new killer scope in town - a true game changer from R&S - RTB2002 & RTB2004

[MikeP]

 And how this signal looks on the FFT screen. Settings and timebase do not matter. We see nonexistent distortions. In one form or another.

 Dear Rich and Markus - whether there is an opportunity to correct a situation and to add a logarithmic scale for mode FFT?

 Thanks.
 Mike.

[Fgrir]

These are some examples of FFT function. I think for precision measurements this device is unsuitable.
 The signal source is G3-118. The first harmonic was suppressed by rejector (- 100db). The second harmonic - 106 dB. For me it is physical vacuum.

The channel-to-channel  isolation spec is only 50dB, so looking at the low signal on channel one with the 10V source on channel 4 is bound to give you problems.  For the rest, it looks to me like you are getting 60dB+ dynamic range out of a 10-bit device, just how good do you expect it to be?

[mtdoc]

These are some examples of FFT function. I think for precision measurements this device is unsuitable.
 The signal source is G3-118. The first harmonic was suppressed by rejector (- 100db). The second harmonic - 106 dB. For me it is physical vacuum.

The channel-to-channel  isolation spec is only 50dB, so looking at the low signal on channel one with the 10V source on channel 4 is bound to give you problems.  For the rest, it looks to me like you are getting 60dB+ dynamic range out of a 10-bit device, just how good do you expect it to be?

It says a lot that the expectations for FFT functionality in an oscilloscope are so high (and perhaps unrealistic).  If frequency domain information is that critical to someone, an oscilloscope is the wrong tool.  But it is a nice bonus that DSO's have come so far and are now including usable (if not perfect) FFT functionality.   There are some bugs here and I'm sure R&S will continue to fine tune this scopes FFT software, but overall I'm pretty impressed with it.

[Robaroni]

These are some examples of FFT function. I think for precision measurements this device is unsuitable.
 The signal source is G3-118. The first harmonic was suppressed by rejector (- 100db). The second harmonic - 106 dB. For me it is physical vacuum.

The channel-to-channel  isolation spec is only 50dB, so looking at the low signal on channel one with the 10V source on channel 4 is bound to give you problems.  For the rest, it looks to me like you are getting 60dB+ dynamic range out of a 10-bit device, just how good do you expect it to be?

It says a lot that the expectations for FFT functionality in an oscilloscope are so high (and perhaps unrealistic).  If frequency domain information is that critical to someone, an oscilloscope is the wrong tool.  But it is a nice bonus that DSO's have come so far and are now including usable (if not perfect) FFT functionality.   There are some bugs here and I'm sure R&S will continue to fine tune this scopes FFT software, but overall I'm pretty impressed with it.

Me too....

[hwj-d]

I know, you have never seen this on a modern scope ...
 

[mikeselectricstuff]

I know, you have never seen this on a modern scope ...

https://twitter.com/mikelectricstuf/status/846419601529430019?lang=en-gb

[hwj-d]

I know, you have never seen this on a modern scope ...

https://twitter.com/mikelectricstuf/status/846419601529430019?lang=en-gb

Thats a real bug. In xy-mode switch from sample to average, and see what happens.
edit: this scope is very useful... 

[MikeP]

These are some examples of FFT function. I think for precision measurements this device is unsuitable.
 The signal source is G3-118. The first harmonic was suppressed by rejector (- 100db). The second harmonic - 106 dB. For me it is physical vacuum.

The channel-to-channel  isolation spec is only 50dB, so looking at the low signal on channel one with the 10V source on channel 4 is bound to give you problems.  For the rest, it looks to me like you are getting 60dB+ dynamic range out of a 10-bit device, just how good do you expect it to be?

 Thanks! Good idea. To my surprise, I found a signal from the built-in generator - a spurious signal is present even when the generator is disconnected from the inputs! Otherwise, the isolation of the channels is absolutely perfect. I hope that it is possible to get a dynamic range 90+ dB.

 I wanted to show INTERNAL distortions. And I want to know - what is the nature of this phenomenon. In other matters, the RTB surpasses my requirements. Every contact with the device is fun. 

