EEVblog
HTML-код
- Опубликовано: 2 ноя 2024
- Teardown of the new Keysight InfiniiVision 1000 X-Series Oscilloscope released TODAY!
Teardown Photos: www.eevblog.co...
Forum: www.eevblog.com...
EEVblog Main Web Site: www.eevblog.com
The 2nd EEVblog Channel: / eevblog2
Support the EEVblog through Patreon!
/ eevblog
EEVblog Amazon Store (Dave gets a cut):
astore.amazon.c...
T-Shirts: teespring.com/s...
💗 Likecoin - Coins for Likes: likecoin.pro/@...
Sub board will be because it needs a higher layer count than the large main PCB
Mike could you come up with an assembly and test cost estimate for these options:
1) single large PCB with front end, ADC and processor
2) simplified mostly-analog plus higher density digital PCB as shown here
would much appreciate your input in the discussion
It won't just be layer count - will also be other factors like feature size, blind/buired vias, yield etc. but the bottom line is the sub-board will be a lot more expensive per square foot then the main one, so for the cost of a few connectors you aren't wasting money on expensive PCB area where not needed. May even be assembled on different machines & use a different reflow profile, and have x-ray inspection.
Thanks for following up. Assembly and inspection - that's an important aspect I had overlooked.
I've also put in some estimated dimensions and specs for the PCBs, $75 vs. $15 for the main PCB (8/4 layers, 0.2mm / 0.25mm holes), $9 vs. $3 for the small one.
They probably spent a lot of time on cost reduction, much bean counting was done.
How many cents per connector, how much per square mm of PCB, what part can we save money on?
Thank you Mike, your opinion meter and always is correct.
Maybe the separate logic PCB has more layers and finer geometry? It might be cheaper to make a small high density board, if that means the larger main board can have fewer layers.
cheap 2/4 layer plus 6-8 layer, makes total sense - and the added opportunity to use blind / buried / tented vias on the small add-on board. I too see that being cheaper than doing the whole PCB in the expensive process. Worth those eight board-to-board headers.
I would like to know the entire system cost breakdown going that route
AndyC772
That was my first thought, too. Plenty of space on the main PCB, so go for a cheaper option there.
Surely the additional PCB assembly will add cost during production and handling, make testing easier or more involved and add steps during assembly.
Numbers below:
A quick check against PCBWay pricing for lots of 100 boards shows how much cheaper it will be to split the PCB specs (other mf options are default):
70mm x 140mm 8-layer 2.0mm FR4, 5/5mil spacing, 0.2mm hole size, $9.00/pc
70mm x 140mm 4-layer 1.6mm FR4, 5/5mil spacing, 0.25mm hole size, $3.06/pc
400mm x 200mm 8-layer 2.0mm FR4, 5/5mil spacing, 0.2mm hole size, $73.34/pc
400mm x 200mm 4-layer 1.6mm FR4, 5/5mil spacing, 0.25mm hole size, $15.54/pc
I have no idea how that scales as you go to lots of 100.000 and beyond and gear up for production with ATE but it appears possible to add immersion gold finish and 100% e test and still save enough to more than offset the added fuss in handling and testing.
If you know the layer counts and feature sizes, you could do a first-order approximation using PCBCART
It may actually be cheaper to put the processor on a daughter board. The daughter board may be more expensive per square inch because it has more layers, high-capacitance dielectric layers, blind vias, etc. Also, it may be easier to test the processor board separately. These factors could easily offset the cost of the extra connectors and assembly labor.
2:27 "May contain foreign or domestic components" - as opposed to WHAT? Extraterrestrial? :D
as opposed to domestic or foreign perhaps?
As opposed to only foreign or only domestic, maybe?
We can't talk about it...
Oh sure, how convenient... :P
+Keysight Dave mentions in this model, "differential pairs buggering off". Comment on that.
Just waiting for Scope Month....😊😍
The separate PCB can be in fact a cost cutting measure. The daughterboard, while much smaller, have to be a 6-layer board or more. However the motherboard can totally be a 2-layer or 4-layer board. A 2-layer board, while larger, is still much cheaper to produce than a 8-layer board.
This is also why a lot of Linux SBC projects uses a module for processor and RAM - those are 8-layer modules covering 5% of the total board space, while the bulk of the board is made using only 2 layers.
