Вставка
- Опубліковано 29 вер 2024
- In this episode Shahriar repairs a Noise-XT 2MHz - 7GHz Ultra-Low Phase-Noise Clock Synthesizer. The instrument’s history indicates that it has suffered a fall from a work table. After a lot of difficulty, the instrument control software was found so that the unit can be tested.
Initial tests reveal that a very small output is present, however nearly 40dB below the desired value. The unit teardown shows an architecture based on dual Analog Devices DDS 14-bits DACs in conjunction with a Z-COMM VCO and a series of dividers & frequency multipliers. A detailed analysis of the unit is presented.
RF and analog measurements indicate that the RF connector is not making adequate contact with the final trace of the RF output. This explains the low output power. This problem is resolved by replacing the SMA connector entirely. The device output power, frequency accuracy, stability as well as cross-correlation phase-noise measurements are presented.
www.TheSignalPath.com
/ thesignalpath
www.Patreon.com/TheSignalPath
Good evening sir! :^)
Mr Carlson's Lab Hi! Nice to see you here. :)
Collaboration? :)
Looks like a pre release board that fell not off the table but also into the pocket of the seller 😀
You talked many times about "frequency doublers".
How they function? What is the active component?
No product support means no business, I think that’s a pretty simple concept....
There is customer support but you just have to talk to the right people.
Wow. This amount of bodges makes me feel way less bad about some of my designs.
Were you doing crap designs like those ??!!
So, the center conductor inside of the connector was broken? Or was there a cold joint between the pad and the pin? Dropping the unit probably knocked the tin whisker loose that connected the pad to the center pin. :) This can't be a production board. The sales guys probably "borrowed" the first prototype for a demo and "somehow" it ended up being sold.
Nice design! So, if I understand it right, the low phase noise comes from the high quality VCO oscillator in combination with a very narrow PLL output filter. I really liked the concept of mixing 2 DDS signals as the reference of a narrow band PLL! 🤩😊
Yes, a low noise VCO in conjunctions with a wide PLL loop bandwidth, which requires a high frequency reference, hence all the paraphernalia prior to the PLL.
Great to hear you again Shahriar, I enjoy your voice, comments and your instructive lecture of the finest on UA-cam. Take care, thank you.
This has shown me how you really have to think different with RF equipment...with most gear I work on, a broken trace near or at the output means nothing at the output.
Many years ago when I was first getting started in electronics, I breadboarded a circuit with a headphone output.
I was really confused as to why my output was so quiet, so I started poking around and debugging. Eventually I found my headphone jack was wired one column away from the opamp output, coupling through parasitic capacitance of the breadboard.
Gave me a new appreciation for parasitics even at audio frequencies!
How can I detect gps car tracker ?
Interesting video as always, keep up the good work.
At first power on you mentioned that the device dumps a lot of heat into the heat sink. After taking it apart, do you know what is the main source of this heat?
The heat comes from the two DDS ICs.
"This design is six years old and the production issues should be fixed". yeah, probably after a few years they got around to fixing it *cries onto hot soldering iron*.
Just saying "HI" and thanks for another great piece of content... very interesting indeed. Commented since you said "see you in the comments section" and I watched this within seconds of you posting it!
Thank you. :)
Random thought; if you sold that wall of test equipment, in which city could you buy a 3 bedroom detached house? .. "Let's unplug it from the incredibly high-end spectrum analyzer and go ahead and connect it to the Rubidium standard..";-)
Holy crap just that Agilent E5052b alone is... $75k _used_ on ebay. Such gear pron from TSP, as ever...
Just a a beginner questions, how did you acquire all those expensive lab bench instruments? I recalled that you mentioned you mostly trying to buy used or broken ones and repair them. Did you get those used and broken units on eBay? Or there is some place one can visit for shopping? Thanks
Hi, Signal Path, Very interesting about the dual DDS design, I guess the output signal's frequency of the dual DDS is the same in order to get 3db better phase noise, and avoid the M*N spur caused by the mixer.
So could you please help measure the output signal of the dual DDS ( AD9912???)
1. What's the output frequency range of the DDS?
2. Does the frequency is the same of the dual DDS?
3. What's the fractional spur performance of this generator @ fractional frequency?
It is likely that the dual DDS is for phase noise improvement indeed (at the cost of doubling the power consumption). I may do another video on this.
@@Thesignalpath Thanks, Also very interesting how about the spur performance comes from the dual DDS chip, because it will be the reference of the PLL, will be 20logN amplified at the final output.
Question: How did they filtered PLL / Divisor etc... harmonics ?
Because at 10MHz it should have huge amount of harmonics related to frequency division.
Am I making a mistake ?
21:00 As I see below 10k it is worse than the spec. Eg @1k it should be -130 but it is -110. Somehow the 1/f^3 part is higher than it should be.
Do you have a guess why is that?
Thank you for pointing that out. I think it might be because of how I had setup the phase noise analyzer. I will go back and try again with higher correlation terms.
