency is comparable with samplerate,
etc.
In ALL these cases, the waveform on the screen should be as correct as possible!
That's the difference between ''sampling machine'' and oscilloscope!
Sampling machine samples data. Oscilloscope represents them.
If I see 100% modulation when a simple sine wave signal is applied, it's a BUG to me. If I see incorrect switching between ''raw'' and ''analysed'' data in Roll mode, it's a bug to me. 
And, yes, everyone can try the old Agilent 6000 scope as a reference. Starting with knobs (how scroll functionality is implemented) and ending with how the signal is represented. Everything works perfectly!

[S13]

This might be true if your trigger were set correctly.

Making this argument with the trigger set too high is the same as thinking there's a bug in the scope when the waveform looks weird using AM modulation with a trigger anywhere off the vertical center/50% point of the waveform. The attached images are all the same waveform.

Well i gotta admit the Magnova does produce some nice looking waveforms on screen! That intensity color grading is gorgeous! 
Even the Micsig cannot quite reach that level of prettiness...

[kaspich]

And one of my Siglent scopes. 50MHz at 200ms/div.
Like I said, Peak Detect works well...

You can use equipment how it is designed to be used, or you can be stubborn and expect the equipment to adapt to you.
I haven't time for lamenting what would be if it would be. If it can be done, but in different way that what I am accustomed to, I change my habits.
World changes, and I adapt.
And my SDS6000A (the one from snapshot) is so superior in every aspect to old Agilent that I would not take the old one ever again.

I checked, and MSOX3000T does what you talk about. But that is because it was very necessary. It has so little memory, it start aliasing immediately as soon you go longer than 100µs/div.  I basically don't use MSOX3104T at all anymore, and am thinking of selling it..

Maybe I misunderstood something, but I was sure that this forum is to report bugs, place some recommendations, and make improvements. That's what I am doing/trying to do.
If the basics are not implemented correctly, for analog measurements (at least + for me), all this mega user-friendly interface means nothing.
I can't trust the picture I see! So simple!
But, ok, if the whole community is united, that problem is in me (I have purchased the wrong tool), let it be.

[S13]

Maybe I misunderstood something, but I was sure that this forum is to report bugs, place some recommendations, and make improvements. That's what I am doing/trying to do.
If the basics are not implemented correctly, for analog measurements (at least + for me), all this mega user-friendly interface means nothing.
I can't trust the picture I see! So simple!
But, ok, if the whole community is united, that problem is in me (I have purchased the wrong tool), let it be.

Well you are not entirely wrong, but i think it is important to keep in mind each scope has their own strengths and weaknesses. Undersampling is a well known issue with any digitizing system, and generally most test equipment operators are aware of this problem and will operate their equipment accordingly. And maybe because this issue is quite well understood by operators, not all scope manufacturers will optimize their scopes to prevent undersampling issues in every possible scenario. Comparing it to an Agilent/hp might be a bit unfair, given the many many many years of experience they acquired, so those scopes are obviously going to be ahead in some aspects. I dont know about Siglent, but i doubt that they do it correctly in every scenario.   

[kaspich]

For sure - unqualified operator
1MHz at the input. Sine wave.

So, 5 radically different pictures. Which one is correct?
Answer - NONE

Really? Is this acceptable?
Ok, how much should an oscilloscope with 1.6G samplerate and 12-bit frontend that is capable of correctly acting in this ''well known'' situation cost?

[rf-messkopf]

What you can try to get your filled screen is switch to Peak Detect. Then you should see a filled band like on an analog scope. But even this fails when the sampling clock and the generator are perfectly snychronized.

Shouldnt Peak Detect always run at maximum sampling rate of the scope, and then store min/max values for each sampling interval (in this case 1/50MHz)?
So Peak Detect should be a good solution as long as you are not inputting a wave frequency equal to the maximum sampling rate of the scope.

Yes you are right. Peak Detect will always be a good solution to that problem, even if the generator and sampling clock are synchronized. Will correct my post. Thanks.

[KungFuJosh]

Magnova. 100% modulation. Frequency meter failed (F=2Mhz).

That's not the hardware frequency counter that you're referring to. The hardware frequency counter won't be available until the next firmware update.

[kaspich]

The same 1MHz
Siglent SDS2000 series

As we see, there are problems in the Roll mode. But, in Normal mode (starting from 20ms), everything works correctly!
If the Roll mode is disabled, everything works fine for all timebase settings.

[rf-messkopf]

The same 1MHz
Siglent SDS2000 series

As we see, there are problems in the Roll mode. But, in Normal mode (starting from 20ms), everything works correctly!

