Magnova oscilloscope

[egonotto]

Hello,

Comparison of the noise of the Magnova with R&S MXO4 and R&S MXO5 and LECROY HDO4034A

It is noticeable that the values are given with an inappropriate accuracy.
The values are only roughly comparable because the measurement method is different.   

Best regards
egonotto

[Andre77]

I'm missing the possibility to trigger on the waveform generator. Also, is the waveform generator clock derived from the external reference clock input? And what is the frequency setting accuracy of the waveform generator?

We plan to support triggering on the built-in waveform generator with an upcoming update. The datasheet will be updated with new features as soon as they are implemented and tested.

The sampling as well as the waveform generator are based on the same reference clock. You can choose between an internal or external reference ref-clock source, which then applies to both the sampling and the generator.

The frequency setting accuracy of the waveform generator is 1 µHz.

Edit: another thing that doesn't make sense to me is the deskew resolution. It is limited to the sample time interval which is 800ps minimum. That makes no sense to me; deskew should be adjustable in picoseconds given the 100ps channel-to-channel skew, minimum of 200ps/div and the trigger jitter.

Yes, the deskew function is limited to the full ADC sampling rate. Maybe later we will be able to relate it to the interpolated sampling rate, but this is not yet on our roadmap for the next few month.

Will it be possible to change the sampling rate manually or is just set automatically according to the number of channels used and the horizontal time scale?

The AD converters operate continuously at the highest sample rate (1 GSa/s or 1.6 GSa/s). If the set (or automatically selected) memory depth is not sufficient to save all samples over the selected time scale, the sample rate is automatically reduced before saving. A manual setting of the sample rate is currently not provided.

I'm wondering why there is no picture from the backside. I would like to see which connectors are available and how the optional AWG will be mounted.

I attach two quick photos of one of the prototypes. One picture shows the back of the Magnova with the cover removed before installing the module, one shows the back after installation.

The production units will look better than this prototype.  The nameplates and the final paintwork are missing, but it should be enough to show the assembly of the generator.

I don't understand the thing with the 320 Mpts and the 12 divs. When do you have 80 Mpts with 4 channels?

With 4 channels and 5 ms/div (=50 ms/screen), you have 60 mpts on each channel. With a slightly higher time scale of 6.666 ms/div (~80 ms/screen), you have 80 mpts on each channel. Of course, this time scale is somewhat unusual and is not in the 1-2-5 steps.

Another possibility would be to set the memory from "Auto" to 60 80 Mpts at e.g. 5 ms/div. Then the Magnova will sample and store 16 divs (12 on the screen and 3 to the left and right of it).

But even if you don't want that, 320 mpts helps for example to increase the number of saved waveforms in the history (by 50% compared to a memory limit of 240 mpts, which will fit more to 12 divs/screen).

[2N3055]

@Andre77

Is that a RTC battery? If it is, compliments for designing in easy accessibility !!

Best,
 Siniša

[Andre77]

@Andre77
Is that a RTC battery? If it is, compliments for designing in easy accessibility !!

Yes, the RTC battery is accessible via the generator module cover for easy maintenance.

[core]

It is clear that Andre77 follows this forum and brings useful information to the questions asked here.

According to the published specifications, the input stages are very well designed. It seems that noise floor and ENOB are comparable or slightly better than SDS3000X HD.
It would be interesting to know which PGA is used in the input stage.
Channel-to-channel isolation is very good.

The offset range with automatic early attenuation (reduced accuracy) is very good: ±70 V at > 35 mV/div. But I'm not sure what this means - reduced accuracy.

[2N3055]

According to the published specifications, the input stages are very well designed. It seems that noise floor and ENOB are comparable or slightly better than SDS3000X HD.

Which is normal because you are comparing 350Mhz design with a 1GHz design.
It should be compared in this regard to SDS2000xHD which would be correct equivalent...

[LoneWolf6912]

I think it is more interesting to compare the spec of the scope with high-end brand. Like R&S for exemple.

This scope for exemple make a perfect transition between the rtb2000 series and the rtm3000 series. The first can be limited in lot of way and the second is far more expensive.

For the education market, I don't see the advantage of buying an rtb2004 when you can have this instead for the same price.



The 12 bit architecture is just a nice good bonus.

[pdenisowski]

For the education market, I don't see the advantage of buying an rtb2004 when you can have this instead for the same price.

The educational market, perhaps even more than the non-educational market, takes a lot of factors into consideration beyond nominal price and specs.

For starters, the ability to buy the entire "bench" from a single manufacturer is pretty important.

