EEVblog #470 - Agilent N9344C 20GHz Spectrum Analyser Teardown

[nitro2k01]

[Laidukas]

Dave, were you wearing a wrist strap for ESD protection during this tear down? I think Agilent guys hope you did 

[babi]

Hi Dave

Did you get any high-res photos of the RF micro-strips on the PCB?

[EEVblog]

Did you get any high-res photos of the RF micro-strips on the PCB?

Some photos on my Flickr site, but just 2MP taken with the video camera.

[JoeMuc2013]

Hello Dave,

what a beauty once again. Just one thing I didn't like is the odd placement of the L2 inductor on the input board at 39:53 in the video (left side) and also shown in some of the closeups at Flickr. I guess there is no "RF" reason for doing this? Just a small flaw I'm not used to see from Agilent. Well, it appears to be working though, right?
Did you manage to put it back together?

Greetz
Joe

[icon]

"Looks like a phillips, smells like a phillips" - but isn't. Is therefore, possibly, a Japanese Industrial Standard screw? Found in laptops and other consumer electronics, on the adjustments of Shimano derailleur mechs, on RC helis, on the lids of Nissin motorcycle brake fluid reservoirs, etc. I have a set of Hozan JIS screwdrivers that get a lot more use than you'd think. A phillips screwdriver will rapidly knacker a JIS head, especially as they seem often to be made of some kind of silvery cheese.

John

[SeanB]

JIS is easy to identify, the head has a small depression on the one side to indicate that it is JIS.

[KJDS]

Circuit design by Slartibartfast. Just as RF should be.

[saturation]

GHz RF, always a blast to see layouts, so black magic. 

[free_electron]

Circuit design by Slartibartfast. Just as RF should be.

Could very well be. He always loved the fiddly bits like the fjords of norway...

The reason for having gps onboard this machine is that it most likely is a site analyser. These are used to measure field strength for things like wifi and cell phones. Hook it up to a pc with mapping software and drive around. And it builds you the power map.

[EEVblog]

The reason for having gps onboard this machine is that it most likely is a site analyser. These are used to measure field strength for things like wifi and cell phones. Hook it up to a pc with mapping software and drive around. And it builds you the power map.

Yes, its inclusion was driven from customer requirements for that.

[icon]

JIS is easy to identify, if the head has a small depression on the one side to indicate that it is JIS.

Edited.

I've often encountered 'phillips' (ie not Pozidriv type) screws with no markings, but which no size of phillips screwdriver will properly fit. JIS usually do the job (if I haven't rounded the damn thing in the process of discovery - then it's Dremel-time).

John

[mikeselectricstuff]

The reason for having gps onboard this machine is that it most likely is a site analyser. These are used to measure field strength for things like wifi and cell phones. Hook it up to a pc with mapping software and drive around. And it builds you the power map.

Yes, its inclusion was driven from customer requirements for that.

Can it use it for a frequency reference I wonder?

This reminds me of a story I heard a while ago - in the early days of analogue cellphones, Vodafone had some vans which used to drive around to evaluate call-site performance - because of the test gear they were using, they had to drive round fairly slowly, and on a couple of occsions got pulled over by cops thinking they were kerb-crawling - apparently they were persuaded otherwise by all the gear in the back!

[lewis]

Voodoo black magic shit.

[SeanB]

Dremel not always going to work, but you can get left hand cutting drill bits, or my favourite item, a Ridgid extractor kit. The left hand drill bits both drill and loosen the screw, great in a deep recess. Just need a reversible drill, and tighten the chuck pretty tight.

[senso]

There are some places riddled with what I would call sloppy soldering, and those botches, what a crap job..
And there is lots of flux residues, I was watching and screaming, like how can you call that top notch quality?

[BravoV]

There are some places riddled with what I would call sloppy soldering, and those botches, what a crap job..
And there is lots of flux residues, I was watching and screaming, like how can you call that top notch quality?

Probably that is a working unit but with the calibration way out of whack big times beyond repair, and also might be as a target practice for noob repair/cal staffs to play with ?

