EEVblog #587 - Tektronix MDO3000 Mixed Domain Oscilloscope Teardown

[Omicron]

Hi,

I have a basic question. It was mentioned in the video that this scope does ~200,000 waveform updates per second. But the LCD display refresh rate is likely to be far less. No more than 60-100 hz. How does the synchronisation work? Or am I missing something here?

Thanks,

How it usually works is that multiple acquisitions are combined into one image where the intensity of each pixel is changed according to how often that pixel was "on" in the different acquisitions, so it gives you an "analogue scope" kind of feel.

Alternatively the image intensity of a pixel can also be determined as the average of a number of samples in a larger single acquisition (e.g. this scope can sample 10M points, much more points than there are pixels on the screen horizontally).

In practise I believe it's a combination of these two.

[bronson]

For all those who guessed what that "touch sensor" on the PCB was, it's nothing of the sort. The PCB designer got bored and wanted to put a symbol of a sine wave on an old school round scope screen. The vias connect to the internal ground plane.
FAIL IMO

That's a pulse train if I've ever seen one.  Definitely NOT a sine wave.  Also, that round scope's vertical gain is cranked rather high.

Probably a good thing that guy designed the board and not me.  If I'd got that idea in my head, I might try to turn this into a bitmap:

[Hydrawerk]

TEK has clearly lost the lead to Agilent in the middle to upper range scope sectors  (do you hear me Danaher Marketing Morons?)

But Agilent is unable to produce a scope with true 1mV/div vertical resolution...

[EEVblog]

How did you find this out? Tek surely didn't release a document stating that "PCB Engineer gets bored designing latest product, adds vintage touch"

I was told so by a Tek employee who would know these things.

[EEVblog]

No doubt about it. Tek is playing catchup.  I don't think Tek will ever be on top again, even with integrated mixed domain stuff.

I don't know about that. The new MDO3000 looks to be very competitive.
And like I said, if they threw in the 3GHz RF option for free, they'd sell like hot cakes, so they have that up their sleeve if they find sales are lacking, just like Agilent are now throwing in the kitchen sink.

[w2aew]

No doubt about it. Tek is playing catchup.  I don't think Tek will ever be on top again, even with integrated mixed domain stuff.

I don't know about that. The new MDO3000 looks to be very competitive.
And like I said, if they threw in the 3GHz RF option for free, they'd sell like hot cakes, so they have that up their sleeve if they find sales are lacking, just like Agilent are now throwing in the kitchen sink.

...and I really like what I see coming down the pike on the ultra-high end of things - a new proprietary sampling technique to give 70+GHz analog BW, 200GS/s (called Asynchronous Time Interleaving), with lower noise than the microwave block down-conversion + reassemble in DSP approach used by Agilent's RealEdge and LeCroy's DBI technologies...

[mikeselectricstuff]

TEK has clearly lost the lead to Agilent in the middle to upper range scope sectors  (do you hear me Danaher Marketing Morons?)

But Agilent is unable to produce a scope with true 1mV/div vertical resolution...

Agilent can produce whatever they think the market wants. The fact that 1mv/div is not common suggests few people ask for it.

[EEVblog]

Agilent can produce whatever they think the market wants. The fact that 1mv/div is not common suggests few people ask for it.

Also, in the case of the Agilent, it's actually only a 2mV/div front end. The 1mV is just software fiddling.
Maybe they concentrated on better performance at the high end?
So if you do need a low level front end, the Agilent is certainly not the best choice. Rigol probably have this market sown up (+ Hameg too I think?).

[Hydrawerk]

BTW according to the MDO3000 manual, there is no demodulation in the spectrum analyser feature. :-(

[echen1024]

Agilent can produce whatever they think the market wants. The fact that 1mv/div is not common suggests few people ask for it.

Also, in the case of the Agilent, it's actually only a 2mV/div front end. The 1mV is just software fiddling.
Maybe they concentrated on better performance at the high end?
So if you do need a low level front end, the Agilent is certainly not the best choice. Rigol probably have this market sown up (+ Hameg too I think?).

Hameg go to 1mV/div.

[Rigby]

Hameg go to 1mV.

how's 10uV grab ya?

https://www.eevblog.com/forum/testgear/looking-for-goodinexpensive-bench-power-supply/msg169147/#msg169147

[echen1024]

Hameg go to 1mV/div.

how's 10?V/div grab ya?

https://www.eevblog.com/forum/testgear/looking-for-goodinexpensive-bench-power-supply/msg169147/#msg169147

*drools over keyboard* I just love the Tek 24xx analogs and the 7000 series.

[Steffen]

I'm curious if the MDO3000 run some kind of embeddet Windows or if they run some other software. I think TDS3000 series run on VxWorks with Tek UI. MSO/MDO5000B scopes run Windows 7 Ultimate 64Bits.
What the hell is wrong with the manufacturers shipping them with Microsoft Windows? If you plug it into the network, you need antivirus and Windows Updates. If you do so, you may lose warranty.
Do the companies really believe, that customers replace expensive devices when some software company decides to drop support for a ceratain product?

[Rigby]

They use Windows because it is FAR easier to build up a UI on Windows than it is to come up with an operating system or an embedded platform on those scopes.