If it is true that the input signal is indeed 1 MHz, then what is shown in picture SDS00003.png is as "wrong" as what you show on the Magnova: A 1 MHz signal sampled at 1 MSa/s, which will of course be folded to 0 Hz (or approx. 10 Hz in your case, likely because your generator and scope are not synced).

Really? Is this acceptable?
Ok, how much should an oscilloscope with 1.6G samplerate and 12-bit frontend that is capable of correctly acting in this ''well known'' situation cost?

This is not a question of cost, but about the limitations of sampling systems. If the scope needs to decimate at large time/div settings, it has to discard samples, i.e., throw away information. This doesn't come for free. Therefore the scope gives you options to handle that, e.g., the simple Sample mode, or Peak Detect, or Boxcar.

As I said above, in theory another option might be to dither the sampling clock, which would turn aliasing products into noise. But that would affect the wanted signal components as well, and add jitter to them.

[kaspich]

The same 1MHz
Siglent SDS2000 series

As we see, there are problems in the Roll mode. But, in Normal mode (starting from 20ms), everything works correctly!

If it is true that the input signal is indeed 1 MHz, then what is shown in picture SDS00003.png is as "wrong" as what you show on the Magnova: A 1 MHz signal sampled at 1 MSa/s, which will of course be folded to 0 Hz (or approx. 10 Hz in your case, likely because your generator and scope are not synced).

Really? Is this acceptable?
Ok, how much should an oscilloscope with 1.6G samplerate and 12-bit frontend that is capable of correctly acting in this ''well known'' situation cost?

This is not a question of cost, but about the limitations of sampling systems. If the scope needs to decimate at large time/div settings, it has to discard samples, i.e., throw away information. This doesn't come for free. Therefore the scope gives you options to handle that, e.g., the simple Sample mode, or Peak Detect, or Boxcar.

As I said above, in theory another option might be to dither the sampling clock, which would turn aliasing products into noise. But that would affect the wanted signal components as well, and add jitter to them.

ok, maybe I have 0 understanding, but I see what I see (example with Siglent and, moreover, 20-year-old Agilent).
My major question - why does my expensive oscilloscope become total crap at low sample rates? WHY?

As far as I understand, if/when samplerate is set to 1MHz (for example), it's 1/1000 of the ADC capabilities.
I assume, 99% of the time, ADC is on ''hold'' as well as all remaining FPGA/CPU. So (just idea/proposal), why not change the samplerate periodically (automatically/hiddenly) to a higher rate to capture the whole spectrum/trace of the input signal?
1 waveform per second, 1Msamples/second, 100k points + 5 incorrect pictures, and that's all that Magnova is capable of?

Sorry for my sharp language (today), but maybe, before making your own product, check competitors? And community - do not look for 117 excuses. Sorry, guys, the 20-year-old scope performs as it should perform! But here I am reading total BS, starting with 'wrong operator' and finishing with 'unsolvable problems'.

[KungFuJosh]

You're not comparing apples to apples, and you should pay attention to what shows in your screenshots...and fix your screenshots to show the related data, rather than unrelated settings.

Look at the sample rates you have set in the Siglent scope vs. the Magnova. They're not set the same way.

[kaspich]

You're not comparing apples to apples, and you should pay attention to what shows in your screenshots...and fix your screenshots to show the related data, rather than unrelated settings.

Look at the sample rates you have set in the Siglent scope vs. the Magnova. They're not set the same way.

ok, I'll try very slowly.
Agilent 6000 series. 8M memory option. 300MHz. 20-year-old scope.
Regardless: Normal mode or Roll mode, for ALL timescale settings, completely correct picture.

Siglent 2000 series. 70MHz version.
For Normal mode, correct picture for ALL timescale settings!

Tested on different frequencies.

Magnova: 5+ different pictures (all completely wrong) depending on timescale settings.

[Grandchuck]

This is basic but might be helpful:

https://dn710608.ca.archive.org/0/items/MITRES.6-008/MITRES6_008_demo2_300k.mp4

Not trying to insult anyone!

[KungFuJosh]

You're not comparing apples to apples, and you should pay attention to what shows in your screenshots...and fix your screenshots to show the related data, rather than unrelated settings.

Look at the sample rates you have set in the Siglent scope vs. the Magnova. They're not set the same way.

ok, I'll try very slowly.
Agilent 6000 series. 8M memory option. 300MHz. 20-year-old scope.
Regardless: Normal mode or Roll mode, for ALL timescale settings, completely correct picture.

Siglent 2000 series. 70MHz version.
For Normal mode, correct picture for ALL timescale settings!

Tested on different frequencies.

Magnova: 5+ different pictures (all completely wrong) depending on timescale settings.

Cool explanation. Now look at your screenshots and compare the sampling rates and see if your results might make sense.