Magnova oscilloscope

[RAPo]

Why, oh why, the negative tone for a product that has not yet been released? Every product starts with a first release, and little is known.
We should applaud Batronix for its entrepreneurship and for trying some innovation, and then when the release is out, we should test it  and, if necessary, help to make it better.

[jusaca]

Oha, they are up again with 5 to 10 days. Seems like there still is quite some stuff to sort out with the firmware?

[FloBX]

We understand your concerns, but actual shipping of the first Magnova devices will start tomorrow. From then on, we will gradually process orders that have already been received. New orders will therefore be delivered with a delay of a few days, which is why we have increased the delivery time at batronix.com.

Best regards
Florian
(Batronix)

[jusaca]

Interesting, thanks for that heads-up. I'm exited to see the reviews testing this new gadget.

[Martin72]

Me too.....

[egonotto]

We understand your concerns, but actual shipping of the first Magnova devices will start tomorrow. From then on, we will gradually process orders that have already been received. New orders will therefore be delivered with a delay of a few days, which is why we have increased the delivery time at batronix.com.

Best regards
Florian
(Batronix)

Hello,

I suspect that there is an error in the calculation of the ENOB in the data sheet (version 0.91). It is probably calculated with Vpp instead of Vrms. The influence of the clock signal does not seem to have been taken into account either.

Best regards
egonotto

[Momchilo]

They published the manual today.
Only in German so far: https://www.batronix.com/files/Batronix/Magnova/Manual/Magnova-Anleitung-DIN-A5_DEU_1_0_0.pdf

[woody]

Ohoh, I'll have to brush up my German then.

Although, on the Benutzeroberfläche the 'Schnelltasten' are called 'Quicksave, Autoset and Clear, which indicates Magnova speaks at least some English 

[LoneWolf6912]

In the end they are both germanic language 

[egonotto]

Hello,

do I understand correctly that no USB3 cable is supplied for the connection to the PC?

Best regards
egonotto

[jusaca]

do I understand correctly that no USB3 cable is supplied for the connection to the PC?

I would assume a rather small percentage of customers will use the USB uplink. Most will use Ethernet, I guess.

[CosteC]

I suspect that there is an error in the calculation of the ENOB in the data sheet (version 0.91). It is probably calculated with Vpp instead of Vrms. The influence of the clock signal does not seem to have been taken into account either.

Best regards
egonotto

Interesting fact: maximum input is 30 pk for 50 Ohm input - 18 W. Either Magnova has impressive technology or bug in datasheet.

ENOB is provided even for 5 V/DIV at 50 Ohm, suggesting even higher input.

I find magical that channels 1&2 have 1.6 GSa/s while 3&4 have 1.0 GSa/s and those channels cause drop on channel 1&2. Oscilloscopes with two dual channel ADCs tend to be symmetrical and divide sampling by two not fractional...

[2N3055]

I suspect that there is an error in the calculation of the ENOB in the data sheet (version 0.91). It is probably calculated with Vpp instead of Vrms. The influence of the clock signal does not seem to have been taken into account either.

Best regards
egonotto

Interesting fact: maximum input is 30 pk for 50 Ohm input - 18 W. Either Magnova has impressive technology or bug in datasheet.

ENOB is provided even for 5 V/DIV at 50 Ohm, suggesting even higher input.

I find magical that channels 1&2 have 1.6 GSa/s while 3&4 have 1.0 GSa/s and those channels cause drop on channel 1&2. Oscilloscopes with two dual channel ADCs tend to be symmetrical and divide sampling by two not fractional...

Specification is:
At 50 Ω: 5 Vrms, 30 Vpk

You cannot pick and choose what information you will read and which you will ignore. Also, Vpeak is not same as V P-P.

As for unusual sampling combinations, I would presume it is a limitation of data pump in the scope.
If you pay attention it is not only that 3 and 4 analog channels that impose limit but different combinations with LA.
It is not easy to process all that data in real time.

[nctnico]

I suspect that there is an error in the calculation of the ENOB in the data sheet (version 0.91). It is probably calculated with Vpp instead of Vrms. The influence of the clock signal does not seem to have been taken into account either.

Best regards
egonotto

Interesting fact: maximum input is 30 pk for 50 Ohm input - 18 W. Either Magnova has impressive technology or bug in datasheet.