So... as a hobbyist, using own money, still better off getting the DS1054Z (and having a sneaky little tamper with it at home) yeah? I wonder if Rigol have any plans to release an updated version of that anytime soon...
Yep. And no idea about an upgrade, but I'd say not for a few years, it's still selling like hot cakes.
+Samgab Check Instek, around $370 US, 4 channels, with Separate Controls for Each Channel. No Brainer.
When choosing the PSU caps, they probably just picked the most funny brands lol
LOL !!!
They picked up cheapest once for sure
Interesting, I always like to plan for expansion as well. Always fun to take something apart and tell what the designers were thinking.
2:35 we can see your Windows CE product key!
Sudos quick, sell it on MySpace!
Thats not the actual key. Just a MS product ID
I would have to disagree with you, Dave, on the cost of manufacturing of the motherboard/daughterboard. The technology of making a high density / HDI board for the high pin-count BGA parts with microvias and all, will dramatically increase the price of the entire board where it is not needed. Keep in mind this is only a 100MHz scope, so no high-end materials need to be introduced into the main, cheap board and the overall cost of the entire design could potentially be halfed...
Yep, likely explanation.
Exactly. One more thing: they could design one base board and different HDI boards for their 1000x series model line-up.
That is not really needed. Since the design is fairly basic, only 2CH 100MHz max, with or without function gen. the HDI board should be more than capable to handle all versions.
What they can do, is to introduce a new 2000X and 3000X series with exactly the same board and reduce its costs by the same method
Yup, that makes sense.
Regarding the rust, they seem to buy galvanized sheet metal then cut it and form it.
The proper way would be to cut it from black steel, do all the cutting and bending then hot dip galvanize it at the end, but too expensive it seems.
I don't really understand how the common mode filter can make much of a difference. Why do everybody use it in power supplies?
Dave could you possibly do a video on them to explain their usefulness please? It is bothering me for a very long time now.
It's not too surprising that the CPU board is a separate module. Those parts are more likely to be affected by product obsolescence and supply chain problems, plus it allows for updates, testing, service, and reuse in other products. Many other Agilent/Keysight products have modular CPU sections that are reused across different product designs.
I've developed on WinCE for 6 or 7 years now and some of what people are saying is misleading or out of date. MS has revised the network stack and our tests have shown it to be comparable to Linux if you have the right driver, at least on x86. It's not at all like big Windows and has much better real-time performance than unmodified Linux. The scheduler, driver architecture, and I/O architecture are much more efficient in WinCE. .NET is also available which means Keysight has the option of leveraging desktop and spectrum analyzer code that is also written in .NET. I'm sure it makes more business sense for them to use WinCE rather than Linux in these products and I promise you that you would rather have a WinCE scope than a Win7 or Win10 scope, unless viruses, poor performance, constant updates, and long boot times are your thing. WinCE is a lot more than an RTOS but a lot less than desktop Windows. It's only real drawback is that MS won't put any resources on its development and critical bugs can linger for months to years before something shakes up in Redmond and they put warm bodies on the issue.
Coated sparkgaps?
Two board construction may also suggest a future reuse of the CPU module on other models.
yea those sparkgaps
"Everything has its breakdown voltage' (c) photonicinduction
WOO! The NEC relays, Daive should spend at least 10 minutes on those, you know, why they are colored the way they are, history of relays, you know.
Also, notice that they missed a large area in the lower left of the PCB with the conformal coating.
Not a quality coating job!
i wonder if is something in this universe, that Dave will like 100%
perhaps this one: www.eevblog.com/product/121gw/ 🤔
@@gl3906 yeah yr!
The open holes for ethernet and other connectors may suggest, that there are different types of mainboards for different models. the processor board can then stay the same, with just a config tweak, across different models. The Processor board with its BGAs may require lower tolerances and seperate QA for both boards may be more cost effective.
Seems that oscilloscope internals have fewer and fewer components on the inside. Is that just the advancements of FPGA's making the additional components unnecessary?
Did you ever look inside an Tek TDS220? There are even less parts in these, and they're quite old now.
Interesting thing... When the first run of HTC Vive VR headsets came out, quite a few of the first run had a LOT of dead or stuck pixels in the LCD's.
