EEVblog #587 - Tektronix MDO3000 Mixed Domain Oscilloscope Teardown

[Chipguy]

The chip with the unknown logo at 39:29:

That is a:
Vishay Siliconix DG411L Precision Monolithic Quad SPST Low-Voltage CMOS Analog Switches
http://www.vishay.com/docs/71399/dg441l.pdf

... yes I am a nerd, I have chosen my nickname for a good reason 

[Lukas]

A 10gsps ADC is pretty near alien tech, probably with some wacky-arse exotic process - I doubt you'd put memory control type stuff on the same die, though I suppose it could be a multi-die package.

Check out this guy's website: http://poulton.net/papers.html He seems to be the ADC greybeard at Agilent doing ADCs for almost 30 Years now.

"A 20-GSample/s 8b ADC with a 1-MByte Memory in 0.18-um CMOS"
http://poulton.net/papers.public/2003isscc_18_1_pg_slides.pdf http://poulton.net/papers.public/2003isscc_18_1_pg.pdf Very interesting read.

[ElektroQuark]

That looks more like it's for clearance.

Watch out with this.   No matter how you bend them, the leads are always the same distance apart when they enter the package.

But the clearance between PCB pads is bigger if you bent them.

[EEVblog]

"A 20-GSample/s 8b ADC with a 1-MByte Memory in 0.18-um CMOS"
http://poulton.net/papers.public/2003isscc_18_1_pg_slides.pdf http://poulton.net/papers.public/2003isscc_18_1_pg.pdf Very interesting read.

Just the input buffer draws 1 watt!
It's got a lot of inputs to drive though I guess.

[bktemp]

I did not see any flash memory next to the main processor. Therefore the two devices connected to the FPGA must contain the firmware for the whole system.

[Icarus]

I'm no microwave expert but maybe that's the configuration ?

[RupertGo]

Is this it? Is all test equipment now signal->ADC->CPU->screen, with nothing left to do but a bit of front end and a lot of software?

Slightly sad, really.

(Loved the teardown, btw)

[German_EE]

Icarus, you beat me to it

The signal from the VCO is split into two, maybe a 90 degree difference or maybe 180 degree. These signals are then applied to the transformer and the four small black devices which are almost certainly diodes, together forming an RF mixer.

[cS2014]

Hi, thanks Dave for this tear down and the EEVblog in general it as reawakened my interest in electronics.

I find these tear downs of high end scopes baffling and compelling in equal measures. I find the differential pair layouts beautiful and the waveguide filters and processing magical..yes i have an artistic side lol. I am in the middle of the tear down at the moment and was just struck with a similarity to a product i came across a few years back when looking into shortwave radio receivers and the concept of software defined radio.

Microtelecom do one called SDR / Perseus and Software Radio Laboratory do the QS1R Receiver, which uses the LTC2208 16 bit ADC. I was struck by the analogy of the above receivers sampling the RF input from the aerial and the scope sampling the RF from the connector inputs.

Take it easy Dave,
Colin

[EEVblog]

Is this it? Is all test equipment now signal->ADC->CPU->screen, with nothing left to do but a bit of front end and a lot of software?

As far as spectrum analysers are concerned, no, not really. Only the very basic ones like the MDO3000 can get away with such apparent simplicity.

[mikeselectricstuff]

Maybe that 'touch sensor' on the PCB is some kind of cheap humidity sensor?

Not humidity but perhaps condensation?
Might be interesting to scope the terminals

[c4757p]

Is this it? Is all test equipment now signal->ADC->CPU->screen, with nothing left to do but a bit of front end and a lot of software?

There's definitely a move in that direction - but keep in mind, we'll always need test equipment to test the test equipment.

[Frost]

Not humidity but perhaps condensation?
Might be interesting to scope the terminals

The loops are open, but could it be a small PCB RFID antenna?

[Carrington]

I  noticed something relatively insignificant in the PSU. Notice how the TO-220 on the riser board has one leg bent forward. The only reason I can think of for doing that is mechanical support. They really, really don't want anything flapping in the breeze!

Agilent 4000X series uses the same PSU.

