Magnova oscilloscope

[temperance]

Hi,

Based on a recommendation from user Performa01, I had assembled a cable with Hyperflex 5 myself.
The result was so convincing that I built more cables and disposed of my old ones.
I had bought the plugs and the cable there:

https://kabel-kusch.de/Koaxkabel/HyperFlex-10/HyperFlex-5.htm

https://kabel-kusch.de/BNC-Stecker/BNC-58-155/bnc58-155.htm

Personally I never had much luck with those solder mounted BNC connectors. In my case the shielding tears off where the shielding is clamped under the strain relief after a while (resulting in a broken cable). Since then I buy pre-made cables. The hyperflex cable seems to have nice specs so I wonder if it can be bought with BNC and/or SMA connectors already mounted.

The source of the cable:
https://messi.it/

They offer assembled cables. A crimp tool for BNC plugs isn't expensive either and nothing special and can be bought for less than € 20.

The hyperflex cable a has a 0.212 diameter. Amphenol part 112514 will fit onto this cable.

Edit: The same cable can be found here and they offer assembled cables to your requirements.
https://www.koax24.de/

[Martin72]

Kusch and messi seem to have the same web designers...

I had a quick look again, a triangular signal from SDG2122X, 0dBM, 1khz.
First natively 12bit, then 16bit hi-res.
The 12-bit signal was saved as a reference, then set to 16 bits and the reference was added again, wow, what a difference (Fig. 16)... But wait a minute, did the 12 bits really look like that...
Again 12 bits, no, that looks much finer than the stored reference (Fig. 17) - strange...
Then the other way around, the 16 bits stored as a reference, aha, they look the same, no, not quite, a bit stepped (Fig. 18).
I don't know why that is at the moment, but you can still see a significant improvement 12 bits vs. 16 bits (Fig. 19).

[2N3055]

Kusch and messi seem to have the same web designers...

I had a quick look again, a triangular signal from SDG2122X, 0dBM, 1khz.
First natively 12bit, then 16bit hi-res.
The 12-bit signal was saved as a reference, then set to 16 bits and the reference was added again, wow, what a difference (Fig. 16)... But wait a minute, did the 12 bits really look like that...
Again 12 bits, no, that looks much finer than the stored reference (Fig. 17) - strange...
Then the other way around, the 16 bits stored as a reference, aha, they look the same, no, not quite, a bit stepped (Fig. 18).
I don't know why that is at the moment, but you can still see a significant improvement 12 bits vs. 16 bits (Fig. 19).

It might  depend if it saves it as a screen mapped reference waveform (decimated) or as a full memory channel (full original data).

[Martin72]

Ah, that sounds plausible.

[LoneWolf6912]

Hello,

Kind question André77.

When will it be possible to have the accessory to mount the Magnova in a rack? I would appreciate being able to implement it in various machines in our lab, but unfortunately, we need to be able to rack-mount it to maintain a professional appearance and following our procedures.

[ebastler]

When will it be possible to have the accessory to mount the Magnova in a rack? I would appreciate being able to implement it in various machines in our lab, but unfortunately, we need to be able to rack-mount it to maintain a professional appearance and following our procedures.

Hmm -- how would that work with the probe connectors on the side?

[Andre77]

When will it be possible to have the accessory to mount the Magnova in a rack? I would appreciate being able to implement it in various machines in our lab, but unfortunately, we need to be able to rack-mount it to maintain a professional appearance and following our procedures.

We will soon be offering a rackmount kit. This allows the Magnova to be installed in 19’ cabinets and also allows it to be pulled out, swivelled, tilted and rotated.

I've just taken some quick photos:

[nctnico]

Magnova tests

Today forum member Woody has visited my lab and he brought his brand spanking new Batronix Magnova BMO350 along for a playdate. As I wanted to keep the amount time limited and not all hardware options for the Magnova are available yet, I came up with a small list of tests we could do in a couple of hours to get a feel for the abilities of this new DSO.

