EEVblog #876 - NI VirtualBench Review

[EEVblog]

Dave reviews the 350MHz National Instruments VirtualBench combined 4 channel oscilloscope, logic analyser, arbitrary waveform generator, 5.5 digit multimeter, digital I/O, and power supply.

[Hypernova]

Surprised that they didn't build the UI with Labview.

Considering the efforts the HW team has put in the SW guys really dropped the ball here. For the price and platform possibility (Sequenced or even low freq wave formed PSU, waterfall FFT etc) it's bad value for even universities, you would need the lab techs and teachers put in the effort to build additional software to really make use of it.

[gb243]

This is a big fail for the software but I believe these units have been out for some years. Now there is probably one man and a dog allocated for bug fixes and new features. OK unfair maybe 1/2 a man and two dogs. Either way potentially a brilliant bit of kit in 300 or 100MHZ form but as has been said before the devil is in the detail or rather the software. Great shame.

[Fungus]

Remind me again what that monster $500 FPGA is for...?

Is it possible for dozens of people to connect to it simultaneously in a classroom via WiFi?

[rs20]

Your claim that the power supply is "failing to deliver 22W" is a PEBKAC, I believe. When the power supply turns on, it inevitably has to ramp up the voltage (charging up the output capacitor and all, it's never going to have an infinite slew rate.) So it gets up to 0.052V, your electronic load calculates that 22W / 0.052 V = 423A, and it promptly tries to draw 423A (or similar) from the power supply. Then the power supply's CC limit kicks in, and only provides 1A as requested. What are you expecting to happen? For the PSU to provide 423A to charge up the output cap some more to push the voltage higher, or for the programmable load to not load down 22W as requested? I believe this will happen with virtually any power supply, especially if you connect it and then press the "enable output" button on the PSU. The DC load will win the race.

Constant power mode on an electronic load interacts poorly with CC supplies; the moment the PSU hits the CC part of its V/I curve, the system goes divergent (towards the PSU delivering its selected current and the DC load doing its best impersonation of a short circuit). I suggest using CC or CV instead.

[Towger]

Dave, does it have to go back?
It would make a nice prize for one of our younger members.

[Kleinstein]

The PSU might still work to full power, just the test at constant power is not good. The PSU could really have some more programming features.

For the scope there is an averaging mode, so this might also work as high resolution mode - not other way to get higher resolution with just an 8 Bit ADC. The curve with averaging looked really good - better than I would expect from 8 Bits.

For something like bode-plots I would not mind having a separate application, as this would need a different screen layout anyway.

For an educational use, I would guess the 100 MHz version is way more viable - it's quite an extra price tag for the higher bandwidth.

[isopropilick]

As dave said... it will be a gorgeous debugging tool in addition to LABview, i want to  see that in action (some day).

[hamdi.tn]

that's a rubbish lazy software compared to what the hardware is capable of. 

[rs20]

For the scope there is an averaging mode, so this might also work as high resolution mode - not other way to get higher resolution with just an 8 Bit ADC. The curve with averaging looked really good - better than I would expect from 8 Bits.

Averaging mode is lovely and all, but it's extremely limiting. You'll never see glitch waveforms, it relies on perfectly repeatable triggering, and the 9+ bits of resolution is meaningless unless your waveform is perfectly repeatable to within 0.2% FS vertical with respect to the aforementioned trigger jitter.

Proper high-resolution mode (boxcar averaging on a single capture) suffers from none of these limitations. It's a joke to consider waveform averaging to be an acceptable substitute for boxcar averaging.

[Kohanbash]

I got to use one of these a while back at an NI tour bus show. I think the idea of these all in one devices are nice. However my concern is that if this is how we train students they will not be come familiar with the standard benchtop equipment that is used in industry.

What do people think? Is that a concern, or will people just jump to standard benchtop equipment with no problems.

[Brumby]

Personally, considering the level of the equipment and the skills required, I think those who gain competence will find stepping out into the larger world will be rather straightforward.

They will have picked up fundamental understanding on each of the components as separate devices, so when they sit down at an employers bench with separate boxes, it won't be too hard to grasp at all.  Also, having been exposed to multi-function devices, they will be able to identify such features more easily.  eg. When they see "Output" on a scope, they will look for a built-in signal generator, rather than just sit there scratching their head.