And while the Magnova has a very interesting and innovative design, educational customers often prefer instruments that are more "traditional," at least for introductory undergrad labs. This may be changing as some EE departments move towards "self-purchased instrument per student" (like the Analog Discovery series), but as an educator I would want my students using instruments that are similar to the ones they might be using after graduation. 

[mawyatt]

but as an educator I would want my students using instruments that are similar to the ones they might be using after graduation.

Very important point!! This was the very basis of the start of the WAMI Labs way back in 1997

http://wami.eng.usf.edu

[core]

I think that at school they should learn not an instrument, but what the purpose of that instrument is and the principle of operation, so that after graduation they can use any similar instrument.

So after graduation they will be able to use any oscilloscope, spectrum analyzer, power source from R&S, Keithley, Lecroy, Tektronix, Siglent, Rigol, etc.

In this respect, it is really useful to have instruments from different manufacturers in the school lab.

The manufacturing companies on the other hand, really need to have the same type of instruments in order to be able to standardize the test procedures. Depending on company policy they will choose a particular manufacturer.

On the other hand it is quite possible that schools will have a certain policy for the purchase of instruments, e.g. a local manufacturer if possible and available.

Fortunately, small companies have more freedom in their choice, as do hobbyists.

Otherwise there would only be 5-6 manufacturers.

[core]

Even if many don't know, in Romania we have at least one local manufacturer, R&S 

[egonotto]

...
It should be compared in this regard to SDS2000xHD which would be correct equivalent...

Hello,

this is very dependent on the settings and the algorithm of the measurement (is it averaged somehow?)

Best regards
egonotto

[tggzzz]

but as an educator I would want my students using instruments that are similar to the ones they might be using after graduation.

As an employer, that is the last thing I would want. Even if I used the same tool, what will happen in 3-5 years when it is superseded? I want employees that are familiar with fundamental principles that can (1) be applied to any tool (2) can enable them to select the right class of tool.

As a student, university is the ideal (and possibly last) place to learn concepts that will last my entire career. Employers will (at best) teach employees how to mung their flagets using this year's tool.

Example: last month I went back to one of my university textbooks to understand a subtle point I had come across in a circuit. Why does the amplitude of the "fuzz" on the 1kHz triangle wave vary?

I think that at school they should learn not an instrument, but what the purpose of that instrument is and the principle of operation, so that after graduation they can use any similar instrument.

So after graduation they will be able to use any oscilloscope, spectrum analyzer, power source from R&S, Keithley, Lecroy, Tektronix, Siglent, Rigol, etc.

Precisely.

Isaac Asimov wrote a novella https://www.abelard.org/asimov.php back in 1957. It significantly influenced my career and is as true today as it was then. That "geocities style" hurts my eyeballs, but there's nothing I can do about it!

[core]

As far as Magnova is concerned, the FFT implementation seems to relaunch the standard for oscilloscopes. 4 FFT spectra with up to 8 million data points each,  waterfall display ... it seems to have a very powerful engine behind it.
Finally, Siglent will have a rival in this field, which is not bad at all.

On the other hand, in Siglent's case, FFT can be applied to physical channels as well as M1~M4, F1~F4 and Zoom traces. So it's much more flexible.

If the Bode Plot implementation will also be at least similar to Siglent, it's worth waiting a bit longer before making a choice.

[nctnico]

I'm missing the possibility to trigger on the waveform generator. Also, is the waveform generator clock derived from the external reference clock input? And what is the frequency setting accuracy of the waveform generator?

We plan to support triggering on the built-in waveform generator with an upcoming update. The datasheet will be updated with new features as soon as they are implemented and tested.

The sampling as well as the waveform generator are based on the same reference clock. You can choose between an internal or external reference ref-clock source, which then applies to both the sampling and the generator.

That sounds good!

The frequency setting accuracy of the waveform generator is 1 µHz.

Unfortunately that says as good as nothing about the accuracy of a waveform generator.  It is a setting resolution specification, not accuracy. With a maximum frequency of 90MHz, you'd end up with a frequency setting which has 8+6 = 14 digits. As any digital waveform generator is somehow build around a DDS architecture as the basic operating principle, the number of meaningfull digits for the frequency setting depends on the numeric resolution of the DDS core. I doubt that resolution is going to be 14 digits. Therefore you'll find a frequency resolution accuracy specification on higher end waveform generators (like the Tektronix AFG31000 for example which guarantees 6 digits). This number says what kind of frequency accuracy (in respect to the reference clock) is to be expected using the worst case combination of waveform shape and modulation type.

