Vložit
- čas přidán 6. 09. 2024
- In this episode Shahriar repairs a non-functional N4901B 13.5Gb/s BERT mainframe. This instrument is equipped with both the generator and analyzer modules which makes it fairly valuable. The instrument does not power on and does not react to the soft power button.
Teardown of the unit reveals a TD-Lambda Vega power supplies series which is a fully configurable switching power supply design. Even though the power supply in the instrument is completely custom, it is made from generic sub-blocks readily available on eBay. A replacement front-end is purchased to repair the power supply after detecting the problem.
The instrument also shows a minor problem with falsely detecting an overload condition on the delay control port which is by passed by installing an overwrite switch. The instrument performance is verified by generating and measuring various eye diagrams through a back-plane PCB board.
The Signal Path
www.TheSignalPath.com
/ thesignalpath
www.Patreon.com/TheSignalPath
Back in 2002/2003 I build a 43Gbps BERT when I worked at Ixia (now part of Keysight). I lead a team of about 10 people (mostly software people) and I did all the analog circuit design. It won some award for best new test equip of the year, and made the cover of EE Times magazine (Dec 2, 2002). But just as we were rolling it out in 2003, the US invaded Iraq. This lead to a recession, all the start-ups in the emerging 40G market went bust, and the project was cancelled. The last time I saw the prototype it was sitting on a shelf in the lab, gathering dust. Bummer.
@@aphenioxPDWtechnology It was NRZ. It used an SF5 interface (I hope I remember the right name) there were 17 lanes, at 2.5Gbps each: 16 data lanes, and one alignment lane. I designed plug-in modules for series (combine all the lanes into one) or parallel interfaces on both inputs and outputs. Part of the analog design was the ability to adjust the phases of all the lanes, which I did with PLLs. Each lane had its own PLL that multiplied-up the 125MHz reference clock, and I used a DAC and transconductance amplifier to feed a bias current into the loop filter of each PLL to adjust its phase. The SF5 mux and demux chips were GaAs, and made by some start-up whose name I can't remember now. But the demux had bias inputs that allowed me to adjust the voltage threshold and sample time, so I could move them around and see how the error rate changed; you need that to produce the contours of the eye diagram. A lot of fancy math there too: basically you decompose the jitter distribution into a series of best-fit normal distributions, kind of analogous to a Fourier series, then you can use this series to extrapolate down the skirts, into the really low bit-error contours in the middle of the eye.
I'm probably going on too long, but you might find this next part funny. We unveiled the BERT at a small trade show in San Jose (I think it was called Opticon) and our competitor, Spirent, unveiled theirs too at a booth a few feet from ours. A Spirent engineer came into our booth and the first thing he did was put his hand over the air exhaust vent to feel how hot it was. The air was barely warm, and so he proclaimed that our demo was fake. But I eventually convinced him it was real. I guess they had thermal issues, but I was too polite to feel their exhaust. And our unit was less than half the volume of theirs.
@@steverobbins4872 That is so cool, I long to being able to design instruments like that. And by the way I would have read a book written by you about stuff like that.
@Mai Mariarti Our target market was companies that made fiberoptic transceiver modules, and 40G network gear..
They are full on data converters with sample rate/bandwidth higher than the channel being tested with DSP tacked on
today you can download an app on your phone that does the same thing :D
that modular transformer in the PSU is beautiful. I'm surprised I've never seen one used in a bench power supply, with the flexibility in ratios and isolation between channels it seems like a perfect choice.
@MichaelKingsfordGray What about it gives it poor magnetic coupling?
I repaired the same power supply of the same equipment some years ago. Although it is a very complex and dense design, it was easy and fun to fix it.
That small PCB bar that connects all the secondary boards is the fault signal. In my case, one of the boards was asserting that signal during power-up, telling the primary side to shut down. I removed one secondary board at a time to find out which one had the problem. Then I found a couple of zener diodes in short-circuit on the damaged board. After replacing them, it started to work again.
Excellent video. You are definitely a genius the way you got that BERT to work!
Have to hate when a "Cat Scan" dosent reveal the problem. :-)
Completely irrelevant, but: You missed the +24V fuse
He missed 5 fuses, the 24V was one of them.
Nice bit of kit.
I'm working with a cPRI project right now, that Bert would come in handy to check on the clock recovery system... We are not really going over rate 8 right now so 10 Gbps would be fine.
Cheers,
If this were a design from just a few years later the front end would had probably been implemented in a couple of FPGAs.
I laughed quite a bit at cat scan. Well done, sir. ;)
Now all you need is the Agilent ERNIE
V cables for 13 Gbps signal? That's a bit of an overkill, but the customer gets what the customer specs (hopefully :) ). It is a nice and compact piece of equipment.
Around 5:43... I thought I was having audible alarms from my machines (servers and security system) and had to check logs... etc. for 15 mins or so. After another 15 mins to check what can trigger audible alarms (and finding nothing out of the ordinary!) I came back and resumed the video to find out the "alarms" were coming out of the video. The distance of the meter to the microphone and microphone itself makes EXACTLY the same sound with pitch, depth and distance from where I'm watching the video and where my machines are. That's 30 mins of my life I'll never get back :) Sucks to be a dumb ass, I guess.
At least you get to laugh about it afterwards, wild goose chases suck!
Had a similar condition with my PC -- it beeped a couple of times -- checked the logs no problems, while later it beeped again -- now its driving me nuts. Time for bed so I turnd the pc off, few minutes later I was walking past my office and it beeped again -- no way the dam thing is off -- turns out it was a pager in my desk drawer with a dying battery with the same exact sound.
Nice hack! Super interesting power supply - thanks for the great contend!