[Rasz]

A 10gsps ADC is pretty near alien tech, probably with some wacky-arse exotic process

"2x64-GS/s ADC core generates 128 billion analog-to-digital conversions per second, with a total power consumption of 2.1 W; the dual DAC consumes 1.7 W. The cores achieve 5.8 ENOB value of up to 10 GHz and spurious-free dynamic range greater than 43 dB"

http://www.lightwaveonline.com/articles/2014/03/semtech-announces-ultra-high-speed-adc-dac-cores.html

 

[free_electron]

take along hard look at the balun. the gold wire goes form RF-in to the VCO. ( top left to top right pin)
the black wire sits on the bottom.

So the Balun is the mixer. the bottom path coming form the vco is a delay line.

you mix incoming with vco in the top coil, and then mix it again with a delayed vco signal and pick it off

so they are mixing in the balun.

[Macbeth]

Gotta say, I'm not a fan of that airflow...  Those narrow heatsink fins will be clogged with gray gunk in two years.  Might be an issue since the video hints that the Tek is running near the edge of the heat envelope anyway.

The boards that the air flows across will get gunky too.  But, since they're all digital, that seems a minor aesthetic concern.

If it must blow in, would a filter have added so much to the cost...?

Sadly, I work in a terribly dusty environment (My House) - I have 4 dogs and the shedded fur and dust is nay on impossible to control. My PC's start overheating after a few months, and what do I find? GFX and CPU coolers chock full of dust and fur.

Thankfully I have an air compressor and blast the insides out every 6 months or so and get a whole new lease of life. I'm sure my bench equipment could survive the odd air blast too.

[Monkeh]

You might consider using filters..

[PA0PBZ]

so they are mixing in the balun.

You need something nonlinear to mix, I don't think the balun qualifies.


And by the way, who is this Alan (Wooky?) Dave is talking about, where's his youtube channel?

[cosmos]

I think the balun and the 180deg delay is there simply to reject the dither signal (its fixed frequency so cancellation should be good)
Probably so the VCO signal does not propagate the wrong way and reach the frontend LNA where it would cause (unwanted) mixing.

[SeanB]

W2AEW on Youtube. The real work is done digitally in the firmware after the ADC stage. The mixer could be those small black chips on the board after the mixer, looks like a set of PIN diodes there. About the right configuration for a mixer, and they are unmarked. and there is a quarter wave delay along with the transmission line splitter as well. Seems they removed input protection diodes as well, the pads are there, the labels are there and there is solder screened on the points but they are unpopulated.

[EEVblog]

I think the balun and the 180deg delay is there simply to reject the dither signal (its fixed frequency so cancellation should be good)

No, the dither signal is removed in the digital down conversion process.

[cosmos]

I think the balun and the 180deg delay is there simply to reject the dither signal (its fixed frequency so cancellation should be good)

No, the dither signal is removed in the digital down conversion process.

The dither is supposedly just added to the input signal to increase the dynamic range after processing.
If that addition is done directly to the output signal of the LNA in front of the VCO then I suspect it could have unwanted non linear effects and that this would cause unwanted mixing to happen in the LNA.
So I view the balun and the strip lines as a trap for VCO signals so they can not propagate backwards to the LNA in the input.
Alternatively it might be useful to not have VCO emissions leaking out of the coax connector on the frontpanel when the amplifier is bypassed.

Edit: never mind that theory about rejection, on second looks the delay line is just the balun for the signal from the VCO to make it differential (the ADC wants it like that)
Notice how the signal takes of on an inner layer up and to the left just after the balun just in the same place where it met the signal from the VCO.

[tom66]

What are the main disadvantages of doing this very simple RF frontend?
I imagine you're trading reduced RF costs for increased ADC costs -- but if your scope already needs 10GSa/s ADC, then you may as well?

[jancumps]

What are the main disadvantages of doing this very simple RF frontend?
I imagine you're trading reduced RF costs for increased ADC costs -- but if your scope already needs 10GSa/s ADC, then you may as well?

That you can't use scope and spectrum analyzer together as a true mixed domain tool?