Magnova oscilloscope

[king2]

After cold calibration everything looks to be fine.
While warming up, CH4 shifted up by 1-2mV:



CH2 and CH3 was near zero, CH1 was shifted up a little bit too (but much less than CH4).

But after I tried to shift channel zero point from middle of dispay, CH4 was jumped back to near zero point.
Any next shifts (by touch, by encoder rotation, by encoder pressing) did not affected CH4 offset, which was like normal from this moment.

Maybe, thermal compensation applying to channels not while temperature  changes, but while you trying to change channel line offset on display. In my scope this affected CH4 only, but I will try to test again later.

Some first impressions:
1. VERY responsive interface. Simply the best.
2. This scope definitely deserved its own bootup logo
3. Maybe it will be nice to make temperature compensation with piecewise line approximation and 'long calibration' mode, in which we turn on scope on cold mode, then enter such mode, then it warming up, making calibration each X degrees (maybe more frequently near working range).

[Martin72]

After cold calibration everything looks to be fine

After calibrating when cold, you have to wait until it shows warm and then calibrate again.
That should be enough.
This is then roughly what you wrote under point 3.

[king2]

After calibrating when cold, you have to wait until it shows warm and then calibrate again.
That should be enough.

I will do this today later.

This is then roughly what you wrote under point 3.

Difference is that instead of one set of approximation coefficients fentire temperature range we will have one set for each temperature range, what can do its task better (if ADC and input schematic temp offset is not linear).

[king2]

How (and when) thermal offset compensation is applied?

If I turn on the scope (calibrated both cold and warmed) at a cold state - first 3 channels works like a charm (at first glance), but CH4 gradually climbing up while scope's temperature raises.
I think CH4 has something in its electronics located near hot parts of a scope, so temperature changes inpact it more then other channels (CH2 and CH3 are most stable channels in my scope).

But if I change time scale or offset - this raised channel immediately drops to its initial (calibrated near zero) position.



In video I told that this can be a bug in loop condition, but no - this effect can be seen on other channels too (but you have to wait a little bit more).

Definitely, thermal compensation applied while changing time scale or offset of any channel.
Maybe it needs to be applied also from time to time without touching buttons, because even warmed up scope will change its temperature when turning on/off its cooling fan.

[Andre77]

How (and when) thermal offset compensation is applied?

Outside of roll mode, an automatic offset compensation is carried out each time the temperature changes. The process takes place in the ‘blind time’ between two recordings and is so fast that it goes unnoticed. After switching the oscilloscope on (cold start), this initially occurs nearly every second and then gradually becomes less frequent.

In roll mode, the behaviour is different: The data is recorded in a continuous stream. Intermediate offset compensation does not take place in this mode. To ensure a stable offset position in roll mode (especially in combination with a fine vertical scale), the oscilloscope should therefore be warmed up to operating temperature beforehand.

[ZPeterZ]

My impression is, that I have to do the offset calibration each time I want to look at sub mV DC levels.
Yesterday evening I did the cold calibration and 90 minutes later the warm calibration. Today morning powered on and the offsets I see are up to 120µV. An hour later these measure up to 80µV - I guess this is normal. While changing vertical scale the offsets stay consistent and of course there is no influence from changing time scale.
I will compare to my Siglent, which offers 500µV/div as finest vertical scale setting.

edit: I did not use roll-mode.

[ZPeterZ]

I did the offset test with the Siglent SDS1104E-X which is an entry level scope so it is not a real challenge.
The scope was unused for some weeks, but IIRC I did a self cal last time I used it. This scope has only one temperature for the cal, warmed up.
After powering on it started with offsets up to 900µV in the 500µV/div setting. Waiting for an hour the offsets were reduced and measured up to ~ 250µV. Changing the vertical scale to 10mV/div seemed to double the offset. 90 minutes later I did a self cal. That took some minutes. After that, the offsets were still there and some looked even worse than before, so I can leave out the "next day test". What was better after the self cal was the consistency over different vertical scale settings. The "quick cal" on or off feature did not change the offsets.

edit: With 500µ/div the zero lines of all 4 channels are only met, when the input coupling is set to GND on the SDS.
edit2: No, channel 3 hits the zero line also in the AC setting. Some channels invert polarity of the offset when changing from DC to AC coupling.

[Martin72]

The offset behavior of an SDS1000X-E is more something for the SDS1000X-E thread.

[joeqsmith]

I am envious of all this drift stuff.   My scope doesn't have enough resolution to detect such small changes.  When it drifts, it's several mV, not uV.   It constantly autocals, trimming that offset to zero, and loosing my data. 

[ZPeterZ]

> The offset behavior of an SDS1000X-E is more something for the SDS1000X-E thread.