I got one of the first 2000 and mine was luckily fine, but when everyone started talking to tech support, they were telling people they would only fix it if they had more than 5 pixels per eye display that didn't work!! Thankfully things were quickly changed with their policies, but it makes me think if they were going off what the data sheet said....
Stupid in a scope to have such a thing, even crazier when its a monitor an inch from your eyes!
Impressed with the review -- I was expecting more of a sales-pitch-type review glossing over the imperfections -- as I saw the rust before it was mentioned; but I found the review to be quite fair. I do not think this scope competes favorably with the Rigol 1000Z series (4-channels + logic analyzer), but Keysight certainly seems to be feeling the pressure from Rigol.
The red goop on the regulator is called Glyptol (Glyptal). It's been around for over 100 years. It was used on coils and trimmer pots to lock them for calibration purposes, and since it's brittle, as a tamper indicator. Older cast iron machine tools were painted with Glyptol, internally, to seal the metal's pores so lubricating oil wouldn't migrate (leak).
Wow, you are definitely have an ability to perform chemical analysis by just watching a video on YT.
double board construction is for cost savings. The board with the BGA chips has to have more layers and buried vias, it's a break even for PCB cost and connector cost
Would be interesting to know the actual cost breakdown.
I did this recently for some lower volume. It's an annoying game of quoting the PCB and connector manufacturers, juggling with your project managers and ending up with a clean tie :-) Not engineers work, more for bean counters.
An engineer would start from system level and end up with the same solution for several reasons (cost of the PCB beeing one of them, others mentioned here like reusability and upgradeability). Bean counters are not able to convert these into cost items, so it's a really sad game for an engineer.
Did you notice the footprint for U39? (Time 23:40) It looks like the footprint is way too big for the part, and there is very little solder on the lands. Looks like the pitch is wrong too.
It's a dual purpose footprint, nothing wrong with that.
eh, the rust is just flash rust from the cutting process. it won't get any worse unless your lab is humid or on a boat in the ocean. if you were really worried about it you could swab some dielectric grease on those cut edges or something i guess.
26:51 "high risk photos" :D In australia, everything tries to kill you.
You'd be amazed how few people in Australia actually die from venomous animals. The European Honey Bee is the number one venomous killer in Australia by a huge margin.
A bit harsh on the power supply output wiring Dave.. Did you forget the loom terminates 200 mm further on in a single connector? What reliability economic or convenience concerns would be served by adding another connector?
i have no idea what all this is but i love teardowns :D
Besides difference in layer count of PCB, you can predict they can upgrade the aging ASIC to a newer iteration during the lifetime of this oscilloscope.
Another possibility (or perhaps a secondary advantage) of the second board is they can manufacture it elsewhere and not have to risk a technology/IP transfer to a firm in Mainland China.
LCD in not noname, you just had not removed the metal holder/shield.
- my guess - keep the heat away from the main pcb - also better cooling, as the air also flows between the pcbs - great idea
was there enough room in the back, to put a 120 mill fan running low RPMs? and 3d print a vent from large 120 fan to the smaller hole, so the air doesnt just stay trapped heating up more and more.
Perhaps they use a daughter board based approach for the Megazoom chip etc, so that only the smaller PCB needs to be 12+layers, with the larger main board being a smaller number of layers.
And also with the 'inteligent stuff' on a seperate PCB, you have to redo less stuff if you want to update the unit as well.
And I think that might be a thing too...
Are the spark gaps still efficient with the conformal coating?
Anyways regarding the two board construction, is that maybe because if one big board with everything fails, you have to throw all away, and in this case you can likely reuse one of the two? I mean at the production stage? Kind of like how bigger silicon dies have higher failure rates
just wanted to add that the pixel clause mentions 'sub pixels', as in the individual r/g/b elements of each pixel rather than an entire pixel element. which yeah is just as bad rally but at the same time its still less of a a defect
At 9:45 the spark gaps are coated- I always thought the air should be conducting, not the PCB.
After been through hell desoldering front end shieldings in vintage oscilloscopes when i saw on the silk screen written "nut" i nearly cried
"Differential pair buggering off". Hey Daive, (pronounced Dyve) load up the Toyota land cruiser and "bugger off" to Pine Gap and characterize the electromagnetic radiation "buggering off" that site. Looks like we are getting closer to a 100mhz, 4 channel (separate controls for each channel, please) scope for less than $600. That would be a game changer.
