<awygle>
azonenberg: do you happen to know if the RX clock from an SGMII PHY will be synchronized to the TX clock it recovers from the TX pair?
Degi_ has joined #scopehal
Degi has quit [Ping timeout: 264 seconds]
Degi_ is now known as Degi
bvernoux has quit [Ping timeout: 250 seconds]
<azonenberg>
awygle: implementation defined
<azonenberg>
Depends on how the rx cdr works etc
<azonenberg>
i don't think the spec says either way
<azonenberg>
there may well be a clock domain crossing fifo somewhere in the rx path in which case all bets are off
<awygle>
lame. Thanks
<awygle>
The spec is some Cisco thing right? It's not in 802.3?
<azonenberg>
Yes. SGMII is not an IEEE standard although the spec is open and multiple vendors make compatible hardware
<azonenberg>
the spec i have is rev 1.7, Cisco ENG-46158
<azonenberg>
i believe that is the most current version
<azonenberg>
There is also QSGMII, which is a... broadcom or cisco, i forget, extension that runs four lanes of 1g ethernet over a single 5G serdes lane
<azonenberg>
i've never seen a PHY that supports it sold openly in qty 1 with no NDAs etc. So i've never tried to get the spec
<azonenberg>
then on the opposite direction there's XAUI, which is 4 lanes of serial -> 1 lane of 10G (IEEE standard)
<azonenberg>
and RXAUI, a proprietary extension of 2 lanes @ 5 Gbps -> 1 lane of 10G
<azonenberg>
this would be awesome because i could make more efficient use of artix7 GTPs, but i havent found a part supporting it that i can actually get / get docs for
<azonenberg>
AFE board shipped
<azonenberg>
current ETA tuesday
bvernoux has joined #scopehal
bvernoux has quit [Quit: Leaving]
<Degi>
Neat
<Degi>
Woo my PCBs got shipped today
<Degi>
Does "Shipment not arrived; customs status updated" mean anything
<Degi>
Oof the ETA is April 21 in Düsseldorf and then it needs to get shipped to my place from there. I guess flight quantity has been reduced? I mean I selected the cheapest shipping too...
<azonenberg>
i've seen it before, not sure of the exact meaning
<azonenberg>
my guess is, it means that the shipment is still in customs and DHL doesn't have it physically
<Degi>
Well in hong kong heh
<Degi>
Wow it took them 2 minutes to get it through the shenzhen facility
<azonenberg>
lol probably stayed on the same vehicle and just got scanned as "about to leave"
<azonenberg>
also i have the case for BLONDEL now, assembling it to see how things look/feel
<Degi>
Hm I think sometime I may wanna design a handheld mainboard for this ADC/AFE board... I mean you could even use that as a SDR to listen to radio heh
<azonenberg>
the enclosure is made of aluminum extrusions with separate top and bottom panels. This means it's fairly easy to remove a panel in order to machine cutouts into it, etc
<azonenberg>
however, it also assembles with self-tapping screws that need careful cleaning to avoid leaving metal shavings that can short things
<azonenberg>
it also uses cage nuts for holding the top/bottom panel on, and the screws stick up a decent bit
<azonenberg>
None of these are fatal problems, just things to consider in the mechanical engineering of the system
<azonenberg>
we also need to decide where to put the vent holes (front or back)
<azonenberg>
the case is otherwise fully symmetric
<azonenberg>
top/bottom and front/back are interchangeable
<azonenberg>
and we need to consider that we can't put standoffs in the vent hole field - or if we do they have to either be tiny and between holes, or large and threaded into one of the vents
<azonenberg>
again, not serious problems but things that need consideration when doing final layout
<azonenberg>
So this is an important design consideration, when assembling these cases we need to be sure all shavings are removed after screws are put in
<Degi>
Self tapping as in there was an unthreaded hole already there? Or like wood screws?