Basically your are right but if it comes to a comparison with the BMO, should I write "this is my result with the BMO and the result of my other scope can be read somewhere else"?

[ZPeterZ]

Hi,

I wanted to get a feel for the frequency response of my BMO200.

The fastest way to get the bandwidth is to apply a squarewave with a fast edge and measure rise- and fall-times. These are <= 1.75ns according to the datasheet. The formula to calculate the bandwith is :
     
bandwidth = 0.35 divided by rise-time

This is for the BMO200: 0.35 / 1.75ns = 200MHz in the worst case.

First I setup a pulse generator to output a squarewave with 100MHz and 5Vpp. Rise- and fall-times were set to the minimum.  Because the generator is old and not calibrated I checked the signal with a fast scope and measured less than 1ns risetime and falltime. So the signal should be sufficient for this test.



With the BMO200 the measured rise-time of 1.405ns is equal to a bandwidth of 249MHz. The slowest rise-time I measured was 1.45ns with a 1Vpp signal. This still results in a bandwidth of 241MHz. Not bad I think.


Now I have the bandwidth but no idea, how steep the filter slope might be.

I set the scope to display the FFT up to 800MHz in the max-hold mode. The 0dBm stimulus came from a sweep generator. The generator does discrete steps, so it took a long time to fill the curve and even where it looks smooth and solid, it is in fact not. Moving a marker one step further can result in a complete different level reading. So I looked carefully at the horizontal gridlines (6dB/div), to place the markers right. edit: It is all about peaking a value from the top of the curve.



The -3dB frequency given from the FFT is at about 292MHz. One octave higher the curve is roughly another 12dB down.


Because the last measurement took a lot of time I thougt about a faster method. Then I remembered that the R&S VCA on the shelf has an option installed that provides a noise source (5MHz - 1GHz). The output is 75 ohms so the signal was applied with a minimum loss pad for 50 ohms conversion.



As can be seen on the photo, it would have also taken a lot of time to get a smooth curve.

Last I tried the FFT average function which seems to be suitable - maybe an average of 1000 would be fine.




- Peter

[ZPeterZ]

I repeated the last measurement with noise applied to channel 4 at 1Gs/s. The timescale was set to the maximum value that allowed the FFT to be filled up to 500MHz, 200µs/div. With a curve built from an average of 1000 sprectra is still not easy to place markers right; even tenfold more spectra leave you space for variation in values. In this case the min and max tables did not help either.

This lets me think if it would be possible to apply some post processing to the FFT curve, to draw some average line that the markers can be related to? (this would be some kind of low-pass over the data series). Or do I miss some other possibility, i.e. like increasing the RBW to smoothen the curve?

- Peter

edit: I forgot to post the result of the test: The -3db frequency of channel 4 is about 215MHz. At 430MHz the signal is lowered by further ~9dB.

[2N3055]

I repeated the last measurement with noise applied to channel 4 at 1Gs/s. The timescale was set to the maximum value that allowed the FFT to be filled up to 500MHz, 200µs/div. With a curve built from an average of 1000 sprectra is still not easy to place markers right; even tenfold more spectra leave you space for variation in values. In this case the min and max tables did not help either.
This lets me think if it would be possible to apply some post processing to the FFT curve, to draw some average line that the markers can be related to? (this would be some kind of low-pass over the data series). Or do I miss some other possibility, i.e. like increasing the RBW to smoothen the curve?
- Peter

Just a word of advice, if I may.
When plotting BW plot, there is no need for a RBW of 2 kHz.
Set RBW to 1 MHz and less averages.
Frequency response deviations are not that peaky.
You'll get the same result but a lot faster.
And you'll get more continuous plot where cursors will be easier to work with..

[ZPeterZ]

Just a word of advice, if I may.
When plotting BW plot, there is no need for a RBW of 2 kHz.
Set RBW to 1 MHz and less averages.
Frequency response deviations are not that peaky.
You'll get the same result but a lot faster.
And you'll get more continuous plot where cursors will be easier to work with..

You are right and setting the RBW can be done directly on a spectrum analyzer. IMHO here it is implied by the number of sampled values. If I want to cover the range up to 500MHz, I do not know, how to get larger RBW values. I set the memory depth to auto(max) which results in 2.4Mpts/ch. Lowering this value results in a FFT that ends well below 500MHz.
Any hint?

[2N3055]

Just a word of advice, if I may.
When plotting BW plot, there is no need for a RBW of 2 kHz.
Set RBW to 1 MHz and less averages.
Frequency response deviations are not that peaky.
You'll get the same result but a lot faster.
And you'll get more continuous plot where cursors will be easier to work with..