I really like the bode plot function in the scope.
How do spark gaps perform with a conformal coating over the top of them?
guess it will work as would have a slightly lesser breakdown voltage compared to air because insulation resistance provided by air is higher than that of a conformal coating material.
As best as they can. :-)
9:43 Nice spark gaps, but unfortunately coated by insulating lacquer...
HILARIOUS!!!! at 28:45 you can see the spark gaps are also covered with conformal coating!!!! they are useless!!!
There could be 2 reasons for the daughter PCB:
1. ST had decalred the Spear600 as NRND, so they might change the board soon as it will be EOL soon:
www.st.com/content/st_com/en/products/microcontrollers/legacy-mpus/spear-arm-926-microprocessors/spear600.html
2. With BGA devices, Memory etc, there is a greater change of PCB layout issues, so it is more easy to have a small PCB for that section. Many customers today buy a ready made one called SOM ( System On Module ) and save the design fase.
I'd say they are using an 8/10 layer hdi ( small pitch, blind/buried/laser vias and all) for the digital section, for the rest of the scope where fine(ish) pitvh bga is not used they don't need that and they use a simple 2/4 layer board much cheaper I'd guess
Yes, I would like to know the cost breakdown with the connectors and the large 4 layer vs smaller 8-10 layer, including assembly overhead etc.
Great video Dave!
27:56 They have a lot of "scope" to move on the pricing... No pun intended? :)
I noticed that during editing
What is the layer count of the 2 boards? It can get expensive to make a large 8 layer board. Also it makes the processor transplantable to other designs. When you factor in the cost of designing, making the die, inventory it is very cost effective to make a chassis with extra holes and be able to use it in multiple models.
Also I was thinking the yield of the PCBs would go up. The main one looks pretty simple so that would run off the production line at probably 99-100% accuracy, the CPU one has 7 CPU-like BGA IC's and there could theoretically be reflow problems or something that causes yield to go down.
Could there be another way to "be in" besides flynn? Just for variety??
9:45 - The "conformal coating" doesn't conform to the entire PCB. Check out the lower left corner! Not impressed!
Waah, I really hate fans in lab equipment. The noise is really annoying.
Yepp 4 Ch will come. Have a look on the ASIC ADC board. 4 pair of differential line!
I missed that, will take another look.
12:41 if My eyes are correct there is 4 differential pair
+EEVBLOG WHAT? You missed that, Daive? In this same review, you specifically mentioned, "Differential pairs buggering off". Keep up, man. I am, and will continue my deluge regarding your phrase. It's fun. You know what would make me stop? Yes, that's right, award me an oscilloscope.
He said: " EEVblog1 day ago I missed that, will take another look." So just wait until He will take a look. I think his scope is the best choice for young players. Even only with 2 channel.
I wonder how long it'll be before someone finds a way to run DOOM on this oscilloscope? Because it's not a matter of "if", but "when."
In the power-supply board; RX1, RX2, RX3, RX4 reduce the clearance, don't they?
Hey Dave from Memphis, TN 🇺🇸
AsiaX capacitors were prevously named Fujiyuu (according to Badcaps forum)
17:05 Fujiyuu/AsiaX RVT series (life of only 1000h @ 105C)
8:00 Is it just a funky angle, or is that larger/largest cap on the bottom (68u/450v, gold markings) ever so slightly bulged?
PS: 105C rated "only" means that it's not bottom-of-the-barrel general-pupose (85C rated). There are plenty of general-purpose 105C caps, without them being low-ESR. Just FYI :P
Noticed that one too.
Yup, it is ever so slightly bulged!
Now I'm really curious. in-circuit ESR meter maybe?
Damn i'm good! :D I wouldn't (even) bother measuring ESR or anyting else, when crap-brand or no-name caps are involved. Asia-X and Aishi are known-to-fail-sooner-rather-than-later brands, same as CapXon and Fuhjyyu.
Is it wrong to put conformal coating over a spark gap?
...and of course there would be six to choose from, four 2 channel and two 4 channel models...