If you read further, the maximum is 5Vrms. So for as long as the 30V peaks are not adding up to >5Vrms, it is OK. BTW, my old Tektronix TDS510A had a 10V/div setting while in 50 Ohm mode while allowing +/-30V peaks as well. Most modern day scopes seem to cut out at 1V/div in 50 Ohm mode while the Magnova allows up to 5V/div in 50 Ohm mode. Nice!

[Andre77]

Interesting fact: maximum input is 30 pk for 50 Ohm input - 18 W. Either Magnova has impressive technology or bug in datasheet.
ENOB is provided even for 5 V/DIV at 50 Ohm, suggesting even higher input.

The 50 Ω input is specified with 5 Vrms, 30 Vpk (0.5 watts). Short peaks of up to 30 Vp are permitted as long as the 5 Vrms are maintained.

Vertical scaling is not limited even with 50 Ohm input termination. This allows, for example, that two channels can be displayed with the same scaling (5 V/div), even if one of the two is terminated with 50 ohms. However, regardless of the vertical scaling, you must observe the maximum voltage when using the 50 Ohm termination.

I find magical that channels 1&2 have 1.6 GSa/s while 3&4 have 1.0 GSa/s and those channels cause drop on channel 1&2. Oscilloscopes with two dual channel ADCs tend to be symmetrical and divide sampling by two not fractional...

We use a 4 channel 1.6 GSa/s 12 bit ADC from TI. The limitation to 1 GSa/s with 4 active channels does not come from the ADC, it comes from a limitation of the maximum data processing rate in the FPGA/RAM. Rate limitation is achieved directly through reduction of sampling clock frequency.

I suspect that there is an error in the calculation of the ENOB in the data sheet (version 0.91). It is probably calculated with Vpp instead of Vrms. The influence of the clock signal does not seem to have been taken into account either.

The ENOBs were calculated on the basis of the Vrms noise values. Where do you see a mistake? I have validated this briefly at 100 mV/div and come up with the same result as in the data sheet.
SNR: (8 div *100mV/div) / 502µVrms =1593.6
SNR [dB]: 20*LOG10(1593.6)=64.05 dB
(64.05 dB - 1.76 dB) / 6.02 =10.34 bit

do I understand correctly that no USB3 cable is supplied for the connection to the PC?

It's true, we don't currently include a USB 3 cable. I will address this internally.

Ohoh, I'll have to brush up my German then.

The English-language manual is almost ready and will also be made available online in a few days.

[egonotto]

...

I suspect that there is an error in the calculation of the ENOB in the data sheet (version 0.91). It is probably calculated with Vpp instead of Vrms. The influence of the clock signal does not seem to have been taken into account either.

The ENOBs were calculated on the basis of the Vrms noise values. Where do you see a mistake? I have validated this briefly at 100 mV/div and come up with the same result as in the data sheet.
SNR: (8 div *100mV/div) / 502µVrms =1593.6
SNR [dB]: 20*LOG10(1593.6)=64.05 dB
(64.05 dB - 1.76 dB) / 6.02 =10.34 bit

...

Hello,
 
(8 div *100mV/div) = 800 mV. If you feed in a sine wave with 800 mVpp, you have full scale. But that is about 290 mVrms. That gives an SNR of about 570, so about 55 dB. That gives an ENOB of about 8.9 bits

Best regards
egonotto

[nctnico]

AFAIK ADC ENOB is calculated using ADC full scale. Which in this case is 800mV to cover 12 bits of resolution (assuming there isn't a little bit of headroom left). Think about how to get back to calculating the ENOB to 12 bit without noise.

[egonotto]

AFAIK ADC ENOB is calculated using ADC full scale. Which in this case is 800mV to cover 12 bits of resolution (assuming there isn't a little bit of headroom left). Think about how to get back to calculating the ENOB to 12 bit without noise.

Hello,

If you have no noise, SNR is infinite. Then you get an infinite number of bits.
ADC bits do not appear directly in the formulas for ENOB. This is why an ENOB can result that is larger than the ADC bits if the noise is assumed to be small enough..

Best regards
egonotto

[Martin72]

How do the 502µVrms come about?
Is this the value measured at 100mV/div without an external signal?

