<azonenberg>
not sure what the dip at 5.3 GHz or so is, i didn't see that in my VNA measurements of the actual board
<azonenberg>
But otherwise the curve shape is quite close to the actual board. These two sims do not include ENIG effects so the loss is a bit higher on the actual board, but i've modeled that separately and that delta is consistent with what i've seen in other tests
<azonenberg>
Red is the board as is (resistors at 2mm pitch)
<azonenberg>
blue is at 1.3mm pitch. mind you an EIA 0402 resistor is ~1mm long so there's about 300um between the packages now and the 500um square SMT lands actually overlap
<azonenberg>
I could go a tiny bit closer, but the max material dimension on the resistor at the tolerance extreme is about 1200um so i wanted to allow some margin for that tolerance
<azonenberg>
This is using the same schematic, same general layout floorplan, literally the only difference is moving parts closer together
<azonenberg>
And the difference is staggering
<azonenberg>
wooooo RF black magic :D
<NeroTHz>
yeah :p I´m all to familiar with RF black magic
<NeroTHz>
I spent 2 weeks optimizing the placement of 4 vias on my CPW-to-SIW transition
<azonenberg>
lol
<azonenberg>
I'm trying to milk as much as i can out of this sonnet pro demo
<azonenberg>
I've used about 1/5 of the time on it already, time flies
<NeroTHz>
and moving them a total of 120 um had 10 GHz of bandwidth and 1 dB of IL impact (mind you 10 GHz bandwidth at 120 GHz is <10% but still)
<azonenberg>
lol wow
<NeroTHz>
yeah, I can imagine
<NeroTHz>
but this is why I think EM solvers are so important
<azonenberg>
But i can say having had time on pro now, i'm SOLD
<monochroma>
RF: we take away all of your "wires/traces", you are left with nothing but inductors and capacitors to connect everything together :P
<NeroTHz>
and why once most people get used to them in their design flow they can´t go back
<azonenberg>
even as a current user of the lower edition
<azonenberg>
The things i can just... DO, rather than having to spend hours simplifying my design or splitting it up into little bite-sized tidbits
<NeroTHz>
like, you get to this point where you don´t trust anyone using rules of thumb anymore and you are like ´wait you don´t simulate that?´
<azonenberg>
Oh i've been there for a while now
<azonenberg>
But there were a lot of things i wanted to sim that i couldn't
<azonenberg>
it's gonna be tough going back to the basic edition next month lol
<azonenberg>
If i had $20K to spare i'd have forked it over already
<azonenberg>
I think with some effort, ~95% of what i want to do can be modeled in gold
<azonenberg>
the stuff i'm doing now using 8+ GB of RAM and the 32-thread solver is cool and all, but if i put in some time i think i could simplify the geometry by enough that i could get a good-enough result within the capaibilities of gold
<azonenberg>
and gold is less than half the price of pro so i think that's going to be the sweet spot for me
<azonenberg>
but less than that is just not an option for my work
<azonenberg>
Just gonna guess you've got the full uncapped HFSS?
<azonenberg>
with all the bells and whistles?
<NeroTHz>
me? Amongst other things, yes
<NeroTHz>
keep in mind as academic institutions we get big discounts on that stuff
<azonenberg>
Yeah
<azonenberg>
I do get a small discount on sonnet for being a small business, actually
<NeroTHz>
though I haven´t used HFSS for a while, there was a hickup in the licensing agreement for a few months at the start of this year, so I switched to EMPro back then, and stuck with it for the project just because I didn´t want to re-fix my design flow
<azonenberg>
They're a tiny company too, a few dozen employees probably. So when i told them i was a one-man shop the sales rep knocked the price down a decent bit
<NeroTHz>
Sonnet has a bit more than just a few dozen employees I think
<NeroTHz>
but I´m not sure
<NeroTHz>
still gonna be tiny compared to the big names
<azonenberg>
well as far as i know the whole company is a single office in upstate NY
<azonenberg>
And when i look it up on streetview i see multiple companies at that address. And the building is, well, not that huge
<azonenberg>
i get the impression they have one floor or less of a single building
<azonenberg>
and that's it
_whitelogger has joined #scopehal
<NeroTHz>
Why does keithley include a screwdriver in their packaging of the 7000 series scanner cards
<NeroTHz>
I can´t imagine someone buying those cards and not already having screwdrivers
<NeroTHz>
they don´t include cable snips either, but you need those to do anything usefull with the card
<monochroma>
NeroTHz: is it golden? :P
<azonenberg>
Lol
<azonenberg>
diamond encrusted?