<azonenberg>
There was an unthreaded hole all the way through the extrusion
<azonenberg>
it cuts threads into that as you tighten it
<azonenberg>
Which makes shavings
<azonenberg>
Nothing inherently wrong with it, but it does mean you need to clean after
<azonenberg>
The case comes with rack ears which i didn't bother installing at this point since i'm going to have to dismantle the case in order to cut holes for connectors etc
<Degi>
I think we should rotate one of the plates, so that there are holes at the front and at the back
<azonenberg>
The bottom plate matters more
<azonenberg>
because that's what stuff is mounted to
<azonenberg>
We can position the top pretty arbitrarily
<Degi>
Hmm
<azonenberg>
we also need to figure out if we want to add any vent holes on the rear panel
<Degi>
I mean the AFE's are screwed to the front plate with SMAs anyways right? I'd put the bottom vent holes on the front
<azonenberg>
one possibility is to add a few fans to the rear of the case
<Degi>
Hm if this is mounted in a rack, is it only supported by the ears?
<azonenberg>
so air will be drawn in from top and bottom, over the boards, then exhausted on the rear
<azonenberg>
Depends on if you install rails or not
<azonenberg>
i intend for this to be used in a 2-post rack supported only by ears on the front
<Degi>
Ahh rails
<azonenberg>
as that's what i have on my bench now
<azonenberg>
but if you had a 4-post rack you could certainly put it on rails
<azonenberg>
8 inches is a very shallow case
<azonenberg>
i have zero concern about using the ears as the sole mounting
<Degi>
Hmm is there a chance of up/down vents being blocked by other rackmoutn devices?
<azonenberg>
Yes. There's a few ways around that
<azonenberg>
one is to drill our own vent holes in the sides, and possibly even buy the variant of this case without vented top/bottom panels
<Degi>
Maybe add vent holes on the front panel and a fan on the back panel
<azonenberg>
Another is to do what R&S did with my power supplies, and simply include a requirement in the manual
<azonenberg>
that states you shall leave 1U of free space above the unit
<Degi>
Hm yeah
<Degi>
I have a PSU like that too, though not rackmount
<azonenberg>
These are things to think about, but we can't make final decisions yet
<azonenberg>
once we have more of the pcb design done and have a power budget for the whole system, we can figure that out
<azonenberg>
it's entirely possible a fanless design will be viable
<Degi>
Hmm
<Degi>
We need to consider temperature drift too
<Degi>
Like if you need to wait an hour for it to warm up till everything is stable
<azonenberg>
Yes. We will want multiple internal temp sensors so we can measure drift as temp changes
<Degi>
Ideally temp could be held constant with fans
<azonenberg>
Yeah. Variable speed fans on a relatively low heat design would probably work well
<azonenberg>
but again, details TBD. I want to minimize use of fans for noise reasons
<Degi>
Just stick sensors on ADC and AFE...
<Degi>
Yeah
<azonenberg>
so if we can pwm them really low, that would be nice
<azonenberg>
Having them and only turning them on if things get hot, etc
<Degi>
Well there are quieter and louder fans too, usually a question of size and price...
<azonenberg>
IME 40mm fans are loud in general
<Degi>
We could use those centrifugal fans or the ones in laptops
<Degi>
Though that would require some 3D printed parts for airflow routing
<azonenberg>
no i want to use an axial fan like 1U servers do. just keep it at low rpm or even off as much as possible
<Degi>
Hm ok
<Degi>
Still there should be fans which are louder and quieter for a given airflow, depending on bearing, blade shape and position etc
<azonenberg>
yes, correct
<Degi>
Ion fans haha
<Degi>
Ugh why does mouser use CFM...
<Degi>
Well there are 40 mm fans with as low as 10 dbA noise which give sufficient airflow (2 L/s)
<azonenberg>
meanwhile my nexus 3064 spits out 71 dBA
<Degi>
Yeah that's like a million times louder
<Degi>
Do two fans increase the dbA by 3 too?
<Degi>
Like if I have a million 10 dbA fans I get approx 70 dbA?
