azonenberg changed the topic of #scopehal to: libscopehal, libscopeprotocols, and glscopeclient development and testing | https://github.com/azonenberg/scopehal-cmake, https://github.com/azonenberg/scopehal-apps, https://github.com/azonenberg/scopehal | Logs: https://freenode.irclog.whitequark.org/scopehal
Degi_ has joined #scopehal
Degi has quit [Ping timeout: 256 seconds]
Degi_ is now known as Degi
<Degi> Huh multech can do 36 layers
<Degi> Huh 8 layer ceramic pcb
<azonenberg> Degi: yes
<azonenberg> They've got much lower prices than an EU/US fab, but are very much a high end shop
<azonenberg> they can do moderate volume but it seems like prototyping is one of their specialties
<azonenberg> I've used a fair number of fabs over the years and i keep coming back to them
<azonenberg> great customer service, the salespeople speak decently good english
<azonenberg> they don't make changes to your boards silently, they always message you back with questions if something seems off or is difficult/impossible to manufacture
<azonenberg> one time they caught that my board wasn't quite square, one corner was like 100um to the right of the opposite one
<Degi> lol
<azonenberg> i caught their CAM engineering team in a mistake *once* in all the time i've been working with them, they missed that a trace had no soldermask and calculated impedance with mask
<azonenberg> so the suggested changes to trace width were massive compared to what i had in my design
<azonenberg> i re-ran the numbers in sonnet and asked them to double check
<azonenberg> they apologized, said my numbers were correct, and proceeded with the order
<Degi> Hmh they recalculate traces for you?
<azonenberg> if you say controlled impedance, they'll double check your parameters are sane
<azonenberg> then they TDR a sample trace on each panel and verify it's within tolerance
<azonenberg> they also cross section a board or two on each order and take measurements of copper, soldermask, and dielectric thicknesses, via sidewall plating thickness, etc
<azonenberg> and do a solderability test where they dip a board into a solder pot and make sure all of the pads take the solder properly
<azonenberg> This is what happens when you work with a high end fab instead of one like jlc :p
<Degi> neat
<Degi> How did they get 52.22 max 46.56 min 53.62 average for impedance lol
<Degi> whats float solder
<azonenberg> I think the translation just has the wrong columns
<azonenberg> min 46.56, max 53.62, average 52.22
<azonenberg> i found one or two small typoes in the QA report but the data looked good
<azonenberg> Float solder i think means they have a pot of molten solder and float the board on top of it
<Degi> How big of a pot? They can make PCBs nearly a meter square
<azonenberg> No idea. they may just cut a piece off for huge boards
<azonenberg> generally the cross section comes from the board they do the solderability test on
<azonenberg> so they are cutting it up anyway
<Degi> Hm what do pcbs cost there like
<azonenberg> The probe characterization board is 75 x 25 mm, 2 layer RO4350B, immersion silver, controlled impedance on one layer
<azonenberg> 10 board was $10/board, $220 setup fee, $45 shipping, $15 currency conversion fee, $380 for the order
<Degi> huhh
<azonenberg> So, not JLC pricing (they have i think a 200 USD minimum for all orders)
<azonenberg> But for what you get, it's not bad at all
<azonenberg> Like i said they're a high end shop
<Degi> Cool jlc now has red, blue and black soldermask
<Degi> available at no extra csot
<Degi> Hmm I think I used Tg 130-140 °C instead of Tg 155 °C for my 4 layer board, hope its ok
<azonenberg> Yeah i always spring extra for higher Tg to ensure good performance. Even 155 is a bit light IMO
<azonenberg> My minimum material for new designs is Isola 370HR which is 180 Tg
<azonenberg> That's as low end as i go
<Degi> Hm the problem is that during reflow it may get damage?
<azonenberg> Yes. Or flex
<Degi> Hmm theres excess soldermask and a bit missing soldermask on one of the 200x boards
<Degi> Yeah maybe I should do a dry run with the reflow device
<Degi> And excess HASL solder on some palces
<azonenberg> For higher speed stuff, i use Isola FR408HR (Tg 190, this is what OSHPark uses by default)
<azonenberg> or Rogers RO4350B, which has a Tg of >280C
<azonenberg> Which makes sense because 4350B is a hydrocarbon-ceramic material and not epoxy-glass based
<Degi> 4 layer pcb looks good on inspection
<Degi> Soldermask pretty even
<Degi> Hmm maybe next time 0201 for PCIe decoupling
<Degi> Hm meh I couldve routed the RF traces differently, then they'd be a bit more length matched though arrive at different receivers. I think that shouldnt be a problem since pcie doesnt care about length matching
<Degi> I wonder why smaller vias have soldermask but bigger vias dont
<Degi> Hmm some traces are really close by, hope they dont get shorted in reflow lol
<Degi> There is no ground plane on the backside of the PLL, only on one of the interior layers. Meh its not gonna make that much heat anyways
<Degi> Maybe next time no soldermask above RF traces or at least calculate impedance with soldermask... Not sure if kicad pcbcalc can do that
<Degi> Lol I should use BOMs, this PCB has many components which I didnt order (but it can work without them)
<azonenberg> yeah as part of my layout review i verify all packages are right for the component
<azonenberg> so as i do that i check that i either have them in inventory or add to my digikey cart
<Degi> I mean the packages are right, I just forgot to order LEDs and a fuse
<Degi> Hm oh, I need a 3.3 V ISP device too
<Degi> How to align soldermask
<Degi> The metal thingie
<Degi> Meh i should watch a vid
<monochroma> carefully :3
<Degi> Like damn it needs to be less than like 0.2 mm it took a while, but how to keep aligned during paste application?