UI performance
The UI operation is very snappy and the touchscreen and knob response is instant. The only situation where the UI slows down is when large amount of memory needs to be shown compressed onto the screen. In that case it is easy to turn the horizontal knob a few clicks extra before the Magnova responds.

I do find it a downside that 'clickable' items are not easely identifiable as such. I know that the 'flat' user interfaces are modern but I think adding some decorations would help to show which items bring up menus and which don't.

What is nice though is that the colors of traces in general can be adjusted to allow extra clarity. Better than having (for example) four math traces in the exact same color.

Bandwidth (cursory measurement)
Using an R&S SMIQ and a good piece of coax cable the bandwidth was tested with one channel enabled and three channels enabled (set to 50 Ohm termination) with a level of 220mV RMS. With one channel enabled (1.6Gs/s) the bandwidth is 429MHz, with three channels enabled it is 380MHz. It is very nice that with a higher samplerate there is more bandwidth available.

But this test isn't very complete as there are likely pre-amplifiers in the signal path which are enabled at lower V/div settings so it would require more tests at different levels and V/div settings to paint a complete picture.

Low level signal triggering
Low signal level triggering is excellent. If the signal is visible, it will trigger. Triggering worked fine down to 350uVp-p. Below that the signal started to drown in the noise.

Sin x/x performance
Sin x/x reconstruction is vital to visualise a signal on screen. As a rule of thumb this should work for signals up the samplerate divided by 2.5. On the Magnova the sin x/x reconstruction works as expected. With 1.6Gs/s it has no problem showing sine waves up to 640MHz (although attenuated ofcourse).

Filtering?
There has been some talk about filtering however this seems to be the standard bandwidth limiter as found on other oscilloscopes. I think this should be renamed to 'BW limit'. Filtering suggests having user configurable lowpass/highpass/bandpass filters.

Overdrive recovery
Overdrive recovery is a measure how well an oscilloscope can deal with an input signal exceeding its range. Having a large margin can help to analyse steep edges and trigger on those edges more accurately. Unfortunately the signal starts to distort rather soon. Taking the vertical sensitivity two steps down starts to cause visible distortions already.

Noise level
The RMS noise level (measured using standard deviation!) at 50mV/div looks to be 3 times better compared to the R&S RTM3004 I have here so the Magnova definitely scores some points here. From a cursory check it does appear as if the RMS noise level expressed as a percentage of the V/div scale doesn't seem to be constant though. At 50mV/div the noise level is 415uV (around 1%) while at 100mV/div it is 567uV (around 0.5%).

Deep memory (zooming out)
The Magnova supports up to >300Mpt/s of memory with a single channel enabled which is quite a lot. When the memory is set to a fixed amount and the time/div is short enough so the memory won't fill the screen entirely, it will capture beyond the screen allowing to zoom-out / scroll vertically after the acquisition. An interesting option is to enable 'extended acquisition'. With the memory depth set to auto and 'extended acquisition enabled', the Magnova will acquire 1 or 2 screen widths worth of signal extra so you can have some extra signal to look at (through zooming-out / scrolling vertically) if necessary but you don't have to set a fixed memory depth. Neat!

Reference traces
The reference traces are basically screen-grabs of how the trace is displayed on screen at the moment of copying a trace to a reference trace. Magnification of reference traces is limited because the trace starts to lack detail quickly. Something else I noticed is that the reference trace can not be brought to the foreground. Maybe it is possible but we have not found out how.

Trigger jitter
In short: I'd say the trigger jitter is excellent. With the timebase set to 200ps/div a 1MHz square wave is triggered at the exact same spot on the horizontal (time) axis. I've seen this much worse on other DSOs.

CAN decoding (decoding options, triggering on packet data)
Next up: decoding. We've tested decoding CAN messages. It works but that is about all there is to say about it. The messages are shown in a very wide format which does not allow to have several messages with readable decoding on screen. Far from ideal if you want to see a couple of messages and determine their relationship to other messages. Compacting the decoded data horizontally by using a higher width, multiple lines and rotated text plus the use of colored rectangles to identify sections of the message (in case of CAN: ID, DLC, payload, length, crc and acknowledge fields) would be very benificial.