[2N3055]

I believe confusion comes from fact that formula is using both RMS and P-P values.

But they are not mixed.

First SINAD is calculated from RMS of signal applied and residual noise+distortion with test signal removed from calculation. On scope that can be well approximated in FFT and delta from cursors. Ideally we would calculate FFT and remove bin with test signal and then calculate RMS on reminder and use that in calculation.

But since we usually do not measure completely full range (to optimize for less distortion) there is a term at the right side of equation that serves to normalize results to full range of ADC.  That one uses full range of scope range (that is by definition P-P) and P-P value of the signal applied.

[egonotto]

How do the 502µVrms come about?
Is this the value measured at 100mV/div without an external signal?

Hello,

in the data sheet of the Magnova BMO350 it is the noise at 100 mV/div and 50 Ohm.

Best regards
egonotto

[Wolfgang]

We understand your concerns, but actual shipping of the first Magnova devices will start tomorrow. From then on, we will gradually process orders that have already been received. New orders will therefore be delivered with a delay of a few days, which is why we have increased the delivery time at batronix.com.

Best regards
Florian
(Batronix)

Hello,

I suspect that there is an error in the calculation of the ENOB in the data sheet (version 0.91). It is probably calculated with Vpp instead of Vrms. The influence of the clock signal does not seem to have been taken into account either.

Best regards
egonotto

When I look up the Magnova datasheet their best ENOB is 11.6 (!) for 20MHz and the 100MHz scope, on a 12Bit ADC. This is more than a whole bit better than RS MXO and RIGOL HDO. I call bullshit on these numbers, as they are not supported by the underlying ADC data used. The difference to a realistic value (10.something bits) could be a Vpp/Rms confusion.

[Martin72]

Hello,

in the data sheet of the Magnova BMO350 it is the noise at 100 mV/div and 50 Ohm.

Hi my friend,
I know that...
I was wondering where the value came from in the first place.
Is it the inherent noise without an external signal feed, that wasn't clear to me.
I can't imagine that it's that “simple”, otherwise Wolfgang wouldn't go to so much trouble to determine the ENOB.
https://electronicprojectsforfun.wordpress.com/to-enob-or-not-to-enob/

[nctnico]

We understand your concerns, but actual shipping of the first Magnova devices will start tomorrow. From then on, we will gradually process orders that have already been received. New orders will therefore be delivered with a delay of a few days, which is why we have increased the delivery time at batronix.com.

Best regards
Florian
(Batronix)

Hello,

I suspect that there is an error in the calculation of the ENOB in the data sheet (version 0.91). It is probably calculated with Vpp instead of Vrms. The influence of the clock signal does not seem to have been taken into account either.

Best regards
egonotto

When I look up the Magnova datasheet their best ENOB is 11.6 (!) for 20MHz and the 100MHz scope, on a 12Bit ADC. This is more than a whole bit better than RS MXO and RIGOL HDO. I call bullshit on these numbers, as they are not supported by the underlying ADC data used. The difference to a realistic value (10.something bits) could be a Vpp/Rms confusion.

Actually, the ADC which is used in the Magnova has an ENOB of 9.3 bits at best so any higher number is suspicious.

Maybe the noise floor in Vrms can be used to take a stab at how many bits are missing and get some best case ENOB based on SNR only. AC RMS is standard deviation which is 1 sigma in statistics. 3 sigma covers 99.7% of the values and is removing the outliers. Using 3 sigma we would have an apples to apples relation with full scale.

If I take the numbers used earlier for 800mV full scale and  502µVrms of noise, the SNR would be 20 * log(0.8 / (502uV * 3)) = 54.47dB
Number of bits would be around (54.47 - 1.76) / 6.02 = 8.8

The SINAD values for the ADC used seems to be equal to SNR at the bandwidths of the Magnova scope so if the analog front-end doesn't add much distortion the SNR should be dominant.

[joeqsmith]

The English-language manual is almost ready and will also be made available online in a few days.

Looking forward to reading it. 

Please note: Due to pending product certifications, the sale of the Magnova oscilloscope is currently not available in the United States and Canada.

Any estimates for when it will be available in the USA? 

***
Attached after running it through Google translate (English).   Good enough while we wait for the real manual.