<Degi> I guess just glue it down with tape
<Degi> And then hope for the best and rework it
<monochroma> that's the tricky part, you usually build a jig (commonly out of other PCBs) for your PCB, then spend forever aligning the mask and taping it down
<Degi> I have 5 of this PCB
<Degi> So yeah I could do that
<Degi> Hm idk bga would probs be easier
<Degi> Since that would at least have 0.8 mm instead of 0.5
<monochroma> mask liftoff needs to be quick rather than slow (but yaknow.. not too fast...) to have a smol vacuum
<monochroma> (also assuming this is a metal stencil)
<Degi> Why a vacuum
<Degi> Wont that make the paste flow aprt
<Degi> Yeah metal stencil
<monochroma> IPC guidelines say you want a good gasket to the board, so you get a slight vacuum during liftoff so the paste releases better from the stencil, and smears less
<azonenberg> Degi: what i do is i tack the stencil down with a tiny piece of tape on one side so it's easy to move and align
<azonenberg> get it lined up, tack the other side, then put a longer strip of tape along the first side to firmly secure it
<azonenberg> then apply the paste, carefully lift under the tape on the far side with tweezers, and rotate it up using the long tape strip as a hinge
<lain> TIL digikey has products NOT shipped by digikey o.o
_whitelogger has joined #scopehal
awygle has quit [*.net *.split]
azonenberg has quit [*.net *.split]
azonenberg has joined #scopehal
awygle has joined #scopehal
futarisIRCcloud has quit [Quit: Connection closed for inactivity]
bvernoux has joined #scopehal
<bvernoux> hi
<bvernoux> I have finally found a spanner set ultra thin for SMA ;)
<bvernoux> BGS 8646 | Combination Spanner Set | Stubby Type | 4 - 10 mm | 8 pcs
<bvernoux> Advantage is there is 5.5mm and it is thin as 1.6mm
<bvernoux> very nice for my VNA calibration kit Female-FEmale thru
<bvernoux> as the ultra thin 5.5mm spanner is hard to find especially here it shall be less than 2.5mm
<electronic_eel> bvernoux: BGS tools are usually not the best quality
<electronic_eel> have a look at the Hazet 440 series instead, they are nice
<electronic_eel> I have a full set of those for accessing tight spaces
<electronic_eel> the 5.5mm one is exactly 2mm thick
<bvernoux> ha interesting
<bvernoux> hmm Hazet will take more than 30days ...
<bvernoux> but interesting for next time I will buy spanner especially 8mm & 5.5mm mainly for my VNA Calibration Kit
<bvernoux> mainly to unscrew the SMA ...
<electronic_eel> may not be the cheapest option, but they have the full set in stock
<Degi> Thanks, yeah I think that sounds good with using tape as a hinge
<Degi> Do you put PCBs of the same thickness under the tape because otherwise it would bend right
<Degi> And I think I found a usecase for bandwidth >> sample rate
<monochroma> correct @ same height PCBs
<Degi> Hm do people still use EPROM?
<Degi> Like the ones with a quartz window
<monochroma> nope
<monochroma> not for anything modern
<Degi> I think I have two of them lol, maybe 3
<Degi> You can use them as a camera hehe
<monochroma> lol
<monochroma> there were a bunch of digital camera projects back in the 80s that used decapped DRAMs as the sensor
<Degi> I found an instruction somewhere to build a camera from a quartz topped EPROM and a lens xD and it makes shitty images but kinda nice demo. I guess back then CCD/CMOS was pricy
<Degi> I even have two vidicon-like tubes
<Degi> Hm why is the HMCAD so much cheaper than comparable ADCs and has so much more features
<Degi> Like I guess nobody buys that stuff and thus companies dont care?