A bit of background: Some of the tests I do deal with very high precission time (picosecond level) and frequency distribution which typically revolves around 10MHz + 1PPS (1Hz) signals and sometimes with a slow drift modulated on top for test purposes. Where the lower end (<1000 euro) waveform generators fall short is the inability to generate a precise 10MHz and/or 1PPS from the outputs. In many cases the 1PPS is slightly off and/or modulation starts to accumulate rounding errors. These errors are easy to check by using an oscilloscope and trigger on the output signal on one channel and look at the reference clock on another channel. Normal (y/t) or X/Y mode can be used to easely spot phase shift / drift.

I'm not saying to shoot for the moon but to give an idea on how to setup / define realistic expectations.

[Martin72]

You're taking it to the extreme again...
This is just a generator for little money, although I would have expected 2 channels for the price.

[ebastler]

A bit of background: Some of the tests I do deal with very high precission time (picosecond level) and frequency distribution which typically revolves around 10MHz + 1PPS (1Hz) signals and sometimes with a slow drift modulated on top for test purposes. Where the lower end (<1000 euro) waveform generators fall short is the inability to generate a precise 10MHz and/or 1PPS from the outputs. In many cases the 1PPS is slightly off and/or modulation starts to accumulate rounding errors. These errors are easy to check by using an oscilloscope and trigger on the output signal on one channel and look at the reference clock on another channel. Normal (y/t) or X/Y mode can be used to easely spot phase shift / drift.

I'm not saying to shoot for the moon but to give an idea on how to setup / define realistic expectations.

As it turns out, the bosun's whistle which was included in Cap'n Crunch packages for a while will miserably fail this test as well.

You stated yourself that the level of performance required by your specific use case is found in > 1000€ generators. Why would you expect to find it in an oscilloscope add-on which is integrated in the scope for convenience and for automated Bode plots? 

[Antonio90]

but as an educator I would want my students using instruments that are similar to the ones they might be using after graduation.

As an employer, that is the last thing I would want. Even if I used the same tool, what will happen in 3-5 years when it is superseded? I want employees that are familiar with fundamental principles that can (1) be applied to any tool (2) can enable them to select the right class of tool.

As a student, university is the ideal (and possibly last) place to learn concepts that will last my entire career. Employers will (at best) teach employees how to mung their flagets using this year's tool.

Example: last month I went back to one of my university textbooks to understand a subtle point I had come across in a circuit. Why does the amplitude of the "fuzz" on the 1kHz triangle wave vary?

I think that at school they should learn not an instrument, but what the purpose of that instrument is and the principle of operation, so that after graduation they can use any similar instrument.

So after graduation they will be able to use any oscilloscope, spectrum analyzer, power source from R&S, Keithley, Lecroy, Tektronix, Siglent, Rigol, etc.

Precisely.

Isaac Asimov wrote a novella https://www.abelard.org/asimov.php back in 1957. It significantly influenced my career and is as true today as it was then. That "geocities style" hurts my eyeballs, but there's nothing I can do about it!

Apologies for the off-topic but, ¿any chance you could explain the "fuzz", hint at the type of cricuit or somehow give any direction to replicate it?
Just ignorant and curious.

[nctnico]

A bit of background: Some of the tests I do deal with very high precission time (picosecond level) and frequency distribution which typically revolves around 10MHz + 1PPS (1Hz) signals and sometimes with a slow drift modulated on top for test purposes. Where the lower end (<1000 euro) waveform generators fall short is the inability to generate a precise 10MHz and/or 1PPS from the outputs. In many cases the 1PPS is slightly off and/or modulation starts to accumulate rounding errors. These errors are easy to check by using an oscilloscope and trigger on the output signal on one channel and look at the reference clock on another channel. Normal (y/t) or X/Y mode can be used to easely spot phase shift / drift.

I'm not saying to shoot for the moon but to give an idea on how to setup / define realistic expectations.

As it turns out, the bosun's whistle which was included in Cap'n Crunch packages for a while will miserably fail this test as well.

You stated yourself that the level of performance required by your specific use case is found in > 1000€ generators. Why would you expect to find it in an oscilloscope add-on which is integrated in the scope for convenience and for automated Bode plots?

Clearly you overlooked the last line you even quoted: all I'm saying is to have a good set of specifications so people know what to expect. Having the internal generator locked to an external reference clock, opens all kinds of possibilities. Without proper specifications, you can't tell suitability without testing.

[ebastler]

Clearly you overlooked the last line you even quoted: all I'm saying is to have a good set of specifications so people know what to expect.