<NeroTHz>
hehe
<azonenberg>
NeroTHz: So this is interesting. Sim is still running but right now it looks like I have about a dB of peaking at 3.5 GHz with the new resistor array
<azonenberg>
then it starts rolling off
<azonenberg>
I have some loss from the probe tips that i still want to try improving and this should help cancel that out so the peaking is actually good
<azonenberg>
this is before adding probe tip or termination effects, the port is at the edge of the PCB
<NeroTHz>
that looks good
<NeroTHz>
nice improvement, just squeezing resistors closer together?
<azonenberg>
I also made a small change to the ground plane near the tip to reduce capacitance. But that had a minimal change compared to the resistor placement
<azonenberg>
Once it's done i'll do a circuit theory sim combining this with my best guess model of the probe tip
<azonenberg>
And we'll see what i end up with
<azonenberg>
It may not be the answer but it's a step in the right direction beyond any doubt
<NeroTHz>
that would be itneresting to see indeed
<azonenberg>
This is the biggest EM sim i've done yet. 7.5 GB of memory, about 20 minutes per frequency on 32 cores
<azonenberg>
28604 mesh cells
<azonenberg>
Actually giving this rig a workout, i pretty rarely hear the CPU fan get this high lol
<NeroTHz>
need a bigger CPU fan then :p I´ve never heard mine spool up
<azonenberg>
Lol
<azonenberg>
well normally it runs near idle
<azonenberg>
Now i'm running up to ~70C on the hottest cores so the fan is actually cranking
<azonenberg>
And according to the ACPI power meter my computer is using 4.29 megawatts
<azonenberg>
i think there's a scaling factor missing :P
<azonenberg>
either that or it's just garbage, because that sounds suspiciously like 2^32
<NeroTHz>
yeah :p I have my doubts about that number
<azonenberg>
well i mean the fact that my feet aren't on fire is probably a good hint as well :p
<azonenberg>
Or that my 6 kVA UPS hasn't exploded
<_whitenotifier-3>
[scopehal] azonenberg pushed 1 commit to master [+0/-0/±4] https://git.io/JJd15
<_whitenotifier-3>
[scopehal] azonenberg 06cd31a - FFTDecoder: now output dBm instead of volts
<_whitenotifier-3>
[scopehal-apps] azonenberg pushed 1 commit to master [+0/-0/±4] https://git.io/JJd1d
<_whitenotifier-3>
[scopehal-apps] azonenberg 7d279ab - FFT now outputs dBm directly so no conversion needed in rendering code
Nero_ has joined #scopehal
NeroTHz has quit [Ping timeout: 272 seconds]
<_whitenotifier-3>
[scopehal] azonenberg pushed 1 commit to master [+0/-0/±3] https://git.io/JJdQI
<_whitenotifier-3>
[scopehal] azonenberg d6b0fd8 - FFT decode now has proper gain/offset internally. Not settable, but allows better integration w/ glscopeclient UI.
<_whitenotifier-3>
[scopehal] azonenberg pushed 1 commit to master [+0/-0/±2] https://git.io/JJdQn
<_whitenotifier-3>
[scopehal] azonenberg closed issue #224: FFTDecoder: cache plan rather than recreating each waveform - https://git.io/JJMwu
<_whitenotifier-3>
[scopehal] azonenberg commented on issue #224: FFTDecoder: cache plan rather than recreating each waveform - https://git.io/JJdQl
<_whitenotifier-3>
[scopehal] azonenberg commented on issue #225: CTLEDecoder: cache FFT plan instead of recomputing each waveform - https://git.io/JJdQ0
<_whitenotifier-3>
[scopehal] azonenberg closed issue #225: CTLEDecoder: cache FFT plan instead of recomputing each waveform - https://git.io/JJMwz
<_whitenotifier-3>
[scopehal] azonenberg labeled issue #203: Rewrite header comments to say "libscopehal" and not "Antikernel" - https://git.io/JJ2fl
<azonenberg>
noopwafel: ping re scopehal:#175
<azonenberg>
did you get in touch with whitequark?