What is also missing is triggering on CAN addresses and payload. Likely the support for CAN decoding / triggering hasn't been fully finished yet.

What does work is showing a bus decoding table. When scrolling through the trace, the part which is on screen is highlighted in the decoding table. Vice versa, the message selected in the bus decoding table is shown on screen.

Something else I noticed is that several screens need to be brought up in order to configure decoding. It would be nicer not having to dive through several layers of menus in order to configure decoding.

FFT speed. Spectrum analyser-esque options?
The FFT speed is quick. Even at 8Mpts you get a few updates per second. What is confusing about FFT mode is setting the number of points. The FFT settings allow to select a number of points but this has no effect. Only when setting the memory depth manually and adjusting the time/div scale, the number of FFT points changes. I'm not sure the actual number of FFT points is shown anywhere. It would be nice to get a feedback on the actual number of FFT point being used. An improvement would be to force the memory depth so the number of configured FFT points can actually be obtained from the memory. The same goes for the time/div setting.

I'm also missing a way to set the start & stop frequency. Personally I rarely use center and span on a spectrum analyser because most of the measurements I do have the frequency range defined as start / stop frequencies.

Measurements? Statistics? Histograms?
Onto measurements... The measurements are performed on acquired data so even when the shape of the waveform isn't visible, you still get the right value. So that is good. There is an option to show labels to indicate where the measurement has been taken. It looks like that a measurement is taken closest to the trigger point and per acquisition only 1 measurement is made. So statistics are gathered over multiple acquisitions. There is no histogram function to show the statistic distribution of the values. On the other hand there is a neat feature on the Magnova scope and that is the ability to plot trend charts (value, min, max and average) so you can observe variations in the measurement results in several ways. The trend chart can also be averaged but when an FM modulated sine wave (1kHz +/-100Hz modulated at 1Hz) is fed into the scope, the average trace shows a start-up artefact due to lack of historic data. Not sure how you'd fix something like this though.



Another issue with the measurements is in the measurement selection screen. The contrast is very low so it is difficult to see/read the measurement methods. It would be better if the selection had some way to tell what is a 'button' to click / dividers between the various measurement. Maybe have a rectangle around the active measurements or something like that instead of making the text of the active measurements brighter.

Math channels. Operations? Stacked? Freeform formulas?
The math channels offer basic operations on the waveform data using free-form math formulas. This is nice but there are features missing like filtering and doing tracking math operations. One application of 'tracking math' is to make a plot of duty cycle variations in a signal and thus recover a PWM modulated signal.

Cursors (outside screen?)
One of the things I like to check is whether the cursors can be placed outside the screen or not. One application for having cursors outside the screen is to be able to place the cursors precisely on two edges which are far away in order to make an accurate time measurement between the two edges. The Magnova oscilloscope doesn't allow this though. The cursors are bound to their screen locations and don't move along with the signal position.

Issues / remarks
During testing we encountered one issue: the samplerate got stuck at 1Gs/s even with only one channel enabled. A power cycle didn't help but a reset to factory defaults rectified the problem.

Temperature has been discussed before. The Magnova gets quite warm on the outside and I have doubts the heatsink at the back is doing much as it looks like a heatsink inteded for use with a fan instead of natural convection. The fin spacing is a few mm only. The internal fan isn't very loud though but I have not listened to it long enough to determine whether it would irritate me or not.

Summary
Not mentioned before but during this test we didn't consult the manual. As experienced DSO users everything is pretty obvious and self-explanatory to use. Then again, this test session wasn't exhaustive or intended to push the limits.

There is some room for improvement and I think plucking some of the low hanging fruit can make a big difference. I'm looking forward to seeing what the digital channels and function generator module are going to add in terms of functionality.

At this moment the Magnova doesn't have much to offer in terms of die-hard signal analysis (like making measurements over all the acquired data and producing statistics based on an entire acquisition). Not sure whether Batronix wants to go this route though as I don't know the roadmap. OTOH the trend chart is a rather unique feature but it can only update information as fast as the waveform (acquisition) update rate as there is only one measurement result per acquisition.