No, I didn't overlook that sentence, but apparently misunderstood it. I thought you were saying that your own use case is an example of "not shooting for the moon" and "defining realistic expectations".  Alright then, we are largely on the same page.   

[2N3055]

I doubt that resolution is going to be 14 digits...

There are off-the shelf DDS chip solutions that have 48bit tuning word. That is 281,474976710656e12 to you ...
A 35,5271367880050092936E-9 Hz resolution on 10MHz signal.
There is no technical reason why tuning word cannot 64 bit..

Except it is not needed for general purpose AWG. Or any purpose really..

And your statements about accuracy are misleading.

DDS synthesiser will have perfect accuracy (clock phase noise/accuracy excluded) if frequency is multiplier of resolution. In case above with 10 MHz (accurate) clock, any frequency that is multiplier of the 35,5271367880050092936E-9  (Clock/2^48) will be dead accurate.

If you feed it 281,474976710656E6 Clock, your setting resolution will be one 1µHz. And it will be dead accurate if clock is.

One way of achieving this is to use NCO (no need for full DDS) as a tunable "weird number" clock source that you then feed into DDS to create nice "round" decade numbers.

[nctnico]

I doubt that resolution is going to be 14 digits...

There are off-the shelf DDS chip solutions that have 48bit tuning word. That is 281,474976710656e12 to you ...
A 35,5271367880050092936E-9 Hz resolution on 10MHz signal.
There is no technical reason why tuning word cannot 64 bit..

If it actually was that easy, then the low cost AWGs would be way more accurate than they are (and I have tested a whole bunch of these). Creating a single sine wave is relatively easy but when you throw different (arbitrary) waveforms, modulation and low jitter into the mix, things become way more complex. But that is beside the point I wanted to make about having proper function generator specifications.

[joeqsmith]

I received the data sheet today.  ...

I had joined their mail list as well.  I was surprised to see that they had passed 61326 but the website shows the product is still no offered in the USA due to lack of cert.  I suspect there is more testing to come.

That coin battery so easily accessible compared to my scope where I have to literally remove the entire data acquisition system to get to it, big

Like others, look forward to seeing more details added.   I'm currently looking at the Siglent 3000 series as a possible replacement to my aging scope problems but would sure like to see this new scope of yours.

[2N3055]

I doubt that resolution is going to be 14 digits...

There are off-the shelf DDS chip solutions that have 48bit tuning word. That is 281,474976710656e12 to you ...
A 35,5271367880050092936E-9 Hz resolution on 10MHz signal.
There is no technical reason why tuning word cannot 64 bit..

If it actually was that easy, then the low cost AWGs would be way more accurate than they are (and I have tested a whole bunch of these). Creating a single sine wave is relatively easy but when you throw different (arbitrary) waveforms, modulation and low jitter into the mix, things become way more complex. But that is beside the point I wanted to make about having proper function generator specifications.

How convenient is that your comment consists of part of my post you chose to ignore and not copy. 

And I told you before.
You error is that you are searching for AWG with NCO that will have accurate round (not fractional) divisions of external clock.

In your example if you slave an AWG to 10 Mhz Clock IN (to make it long term stable) and then use one channel of AWG as 10MHz reference for DUT and measurements, and other channel at 1Hz,
my two AWG i have here will hold that 1s edge within 3.4ps RMS (25ps P-P) and 6 ps RMS (48ps P-P) of 10MHz edge from CH1.  And that is at edge of what I can measure. It might as well be better.

It is about learning how to create results with what you have, not searching for equipment to fulfill your way of thinking.

[moerm]

Or "The Signal Path" for in-depth analysis, including software implementation.

Unlikely to happen I guess.

I liked to watch his videos. A lot. But since quite a few months I'm under the impression that his videos mostly are about stuff light years beyond a hobbyists reach, 20 Ghz and higher. In fact, he himself seems to have noticed that he is losing the mere mortal audience and made a weird video about why he does the kind of videos he does now.
I respect the man for his profound knowledge and have learned a lot from him (well, from watching his videos) and am grateful for that, but sorry Shariar, nowadays I do not even look at TSP from time to time anymore.

So, for what little weight my vote carries, it clearly goes to Dave. While I have some (mild) reservations re. Dave's reviews (Siglent = rust, Agilent hurray, and his enthusiasm about the (back then) new Tek 2 toy (to put it politely)  scope, reminded me that, oh well, humans are humans and not really unbiased and objective) a Dave review *is* worth watching and would be very welcome, especially if combined with a teardown.