Spare board to reduce cost, because is cheaper to make small board with more layers
You'd be lucky to add components and get the functionality of a higher model. Even if it electronically worked, I'd hate to see it come down to proms or firmware versions that ended the day. Or maybe some other component that isn't what it should be because they didn't think they'd need the bypassing or something.
maybe the processing logic board was manufactured elsewhere!
still waiting for yokogawa dlm2000 series review.............
You'll be waiting a long time, I don't have one.
lets launch a patreon or paypal process so that you can buy one!, im in!, i cant believe that bastard didnt send you one!
Great teardown!
Would Keysight use a processor board because it is made in Malaysia instead of China so their technology would not be stolen or to avoid ITAR restrictions on some technology?
does conformal coating (specifically on the back of that power supply board) reduce the effectiveness of the spark gaps they put on the board? Based on what dave said in his spark gap video thats no good
When the shielding comes off with a screw
>>NUT
I wonder, why can't these companies just give the oscilloscope a 120mm case fan?
Also, is it possible to overclock an FPGA, ASIC, along with the ARM CPU on the oscilloscope?
why would I buy the 50mhz version of this, when I can get a 4 channel 70 mhz with dual channel function gen and logic analizer for the same money?
I bought HP 24es LED monitor last month. And have two dead subpixels. I've checked docs on HP website and it stays that it is within limits. Only highest end models have zero pixel fault policy. So 90-s :/
Would it be worth it to recap that psu so it can have some proper cemicon (Other brands are available) caps so you wont have to recap it in the future when the cheap ones die and I would maybe change the fan and put it on a rubber mount
As a customer of Keysight i would have loved to see chemicon / Rubycon capacitors and would not have mind paying little extra for a quality stuff, majority of the time i have seen these die and ruin the whole unit, always even the best once electrolytes caps die, Dave is right about Shenzhen capacitor, hope the design team listen of customer request
Real interesting how they laid out the JTAG pins! But for me personally (5th year EE @ university), no option for 4ch is a huge dealbreaker. 2ch too often is just not enough even for basic designs. It's just such a hassle to move your probes around every 5 minutes. It's almost like don't bother putting in serial decoding if you don't have 4ch or at least 2ch+8ch analog/digital. Anyway, great quick little teardown there!
Lack of 4 channel will kill it for meany people. But the Tek TBS1000 only has two channels and that's what it's competing against.
Seriously. 2 channel and a sperate trigger input was more than sufficient for a century or so. Remember that you need serious real estate on your screen to make full use of four channels. So, long debate cut short, I prefer 2 channels and a function generator over 4 channels no function generator.
+EEvblog Quite Right. In two years, we will see a 4 channel, with separate controls, at 100mhz or more, for less than $600. Uncle Bob's Oscilloscope, or UBO. See what I did not mention here Daive, something about "differential pairs" and the particular way you mention these pairs going somewhere?
+Sven Starson "more than sufficient for a century or so". It was sufficient because a century ago, they were using Dave's Grandad Cad.
To your surprise Dave still uses DaveCAD. (I for myself use the inferior SvenCAD). Why? Because it is quick and efficient for some tasks. I have a four channel scope on my workbench. I use just one or two channels of it most of the time. I don't want to spend all the time setting up four probes, not to mention that two probes is a natural fit for two hands (Dave's Grandad already had two hands, I guess).
Will we have a eevblog #1k special episode?
He doesn't usually bother with milestone numbers, too many of them
What happened to that giveaway?
They certainly seem to be pushing the hack-ability factor by giving clear labeling of key system interfaces.
Do the spark gaps in the power supply even work properly with the conformal coating on it?
+PyroShim Spark gaps? What is this, Tesla coil stuff, Jacob's ladder stuff?
NeverTalkToCops1 You can see them at 9:41 on the side of two of the optocouplers. They will ark over when the voltage across them gets too high.
How come the power supply is "beautiful" ? It's kind of on the lower end side, the inductance things are supposed to generate some decent interference regardless of shielding, in such placement. Not to mention other faults.
The processor part may be some typical design, to cut design costs.
Just from a layout aesthetics point of view
At that power consumption, it doesn't need a fan at all.
cool osillyscope
I smell an AvE fan...
osmelloscope
Robert Bates Focus you fuck!