All in all I am impressed though with what the Magnova has to offer. What is there works, looks & feels solid. IMHO it is safe to say Batronix has entered the DSO stage as a serious oscilloscope manufacturer; one to be reckoned with.

I also want to express my thanks and gratitude to forum member Woody for driving all the way to my lab and allowing me to get some hands-on experience.

[Martin72]

It's great that you were able to take a look at it.

What is nice though is that the colors of traces in general can be adjusted to allow extra clarity.

I'm used to this from Lecroy and Siglent and wouldn't want to be without this feature anymore.

[hpw]

Any reasons, why using quick-save, gets the traces as with large Y axis steps, as far from 12 bit resolutions.

Seen on Martin & your given plots... see pictures as 3x optical zoomed.

 

[woody]

Screen resolution?

[FloBX]

Any reasons, why using quick-save, gets the traces as with large Y axis steps, as far from 12 bit resolutions.

Seen on Martin & your given plots... see pictures as 3x optical zoomed.

I just attached two screenshots showing the behavior you were referring to combined with a pixel grid. I assume what you observed was based on the rescaling behavior of your graphics software. Please let me know, if I did not get the actual point.

• Your first example shows a screen with multiple records/traces combined (I guess three records based on a FM modulated signal). Nonetheless the trace has a thickness between one and three pixels.
  
• The second example seems to refer to a combined trace formed by overlaying even more record traces. Also here we have a minimal thickness on pixel level.
  

For the Magnova, there is no reduction of the waveform display resolution with regard to the actual screen resolution (also screenshots are a 1:1 copy of the screen contents). We would like to encourage you, to compare this behavior to other oscilloscope devices available.

Best regards,
Florian

[ebastler]

Any reasons, why using quick-save, gets the traces as with large Y axis steps, as far from 12 bit resolutions.

Of course there can be vertical steps which are much larger than the 12 bit ADC steps. If the slope of the input signal is steep, i.e. the signal changes significantly between time points that are rendered, what else would you expect?

[2N3055]

Any reasons, why using quick-save, gets the traces as with large Y axis steps, as far from 12 bit resolutions.

Of course there can be vertical steps which are much larger than the 12 bit ADC steps. If the slope of the input signal is steep, i.e. the signal changes significantly between time points that are rendered, what else would you expect?

Not only that. Any bitmap image will show steps once you magnify enough, and single pixel width straight lines plotted on screen will have exactly one pixel per step only at exactly 0°, 45° and 90° angles.

That particular person is known to asks for some outraging requirements, without understanding the topic.
I would point people to get acquainted with Bresenham's or Xiaolin Wu's line algorithm for better understanding of this topic.

[FloBX]

Magnova tests

As I skipped this before:
Also many thanks to Nico and Woody for sharing your impressions and evaluations.
"Entering the stage" as new oscilloscope manufacturer comes with the requirement of having to gain the trust of the community, we therefore sincerely value all customer feedback and are especially grateful for feedback supporting our claim: That we have delivered a serious measurement device that can (at least) stand honest comparison against the competition. We know that further extended features are expected to follow in the foreseeable future and we are committed to deliver.
We will (and already did) discuss your suggested changes/optimizations. Please understand that some device properties are based on actual design decisions (e.g. cursor behavior, visibility of reference traces), which does not necessarily mean that we will not reconsider certain behavior based on all customer feedback in the medium term.
We have, however, "reserved" the immediate post-release time (now!) to be able to address commonly requested optimizations and of course potential undetected stability issues we were also not completely immune to. This is what the software/firmware team is currently primarily dedicated to.

Nonetheless, there is some clarification, we would like to add to some of the points elaborated:

Issues / remarks
During testing we encountered one issue: the samplerate got stuck at 1Gs/s even with only one channel enabled. A power cycle didn't help but a reset to factory defaults rectified the problem.

If you have selected channels 3 and/or 4 as FFT, trigger or measurement source, they will be enabled implicitly, even if they are not displayed in the waveform display area. We assume that this is what prevented the device from switching back to 1.6 GSa/s by default.