[senso]

Some examples, there is more top notch quality showing all over those boards, it seems really smart having expensive controlled dieletric pcb's and then just apply some dabs of random flux all over the impedance controlled traces, must be the secret sauce 

[SeanB]

Those little white chips are Mini circuits amplifiers, some are capable of a flat response from close to DC to close to light. Give away is the power feed, a feedthrough ceramic cap, a chip inductor, a decoupling cap to ground then a feed resistor to the output of the chip. The numbers are an indication of the maximum signal level they can linearly handle. On the input PCB the funny thing with 2 lines next to each other and a resistor on the far end is a sampling loop, probably going to a simple rectifier and a signal strength indicator, though it probably has a connection to the gain stages so as to give a constant level to the ADC stages as it is scanned through the passband.

[Corporate666]

Always interesting to see PCB design at a level light years beyond what I do (or ever will do).

Obviously these guys get special training to be able to do that - they aren't just first year EE grads doing those layouts.  Is high-end RF design something that is taught at the masters level, or (as I would guess) based more on in-house knowledge that is passed from generation to generation within a company like Agilent?

Anyone else notice the Chinglish on the stickers?  Left Borad and Right Borad? 

[Watermelon]

Does anyone has a datasheet available for those odd 4 leaded RF amplifiers? Cannot find any at first glance 

[Hydrawerk]

Did you reassemble this insturment? Does it work again?

[KJDS]

Always interesting to see PCB design at a level light years beyond what I do (or ever will do).

Obviously these guys get special training to be able to do that - they aren't just first year EE grads doing those layouts.  Is high-end RF design something that is taught at the masters level, or (as I would guess) based more on in-house knowledge that is passed from generation to generation within a company like Agilent?

Decent RF design is mostly learnt on the job. My bachelors and masters degrees both had RF projects which provided some basic understanding, but I didn't really know what I was doing until I'd spent ten years or so learning on the job.

I remember one of my undergraduate courses had 20 hours of lectures on the Smith chart and I barely understood how to do something basic with it, whereas now it's second nature. That only comes with spending a lot of time on it and that's just a tiny part of RF. A lot of RF engineers are specialists, so will only play seriously with say LNAs or high power or filters or mixers rather than being bale to readily pick up things from a broad spectrum of the building blocks used in RF. There are certainly very few that are genuinely good at say, both high power RF amplifier design and cavity filter design, certainly if they had to do something from first principal rather than relying on synthesis software.

[krenzo]

Does anyone has a datasheet available for those odd 4 leaded RF amplifiers? Cannot find any at first glance 

Look for "4-Micro-X" as the Package / Case.  There are a few on DigiKey that do 0-12 GHz amplification from RFMD: http://www.digikey.com/product-detail/en/NBB-310-T1/689-1001-1-ND/1302337

This one has "N4" on the package like in the picture above.

[robin_]

I’m wondering what these bodge wires are for. What kind of mistakes in the board layout remain unnoticed up until the board is already in production? And in which kind of situations is it even possible to still being able fix it with hacking in small wires afterwards?   

[mikeselectricstuff]

I’m wondering what these bodge wires are for. What kind of mistakes in the board layout remain unnoticed up until the board is already in production? And in which kind of situations is it even possible to still being able fix it with hacking in small wires afterwards?

It isn't always mistakes - for example it can be to adapt to use different parts if the original isn't available.

[EEVblog]

Did you reassemble this insturment? Does it work again?

Yes, seems to work fine.

[Dongulus]

Top-notch session of PCB and systems engineering porn.

I was confused by Dave's comment about how the 10 Mhz reference oscillator would not work after being turned over. I've never worked TCXOs before so I had no idea that they were sensitive to orientation. How can a robustly designed instrument intended for potentially rough field work be expected to operate containing such a component?

[c4757p]

I suspect it was a joke, similar to his usual "the electrons will fall out!" 

[EEVblog]

I suspect it was a joke, similar to his usual "the electrons will fall out!" 

 

[edpalmer42]

I suspect it was a joke, similar to his usual "the electrons will fall out!" 

Not to mention the 'right tongue angle'.

Wait, I mentioned it, didn't I?   

Ed

[SeanB]

I suspect the bodge wires connecting ground planes had more to do with a faulty via only detected during final test as being either open or intermittent, probably going open during the IR reflow of the board. Would have been better to have used a thicker wire, but hopefully the voltage regulator is happy with the added ground inductance.