@@Thesignalpath According the DS of the 5052 you are fine measuring the spec without anyXCor.
i.imgur.com/5QoKTR4.png
Might be something related to ZCOMM
www.zcomm.com/pdfs/datasheets/V600ME14-LF.pdf -60@1k, -90@10k
As the ref is way better than the spec. It interesting to see that they only have 30dBc/Hz budget.
www.nelfc.com/pdf/1319a.pdf The ultimate PN spec: -162@1k, -175@10k
The connector might be this:
www.amphenolrf.com/901-10511-2.html
Quite "günsting" performance/price and simulation model "freely" available.
Interesting bit of gear, nice easy repair too, I suspect someone removed the OCXO from it due to the fan connector being ripped off.
I thought that too, but there's not much solder left behind. Some pins have changed color from the ENIG gold, so maybe that indicates there was one previously installed but they cleaned up well after removal, but some contacts are still clean gold. If its one of those general footprints that can take a range of OCXOs then that might explain the clean pads too.
Btw. those enclosures are made by a German company, called Fischer Elektronik.
www.fischerelektronik.de/web_fischer/de_DE/Geh%C3%A4use/M1.09/Designgeh%C3%A4use/$catalogue/fischerData/PG/ET1/search.xhtml
They look really nice and are reasonably priced.
When hardware is taken hostage by the software. Yes, I've been there. Are they selling bricks?
Anyway, it is very amateur for a finished product. They should have built just a couple of PCBs using a cheap service, find the kinks, redesign, repeat the cycle again, and if no kinks are found, then ramp up production. I'm no professional manufacturer, just an hobbyist, and I do that often. Many green wired PCBs went to the bin.
I don’t think those cheap prototyping PCB services would accept a complex RF board, or that anyone would trust them with it anyway.
@@TheXGamer969 It depends on the frequencies you need. For example, OSH Park is great for prototyping in RF. I've made successfully a working USB 3.0 test jig, with all the right controlled impedances. It is a matter of finding a service having the same PCB stackup and materials that will be used in the final product, if possible. Or you can use the prototyping service of production fab, as many fabs support that. For instance, Eurocircuits can be quite cheap.
Fab services accept any designs involving RF, as that is a matter that has only to do with the layout/design aspect. When submitting a design, one has only to make sure that it complies with or exceeds the minimum distances and trace widths, and that the vias can be fabricated. Fab services won't impose restrictions on any designs involving RF or high-speed signals.
As for other services, like JLC PCB, or PCB Way, I don't advocate them, not even for prototyping. However, I did used them because one of my clients required a very cheap and bare-bones service.
Dear Shahriar, thanks you for the video! I suspect both DDSes operate at the same frequency and mixer in such case works as doubler. This will result in 3dB lower DDS noise and two times higher comparison frequency in the output PLL (which should also lower PFD noise level at the synthesizer output). Can you verify my assumption? Do both DDSes operate at the same frequency?
@@Thesignalpath The signals will be correlated and will add by voltage resulting in 6dB increase, the DDS noise components will be uncorrelated and will add by power resulting in 3dB increase. So, the signal to noise ratio will be better by 3dB.
@@Thesignalpath I bet the internal PLL of the DDS in this synthesizer is not used. There is no sense to multiply 100MHz signal up to 800MHz and then divide it down (inside the DDS) to use internal DDS PLL (to match internal PLL PFD max. frequency requirements). Skipping internal PLL and using external low noise frequency multiplier is the classic solution for the low noise DDS clocking. As for the correlation of the signals generated by PLL - signals and noise inside the PLL bandwidth will be correlated and outside PLL bandwidth will be uncorrelated (it is a bit simplified view of the things). In any case the tones generated by the DDSes will be correlated and the noise generated by the different parts of the DDSes will be uncorrelated. Can you measure the DDSes output frequencies to check if they are equal?
@@UR3IQO Sorry, I think I messed up the comments. I was saying that you are correct. I originally misunderstood your comment. Indeed if they set the DDS to the same frequency, this will improve the phase noise by 3-dB. I can open the unit and try to measure to see what is going on. I am now very curious. Thanks for the comment.
@@Thesignalpath The frequency combination of the two DDS is a key feature of the synthesizer and can be the reason for creating a new video )
@@Thesignalpath " I can open the unit and try to measure to see what is going on." That would be great! I will be looking forward to see it. Thanks!
Good morning, sir thx for sharing all those info in your channel can I respectfully ask you to slow down a bit your explanations as it is way too fast for a non English native to understand everything at first watch :-D thx
Nice little synthesizer, I like that it is so compact. The fault was pretty obvious but it was a very interesting and in depth video non the less, as all your videos.
I came across a very nice S-Band synthesizer once, that was pretty nice spec vise and had all kind of modulation capabioities built in, however to control it you needed some software that was only provided pre installed on a notebook that was sold with the unit.
When I tried to get the software the manufacturer told me it "isn't in their interest to help anyone but the original customer" so the only thing left was to use the unit for parts, which was very sad.
The beauty of RF is that a broken connection doesn't mean no signal path. Very nice fix.
Great to see the architecture as well as you use the equipment most of us won't be able to access like the E5052B. Your Keithley sales rep is sending his kids through college thanks to that lower rack of DMMs, SMUs, and DAQs....