<azonenberg>
I'm not sure. it probably depends on a lot of factors like enclosure design etc. also depends on how far away the measurement is taken
<azonenberg>
wrt size of enclosure
<azonenberg>
these are not point, isotropic radiators
<azonenberg>
my point is if you just search fans by audio volume on digikey
<Degi>
But that fan doesnt support PWM, though the MFG can be contacted for that option
<azonenberg>
it might not be a meaningful sort
<Degi>
Hm es
<azonenberg>
unless they all measure at the same distance in the same direction
<azonenberg>
ideally i would like both tacho and fan capability
<azonenberg>
tacho and pwm*
<Degi>
Heh I have some server fans here, they take like 1.4 A and have 8 pins, which sound like a mini jet engine, probably louder
<azonenberg>
have you seen the little radio transcript i tweeted a while ago about my cisco switch?
<monochroma>
i like the double stacked 1U server fans
<Degi>
radio transcript?
<miek>
the hotswap ones are fun, take them out and they're still spinning a minute later
<azonenberg>
"Good morning Datacenter Tower, this is Nexus 3064 ready for pushback"
<Degi>
Yes its a double fan with 4 pins for one fan, the others for the other
<Degi>
Lol
<azonenberg>
"N3064, pushback approved, push towards rear of rack"
<azonenberg>
"Nexus 3064 requesting takeoff clearance as filed"
<azonenberg>
"N3064, cleared for takeoff. Climb and maintain 12U, expect 40U 5 minutes after departure"
<azonenberg>
why do you think these things take so long to boot? :p
<Degi>
Mine takes a few minutes till bios...
<Degi>
I think the fans use a 100 W or so
<azonenberg>
100W of *fans*?
<azonenberg>
what is this thing
<Degi>
Well 8x server fans
<Degi>
It uses 140 W in total before BIOS
<Degi>
Ancient tech, from 2010 or so
<azonenberg>
because i think my 3064 is only maybe 200W
<Degi>
It uses 500 W full load
<azonenberg>
ok here we go, with SR optics at 100% load, 177W
<Degi>
Well its a dual CPU server with CPUs from 2010 what do you expect lol
<azonenberg>
Lol fair enough
<Degi>
My whole PC area uses like 300-400 W full load with GPU load
<azonenberg>
the 3064 is a relatively recent high end cisco datacenter-grade switch meant for top of rack applications etc
<azonenberg>
It's no longer sold, but not end of life (still software support available)
<azonenberg>
i got mine on ebay for like $500
<Degi>
What if you put it on the botton of the rack
<Degi>
I got mine for free from the physics department
<azonenberg>
lol
<azonenberg>
i have mind about the middle actually
<azonenberg>
i have patch panels at the top terminating all of the cabling going to wall ports elsewhere in the building
<azonenberg>
then switches in the middle and servers at the bottom
<Degi>
Hm where to buy racks? I think I want one
<azonenberg>
just one? :p
<Degi>
Well I could fit a 48U here
<azonenberg>
i have owned (not all at the same time) a total of... seven
<Degi>
Inb4 stack of IKEA tables
<azonenberg>
there was a sgi origin and a 24u 2-post i gave to monochroma, an old DEC 4-post with weird dimensions i gave to rqou years ago
<azonenberg>
then the two 45U 4-post that i have my core infrastructure on. One is switches, patch panels, and servers
<azonenberg>
the other is my UPS and a bunch of FPGA devkits plus a little x86 box that serves as a head end for all of the jtag dongles plugged into said devkits
<azonenberg>
then i have two 16U 2-post racks on the soldering/test equipment bench with my scopes etc
<Degi>
Hm my room is only 307 cm... Only space for 60 U
<Degi>
heh
<azonenberg>
you want clearance on top, not touching the ceiling lol
<Degi>
Mom would be pissed
<azonenberg>
Lol
<azonenberg>
in my case i have 12 inch wide cable trays hanging about 12 inches below the ceiling running just above the top of the racks
<Degi>
Like she already is but if there was suddenly a 3 m high server rack lol
<azonenberg>
lol
<Degi>
I kinda find it funny how I can configure things in the BIOS like air pressure, outside temperature, distance between RAMs etc
<Degi>
Hmm I managed to dissolve copper in isopropyl alcohol...