Wish I knew what reason we needed to look at stuff like this. (Obv I’m not an EEngineer. lol) but it looks super cool any way
I can imagine that in the software there is somewhere a calibration setting for the error theshold voltage, you just need to find it and reconfigure it to whatever is generated. In fact I would guess that something is really wrong in that area, because why would both modules do the same?
Because they're both connected to the same power rails. When it spiked(?) it likely fried both of them.
jfbeam But the cable from the front panel was only connected to one of them? Or do you mean overvoltage from the PSU?
Because the modules aren't the final step in checking the input threshold voltage. If the circuit after the modules got whacked, even two 100% working modules won't fix the problem. And as Shariar said, without documentation there's not really a chance to figure that out.
That transformer design is neat.
Cat is not interested in BERTs. Great.
Useful video 👍
Thanks for another great video!
Do you by any chance happen to have any more information / pictures on the cables attached to the lambda power supply?
I got one of these units with a broken power supply which I've repaired, however the person who tried to repair it last disconnected all cables from the power supply without marking any of them...
I was able to trace most of the cables on the top two rows of the modules, however the ones below are quite hard to make out from the video.
Just to have an idea, this used instrument costs over 20k euros !
I think he got the on from eBay that sold for about 130$, so quite a deal
👍👍
Great video friend!
Always gotta love switching power supply's (:
Alex.
Hi Thank you, great video -but can I ask you, "What happens if you ask it to test a really nice hot cup of earl grey tea" ?
No expense spared... As I glance at the floppy disk drive in the corner of the unit Gotta poke fun at that :)
@18:38 when you touch the Delay Ctrl In terminator, the status lights on the display screen start flashing, maybe its a bad connection.
nevermind i see later on @ 30:37 you changed out most of the terminators for different ones, and also did the identical module swap so it wouldnt be that.
He said catscan when the cat walks by
42:16 red regions are not BER of 1, they're BER of 0.5.
Compare and contrast: Chesterton's Fence and Chesterton's Fuse.
Way to go
25 amps at -5.2 volts...must have a lot of ECL circuitry in it.
Si-Ge bipolar can be current hungry too ...
@@uploadJ Yes but I don't think this uses SiGe. More likely GaAs or InP. And -5.2V is a very common supply voltage for ECL logic. It also has to be tightly regulated since such circuits have a very small switching threshold; this is why the -5.2 supply is adjustable.
You would be surprised what is being done with SiGe today; I don't know of any designs in telecom equipment that uses ECL today. At Cisco/Navini (10 yrs back w/WiMAX radio basestation) we used several SiGe chips. I used to work in TI GaAs Facility before TI sold it to TriQuint; I was DC test engineer so saw all the new designs the fab produced during design reviews. Comms related designs e.g. LNAs, WBDAs, cascaded Amps/Pwr Apps, switches, attenuators etc. nothing really power hungry. -5.2 VDC does sound like old ECL though. Was familiar at one time with Motorola (MECL) their 10K series. LVDS used in new SiGe fabbed chips, ft (transition freq) in the double-digit into triple-digit GHz range now for SiGe ...
www.google.com/search?rlz=1C1AOHY_enUS708US708&ei=Lg-sXMKlOJK0jgT5_KnABw&q=SiGe+GHz+fiber&oq=SiGe+GHz+fiber&gs_l=psy-ab.3..33i160.520266.540641..540787...3.0..0.336.2499.9j10j1j1......0....1..gws-wiz.....0..0i71j0i131j0i67j0i10j0j0i131i67j0i22i30j0i22i10i30j33i22i29i30j33i299.mokP-tfVeWY
"ECL in its true form is pretty much history. Single-ended ECL is almost extinct. High speed paths are typically implemented within deep sub-micron CMOS integrated circuits. The challenge that electronic desingners face today is high speed paths between chips. These paths are typically clock lines and data lines. Even though traditional ECL logic is no longer used, some of the advantages of ECL are so critical to high speed that designer just don't want to give up. "
From: www.quora.com/Is-Emitter-Coupled-Logic-in-use-today
@@uploadJ yeah but this wasn't made recently, its an old instrument
always green wire to turn on the PSU in these babies. Same like with every single PSU would be my guess. Just connect to GND.
Hi My Friend!
This Op. System is Windows Xp?
16:40 Cat
It's made in Germany and obviously it's gonna be good 😂
If you send me the executables I can try to get rid of the error for you
Serban, Best offer I've seen for his woes! Go Serban!
.
Hi I hope that you take this comment as a friendly comment.
You do make interesting videos that are entertaining, but I find your selection of gear, fare outside my budget and maybe also outside most other's budget. The unit in this video, is about 24,000$ used on ebay and that is not uncommon for the gear you talk about! Why do you go trough gear that nearly no one can buy? As I started to say, no critique but a friendly question:-)
I, for one, really appreciate the deep coverage of higher-end equipment than is normally accessible. We already have lots of videos on hobbyist use of what's easily affordable and accessible, but there's far fewer (dare I say only one) channels that discusses higher end test equipment in an approachable and understandable manner. Moreover, I especially appreciate the projects; I would never be able to reproduce them but I appreciate seeing them done and understanding more about how the domain works in a way that would be impossible without high quality test and measurement equipment.
@@benjaminchung991 Great comment, thanks, I do agree! :-)
Keep in mind that nearly all my repairs are done on instruments I paid
Seeing an instrument like this running Adobe Reader and IE makes me sad. IE is probably there by default. But Adobe Reader somebody put there. This isn’t a PC. It doesn’t need people to install crap on it.
Philip Hofstetter Adobe is there because of the instrument software. It loads it’s manuals in Adobe.