As for coping with differences between specific functions and features of varying types, makes and models of equipment - such as scopes - that is a challenge they will need to master no matter what they have used in the past.

[Kleinstein]

...
Proper high-resolution mode (boxcar averaging on a single capture) suffers from none of these limitations. It's a joke to consider waveform averaging to be an acceptable substitute for boxcar averaging.

I would consider boxcar averaging an single capture a joke - averaging a single value is just that value, so not averaging at all. The point of boxcar averaging is to use several periods to reduce noise (not only from the ADC itself) and higher resolution. Waveform averaging is just the more modern form of boxcar averaging. Old style analog boxcar averaging, getting one point at a time is really slow (long measurement time) and thus largely replaced by waveform averaging and digital integration.

[rs20]

...
Proper high-resolution mode (boxcar averaging on a single capture) suffers from none of these limitations. It's a joke to consider waveform averaging to be an acceptable substitute for boxcar averaging.

I would consider boxcar averaging an single capture a joke - averaging a single value is just that value, so not averaging at all. The point of boxcar averaging is to use several periods to reduce noise (not only from the ADC itself) and higher resolution. Waveform averaging is just the more modern form of boxcar averaging. Old style analog boxcar averaging, getting one point at a time is really slow (long measurement time) and thus largely replaced by waveform averaging and digital integration.

We're clearly at crossed purposes when it comes to nomenclature. When I use the term "High resolution mode" or "Boxcar averaging", what I mean is that the ADC runs at a much higher sampling rate than is "required" for the screen. For example, if each pixel wide on the screen corresponds to 1 microsecond, the ADC might nevertheless be running at 1 Gigasample per second. That means there are one thousand samples for each microsecond. So those thousand samples are averaged together (thus giving much more than 8 bits worth of potential resolution), and that's how that pixel is determined.

This is what "High resolution" mode means on my DS2200 series scope, so I don't think I'm just making this up. Maybe "boxcar averaging" normally means something different, but Dave has certainly used it to mean the same thing.

So, once we're agreed that that's what high resolution mode means, then we can see that it's perfectly feasible for use on a single capture.

[Howardlong]

I am struggling to understand why any establishment, educational or otherwise, would purchase this rather than traditonal bench gear.

The only benefit as far as I can see is that it's a single integrated unit, but it's a very expensive one. Possibly LabView integration too? I don't know, but personally I take a great dislike to the bloatware from NI that gets installed on my PC just to run a few SCPI commands. I would imagine that hardware designers are not their target audience though, LabView seems more targeted to production and process control systems rather than electronics R&D.

I can imagine the marketing model is much like Matlab. Almost give it away to education, then the studes become dependent on it. While big corps might consider it, it'll be an almost impossible sell to an SME who'll see value elsewhere. Having said that, I love their MultiSim Touch tablet app for circuit simulation which is cheap as chips.

[Kleinstein]

Boxcar averaging is clearly using multiple periods. In the classical analog implementation only a few point per period like a sampling scope and averaging over multiple periods. So that is the averaging mode, and the only way to get higher resolution without an extra bandwidth limit.

Running the ADC full speed and than averaging adjacent points is a kind of oversampling and digital low-pass filtering. I don't know if the instrument uses it - might be without notice, especially if the 20 MHz BW filter is enabled. But only having full BW and 20 MHz somewhat makes me think not much digital filtering is used. Still the effect is somewhat limited, as this means lower bandwidth. The hardware should be definitely capable to do even full adjustable filtering - so other BW limits as well.

I would not expect the 350 MHz version to be really used much in education - it's more like the upgrade option for those who need more than the 2 channel version can offer. The 2 channel version is more like the one for the students.

[rs20]

Running the ADC full speed and than averaging adjacent points is a kind of oversampling and digital low-pass filtering. I don't know if the instrument uses it

This is what everyone means when they say "High resolution mode", this is what Dave means when he says "Boxcar averaging", and this is the feature that Dave (quite rightly) complained is missing from the NI VirtualBench. Because it is definitely missing; there's no button to turn it on anywhere, and the line is fuzzy-looking.

[EEVblog]

For the scope there is an averaging mode, so this might also work as high resolution mode - not other way to get higher resolution with just an 8 Bit ADC. The curve with averaging looked really good - better than I would expect from 8 Bits.