 I think this is HW problem. If in the future the defect is eliminated, then I will buy another RTB.
 

[Fgrir]

I hope that it is possible to get a dynamic range 90+ dB.

I think that may be an unrealistic expectation for standard oscilloscopes.  I would suggest that you might really want to look for a dynamic signal analyzer or a soundcard-based solution, although this picoscope looks like it might be up to your needs:
https://www.picotech.com/oscilloscope/4262/picoscope-4262-specifications

[LaurentR]

Note: New user manual v04
https://cdn.rohde-schwarz.com/pws/dl_downloads/dl_common_library/dl_manuals/gb_1/r/rtb_1/RTB_UserManual_en_04.pdf

347>509 pages (!)

Haven't looked in details, but there is quite a bit of reorganization, some more general setup, some beefed-up sections and new sections on mask testing, FFT, voltmeter and counter apps, logic analyzer and the function/pattern generators.

I don't see anything obviously missing anymore 

[vokars]

"These are some examples of FFT function. I think for precision measurements this device is unsuitable."

Hm .. precision here means dynamic. The limit for that are the usual noise sources of the scope especially quantization noise. Therefore an 8 bit scope has a max. dynamic of 48dB, a 10 bit scope like the RTB2000 60dB. How much less the real dynamic is below these max. values depends on the actual ENOB. Typically this is 6.5 bit for an 8bit scope having therefore about 40dB dynamic limit only due to quantization noise.

That means that the RTB2000 actually has the best dynamics than any other DSO in that price class (how much depends on the ENOB that I don't know) because it is the only 10bit scope in this price class in the market (and latest Owon models with up to 14bit). We had to wait for more than two decades that a vendor breaks this 8bit curse. It could be compared to the invention of the fastest car and then making it look bad because it cannot fly. The aeroplanes in this example are the spectrum analyzers that have much better dynamics.

But also with a DSO you can get better dynamics:

1. At a few kHz you can use the decimation/boxcar mode, that increases vertical resolution / reduces quantization noise and therefore increases your dynamic. I have no RTB2000 but I expect at least from theory that you can use 16bit vertical resolution mode at these low frequencies reducing quantization noise (theoretical upper limit not regarding unknown ENOB) to 96dB. In general boxcar reduces samplerate by a factor 4 per 1 bit less quantization noise.

2. Additionally you can reduce noise / increase dynamics by averaging several consecutive measurements, which should be possible when using a function generator as signal source. This reduces noise by sqrt(count). E.g. 100 counts give you 20dB lower noise.

Because I don't know the actual ENOB it would also be interesting for me, how far away the RTB2000 models are from these theoretical upper limits.

[MikeP]

 Hello Rich and Marcus.
 
 This is another small bug. When the Russian language is active, then the keyboard loses the MIN and MAX parameters.

 Thanks.
 Mike.

[Hydron]

Some more bugs to report (none are show stoppers):

- Probe de-skew doesn't apply to trigger point, only waveform. If triggered off a de-skewed channel then the trigger appears in the wrong place.
- Cursors move the wrong one (Y) in X cursor mode at first
- Cursors menu sometimes shows wrong channel selected
- FFT max hold corruption on boot (very minor, just looks bad when booting and FFT max hold is enabled)
- Ref waveform doesn't seem to line up exactly with trace it is copied from (vertically with spectrum, horizontally with time domain view)
- When going into quickmeas and changing acquisition mode for one channel, doesn't seem to always apply to other channels when exiting quickmeas
- Pattern gen output can get stuck high when turning off
- In history mode, I disabled channels 1, 3 and 4 to get a better view of channel 2 in a past acquisition, but when re-enabling the channels to look at other captures only channel 1 appeared again correctly, 3 and 4 were flat (see "history bug 2.png")

This list is just my notes I made at the time of finding each issue; Rich/Markus - please PM me if clarification is required for anything.

When the next firmware is released a comprehensive list of fixed and remaining known issues would be very useful to let users know what they should report and what is already known (first firmware release had a list but it was much shorter list than expected).

