<_whitenotifier-3>
[starshipraider] azonenberg pushed 1 commit to master [+1/-0/±2] https://git.io/JvdvV
<_whitenotifier-3>
[starshipraider] azonenberg 35adc91 - Continued design review. Added soldermask apertures over 50 ohm traces.
<azonenberg>
ok so i'm almost done with the signoff review. Just have to triple check footprints for all of the not-board-proven components against the datasheet drawings
<azonenberg>
Just ordered components for one afe board
<azonenberg>
Hmmm
<azonenberg>
so apparently i *slightly* screwed up the stm32 footprint
<azonenberg>
as in the four rows of pads are each 50 μm further out from the center than they should be
<azonenberg>
which makes the whole footprint 100 μm too big
<azonenberg>
Debating whether this is worth fixing
<azonenberg>
i probably should
<azonenberg>
yeah and the corner pads are slightly off too
<azonenberg>
on the plus side, i'm clearly double checking in enough depth
<monochroma>
:D
futarisIRCcloud has quit [Quit: Connection closed for inactivity]
yourfate has quit [Remote host closed the connection]
yourfate has joined #scopehal
bvernoux has joined #scopehal
<_whitenotifier-3>
[starshipraider] azonenberg pushed 1 commit to master [+0/-0/±4] https://git.io/Jvdmg
<_whitenotifier-3>
[starshipraider] azonenberg 4802ffb - Final version of entry-afe-characterization-v0p1 as sent to fab
<azonenberg>
ok so i'm off to bed. tomorrow's project is to start work on the hmcad1520 board
<Degi>
Lol do 50 µm further out even matter for a one-off board?
<azonenberg>
Probably not. But i didn't want to consider that a board-proven footprint if it had known bugs
<Degi>
Hm okay
<azonenberg>
And 50um on an 0.5mm pitch QFN is a 10% error
<Degi>
I mean it wasnt sideways but outwards, right?
<azonenberg>
and the actual lands are less than 0.5mm wide so its more like a 15-20% error
<azonenberg>
Yes the error was in the radial direction
<azonenberg>
The corner pads with the mitering were also slightly misshapen
<Degi>
Welp
<azonenberg>
the miter was either too big or too small, i forget
<azonenberg>
anyway it's fixed now
<azonenberg>
it almost certainly would have worked
<azonenberg>
but the point of a final layout review is to catch as many bugs as possible
<Degi>
asdaNice
<azonenberg>
anyway i'm off to bed for real now :p
<Degi>
Hm yeh for the PCIe board I think I forgot to remove solder mask or something? lol
<Degi>
Good night :)
<Degi>
(Which I catched in design review of couse. Would be kinda bad if I didnt)
<azonenberg>
did you use my checklist?
<azonenberg>
if not, do you at least know it exists?
<Degi>
Hm I think I skimmed over it
<Degi>
But I didn't do a very fine check
<azonenberg>
the intent is for you to copy the checklist into your project's repo and check items off as you do the review
<Degi>
Hm oh that makes sense, neat
<Degi>
Hm kinda hope that the soldermask doesnt change the impedance by too much. On the other hand I have like 13 dB loss budget
<Degi>
Hope that I didn't forget to check the thing where they cut one edge with 45 ° angle though worst case that isnt really required and is mostly just a beauty feature
<bvernoux>
re
<bvernoux>
I have some news about my crystal issue ;)
<bvernoux>
with 10pF it work fine
<bvernoux>
Conclusion do not use formula from Application Note they are wrong ....
<bvernoux>
Test shall be done and margin shall be checked
<bvernoux>
I have capture the XTAL startup with Picoscope
<bvernoux>
now even with scope probe which add 10pF it start correctly
<bvernoux>
next test is to test with a cold spray -40°C to check startup
<Degi>
"Traceability markings (rev, date, name, etc) provided" Well whatever I think I didnt even label it hehe
<Degi>
bvernoux: Do you check the temperature?