You are right and setting the RBW can be done directly on a spectrum analyzer. IMHO here it is implied by the number of sampled values. If I want to cover the range up to 500MHz, I do not know, how to get larger RBW values. I set the memory depth to auto(max) which results in 2.4Mpts/ch. Lowering this value results in a FFT that ends well below 500MHz.
Any hint?

I am sorry but I don't have Magnova scope so don't know how can you set it.

Usually you just want smaller number of FFT bins, and shorter timebase, while keeping sample rate high.
Also for best amplitude measurements use Flat top window.

[ZPeterZ]

Thanks for your reply! In deed I did not change the window type so far and I will see if flat top alters the results significantly. How the lowering of the FFT bin number can be done while maintaining the overall bandwidth is still not clear to me.

[Martin72]

I had also looked into it:

https://www.eevblog.com/forum/testgear/magnova-oscilloscope/msg5675673/#msg5675673

[ZPeterZ]

I had also looked into it:

https://www.eevblog.com/forum/testgear/magnova-oscilloscope/msg5675673/#msg5675673

Hi Martin,
the Quicksave problem seems to be gone - I did not experience any effects with v1.1.7.
What I can`t see: Do you managed to do synchronized sweeps?
As your curves look much smoother: larger RBW due to a shorter time scale? Unfortunately I do not have the time to test it now.
And of course I should use the horizontal cursors.. I guess it was too late at night...  .
- Peter

[Martin72]

I generally use these settings for the generator and the scope:
https://www.eevblog.com/forum/testgear/bandwidth-limit-on-siglent-sds2000x-plus-oscilloscope/msg5208711/#msg5208711

Depending on the scope model, something still needs to be changed in terms of the number of points.

the Quicksave problem seems to be gone - I did not experience any effects with v1.1.7.

Ah, that would be good.
I can't test it at the moment, the scope is still at work.

[laminarlade]

Hi Andre77,

I wonder if you can share some data about the achievable speed on the LAN and USB interfaces once the SCPI interface is available. Assuming the scope can be triggered remotely and the waveform buffer can be read out .

[FloBX]

Hi laminarlade,

we are are currently "wrapping" the next update package, which will come along with the initial documentation for remote control via SCPI commands through various interfaces.

Also, we are preparing a corresponding performance overview.

Best regards,
Florian

[laminarlade]

Hi Florian,

thanks for your reply.

I'm in a bit of trouble now  . At the moment I am writing a driver for ngscopeclient for my old oscilloscope (HMO-3034, 4Gs/8Mpts). The scope is sufficient for my work, but the transfer rates are painfully slow (~50kB/s).

The current discount is tempting, but without acceptable transfer rates there is no justification to buy a new Scope. Apart from the fact that you can never have enough test equipment  .

Michael

[FloBX]

Hi,

as promised, we can now inform you, that Magnova Firmware Package version 1.2.1 is available.
It comes with a documentation for SCPI-based remote control functionality as well as extended file browser functionality.

See a full list of changes below. All customers have been informed.

New functionality:

• Added official support for SCPI-based remote control functionality.
  Note: Corresponding introduction and documentation is provided at https://www.batronix.com/magnova/en/downloads.
• Added support for multiple selection as well as cut, copy and paste file browser operations.

Optimizations:

• Updated default hysteresis for Runt and Window trigger types to 2.5% (previously 0%).
• Raised maximum baud rate for CAN/CAN-FD decoding.
• Various minor optimizations.

Bugfixes:

• Fixed freeze issue when connecting to SMB shares as a guest.
• Fixed analog channel probe divider handling for math channels.
• Fixed inaccuracies in decoded data timestamps when exporting to CSV.
• Fixed changes in spectrogram source not being applied to the full extend.
• Fixed analog channel probe divider usage for the FFT.
• Fixed issue where files were not displayed upon initially opening the file browser.
• Several minor bugfixes.

We have conducted a simple transfer speed test and have achieved a performance of about 5.8 MPts/s via USBTMC and 23.2 MPts/s via Ethernet for this firmware package version.

At https://github.com/Batronix/MagnovaExamples/ we provide a list of corresponding examples and are planning to expand it in the future.

Best regards,
the Batronix Team

[egonotto]

Hello,

how large is the memory for arbitrary in the generator module (option)?

Best regards
egonotto

[king2]

Andre77 and FloBX, thank you for your support!

Can you answer a few questions:
1. When you planning to add histogram functionality into Magnova?
2. Same question about web interface?
3. Same question about desktop software? I can turn its support in Magnova's interface, but does such software exists?
4. Do you have some development roadmap that you can share with your customers?

Thank you in advance!