<azonenberg>
wut
<Degi>
Yeah like ??? but cool
<Degi>
A while ago I dissolved copper in acetone and apparently thats soluble in IPA too
<Degi>
Like pure IPA, not the 70% stuff
<azonenberg>
i clean boards in 99.8% IPA
<Degi>
I think the solution is halogen free
<azonenberg>
or is it 99.9? it's USP and ACS reagent grade
<Degi>
For example CuCl2 dissolves in acetone, but I think I made this one from acetic acid and rosin
<Degi>
Huh I have some 99% cheap stuff
<Degi>
And 70% apothecary grade stuff
<azonenberg>
never seen it dissolve copper
<Degi>
*pharmacy
<azonenberg>
yeah this is actual lab grade solvent
<Degi>
I think I dissolved copper in acetic acid, added rosin, heated it etc
<sorear>
ipa is a much weaker acid than water
<Degi>
Actually I should try extracting the CuO2 from the NaOH thingie...
<bvernoux>
azonenberg, what do you think about the measurements ?
<azonenberg>
you saw the layout i ordered right?
<bvernoux>
yes I think I have seen the latest version
<bvernoux>
we see there is a loss of about 2.8dB at 6GHz
<azonenberg>
has short, open, 50 ohm load, two different length through lines, then probe connections both across a 50 ohm resistor and just at the end of a stub
<bvernoux>
need to compute what is in theory
<bvernoux>
with FR408
<bvernoux>
and the size of the line
<azonenberg>
i believe oshpark uses FR408HR, not FR408
<bvernoux>
yes FR408HR IIRC let me check
<azonenberg>
The smith chart is greek to me, i can't read them
<azonenberg>
Pico TA345 standard with NIST traceable calibration
<azonenberg>
2x TA336 N-SMA test leads
<bvernoux>
0.9Nm on yours
<azonenberg>
?
<bvernoux>
mine is 2 time less
<bvernoux>
0.45Nm
<bvernoux>
it is recommended for SMA
<azonenberg>
5 inch-pounds is the recommendation for brass smas
<azonenberg>
8 for steel
<azonenberg>
Oh, and i also ordered a pair of Rosenberger 53S101-K00N5 N connecter M-F couplers
<bvernoux>
azonenberg, does PicoVNA requires 0.9Nm for the Torque ?
<bvernoux>
strange
<bvernoux>
it is a lost
<bvernoux>
for the connectors
<azonenberg>
gonna stick them over the VNA ports to minimize load cycles on the VNA Ns
<azonenberg>
bvernoux: the pico cal kit has stainless SMAs on it
<bvernoux>
ha ok
<azonenberg>
all of my DUTs have brass, so i got a torque wrench for the torque you use on brass
<azonenberg>
it will be slightly light for steel, but reproducible which should be OK
<azonenberg>
reproducibly light is better than too tight
<bvernoux>
yes clearly
<bvernoux>
IIRc it is recommended 0.45Nm for reproductability in all case and for safety too
<azonenberg>
0.57 is what i see recommended
<bvernoux>
especially when we mix type of connectors ... brass/steel
<azonenberg>
thats the upper end of thr brass torque range
<bvernoux>
because at 0.45Nm it is already strong
<bvernoux>
i see when the wrench do the "click"
<bvernoux>
woo you have TA336 N-SMA
<bvernoux>
they are expensive but very nice
<bvernoux>
on my side I have quite "cheap" SMA cable but from minicircuits
<azonenberg>
you've used them? or just going by specs
<bvernoux>
so they are not the worst
<azonenberg>
I have a bunch of FL086-24SM+
<azonenberg>
that is hardly a "cheap" cable lol
<bvernoux>
the best SMA cable I have is FLC-2FT-SMSM+ from minicircuits
<azonenberg>
I have two different 18 GHz rated cable types
<bvernoux>
it cost > 100USD
<azonenberg>
i have those, and i have a bunch of Crystek CCSMA-MM-086 series
<bvernoux>
ha nice
<azonenberg>
2.18mm hand formable semi-rigid 18 GHz
<bvernoux>
I will by some rigid Crystek they are quite cheap
<azonenberg>
then i have a lot of lower spec cables... rg174tpr, rg188, rg316
<bvernoux>
anyway you will received a nice VNA soon ;)
<bvernoux>
receive
<bvernoux>
fro the price there is nothing better than the PicoVNA today
<azonenberg>
yeah
<azonenberg>
i was hoping darrell would get the harmon instruments VNA ready in time but i couldn't wait any longer
<Degi>
Can it have independent transmit and receive frequencies?