FFT speed. Spectrum analyser-esque options?
[...]
I'm also missing a way to set the start & stop frequency. Personally I rarely use center and span on a spectrum analyser because most of the measurements I do have the frequency range defined as start / stop frequencies.

For configuration of the FFTs, you can easily switch between Center/Span or Start/Stop definition by clicking on the FFT Center(/FFT Start) encoder knob (as displayed, see attached screenshot).

It was a design decision to have the FFT always based on the samples visible in the waveform display area, only. Control and display of utilized FFT points should, however, become more transparent.

Measurements? Statistics? Histograms?
Onto measurements... The measurements are performed on acquired data so even when the shape of the waveform isn't visible, you still get the right value. So that is good. There is an option to show labels to indicate where the measurement has been taken. It looks like that a measurement is taken closest to the trigger point and per acquisition only 1 measurement is made. So statistics are gathered over multiple acquisitions. There is no histogram function to show the statistic distribution of the values.

The amount of measurement results per acquisition is not limited to a single record. The maximum amount of measurement runs per acquisition is currently set to 32 (so 32 up to records "measured" per waveform display update).
All single records can be evaluated separately based on the measurements by using the history functionality.
Additional features like histogram are already on our wish list.

Thanks again for sharing your findings!

Best regards,
Florian
(Batronix)

[2N3055]

Measurements? Statistics? Histograms?
Onto measurements... The measurements are performed on acquired data so even when the shape of the waveform isn't visible, you still get the right value. So that is good. There is an option to show labels to indicate where the measurement has been taken. It looks like that a measurement is taken closest to the trigger point and per acquisition only 1 measurement is made. So statistics are gathered over multiple acquisitions. There is no histogram function to show the statistic distribution of the values.

The amount of measurement results per acquisition is not limited to a single record. The maximum amount of measurement runs per acquisition is currently set to 32 (so 32 up to records "measured" per waveform display update).
All single records can be evaluated separately based on the measurements by using the history functionality.
Additional features like histogram are already on our wish list.

Thanks again for sharing your findings!

Best regards,
Florian
(Batronix)

Florian,

A clarification if I may?

So if you take Single shot of a signal that shows, say, 20 periods on the screen, and enable Rise-Time measurement, it will keep going and make measurement on all 20 periods, and display statistics for 20 from that single capture.
And limit is 32 per one sample buffer (trigger event) ?

Keysight 3000T will measure only one per trigger, and gather stats only across trigger events.
Some other scopes will measure multiple per trigger and have higher adjustable limits..

Best,
 Siniša

[FloBX]

Florian,

A clarification if I may?

So if you take Single shot of a signal that shows, say, 20 periods on the screen, and enable Rise-Time measurement, it will keep going and make measurement on all 20 periods, and display statistics for 20 from that single capture.
And limit is 32 per one sample buffer (trigger event) ?

Keysight 3000T will measure only one per trigger, and gather stats only across trigger events.
Some other scopes will measure multiple per trigger and have higher adjustable limits..

Best,
 Siniša

Currently, we store one measurement result per record and measurement type (all requested measurements are calculated based on the same set of records).
So there will be one per trigger(/per record) and no record-internal statistics, apart from those measurements that are actually based on record-internal statistics, of course.

The limit of 32 therefore defines the amount/subset of records per acquisition, based on which all measurements are calculated (with all measurements being calculated for every record).

[FloBX]

Maybe as general clarification (since I unthoughtfully switched to our "internal" nomenclature):
With acquisition I referred to the process of obtaining a (potentially huge) amount of records/waveforms "as a burst", prior to performing subsequent (non-acquisition) processing (e.g. waveform display, measurements).
Multiple succeeding acquisitions per second lead to the displayed waveform update rate (as wfms/s).

Martin's test setup showed this common behavior:

So I did a quick and dirty measurement with SDG2122X and the BMO350 in default condition.
I connected the other scope to the aux out for control; you have to activate the output accordingly in the Magnova's trigger menu.
A few checks were enough to determine that the waveforms/sec display can be trusted.
The “real” measurements will be done with SDG2122X and SSG3032X and then also with different configurations regarding display and memory.