<_whitenotifier-3>
[scopehal-apps] azonenberg labeled issue #153: Rewrite header comments to say "scopehal-apps" and not "Antikernel" - https://git.io/JJ2f8
<_whitenotifier-3>
[scopehal-apps] azonenberg pushed 1 commit to master [+0/-0/±6] https://git.io/JJd7e
<_whitenotifier-3>
[scopehal-apps] azonenberg 0e0252d - Removed all special-casing for FFT waveforms. They now work just like any other analog signal
<_whitenotifier-3>
[scopehal-apps] azonenberg closed issue #147: Filename parameters to filters should have a "launch file chooser dialog" button next to the text box - https://git.io/JJRx7
<_whitenotifier-3>
[scopehal-apps] azonenberg commented on issue #147: Filename parameters to filters should have a "launch file chooser dialog" button next to the text box - https://git.io/JJddd
<azonenberg>
Nero_: so the sim finished. there's still some artifacts i don't fully understand (in both the old and new sims) which are not present in the measured VNA data
<azonenberg>
however, the new trace seems to have about the same shape as the old
<azonenberg>
... except about 1 GHz to the right
<azonenberg>
actually maybe even a bit more than 1 GHz
<azonenberg>
so basically my ~2.25 GHz probe might have just turned into a ~3.5 GHz probe
<azonenberg>
Re-running the sims now with finer meshing since i had some issues that i think are due to being too coarse
<azonenberg>
woo 46 GB ram with fine meshing
<azonenberg>
I'm a little terrified of how long this sim will actually take to run, lol. Coupling took 37 minutes, matrix fill took 38 mins, and now it's trying to actually solve said matrix
<azonenberg>
adding in some other steps i'm at well over an hour already and that's for one frequency. And it's still actually solving the matrix
<azonenberg>
and i'm doing two, one for the old layout and one for the new
<azonenberg>
wonder if it will be done by monday? :p
<azonenberg>
this is about the point where i wish for a rack of machines to run my sims on :p
<azonenberg>
and a couple dozen sonnet pro seats
bvernoux has joined #scopehal
<azonenberg>
bvernoux: so latest modeling is suggesting that i can turn my 2.25 GHz probe into a 3.5 GHz probe just by packing the resistors a little closer together
<bvernoux>
ha great
<bvernoux>
very big improvement just by packing resistor
<bvernoux>
on my side I have hacked the ST25R3916 ;)
<bvernoux>
and I can use it as unlimited sniffer ;)
<bvernoux>
it was very tricky to set the registers to some specific value on a undocumented mode ;)
<bvernoux>
done by brute force with oscilloscope ;)
<bvernoux>
and setting some registers by spi ;)
<miek>
azonenberg: of course the next evolution is putting every other resistor upside-down so you can mount them pad-to-pad :p
<miek>
bvernoux: nice :D
<bvernoux>
azonenberg, do you have checked also trace width change ?
<bvernoux>
I suspect it is the main issue
<bvernoux>
as the trace is smaller than res pad on photo you have posted
<bvernoux>
is trace was equal to res pad it shall not have such effect I think
<bvernoux>
anyway it is what you are doing when sticking pad together ...
Nero_ has quit [Read error: Connection reset by peer]
<bvernoux>
miek, haha the super sniffer mode allow allow to sniff Type A & Type B without needing reconfiguration and all in live ;)
<miek>
like the rtlsdr of rfid :p
<bvernoux>
yes but here it is better as it output digital demodulation ;)
<bvernoux>
but it can also output IF analog signal to do a SDR NFC ;)
<bvernoux>
haha and the fun is it can also decode Vicinity exchanges ;)
<bvernoux>
all in one with 1 configuration
<bvernoux>
totally amazing ;)
<bvernoux>
to be validated but it smell good at scope ;)