Windows isn't used for acquisition, or math or anything else, just UI.  ASICs do all the work.

[Lukas]

The MDO4000 use some sort of embedded Linux (https://www.eevblog.com/forum/reviews/tek-dpomsomdo-easter-egg/), i guess the MDO3000 as well.

[Steffen]

Windows isn't used for acquisition, or math or anything else, just UI.  ASICs do all the work.

Sure, but in this case you just buy an totally overpriced regular standard performance computer with measurement attachment. I know analyzers with Win NT 4.0 and Win 2000. Price somewhere in the 5 to 6 digit region. Don't even think of installing Win XP or Win7. Insecure as hell, even XP based hardware if you plugin to network. At least you can use a seperate network, but as measurement slave in the field? I'm not amused.

[dr.diesel]

Insecure as hell

I have run into this problem on several occasions.  Some companies absolutely refuse network access to Windows machines that don't run their corporate antivirus, with no exceptions.  Most are oblivious to the fact that test equipment runs an OS, but it's still against the rules and could get you banned as a vendor or contract worker if they found out.

[hikariuk]

Insecure as hell

I have run into this problem on several occasions.  Some companies absolutely refuse network access to Windows machines that don't run their corporate antivirus, with no exceptions.  Most are oblivious to the fact that test equipment runs an OS, but it's still against the rules and could get you banned as a vendor or contract worker if they found out.

Quite a lot of ECDIS and radar display units onboard ships are running Windows.

[dr.diesel]

Quite a lot of ECDIS and radar display units onboard ships are running Windows.

Yeah, some companies are ran by knowledgeable sane people, other Fortune 100 IT departments are managed by Satan himself.

[Rigby]

other Fortune 100 IT departments are managed by Satan himself.

By now, I'm sure Satan has delegated most of the work.  He needs only to look at a few of my former bosses to learn some things.

[hikariuk]

Quite a lot of ECDIS and radar display units onboard ships are running Windows.

Yeah, some companies are ran by knowledgeable sane people, other Fortune 100 IT departments are managed by Satan himself.

Mostly it comes down to allocation of resources and what is actually value adding (i.e. what the customer wants to pay for).  The customer doesn't want to pay for you to write an OS from scratch.  Any resources you expend in that direction are not contributing to the value of your product.  So you don't do it, you find an existing product that gives you what you need and you licence it. 

The decision is also partly coloured by the question "what platforms do we/our developers already know?".  A lot of people know the Win32 API.  It's also coloured by what was available at the time development started; some of these products might have their origins in the 80s and 90s.  If you were starting a product from scratch now you might well looking at using something based on Linux instead (I know we looked at switching once).

[tom66]

It's interesting that Rigol use their own custom OS (probably an RTOS of some form) and they work fine with that. Boot time isn't particularly fast, though. It seems they've written or used pre-existing USB and TCP/IP stacks to do most of the work. Such features would be baked into a Linux kernel, but why overcomplicate it?

[Monkeh]

Windows isn't used for acquisition, or math or anything else, just UI.  ASICs do all the work.

Sure, but in this case you just buy an totally overpriced regular standard performance computer with measurement attachment. I know analyzers with Win NT 4.0 and Win 2000. Price somewhere in the 5 to 6 digit region. Don't even think of installing Win XP or Win7. Insecure as hell, even XP based hardware if you plugin to network. At least you can use a seperate network, but as measurement slave in the field? I'm not amused.

You are aware that under the umbrella of 'Windows' come products with an entirely unrelated codebase to desktop systems, yes?

[vvanders]

Windows isn't used for acquisition, or math or anything else, just UI.  ASICs do all the work.

Sure, but in this case you just buy an totally overpriced regular standard performance computer with measurement attachment. I know analyzers with Win NT 4.0 and Win 2000. Price somewhere in the 5 to 6 digit region. Don't even think of installing Win XP or Win7. Insecure as hell, even XP based hardware if you plugin to network. At least you can use a seperate network, but as measurement slave in the field? I'm not amused.

You are aware that under the umbrella of 'Windows' come products with an entirely unrelated codebase to desktop systems, yes?

But that wouldn't play well into the mantra that windows is evil and could never be a part of a good product. 

[hikariuk]

Windows isn't used for acquisition, or math or anything else, just UI.  ASICs do all the work.

Sure, but in this case you just buy an totally overpriced regular standard performance computer with measurement attachment. I know analyzers with Win NT 4.0 and Win 2000. Price somewhere in the 5 to 6 digit region. Don't even think of installing Win XP or Win7. Insecure as hell, even XP based hardware if you plugin to network. At least you can use a seperate network, but as measurement slave in the field? I'm not amused.

You are aware that under the umbrella of 'Windows' come products with an entirely unrelated codebase to desktop systems, yes?

Yeah.  Windows CE is an entirely separate code base to Windows Embedded, which is based off the current mainline NT kernel products.  You also get access to a large chunk of the source code of portions of WinCE that may need to be customized for your particular hardware, iirc.

Although the fact that WinCE is now called Windows Embedded Compact doesn't really help make matters clearer.