<azonenberg>
i may still buy one of his 20 GHz ones when it comes out
<bvernoux>
azonenberg, yes Harmon VNA are not ready so far they shall be ready in 1 or 2 years
<bvernoux>
me too I'm very interested by the 20GHz 2 Port version if the price is similar to PicoVNA or a bit cheaper ;)
<azonenberg>
i might even want the 6G just for proper open source tools etc
<bvernoux>
as so far my old HP VNA is ultra nice for the price but so loudy and a bit slow... example S2P 801pts take 37s using my own tool to retrieve the data using internal format
<azonenberg>
wow
<bvernoux>
I have quite reversed the format ;)
<bvernoux>
but there is about 10s gain vs Float32 Bin format
<bvernoux>
as the MCU inside the VNA is slow a MC6040 ;)
<bvernoux>
running at 25MHz ;)
<bvernoux>
68040
<azonenberg>
lol
<bvernoux>
similar to what I had on my Amiga(s) computed in old days ;)
<bvernoux>
computer
<azonenberg>
lol ok you were before my time
<azonenberg>
my first computer was a 386 running win 3.1
<azonenberg>
and ms-dos 5
<azonenberg>
it would boot up to a dos prompt and you had to run "win" to start window
<bvernoux>
I started with Thomson TO7-70 then a TO8-D ;)
<azonenberg>
well ok
<azonenberg>
i also had a smith-corona pwp-80
<azonenberg>
but not a GP computer
<bvernoux>
ha never heard about this one
<azonenberg>
80x25 or so monochrome LCD, keyboard, and daisywheel printer
<bvernoux>
from 1988 ok
<azonenberg>
it was an electric typewriter that was able to save documents to a proprietary magnetic disk
<azonenberg>
and could edit stuff in ram or print as you typed
<bvernoux>
ha yes interesting
<azonenberg>
i still remember now, decades later, it would show a counter of ram in the upper right corner
<azonenberg>
50818 bytes free when it turned on
<azonenberg>
presumably out of 64k total
<azonenberg>
each keystroke as you typed it would go down
<bvernoux>
fun I remember somthing like that at school too
<bvernoux>
Tomorrow I will train again to solder SMA with Hot Air to have better results ;)
<bvernoux>
As I have a big TRL Board to solder with at least 8 connectors to solder ...
<Degi>
Hm maybe preheating the board woul help
<bvernoux>
without including Series & Shunt Fixture which are nice to test resistor/capacitor/inductor ...
<bvernoux>
Degi, my main issue so far is to fix correctly the board with the SMA connector
<bvernoux>
my AOYUE 618 is not perfect for that ;)
<Degi>
Hmm glue
<bvernoux>
yes I need to check where to buy glue which resist at > 220°C ;)
<Degi>
Silicone
<azonenberg>
or epoxy
<bvernoux>
do you have a link ?
<Degi>
Sodium waterglass
<Degi>
No but I have tested silicone to 450 °C before
<bvernoux>
ha great
<Degi>
It smokes a little bit, that's probably the anti mold agent and water tbh
<Degi>
The silicone itself seemed fien
<azonenberg>
bvernoux: what are you trying to glue exactly?
<Degi>
I suggested to glue the SMAs
<azonenberg>
i like chipquik ad1-10s for stuff that i will be soldering
<bvernoux>
azonenberg, my SMA connector to be perfectly aligned before to solder them with hot air
<azonenberg>
it's a thermoset epoxy that cures in 2 mins at 150C
<Degi>
Hmm would aluminium oxide PCBs be cool?
<azonenberg>
i normally use it for bottom-side large connectors
<azonenberg>
Degi: they exist
<azonenberg>
it's one of the more common ceramics for RF stuff etc
<Degi>
Heh I see
<bvernoux>
if you have any reference at DigiKey/Farnell it will be great ;)
<Degi>
Because I figured out how to make thin stable substrates with high alumina content, need to figure out how to make them thicker
<bvernoux>
it is definitely what I need to do something clean
<azonenberg>
this is a thermoset, so it won't cure at room temperature
<Degi>
Or fully hermetic hybrid alumina silica devices for a low cost?