<Degi>
Liquid butane can reach like -45 °C those sprays can probably get much colder
<bvernoux>
Temperature is my lab 24°C
<Degi>
I mean temperature of PCB when cooling
<bvernoux>
I shall test up to 80°C also ;)
<bvernoux>
so far I do not have the cold spray I'm buying it
<Degi>
Like butane has a boiling point of 0 °C heh
<bvernoux>
I plan also to do caracterization of the XTAL with my VNA ;)
<bvernoux>
just to check on few XTAL I have a real here I just need to solder pad and to connect it to the VNA
<bvernoux>
I remember azonenberg had similar issue with XTAL on Gigabit Ethernet but it was due to MEMS XTAL which are unstable so it is even harder to check ...
<bvernoux>
I will also do a check of accuracy with my Frequency Counter+GPSDO HP 53150A ;)
<Degi>
Hmh tbh I won't use MEMS ones in a lab enviroment unless they're very good sealed...
<bvernoux>
I have never tested any MEMS Oscillator so far
<bvernoux>
only standard XTAL, TCXO and also VCTCXO which are amazing
<Degi>
MEMS tend to leak helium and then stop working when there's too much helium
<bvernoux>
yes I saw that on an issue with iPhone ;)
<Degi>
Like for example in a low temperature lab where the cryo system leaks some helium
<bvernoux>
the phone does not start at all
<bvernoux>
so it is clear I will never use MEMs Oscillator with all issues there is ...
<Degi>
I bet there are space grade ones which dont have that problem xD
<Degi>
Like those ICs in glass ceramic packages
<bvernoux>
I'm not sure there is an MEMs Oscillator which are space grade as it is a too new technology ...
<Degi>
I mean just sealing it sufficiently should be enough
<bvernoux>
But who know the life time ;)
<Degi>
Like put a drop of lithium or some other small stuff onto it... Dont think that helium can leak thru that
<bvernoux>
Crystal is very well know but MEMs life especially when used as oscillator ...
<Degi>
Ugh I need to do homework but I wanna write more gateware...
somebodyelse has joined #scopehal
<somebodyelse>
any suggestions on which toolchain to use to try building scopehal on windows?
<monochroma>
if you are looking to run glscopeclient, it has not been built on windows yet (afaik)
<somebodyelse>
I understand, however the intent is to minimize any additional work due to wrong toolchain choice
somebodyelse92 has joined #scopehal
somebodyelse has quit [Ping timeout: 240 seconds]
somebodyelse92 is now known as somebodyelse
somebodyelse has quit [Ping timeout: 240 seconds]
<Degi>
Oof JLC wants 40 € more for a PCB I designed a long time ago but ordered now. I think ill just cancel it or so
<electronic_eel>
bvernoux: one additional idea about your crystals: did you check the capacitors you desoldered?
<electronic_eel>
are they really 27pF as you designed it or did they maybe swap them accidently?
<electronic_eel>
you could also try if you can reproduce the problem when soldering on known 27pf again
<awygle>
add a feedback resistor :p
<bvernoux>
yes I have not checked the old capacitor
<bvernoux>
The hard part is I do not have something accurate enough to measure 27pF ;)
<electronic_eel>
shouldn't a lcr meter be able to do this? I think mine can do it, might be inaccurate due to parasitics though
<electronic_eel>
the other thing you could try is solder in a known 27pf and check if it still fails or if it now works
<electronic_eel>
or solder in several caps from 10pf onwards and find the lowest number at which it still works
<sorear>
sealing anything against helium is a losing proposition
<bvernoux>
I do not have LCR meter but I plan to buy one
<bvernoux>
I will do additional test also replacing the expected 27pF by my own to be sure
<bvernoux>
but when I change the capacitor by 10pF I have a big margin
<electronic_eel>
multimeters often suck for measuring caps, they also usually can't measure esr. that is why I bought the lcr mostly for
<bvernoux>
I can add 2 probe 10pF and the crystal still work fine
<bvernoux>
only when i set probe to 10x => 100pF it fail but it is the same with older through hole XTAL
<electronic_eel>
that is what got me thinking about maybe wrong capacitors
<bvernoux>
I have checked the frequency with SDR with RF Probe also and it is quite stable
<bvernoux>
and checked the time the XTAL take to oscillate with at least 300mV amplitude
<bvernoux>
it take less than 700us
<bvernoux>
older one was a bit faster about 400us
<electronic_eel>
once the crystals has started oscillating it is not that sensitive anymore. most often it is just the startup that is affected
<bvernoux>
yes I tried the startup with 2x 10x probe too and it works ;)
<bvernoux>
even with USB DFU with BootRom which is the most sensitive
<bvernoux>
oups I said 10x -> 100pF it is 10pF it is 1x Probe which is 100pF ;)
<bvernoux>
also the amplitude on new XTAL is really higher
<bvernoux>
compared to old one through hole
<electronic_eel>
hmm, I'm not that into crystal tech, but the higher amplitude is maybe lower esr?