Averaging mode is lovely and all, but it's extremely limiting. You'll never see glitch waveforms, it relies on perfectly repeatable triggering, and the 9+ bits of resolution is meaningless unless your waveform is perfectly repeatable to within 0.2% FS vertical with respect to the aforementioned trigger jitter.
Proper high-resolution mode (boxcar averaging on a single capture) suffers from none of these limitations. It's a joke to consider waveform averaging to be an acceptable substitute for boxcar averaging.

Correct.
Waveform averaging is not "high resolution" mode, they are two quite different things. This is why almost all modern scopes have both a high resolution option, and traditional waveform averaging.
http://electronicdesign.com/test-measurement/why-you-should-care-about-oscilloscope-acquisition-modes

[bitwelder]

Googling for Virtualbench it seems that the 100 MHz version has been launched in 2014, and the 350 MHz version that Dave tested came to market at beginning of this year.
Any news whether the 100 MHz version during the last 2 years received any update in its software suite? Are the two versions using the same software package?

[Arjan Emm]

I suggest you call these episodes "first impression". You are very good at this and it shows clearly all the problems where you bumb into when you first unbox it and start playing with it, without any research upfront. But it's not really a review. You are pretty quick at finding out most possibilities without RTFM. But you miss stuff and it doesn't give a complete picture of what the machine is capable off. If you call it a first impression then it's fine, if you call it a review, i personally think you need to do more research before you start thinkering with a peace of equipment.
The video by shahriar 8 months ago of the 100 Mhz version is much more of a review to me.
Both are important. How to control a machine should to be intuitive, and you show how that works.
The positive sides somehow dwinddle in the lack of actual testing like shahriar does.

[Macbeth]

Perhaps this combined front end software is a bit sparse and all the real functionality is given with the LabView drivers and code samples?

A while back I got this TI op-amp dev board when it was on special offer. It comes with an online course using VirtualBench (though that is not required)

You can download the Bode Plot software to go with it from TI for free. The LabView source is available. I fancied hacking it to work with my RealBench  - but I always end up tearing my hair out when trying to "code" with LabView. 

[HighVoltage]

This review definitely showed me that I will never have this in my lab.
Also, I do not see the point for educational purpose.
Students first have to learn on real bench top instruments and may be later make a move to virtual benches.

Hmmm, may be they dropped the ball on making the software perfect after the hardware did not sell the expected quantities.
 

[jitter]

Perhaps this combined front end software is a bit sparse and all the real functionality is given with the LabView drivers and code samples?

That would be my guess too. E.g. one of the code samples is a bode analyzer, something that is lacking in the standard VirtualBench UI.

I wouldn't be surprised that the standard UI has been kept rather basic on purpose to make the step into automated control using LabView more necessary.
It also makes this instrument very usable for both the beginning student that will use the basic functions/UI and the advanced students can continue working with VirtualBench using LabView.

This way graduates will probably find it easier to start a job as a LabView programmer. And using the bigger racks like the PXI shouldn't really be a problem, because they already more or less know how the stuff works. It's win-win for both the students an the companies employing the graduates, and of course also for NI and LabView. There was some clever Happy Meal thinking behind this.

[TheCharels]

I quite like the concept of single measurement instrument that combines an oscilloscope, waveform generator, logic analyzer, multi-meter, power supply and digital I/O.

There are however some compelling down sides to having all of your test and measurement eggs in one basket. This being that a fault in any one measurement mode will strip your bench of all of them when the broken one is sent for repair. For this reason alone this kind of device is not suitable for a class learning basic electronic circuits. From my own experience at least one oscilloscope, waveform generator, multi-meter or power supply was damaged per month while taking a course on basic electronic circuits. More than one by me.

If the course is about using LabView and automation of circuit testing then the VirtualBench seems an ideal teaching device.

[rs20]

From my own experience at least one oscilloscope, waveform generator, multi-meter or power supply was damaged per month while taking a course on basic electronic circuits. More than one by me.

How on earth did you manage that? If I wanted to kill a piece of test gear, I'd have a hard time finding a way that isn't obviously contrived or egregiously physically violent. I'm genuinely curious. It's also relevant, because maybe you just had terribly delicate lab equipment that wasn't up to scratch (e.g., power supplies that can't tolerate a short circuit.)