<azonenberg>
i also suggest repackaging into a smaller syringe as the supplied one is pretty large
<azonenberg>
and using a smaller needle than it comes with
<Degi>
Hm wish I could buy the elegoo mars pro now, but somehow it isnt available anywhere...
<Degi>
Then I could try to make silicone molds for alumina stuff, though I first need to figure out how to make it cure when wet and dry later, because drying while wet causes bubbles when its thicker than a mm
<bvernoux>
hmm I do not find it at Farnell
<Degi>
Hmm buck converter makes weird noises and outputs more volts when loaded but device works now...
<Degi>
Is the red LED on a ESP8266 supposed to flicker?
<Degi>
Wow I can hear the ESP working now
<Degi>
Okay with 10 V on the input it works fine, no noise
<azonenberg>
So i'm trying to design the enclosure for the new probe pcb
<azonenberg>
Making it fit the Pico positioner will be... nontrivial
<azonenberg>
the pico passive probe positioner*
<azonenberg>
i think i can make it fit a Tetris positioner
<azonenberg>
The problem is, the tetris positioner is not nearly as readily availab
<azonenberg>
available*
<azonenberg>
i'd have to buy direct from PMK, which would cost more. So i probably couldn't include a positioner in the student price package
<azonenberg>
yeah it will fit a tetris positioner easily
<azonenberg>
although i will be changing materials
<azonenberg>
The glass filled SLS nylon i used in the original prototypes has a min wall thickness of 0.75 mm
<azonenberg>
and i need 0.6 mm walls in the area the tetris positioner grabs the probe
<azonenberg>
so i'm going to have to switch to pure nylon with no glass filling
<azonenberg>
Get your nicotine fix in 50Ω matched clarity with the new 2 GHz vape pen from Antikernel Labs!
<lain>
XD
<azonenberg>
seriously, you can't tell me it doesn't look like a vape
<azonenberg>
there's one model with an off center mouthpiece i've seen a lot of people at hacker cons smoking
balrog has quit [Quit: Bye]
<Degi>
Lol
<Degi>
Double mothpiece
<Degi>
*mouth
<azonenberg>
anyway what do you think of the new shell design?
<Degi>
Maybe flatten the corners a bit
<Degi>
Hmm is it possible to stack the resistors face by face? That could improve performance
<azonenberg>
what do you mean "face by face"
<Degi>
Like the silvered faces of the resistor touching
<azonenberg>
they're flip chip, they don't have wraparound terminals
<Degi>
Ah yes the graphics is a bit confusing
<azonenberg>
there needs to be a small gap for placement and to allow room for thermal expansion during reflow etc
<Degi>
Still, as near as possible to eachother
<azonenberg>
the cad render uses a standard 0402 footprint
<azonenberg>
i only have like 500 μm of gap
<Degi>
Hm but that gap is like 1 0402 by the looks (though its sideways so I cant rly judge)
<Degi>
Oh
<Degi>
Okay that should be fine nvm
balrog has joined #scopehal
<azonenberg>
i have basically a continuous line of 0.5mm pitch pads
<azonenberg>
0.5mm square*
<azonenberg>
with resistors between every 2
<Degi>
So its basically a continuous resistor strip heh neat
<azonenberg>
well its alternating resistor and short pieces of coplanar waveguide
<azonenberg>
but basically yeah its as close as i reasonably can without violating clearance requirements for placement etc
<azonenberg>
anyway so at this point i just have to spend a little time making tweaks, chamfers, etc to make it look nice, maybe some kind of end stop so the board can't fall out the front before i glue it in place
<azonenberg>
possibly add a text logo engraved/embossed, although i am considering a stick-on label for that instead
<azonenberg>
i need space for a serial number anyway
<Degi>
Hm maybe make a clip in front piece so that no glue is needed
<azonenberg>
my current ones actually friction fit quite well
<azonenberg>
none of them are glued
<azonenberg>
i'd just add a drop or two of epoxy to make sure nothing works loose