<electronic_eel>
also the allowed drive levels on small smd crystals are much lower, so you often have to add a series resistor not to overload it
<bvernoux>
I have a 220ohms resistor
<bvernoux>
like recommended by ST
<bvernoux>
and it seems good to have a good amplitude and nice shape
<bvernoux>
I have pushed the Picoscope capture here
<bvernoux>
old HydraBus with XTAL through hole is rev1_2
<bvernoux>
new one is Rev1_4 10pf
<electronic_eel>
the waveforms look reasonable. but it is hard to debug your problem with the 27pf with just the scope traces.
<bvernoux>
I will do test with cold spray and caracterize both Oscillator with my VNA to compare
<bvernoux>
But I have not found any procedure to check the margin except adding capacitor and testing with hot & cold temperature if the Crystal start correctly
<Degi>
Heh jlcpcb customer service is pretty kind and now I feel bad for responding 4 hours later...
<Degi>
For some reason black PCBs cost 40 € more now
<electronic_eel>
yes, I think margin testing is best done by desoldering/resoldering capacitors. although it might be a bit of work. that is the kind of work when solder tweezers come in handy
<electronic_eel>
Degi: they stopped making any other color other than green after chinese holidays. then they slowly started to allow more colors. but I think they want to discourage people from using them
<bvernoux>
most important is it is stable in cold/hot condition with a margin in pF
<bvernoux>
here I have about 20pF margin with 10pF capacitors C1/C2
<Degi>
Ok I told them to change order to green
<bvernoux>
compared to before with 27pF which are probably 30pF ...
<bvernoux>
Degi, JLCPCB will soon provide new service to choose PCB color with PCB Assembly and also components from other vendor not limited like today
<bvernoux>
It will be amazing as the price and quality is very good
<Degi>
Like other vendors than LCSC?
<bvernoux>
yes
<Degi>
This is my first PCB order ever lol
<Degi>
Hmm so you mean I don't need to solder BGAs myself?
<bvernoux>
yes in future ;)
<Degi>
Neat
<bvernoux>
But price will be higher too for components from outside ...
<bvernoux>
what is the most amazing is their service is ultra fast
<electronic_eel>
when JLC will get this process smooth they will ruin a lot of prototyping manufacturers in Europe
<bvernoux>
without asking for quote which is so slow with other PCB manufacturer
<bvernoux>
electronic_eel, they are already ruined ...
<bvernoux>
electronic_eel, in EU and even USA they are not competitive in time of price/speed ...
<bvernoux>
they are only competitive in term of quality today
<Degi>
Lol yeh
<bvernoux>
better in quality ;)
<bvernoux>
but price 10x more ...
<Degi>
I think jlc doesnt have filled vias for example
<Degi>
Like my PCIe adapter would be 150 € or so on oshpark
<bvernoux>
JLC have very basic features so far
<electronic_eel>
even without components from digikey/mouser - their smt service would be much better if they allowed all components from LCSC (not just some) and also not limit the number of "extended components" to 10
<bvernoux>
yes
<bvernoux>
main issue on LCSC there is no any good SMA ;)
<Degi>
What are extended components?