[ws2812b]

 ;DI guess that's the best documented bug report they ever got 

Hello Rich and Marcus.
 
 This is another small bug. When the Russian language is active, then the keyboard loses the MIN and MAX parameters.

 Thanks.
 Mike.

[vokars]

Did I miss Dave's extensive review concerning the RTB2000? I remember that such an extensive review needs be carefully prepared. Now six month after the release of this scope milestone (not yet another boring 8 bit ...)  this means that we will see the most extensive scope review that was ever made? 

[MikeP]

 Dear vokars.

 Thanks for your thoughts and recommendations. Actually averaging (and other focuses) can reduce noise level very significantly. The main problem is internal distortions of the oscillograph. Also this distortion has changeable structure. I do not understand the nature of this phenomenon.
 The screenshots given above are spectrograms of very clean signal which contains -110dB of second harmonic. And what you see?
 I want to believe - this effect can be eliminated. Let’s make RTB Special Edition!

[vokars]

"The main problem is internal distortions of the oscillograph."

If you need very good THD, I recommend audio hardware. Even my 10 years old RME Fireface 400 has a THD<-110dB with max. sample rate of 192kHz and 24bit resolution, 8 channels.

[vokars]

" Also this distortion has changeable structure. "

1. The quantization of a signal is a distortion by design. The quantization error can be regarded as a sample-dependent change of the amplitude sometimes called granular distortion, because it is larger if the ADC has less bit.

2. This means that a) granular distortion changes over time, because the quantization error occurs as quantization noise and b) is in the order of the quantization noise.

Both a) and b) seems to be in line with your observation.

3. If my assumption is true, could be verified by using the 16bit HighRes (some screenshots have the HighRes label, how many bit?) mode or averaging over time. Of course there are serveral other non-linearities in the input circuits that cause additional distortions but probabely do not cause this time changing pattern as you have observed. Did you give it a try and did it help to reduce the distortions (I cannot do it by myself, because I have no RTB2000)?

Addendum: In http://www.analog.com/media/en/training-seminars/tutorials/MT-085.pdf  chapter "SPURIOUS FREE DYNAMIC RANGE CONSIDERATIONS IN DDS SYSTEMS" similiar effects are discussed for DDS oscillators. It is explained, that the quantization error correlation can be reduced by dithering and by a change of the samplerate/signalfrequency ratio.

[MikeP]

Some more bugs to report (none are show stoppers):

  Mr. Hydron makes a great job. Thanks. This is additional info about the history bug. After the off/on switching we can get surprising waveform in the some segments. The 3 or 4 channels will intact if it trigger's source.

[MikeP]

If you need very good THD, I recommend audio hardware. Even my 10 years old RME Fireface 400 has a THD<-110dB with max. sample rate of 192kHz and 24bit resolution, 8 channels.

 Thanks vokars.

 In general, ALL combinations have been tried - it's impossible to go below about -70dB. The HIREZ mode does not indicate the number of bits and has no significant advantages.

 For measurements I use an external notch. Very cheap and unambiguous method.

 I well understand some groundlessness of my expectations. I do not like a lot of equipment. And the RTB is very-very close to my understanding of the ideal device.

[Hydrawerk]

I know, you have never seen this on a modern scope ...
 

Hello.

[hwj-d]

I know, you have never seen this on a modern scope ...
 

Hello.

Yeah, hello.
The reason behind is, that the RTB makes this by itself. There is no other signal coming in but doing this external. And this shows another real bug from this scope:

Thats a real bug. In xy-mode switch from sample to average, and see what happens.

[hwj-d]

Another bug in cursor settings:
horizontal and vertical setting reversed to actual measurement

[Hydrawerk]

Yeah, hello.
The reason behind is, that the RTB makes this by itself. There is no other signal coming in but doing this external. And this shows another real bug from this scope:

Thats a real bug. In xy-mode switch from sample to average, and see what happens.

Oh, I did not know that.

[Rich@RohdeScopesUSA]

Thanks everyone for the inputs - I’ll add them to our tracker.

-Rich