Great vid!
Would Love to have one of those on my bench!
For PORTUGUESE speakers I did a review on my channel with the DSOX1102G
There are 28 days in every month; the longest being "September" ☺
Can you do software keys for upgrading through the product line? Wonder if the daughter board has anything to do with that.
3:20 On the loud side? How about the turbo-jet inside my HP 3314A sig-gen? x)
You haven't met my Tektronix AWG2021, it sounds like a jet taking-off if you boot it up.
Impossible to use in an office environment. The fan might be old and should probably be replaced...
he doesn't like anything, maybe except for the EEVBlog multimeter
What is the reason for the spark gaps under the common-mode-choke?
EEVblog. Sorry, my english is very basic and I do not know if this review is to say that the device is good or bad. Can you guide me? :(
Can somebody please explain to me why anyone would buy this instead of a rigol ds1054z?
Because Keysight !!!
It's a more responsive scope for everyday use. But yeah, bang-per-buck against Rigol is not great.
Keysight and the other big players can offer a class of service and support that the smaller ones (like Rigol and Siglent) can't (yet). If you're going to spend $100k on scopes for your classroom, you're going to want to know that you can get the support you need when you need it. Also, if you depend on your scope in your work to earn you money, you want the confidence that only the big players can give you. For hobbyists, it's a completely different game.
Hopefully because it's less laggy!
Without having seen this scope in action, I'll stick my nect out and say that I'd be a lot more productive using the Keysight scope. I'd rather have a 2-channel 'scope that triggers reliably on noisy signals than a 4-channel one that doesn't. Also the Rigol is really laggy.
Conformal coating over a spark gap? :D
I have not seen the inside of an oscilloscope in my life
I can't see any ICs in the power supply. strange...
Why not ask Keysight to replace the fan with a Noctua Fan and see if the fan can be made to run slower.
It's the same silicon as the 2000X which has a much larger board and greater conduction capability, so the heat must be dissipated by more forced air cooling. There isn't the space for a larger fan that can run at lower speed which is an easy way to make it quieter. Only thermal improvement I can foresee is to put the three ASICs on the main board and load it with thermal vias, or maybe add variable speed control to the fan, but if the instrument is on all day, it would probably need max cooling all the time anyway.
+Marc Davis Well, the "hackers" could just bolt on some off the shelf Peltier junction stuff.
there are some pretty surprising things going on in there given the name on the outside. im sure its all good and id be happy to have one but still....
So in 1 years time it will need to be de-rusted, and all the power supply caps replaced.
hiya dave hav u thought about makin a vid tellin ur old subscribers whats goin on with the bell next to the sub and to make sure we r gettin the notifications of ur new vids....seen a few of my other channels i watch doin it and i thought i would let u know....trouble is u may hav done 1 and missed it but at least that helps me prove a point...cheers
8:44 You mean they didn't even splash out on CHONGX caps?!
Can't wait to see the reverse engineering of the AFE. Good job.
Do spark gaps still work under conformal coating?
SUB PIXEL, means, the red/green/blue part of it...
I wouldn't say nice earth point, spade terminal ! Should be bolted!
It was as tight as a nun's nasty though.
You must be from the U.S.. Your love for bolted ground points is unsurpassed (and highly ridiculous).
No actually, I'm English, and live in New Zealand. The reasoning behind a bolted earth is that it does not come loose over time, spade terminals gradually relax and get looser, this doesn't happen with bolted connections. You could say the same principle applies about twisting wires together and saying that they are a good connection and an acceptable way of connecting things together, sure they are fine for a while, but eventually it will fail, which is why we solder stuff together or use proper terminals. The earth has to handle a potentially high fault current, so it MUST have a VERY low resistance, not just for the first couple of years of its life, but until the thing reaches the end of its life and gets binned. and dismantled.
Also, if you look at any older equipment, (where corners were not cut to reduce costs to the same extent it is these days), you will see that the earth is usually bolted, the country of origin is irrelevant. Most electrical PAT test regulations require a bolted earth.
Soldering wires is nowadays very much frowned upon. Even in this video. And for good reason. The critical part is rather how the wire is crimped to the connector than the choice of connector. I have seen far more failed crimps than failed connectors. All types of connectors put together.