<bvernoux>
what a shame
<Degi>
LCSC also doesnt have ECP5
<bvernoux>
Degi, components not standard on their line so they ask additional $ to add the component
<bvernoux>
Also I can say they have genuine STM32 ;)
<bvernoux>
with price half of DigiKey ...
<electronic_eel>
bvernoux: soldering on a SMA by hand is easy, compared to big bgas or qfns...
<bvernoux>
electronic_eel, It is boring I prefer to let do everything to manufacturer and to use my time for other things that soldering tons of boards ;)
<azonenberg>
i dont get the qfn hate
<azonenberg>
i prefer qfn to, say, tqfp
<electronic_eel>
it is hard to check for soldering defects, like shorts from pin to exposed pad
<electronic_eel>
with qfp you see them in most cases
<bvernoux>
yes qfn with ground pad are sometime awful to check ;)
<bvernoux>
if you are not careful and add too much solder you are dead ;)
<bvernoux>
I remember with TRF7970A ...
<bvernoux>
they are small but really ugly to have a good yield for some prototypes
<bvernoux>
or it requires very good process ;)
<bvernoux>
my hacked plancha was not so bad but the reflow curve was quite random ;)
<bvernoux>
But it was only to produce proto for me I will never sell such things done like that
<electronic_eel>
yes, but when something with your prototype doesn't work - it can take a lot of time to debug if it is your circuit, your firmware or just a bad solder joint
<bvernoux>
exactly ;)
<bvernoux>
it is why electronic is not fun sometime ;)
<electronic_eel>
if a qfp doesn't work as expected, I can just gunk some flux on, move my iron over all 4 sides. takes less than a minute. then recheck. if it was a soldering defect, it is fixed in most cases
<bvernoux>
as you doubt in tons of things HW, SW ...
<bvernoux>
yes I like QFP/LQFP for that you see everything and it can be fixed
<bvernoux>
qfn and bga good luck for that ;)
<electronic_eel>
with a qfn I have to desolder, clean the pads, put on the right amount of solder and resolder with hot air. not under 5 mins.
<bvernoux>
anyone know good cheap LCR ;)
<bvernoux>
to measure pF, nH ;)
<Degi>
Transistor tester
<Degi>
Not very good but can measure pF
<Degi>
It can do surprisingly much for like 5 €
<bvernoux>
Smart Tweezers LCR-Reader Pro seems very nice
<bvernoux>
about 200Euros
<electronic_eel>
I have the keysight handheld one. works quite well. I think you can hack the higher models with a eeprom rewrite. I didn't do that because when I bought mine it wasn't known yet
<electronic_eel>
hack *to* the higher models
<Degi>
Hm I do wonder if when we put a DAC on the scope, whether that could be used as LCR meter
<Degi>
Or to measure diode speed hehe
<electronic_eel>
you need not dc signals (which the dac is designed for), but ac (around 100khz is common)
<electronic_eel>
and then get the phase info between the dac output and response on the scope
<Degi>
Hm but DC gives extra abilities
<Degi>
Like determining how shitty the X7R you just ordered is
<bvernoux>
KeySight LCR at 800Euros ;)
<bvernoux>
to check a capacitor 5 time per year ;)
<Degi>
LCR and multimeters miss nonlinear / DC offset capabilities tbh
<electronic_eel>
what? 800? no, mine was like 300 when I bought it
<Degi>
Like I dont think that many LCR meters can determine core losses and inductance in an inductor at X current (there are some that can add DC bias at least)
<Degi>
Hm something hookable to a one or two port circuit and then giving a schematic would be nice
<electronic_eel>
Degi: entry level model ;)
<Degi>
I mean if we add a DAC (we have 2 card slots, right?) when we're done with ADC design, that'd be pretty neat
<electronic_eel>
the dac alone isn't enough. what you are talking about is a function generator
<Degi>
Well what do we have an FPGA for?
<electronic_eel>
other than on the prototype board, you can't just break out the signal straight from the dac ic to the user
<Degi>
Oh I dont mean that one...