Is there any reason why you didn't use the Rb standard/counter for the 10MHz ref? You'd have the same basis for the counter and DUT to check the locking, but wouldn't that be best for the phase noise analysis too? You clearly didn't need it so maybe it was just convenience?
Yeah, it was just easier to use the Tek. I also wanted to show how amazingly accurate that instrument actually is!
@@Thesignalpath thanks for the reply, that Tek is indeed very impressive!
After your last video and spectral view I immediately was thinking , ok, another connector repair video :) :D:D
Yeah, I was hoping for a more complicated repair, but alas sometimes they end up being simple. But the reverse engineering was fun and we got to see the performance measured. :)
considering i had a board come in today that has a pin/package mismatch, i feel for the engineer that did that design lol. Luckily my board was probably a fraction the cost of this thing. Thanks for the vid!
No idea what's going on, but good job!
Funnily, that same model of the signal source analyzer, the Agilent E5052B, was used at JitterLabs to measure the jitter from my 10MHz clock reference (named GR10M-S).
The Tek generator is impressive indeed
Yes, really impressive for a synth.
Why don't you analyse the reason for distortion, it has a lot distortion
The instrument's THD is not very good. But it is not designed to have very good THD, just very good phase noise. For clocking applications, THD is not so important. Note how it says that is is a Clock source.
I love these videos :-D
Like I said finding you in an earlier video, I so thank you for allowing me to enjoy things I cant afford! Wonderful! I'm jealous, ha. I'll subscribe and thumbs up, its all I can afford.
I really enjoy your videos Dr . I was always fascinated by RF and Microwave engineering. At first it was completely black magic for me that's until I discover the signal path .
They should have some open source reference software, but what you really need is an API so that you can write your own Matlab or Python code for lab automation.
I knew from the start it was the connect I have work with this connector. The weight of cable or user abuse. Your replacement connector is fine for under 10 Ghz.
Looking forward to seeing an FSWP on that bench!
wow what a horrible design job, it hope they got fired
This is really interesting and informative! Great video!
Thanks for the video! Hope you guys are doing well
Great repair! Where's the kitty? No more visits?
So informative! Thanks
Hey new start with pooch. Nice
Dab dab dab.....and what does DAB mean anyone?...
When did you hear DAB?
@@Thesignalpath tbh I was just testing do u read comments or not yr vids are good
@@Thesignalpath lol
Electronics porn delivered right on time.
This is the weirdest breaking I ever saw in my entire profession life ... How could a fraction happen inside a rugged connector without leaving sign outside ?? .... Shahrayar my friend , you are so lucky to get this problem at this point without digging deeper inside the madness on the circuit board ....
It probably wasn’t a good connection to begin with.
It would have had a plug inserted which would have taken the brunt of the impact, and would thus conceal dents and dings.
As for SMA being "rugged", it isn't.
All the two-way radio manufacturers are now fitting these daft little connectors despite being down between 140 - 460MHz. Every week I get another handset on the bench where the SMA connector has been snapped off because the radio was dropped, leaving the thread inside the antenna.
For really high density bench instrumentation, I can see where SMA would be useful, but for me, N-type will always be king for low loss and ruggedness.
Thumbs up!
How could possibly a several thousand dollars instrument fell off the table!? I‘ll take it to my bed if I own such a device.
the more expensive the equipment, the less likely it is to be owned by the people who actually operate it. when the company is paying, people are less careful.
This thing is very light weight. It also has a USB/Power/SMA connected to it. Probably a cable got yanked and pulled it off the table. This is also how the connector must have broken off.
1:26 Yes there is and it could be a ridiculous reason; if they deem the one who have one to be an individual they don't want that one to be able to control it since that person, as they see it, most likely has not purchased it and thus should not be able to get free stuff from them.
I am afraid this makes no sense. The unit without the software/driver is useless. The software without the unit is even more useless. This is just bad practice. If you could get something done with the software alone, I could understand, but nothing can be done.
Absolutely bad practice. I own for example a brand new probably only used once turbo pump I bought very cheap on Ebay recently. To get to the real good control things inside it I need a software specifically made for these pumps and they have the software downloadable no problem at all but it is totally useless without a key. That key will only be sent if I send a register form to them which they then really check up if it is a company (thus no hotmail or gmail adress and such) and most likely also that that company have bought it, then they might send the key. Thus as an individual I can't use the pump fully which is just ridiculous.
@@Thesignalpath Even worse is the fact that the software literally sends a number via some USB-serial chip to the device so they can't even get away with "there's our IP inside the program".
Note to self: don't buy from this company. They didn't bother to make a new FR4 PCB spin. For goodness sake it is a PCB spin, not an IC spin.
Sorry to see your comments but this company is specialized in phase noise measurement since 30 years and their systems are very efficients.
They most likely had multiple (dozens or hundreds) already pick and placed considering the BOM cost of those things, which is easily a few hundred dollars it most likely was more economical to do the modifications.
Often there is someone in the company just sitting around, a classic candidate to do those modifications.
But yeah, it is better practice to assemble a fully functional prototype first.