<monochroma> it was cheap to begin with
<monochroma> analog devices got the line when they gobbled up hittite microwave
<Degi> Hm I found this http://www.ti.com/lit/ds/symlink/adc08d1020.pdf it has 1.3 GHz input BW and costs approx 190 € per GS, a bit pricier and less features than HMCAD but it has more BW
<lain> the hmcad1511 was crazy cheap, it was only $30 back in 2007 on mouser
<lain> it was by arctic silicon devices, a fabless startup
<lain> then hittite bought them and raised the price, and then analog bought them and raised the price again
<Degi> Hm ugh yeah
<Degi> Hmh why are ADCs so pricy anyways? Is it because of low volume?
<Degi> For example modern CPUs with more silicon area and smaller features cost way less... (I mean in the 1 GS/s range ADCs are mostly silicon, rihgt?)
<azonenberg> lower yield? more R&D effort amortized over less chips?
<azonenberg> just guessing
<monochroma> analog circuitry tends to be large due to the nature of the physics, so large and specialized process nodes are used, so you get less out of a wafer as well
<azonenberg> that too, but doesnt explain how a hmcad1520 is $100+ and is a tiny qfn package while an xc7a100t is 28nm and has a die the size of the whole hmcad *package*, and around the same price
<azonenberg> also btw, shipment from shapeways is coming tomorrow
<azonenberg> it's either the samples i ordered of the old probe shell in a couple of materials, the new probe shells, or both
<sorear> if you took apart the modem in a new high-end smartphone (let’s imagine a magic FIB here, don’t worry about what signals are exposed) how many ADCs would it have and what bandwidth and resolution?
<azonenberg> i would imagine a fair number
<azonenberg> even a modern ethernet PHY will have a fair few
<monochroma> modern PHYs have entire DSPs :O
<lain> I know the hmcad1511 was so cheap back in the day because they leveraged a very clever architecture with lots of DSP to reduce the size and required tolerances of the analog silicon
<lain> most standalone ADCs, as I understand, are still heavily analog, so, lower yield and larger silicon area
<monochroma> need some scope food
<lain> :3
<Degi> Hm yeah, yield...
<Degi> HMC1061 / 661 seem interesting
<Degi> That would be really nice for splitting a high frequency signal into multiple lower frequency signals but its priicy
<Degi> azonenberg: With that IC you could do a 10 GHz sampling scope where it only needs to do a 1D scan (I think? Not sure how long its hold time is tbh)
<azonenberg> Degi: i could do 18 GHz with that
<azonenberg> more importantly, i could also use it for higher bandwidth interleaved oscilloscopes
<Degi> Yes but the cheaper one is like 600 bukcs
<Degi> I'm not sure what better things the fancier one can do? Like it has two SH in series but of what use is that?
<azonenberg> I see it as potentially being useful for MURDOCK
<azonenberg> Not for lower end stuff
<azonenberg> We could push MURDOCK past 6 GHz if we did that
<Degi> Hm yeah
<azonenberg> since we no longer are limited by ADC front end bandwidth
<Degi> Maybe 5 AD9213?
<azonenberg> let's worry about that once i have $50K or more to spend on prototyping :p
<Degi> rip
<azonenberg> Which i currently do not. Lol
<Degi> Also gotta use RF materials
<azonenberg> well obviously
<azonenberg> but a little bit of RO4350B is probably going to be the least expensive part of that scope
<azonenberg> except maybe the screws that hold the board into the enclosure :p
<Degi> Lol
<Degi> Solid gold screws?
<Degi> Tbh solid gold screws would be way cheaper than that SH lol
<Degi> I think it should be possible to use a RF amplifier as a comparator too to get a comparator beyond 10 GHz (though at worse quality)
<Degi> Read the datasheet tho, at higher amplitudes it seemingly has less tracking BW but the sampling BW remains at 18 GHz. not sure what that means
<Degi> Ohh its a dual rank track and hold, that makes that the output is at a constant value (because for a single track and hold, the output would be varying for the track duration)
<azonenberg> ah ok that makes sense
<Degi> Though only 4 GS/s is kinda meh
<Degi> Like the AFE is gonna be big monies if you wanna use the full BW
<azonenberg> For sure
<Degi> Maybe short of 20k
<Degi> Or 30k
<Degi> Much fun getting sufficiently good ESD diodes if at all
<azonenberg> oh i'm going to assume something that wideband will have ~zero esd prot
<Degi> Like it costs 30k and breaks when you touch it
<Degi> Hm on my PCB the holes on the front layer are perfectly aligned but on the back layer theres some ~150 µm misalignment with some holes
<Degi> And some weird stuff in the PCB, I guess prepreg marks (the red stuff on them)
<Degi> I think the drill went sideways since the offset is different for different ohles
bvernoux has quit [Quit: Leaving]
futarisIRCcloud has joined #scopehal