<Degi>
I mean a 1 GS/s 8 bit one or something on the same form factor as an ADC board
<electronic_eel>
yes, as I already wrote a few days ago, you'd need to build a analog frontend for that
<bvernoux>
Degi, hehe KeySight LCR @24KUSD ;)
<bvernoux>
Degi, such LCR are used to do the datasheet for resistor ...
<Degi>
Hm yes I might design one sometime...
<Degi>
bvernoux: Lol show me a datasheet of a resistor characterized with that... Sometimes you'd be lucky to find a temperature coefficient, not to speak of RF data lol
<Degi>
(Though some RF resistors sometimes have a rly nice datasheet)
<bvernoux>
Degi, RF resistor are ;)
<bvernoux>
Degi, there is some nice curve
<bvernoux>
but yes all other standard resistor have nothing
<Degi>
Hm is there a good way to somehow connect 50 ohm resistors where one side is grounded to a big plate together for a 2.4 kW 27 MHz RF load with 50 ohm output? Should I use transformers or can I just arrange them in 2x2 and hope that stray capacitances aren't too bad?
<azonenberg>
and the test setup for that stupidly fast risetime measurement. Foot-long semirigid coax right to the scope lol
<Degi>
Lol is that solid coax
<Degi>
Neat
<azonenberg>
practical for probing? absolutely not
<Degi>
How often can you use that coax?
<azonenberg>
but it was the highest bandwidth cable i had :p
<azonenberg>
Not sure. I haven't observed degradation yet
<azonenberg>
but i try only to use it when i really want a superb cable
<azonenberg>
My minicircuits FL086-24SM+ cable is almost as good
<azonenberg>
and much more flexible
<Degi>
Hm I found a nice 8 1/2 digit DMM but it costs 10k lol
<azonenberg>
I'm happy with my 5 3/4 digit R&S HMC8012 right now
<azonenberg>
Re s-parameters, that is literally the purpose of the probe characterization board
<azonenberg>
a nice solid fixture i can mate the probe to and do vna plots
<azonenberg>
I heard back from my sales rep at the fab and it's shipping this week
<azonenberg>
so probably end of week to weekend i should have boards in hand
<Degi>
Neat
<Degi>
Hm yes R&S... heard good thingsabout that
<Degi>
(The spectrum analyzer I'd have access to when lockdown is over is from them)
<azonenberg>
Same as my power supplies, HMC8042
<azonenberg>
both are Hameg but now branded R&S
<azonenberg>
(R&S bought Hameg, these designs have R&S branding so were built after the buyout, but are still Hameg PCB/firmware)
<miek>
an 8.5 digit DMM is a big rabbithole if you want to maintain it properly
<Degi>
I have a HM312 scope
<Degi>
And a broken TDS520 :/
<miek>
what's broken on it?
<Degi>
Now it refuses to turn on and the relays keep making noises
<adamgreig>
azonenberg: is it deliberate that glscopeclient holds the left-most timestamp constant when zooming? it keeps tripping me up
<adamgreig>
i would have hoped for like 'zoom holding current mouse position constant' or perhaps expected 'holding centre of timeline constant'
<azonenberg>
adamgreig: zoom to mouse position is the intent
<azonenberg>
there may or may not be a ticket for this
<azonenberg>
but fixing it is on the TODO
<azonenberg>
basically, internally glscopeclient uses the x axis offset as the means for setting the point in time
<azonenberg>
so if you only change the scale factor, that's what happens
<azonenberg>
i need to calculate the time point under the cursor, change the scale, and adjust the offset such that the same time point is still under the cursor
<azonenberg>
if you want to try fixing it, WaveformArea::on_scroll_event and OscilloscopeWindow::OnZoom*Horizontal are the functions of interest
<adamgreig>
rgr, cool
<adamgreig>
will have a poke after i resolve my crisis of confidence in swd timing diagrams
<azonenberg>
you'll probably want to convert ev->x to a time offset in on_scroll_event then pass that onto OnZoom*Horizontal as an additional argument
<azonenberg>
or maybe just pass the x coordinate in