[ftransform]

unsolder that gold shielding you coward, I would be proud of you then, like I am proud of war journalists.
put it on a hot plate... agilent said to take it apart......... common I know you want to....

[SeanB]

The copper tape over the hole is a vent for reflow soldering , so the case does not collapse from vacuum after soldering, or blow out the solder during reflow. The copper adhesive is most likely a conductive adhesive.

[bitwelder]

What's the purpose of the arrow on top of the gold shielding? Reflow direction?

[Watermelon]

Does anyone has a datasheet available for those odd 4 leaded RF amplifiers? Cannot find any at first glance 

Look for "4-Micro-X" as the Package / Case.  There are a few on DigiKey that do 0-12 GHz amplification from RFMD: http://www.digikey.com/product-detail/en/NBB-310-T1/689-1001-1-ND/1302337

This one has "N4" on the package like in the picture above.

Interesting read, thanks! Now i understand why they don't need an external supply, it's just 2 transistors and some resistors...

[pomonabill221]

I suspect it was a joke, similar to his usual "the electrons will fall out!" 

Or the electrons will all roll to one side of the can when the board is rotated...

[lewis]

What's the purpose of the arrow on top of the gold shielding? Reflow direction?

So you know which way to orientate the units so the electrons don't fa..... bugger it.

[SeanB]

Interesting read, thanks! Now i understand why they don't need an external supply, it's just 2 transistors and some resistors...

Some have a lot more than the 2 transistors inside, you can make a pretty complex amplifier if it needs to be level from near Dc to light, and has only 2 pins available. Those square ones must be a lot more, even if they have the same power arrangement.

[KJDS]

Having once worked at a test gear manufacturer I can give some insight into why there'll be a few bodges on the boards.

The early demonstrator models are part of the initial pre-production run. These are built and tested using the normal production facilities, but with design support. Some of these will then be used for the qualification testing, full temperature tests, both operational and storage, vibration, EMC and a whole load more detailed testing that would be carried out in production. Sometimes these tests bring up issues that require minor mods to a board and this will often be replicated across most of the batch. It's not unusual for the first couple of units to have components changed by designers who are not expert solderers just to prove that the mod works as intended. Once everyone is happy then some of these units become demo models, others will be used for accelerated and highly accelerated life tests and production will be allowed to ramp up, sometimes carrying on with minor board mods and sometimes waiting for a board respin and any associated requalification.

[ftransform]

I figured out how to remove the can nicely. Wrap it with resistance wire and heat it. Probably more safe for the board then a hot plate.

Aint anyone else dying to know what the fuck is under those cans?

dave should take some risks... he has it made.....

[larry42]

Under those cans? Local oscillators and the front end attenuator.

[ejeffrey]

Some examples, there is more top notch quality showing all over those boards, it seems really smart having expensive controlled dieletric pcb's and then just apply some dabs of random flux all over the impedance controlled traces, must be the secret sauce 

Flux on the PCB has essentially zero effect on the impedance.  For a microstrip geometry the capacitance is almost entirely due to fields inside the dielectric not above it -- which is the whole point.  A small fraction of the capacitance does come from the field above the board, but those field lines are relatively long, so a thin layer of dielectric crap won't really affect it.

[jancumps]

so a thin layer of dielectric crap won't really affect it.

If it would affect it, it wouldn't be there.

[kfitch42]

Did you reassemble this insturment? Does it work again?

Yes, seems to work fine.

But, how many parts were left over?   IIRC you said there were 2 million parts ... so if you had two 'extra' screws at the end that would be 1 PPM, right? That sounds pretty good when you put it that way

[thewyliestcoyote]

I love the tear down! Really shows all the RF design magic that is hard to come by.

I have a couple of 4 port +50GHz PNA's, older VNA's, and about 2 dozen field foxes at work and would look to see what they are in the insides and not get fired for destroying a instrument that costs more then I make in a couple of years.

Dave maybe you could do a teardown of a PNA, field fox, or the PXI unit for a 26.5 GHz VNA?

[rfbroadband]

at 20:25 its is a 1:1 125 MHz RF balun not a relay...

very nice tear town.