[nctnico]

Florian,

A clarification if I may?

So if you take Single shot of a signal that shows, say, 20 periods on the screen, and enable Rise-Time measurement, it will keep going and make measurement on all 20 periods, and display statistics for 20 from that single capture.
And limit is 32 per one sample buffer (trigger event) ?

Keysight 3000T will measure only one per trigger, and gather stats only across trigger events.
Some other scopes will measure multiple per trigger and have higher adjustable limits..

Best,
 Siniša

Currently, we store one measurement result per record and measurement type (all requested measurements are calculated based on the same set of records).
So there will be one per trigger(/per record) and no record-internal statistics, apart from those measurements that are actually based on record-internal statistics, of course.

The limit of 32 therefore defines the amount/subset of records per acquisition, based on which all measurements are calculated (with all measurements being calculated for every record).

I don't quite follow / fully understand this yet.

What we have tried yesterday was using the 'single' button to make several singular acquisitions (as in one trigger event). This resulted in incrementing the number of measurements by one for each acquisition. Hence my conclusion that each measurement result is based on taking one measurement from one acquisition.

You mention a maximum of 32 measurement updates per display update. Am I right when I say there is a maximum of 32 measurement updates per display update? So if you have 10 display updates per second and the Magnova is acquiring 1000 waveforms/second then this would mean 1000 / 10 = 100 waveform updates for each display update from which 32 are used for calculating measurements and thus 100 - 32 = 68 acquisitions are not used for measurements. Is that correct?

[FloBX]

You mention a maximum of 32 measurement updates per display update. Am I right when I say there is a maximum of 32 measurement updates per display update? So if you have 10 display updates per second and the Magnova is acquiring 1000 waveforms/second then this would mean 1000 / 10 = 100 waveform updates for each display update from which 32 are used for calculating measurements and thus 100 - 32 = 68 acquisitions are not used for measurements. Is that correct?

Yes, this is correct, with measurements being performed based on the most recently captured waveforms.

[king2]

Additional features like histogram are already on our wish list.

Will histogram functionality added as software update?
When it can be expected?
Do you planning to make some updates not free?

[Martin72]

Will histogram functionality added as software update?

How else should it be done....

[king2]

Will histogram functionality added as software update?

How else should it be done....

I have no idea. Maybe this will also require hardware upgrade. So my question just to be sure

[shabaz]

Does the Magnova support a 'several-acquisition' form of using the single trigger button? I couldn't see it in the user guide, but I may have missed it.
If it's not present, just a personal opinion, I think it should be added as a feature request (maybe it already is on the list, since the 'scope is so new and new things may be quickly added). On the R&S MXO series (and maybe on other R&S 'scopes but I don't know) the feature is called N-Single. The screenshot explains it.

It's especially relevant to high-resolution 'scopes, because it allows the user to see the waveform shape trend when the single trigger button is used, whereas if there was just a single acquisition, then, simply because you can scrutinize in higher-res, you wouldn't know if what you're seeing could be caused by noise present on the signal, or not.

I'd not seen that feature in the past, but it's really useful now.

[Andre77]

Will histogram functionality added as software update?
When it can be expected?
Do you planning to make some updates not free?

Thank you for your enquiry! The histogram function is a valuable analysis function and we have already added it to our list.

We are currently focussing on improvements and small new functions. I cannot yet say when the histogram function will be implemented. We will publish a roadmap in a few weeks' time, in which we will then prioritise various planned mid+major enhancements and functions.

We do not plan to charge for updates.

Does the Magnova support a 'several-acquisition' form of using the single trigger button? I couldn't see it in the user guide, but I may have missed it.
If it's not present, just a personal opinion, I think it should be added as a feature request (maybe it already is on the list, since the 'scope is so new and new things may be quickly added). On the R&S MXO series (and maybe on other R&S 'scopes but I don't know) the feature is called N-Single. The screenshot explains it.

The function is well known to us and we already have it on our list. Thank you for the suggestion, we use every enquiry to estimate the relevance of the extensions.